Healthcare & Biomedicine Informatics Handbook

Information Science Reference
BOOKS BOOK SERIES JOURNALS PROCEEDINGS TEACHING CASES PAY-PER- REFERENCE E-RESOURCES ABOUT IGI
VIEW
BECOME AN AUTHOR/EDITOR | MAILING LIST | HOW TO ORDER | LIBRARY SUGGESTION | EXAMINATION REQUESTS/COURSE ADOPTION | DISTRIBUTORS
IGI Online Bookstore Reference Exact Search
About This Book

Handbook of Research on Informatics in Healthcare and
• Description &
Biomedicine
Key Features
• Topics Covered Edited By: Athina A. Lazakidou, University of Peloponnese, Greece
• Accolades
Table of Contents:
• Contributors
• Editorial Advisory
Board SECTION I: MEDICAL DATA AND HEALTH INFORMATION SYSTEMS
• Table of Contents Chapter I

Electronic Health Records / Olga Galani and Ageliki Nikiforou
• Preface
ISBN: 1-59140-982-9
• About the Editors Chapter II
Security in Health Information Systems / Christina Ilioudi and Athina A. Lazakidou Hard Cover
• View the Brochure Publisher: Information
Chapter III Science Reference
• View the Excerpt
Development of a Health Information System in a Post-Communist Country / Ranko Stevanovic, Copyright: 2006
Ivan Pristas, Ana Ivicevic Uhernik, and Arsen Stanic
Pages: 479
• Reference Home Page
Chapter IV List Price: $215.00
• Recommend to your Computerized Systems Validation in the Pharmaceutical Industry / Kashif Hussain, Shazia Yasin add to cart
Library Mughal, and Sylvie Leleu-Merviel
Perpetual Access:
Authoritative - Chapter V
Innovative - $325.00
Chronic Disease Registers in Primary Healthcare / M. F. Harris, D. Penn, J. Taggart, Andrew
Comprehensive Georgiou, J. Burns, and G. Powell Davies add to cart
SECTION II: STANDARDIZATION AND CLASSIFICATION SYSTEMS IN MEDICINE Print + Perpetual Access:
$430.00
Chapter VI
Standardization in Health and Medical Informatics / Josipa Kern add to cart
http://www.igi-global.com/reference/details.asp?ID=5571&v=tableOfContents (1 of 5)7/4/2009 11:15:46 πμ

Chapter VII
Basic Principles and Benefits of Various Classification Systems in Health / Dimitra Petroudi and
Athanasios Zekios
SECTION III: VIRTUAL REALITY APPLICATIONS IN MEDICINE
Chapter VIII
Virtual Reality in Medicine / Theodoros N. Arvanitis
Chapter IX
Modelling and Simulation of Biological Systems / George I. Mihalas
Chapter X
Virtual Reality Simulation in Human Applied Kinetics and Ergo Physiology / Bill Ag. Drougas
SECTION IV: VIRTUAL LEARNING ENVIRONMENTS IN HEALTHCARE AND BIOMEDICINE
Chapter XI
Care2x in Medical Informatics Education / Andreas Holzinger, Harald Burgsteiner, and Helfrid
Maresch
Chapter XII
An Object-Oriented Approach to Manage E-Learning Content Using Learning Objects / Andreas
Holzinger, Josef Smolle, and Gilbert Reibnegger
Chapter XIII
Motivating Healthcare Students in Using ICTs / Boštjan Žvanut
Chapter XIV
The User Agent Architecture and E-Learning in Healthcare and Social Care / Konstantinos M.
Siassiakos, Stefanos E. Papastefanatos, and Athina A. Lazakidou
Chapter XV
E-Learning in Healthcare and Social Care / Maria Kalogeropoulou, Maria Bastaki, and Polyxeni
Magoulia
Chapter XVI
Potential Benefits and Challenges of Computer-Based Learning in Health / Athina A. Lazakidou,
Christina Ilioudi, and Andriani Daskalaki
SECTION V: COMPUTER-ASSISTED DIAGNOSIS
Chapter XVII
Brain Mapping in Functional Neurosurgery / George Zouridakis, Javier Diaz, and Farhan Baluch
Chapter XVIII
ECG Diagnosis Using Decision Support Systems / Themis P. Exarchos, Costas Papaloukas, Markos G.
Tsipouras, Yorgos Goletsis, Dimitrios I. Fotiadis, and Lampros K. Michalis
SECTION VI: DATA MINING AND MEDICAL DECISION MAKING

Chapter XIX
Information Processing in Clinical Decision Making / Vitali Sintchenko
Chapter XX
Data Mining Techniques and Medical Decision Making for Urological Dysfunction / N. Sriraam, V.
Natasha, and H. Kaur
Chapter XXI
Spline Fitting / Michael Wodny
Chapter XXII
Parameter Estimation / Karl-Ernst Biebler
Chapter XXIII
The Method of Least Squares / Bernd Jaeger
SECTION VII: CURRENT ASPECTS OF KNOWLEDGE MANAGEMENT IN MEDICINE
Chapter XXIV
The Data-Information-Knowledge Model / Andrew Georgiou
Chapter XXV
Goals and Benefits of Knowledge Management in Healthcare / Odysseas Hirakis and Spyros
Karakounos
Chapter XXVI
Knowledge Management in Medicine / Nikolaos Giannakakis and Efstratios Poravas
Chapter XXVII
Knowledge Management in Telemedicine / Jayanth G. Paraki
SECTION VIII: TELEMEDICINE AND E-HEALTH SERVICES
Chapter XXVIII
Use of Telemedicine Systems and Devices for Patient Monitoring / Dionisia Damigou, Fotini
Kalogirou, and Georgios Zarras
Chapter XXIX
Current Telehealth Applications in Telemedicine / Georgios Economopoulos
Chapter XXX
Mobile Telemonitoring Insights / Pantelis Angelidis
Chapter XXXI
Telepathology and Digital Pathology / Vincenzo Della Mea
Chapter XXXII
Collaborative Environments for the Health Monitoring of Chronically Ill Children / G. Ganiatsas, K.

Starida, and Dimitrios I. Fotiadis
Chapter XXXIII
Electronic Submission of New Drugs in Europe / A. Susanne Esslinger and Daniela Marschall
Chapter XXXIV
Semantic Web Services for Healthcare / Christina Catley, Monique Frize, and Dorina Petriu
SECTION IX: IMAGE PROCESSING AND ARCHIVING SYSTEMS
Chapter XXXV
Imaging Technologies and their Applications in Biomedicine and Bioengineering / Nikolaos
Giannakakis and Efstratios Poravas
Chapter XXXVI
Medical Image Compression Using Integer Wavelet Transforms / B. Ramakrishnan and N. Sriraam
Chapter XXXVII
Three Dimensional Medical Images / Efstratios Poravas, Nikolaos Giannakakis, and Dimitra Petroudi
Chapter XXXVIII
Imaging the Human Brain with Magnetoencephalography / Dimitrios Pantazis and Richard M. Leahy
Chapter XXXIX
Region of Interest Coding in Medical Images / Sharath T. Chandrashekar and Gomata L. Varanasi
Chapter XL
Imaging the Human Brain with Functional CT Imaging / Sotirios Bisdas and Tong San Koh
SECTION X: SIGNAL PROCESSING TECHNIQUES
Chapter XLI
Nonlinear Signal Processing Techniques Applied to EEG Measurements / Christos L. Papadelis,
Chrysoula Kourtidou-Papadeli, Panagiotis D. Bamidis, and Nicos Maglaveras
SECTION XI: USE OF NEW TECHNOLOGIES IN BIOMEDICINE
Chapter XLII
Medical and Biomedical Devices for Clinical Use / Evangelos K. Doumouchtsis
Chapter XLIII
Artificial Intelligence in Medicine and Biomedicine / Athanasios Zekios and Dimitra Petroudi
Chapter XLIV
Comparative Genomics and Structure Prediction in Dental Research / Andriani Daskalaki and Jorge
Numata
Chapter XLV
Genomic Databanks for Biomedical Informatics / Andrea Maffezzoli and Marco Masseroli

Chapter XLVI
Basic Principles and Applications of Microarrays in Medicine / Andriani Daskalaki and Athina A.
Lazakidou
Chapter XLVII
System Patterns of the Human Organism and their Heredity / Manfred Doepp and Gabriele Edelmann
Chapter XLVIII
Evaluation Methods for Biomedical Technology / Maria Sevdali
SECTION XII: ERGONOMIC AND SAFETY ISSUES IN COMPUTERIZED MEDICAL EQUIPMENT
Chapter XLIX
Ergonomic User Interface Design in Computerized Medical Equipment / D. John Doyle
Chapter L
Safety Issues in Computerized Medical Equipment / D. John Doyle
Chapter LI
Alarm Design in Computerized Medical Equipment / D. John Doyle
SECTION XIII: HEALTH ECONOMICS AND HEALTH SERVICES RESEARCH
Chapter LII
Organizational Factors in Health Informatics / Michelle Brear
Chapter LIII
Measurement of Cost and Economic Efficiency in Healthcare / Panagiotis Danilakis and Pericles
Robolas
Chapter LIV
Understanding Telemedicine with Innovative Systems / Irene Berikou and Athina A. Lazakidou
Chapter LV
A Capacity Building Approach to Health Literacy through ICTs / Lyn Simpson, Melinda Stockwell,
Susan Leggett, Leanne Wood, and Danielle Penn
The premier reference source for information science & technology research

VIEW
About This Book

• Description &
Biomedicine
Key Features
• Accolades
Preface:
• Contributors
Board Advances in information and communication technologies (ICT) have provided the tools and the
environment to study, analyze, and better understand complex medical problems. This technological
• Table of Contents development has enabled researchers to provide increasingly advanced services, including computer-
assisted radiology, telemedicine, robotized tele-operating systems, and so forth.
• Preface
ISBN: 1-59140-982-9
In recent years, research in computer applications applied to healthcare and biomedicine has
• About the Editors
dramatically intensified. The Handbook of Research on Informatics in Healthcare and Biomedicine Hard Cover
• View the Brochure aims to provide a platform for researchers to describe and analyze recent breakthroughs in these Publisher: Information
areas. This handbook will be most helpful as it provides comprehensive coverage and definitions of Science Reference
• View the Excerpt the most important issues, concepts, new trends and advanced technologies in healthcare and
biomedicine. This important new handbook will be distributed worldwide among academic and Copyright: 2006
professional institutions and will be instrumental in providing researchers, scholars, students, and Pages: 479
• Reference Home Page professionals access to the latest knowledge related to information science and technology in the
areas of healthcare and biomedicine. List Price: $215.00
• Recommend to your
add to cart
Library
This handbook provides a compendium of terms, definitions, and explanations of concepts,
processes, and acronyms. Additionally, this volume features short chapters authored by leading Perpetual Access:
Authoritative - experts offering an in-depth description of key terms and concepts related to different areas, issues,
and trends in information science and technologies in hospitals and other organizations worldwide.
Comprehensive add to cart
The Handbook of Research on Informatics in Healthcare and Biomedicine is an excellent source of
comprehensive knowledge and literature on the topic of health and biomedical informatics. Print + Perpetual Access:
$430.00
The topics in this handbook cover useful areas of general knowledge including medical data and
add to cart
health information systems, standardization and classification systems in medicine, virtual reality
applications in medicine, virtual learning environments in healthcare and biomedicine, computer-
http://www.igi-global.com/reference/details.asp?ID=5571&v=preface (1 of 4)7/4/2009 11:17:31 πμ
assisted diagnosis, data mining and medical decision making, current aspects of knowledge
management in medicine, telemedicine and e-health services, image processing and archiving
systems, signal processing techniques, use of new technologies in biomedicine, ergonomic and
safety issues in computerized medical equipment, health economics, and health services research.
Speicifically, these useful terms and key words have been included and analyzed in the concrete
sections of this publication.
A healthcare provider’s competitiveness, level of efficiency, and quality of care may be in direct
relationship to the rate of progress toward a paperless system—with digital documentation of all
clinical and administrative care processes. Each small step toward the electronic health record should
be analyzed according to its benefits and costs. Hospitals and delivery networks must share secure
health information and improve processes and efficiency in handling IT. The first section, “Medical
Data and Health Information Systems”, contains chapters related to the current status and future
prospects of the electronic health record systems, the security in health information systems, and
various applications in the area of health informatics.
For more than 130 years the systematic collection and record of medical information has been based
on the use of traditional classifications, nomenclatures, and coding schemes of various kinds. Until
relatively recently, such schemes were used mainly for recording causes of death and gathering
minimal diagnostic information for statistical and epidemiological purposes. Despite their many
limitations, schemes such as the international classification of diseases (ICD) have been successful
in supporting the collation and comparison of national and international statistics on morbidity and
mortality, and advancing our understanding of the distribution and causes of diseases. The second
section, “Standardization and Classification Systems in Medicine”, contains chapters related to
standardization and classification systems in health.
Medicine will benefit from virtual reality. As recorded in the military, virtual reality can provide an
excellent training mechanism when there is no room for mistakes. Doctors will be able to practice
alone or in teams to fine tune their skills for highly sensitive operations without having to risk a
human life. Virtual reality can improve the doctor’s performance during operations by superimposing
vital information on the patient during an operation. Superimposed images can increase the
effectiveness of radiation treatment and reduce the scarring of a surgery. In the third section,
“Virtual Reality Applications in Medicine”, various virtual reality applications in (bio) medicine and
their benefits are presented.
The application of computer technology to education often refers to computer-assisted learning

(CAL), computer-based education (CBE), or computer-aided instruction (CAI). Computer-based
learning has been developed for the beginning medical student and the experienced practitioner, for
the lay person and the medical expert. In the fourth section, “Virtual Learning Environments in
Healthcare and Biomedicine”, examples of actual programs that are being used to support medical
education for each of these categories of learners are presented.
Digital imaging still remains one of the key technologies for progress in healthcare. With further
advances in processing, display, and communication of medical imaging it becomes the key to solve
many problems in diagnosis and therapy. As well as computer-assisted diagnosis, computer-assisted
surgery relies increasingly on some type of image management. Typical examples can be found in
craniofacial surgery, neurosurgery, orthopaedic surgery of the hip and spine, plastic/reconstructive
surgery, otolaryngology, and so forth. The fifth section, “Computer-Assisted Diagnosis”, includes a
large number of examples in the area of computer-assisted diagnosis.
The process of extracting useful information from a set of data is called “data mining”. Data mining
techniques have been used as a recent trend for gaining diagnostic results especially in medical
fields such as kidney dialysis, skin cancer, and breast cancer detection and also biological sequence
classification. Various “Data Mining and Medical Decision Making” are presented in the Section VI.
Knowledge management is a basic tool for all those who work in the health field and in hospitals. It
helps sending the right information, to the right part, to the right person, at the right time, so that
the right decisions can be made, depending on the existing problems. It is certain that with the help
of knowledge management the effectiveness in the health field will be increased through unified
systems, processes and methods, the cultivation of exchanging knowledge, and the promotion of the
effective use of available information. In Section VII, basic principles and theoretical aspects of the
use of knowledge management in medicine are clearly presented.
Rapidly emerging information and communication technologies (ICT) have spurred the recent
escalation of various telehealth applications. It is true that there is an enormous interest in finding
new ways to apply telehealth as much as telemedicine as a special part of telehealth. Section VIII,
“Telemedicine and E-Health Services”, has, along with providing a better understanding of what
telehealth is, investigated the ways in which such an avant-garde, advancing, and newly emerging
technology could be used in order to make an upper healthcare level to be available. This section
aims to help someone to clarify confused terms such as telehealth and telemecine or even telecare
and e-health.
The rapid progress in imaging technologies during the last decades has stimulated many
developments and applications in medicine, biology, industry, aerospace, remote sensing,
meteorology, oceanography, and environment. New developments are continually making the
technology faster, more powerful, less invasive, and less expensive. Imaging technology was
primarily used in medical diagnosis initially, but it is being increasingly used in pure neuroscience,
psychological research, and in many other fields. The quantitative nature of data will be relevant to
the effective diagnostic as well as therapeutic management of patients—whatever disease they may
have. In the ninth section, “Image Processing and Archiving Systems”, various imaging technologies
and their applications in Biomedicine are clearly presented.
In digital signal processing, numerous powerful algorithms, both linear and non-linear, have been
developed during the past three decades. These have given rise to tremendous progress in speech
and image processing. But digital signal processing is not restricted to communications and
information processing. It also plays a leading role in such diverse fields as measurement, automatic
control, robotics, medicine, biology, and geophysics to mention just the most important ones. New
signal processing techniques for use in medicine are clearly presented in Section X.
Medical technology is a science discipline that has been rapidly growing over the last decades. It is
characterized by a constant flow of innovations and a high level of research and development. Many
technological achievements have dramatically changed the way that medicine diagnoses and treats
human disease. Improved healthcare technology has presented many revolutionary medical devices
that reduced mortality and morbidity. New various technologies applied in biomedicine are presented
in Section XI.
As the cost of microcomputer technology continues to drop, computers are being used increasingly
in medical systems and equipment such as ventilators or pacemakers, sometimes with safety-critical
results. “Ergonomic and Safety Issues in Computerized Medical Equipment” are clearly presented
and discussed in Section XII. Health services research is research that seeks to improve the quality,
organisation and financing of health services. Its concern extends from the care of individuals
through health care organisations to national and international policies. Section XIII contains
chapters related to health economics and health services research.
Athina A. Lazakidou, Ph.D.

VIEW
About This Book

• Description &
Biomedicine
Key Features
• Accolades
Contributors:
• Contributors
Board Angelidis, Pantelis / Vidavo Ltd., Greece
• Table of Contents Arvanitis, Theodoros N. / University of Birmingham, UK

• Preface
Baluch, Farhan / University of Houston, USA ISBN: 1-59140-982-9
• About the Editors
Hard Cover
Bamidis, Panagiotis D. / Aristotle University of Thessaloniki, Greece
• View the Brochure Publisher: Information
Science Reference
• View the Excerpt Bastaki, Maria / National and Kapodistrian University of Athens, Greece
Copyright: 2006
Berikou, Irene / Athens University of Economics and Business, Greece Pages: 479
• Reference Home Page
List Price: $215.00
• Recommend to your Biebler, Karl-Ernst / Ernst-Moritz-Arndt-University, Germany
add to cart
Library
Bisdas, Sotirios / Johann Wolfgang Goethe University Hospital, Germany Perpetual Access:
Authoritative -
Brear, Michelle / University of New South Wales, Australia
Comprehensive add to cart
Burgsteiner, Harald / Graz University of Applied Sciences, Austria
Print + Perpetual Access:
Burns, J. / University of New South Wales, Australia $430.00
add to cart
Catley, Christina / Carleton University, Canada
http://www.igi-global.com/reference/details.asp?ID=5571&v=contributors (1 of 5)7/4/2009 11:19:50 πμ

Chandrashekar, Sharath T. / Sarayu Softech Pvt Ltd., India
Damigou, Dionisia / National and Kapodistrian University of Athens, Greece
Danilakis, Panagiotis / National and Kapodistrian University of Athens, Greece
Daskalaki, Andriani / Max Planck Institute of Molecular Genetics, Germany
Davies, G. Powell / University of New South Wales, Australia
Della Mea, Vincenzo / University of Udine, Italy
Diaz, Javier / University of Houston, USA
Doepp, Manfred / Holistic DiagCenter, Germany
Doumouchtsis, Evangelos K. / National and Kapodistrian University of Athens, Greece
Doyle, D. John / Cleveland Clinic Foundation, USA
Drougas, Bill Ag. / ATEI Education Institute of Epirus, Greece
Economopoulos, Georgios / National and Kapodistrian University of Athens, Greece
Edelmann, Gabriele / Holistic DiagCenter, Germany
Esslinger, A. Susanne / Friedrich-Alexander-University, Germany
Exarchos, Themis P. / University of Ioannina, Greece
Fotiadis, Dimitrios I. / University of Ioannina, Biomedical Research Institute—FORTH, and

Michaelideion Cardiology Center, Greece
Frize, Monique / Carleton University and University of Ottawa, Canada
Galani, Olga / National and Kapodistrian University of Athens, Greece
Ganiatsas, G. / University of Ioannina, Greece
Georgiou, Andrew / University of New South Wales, Australia
Giannakakis, Nikolaos / National and Kapodistrian University of Athens, Greece
Goletsis, Yorgos / University of Ioannina, Greece
Harris, M. F. / University of New South Wales, Australia

Hirakis, Odysseas / National and Kapodistrian University of Athens, Greece
Holzinger, Andreas / Medical University Graz (MUG), Austria
Hussain, Kashif / University of Valenciennes, France
Ilioudi, Christina / University of Piraeus, Greece
Jaeger, Bernd / Ernst-Moritz-Arndt-University, Germany
Kalogeropoulou, Maria / National and Kapodistrian University of Athens, Greece
Kalogirou, Fotini / National and Kapodistrian University of Athens, Greece
Karakounos, Spyros / National and Kapodistrian University of Athens, Greece
Kaur, H. / Multimedia University, Malaysia
Kern, Josipa / Andrija Stampar School of Public Health, Zagreb University Medical School, Croatia
Koh, Tong San / Nanyang Technological University, Singapore
Kourtidou-Papadeli, Chrysoula / Greek Aerospace Medical Association and Space Research, Greece
Lazakidou, Athina A. / University of Piraeus, Greece
Leahy, Richard M. / University of Southern California, USA
Leggett, Susan / Queensland University of Technology, Australia
Leleu-Merviel, Sylvie / University of Valenciennes, France
Maffezzoli, Andrea / Politecnico di Milano, Italy
Maglaveras, Nicos / Aristotle University of Thessaloniki, Greece
Magoulia, Polyxeni / National and Kapodistrian University of Athens, Greece
Maresch, Helfrid / Graz University of Applied Sciences, Austria
Marschall, Daniela / Friedrich-Alexander-University, Germany
Masseroli, Marco / Politecnico di Milano, Italy
Michalis, Lampros K. / Michaelideion Cardiology Center and University of Ioannina, Greece
Mihalas, George I. / “Victor Babes” University of Medicine and Pharmacy, Romania

Mughal, Shazia Yasin / University of Valenciennes, France
Natasha, V. / Multimedia University, Malaysia
Nikiforou, Ageliki / National and Kapodistrian University of Athens, Greece
Numata, Jorge / Free University, Germany
Pantazis, Dimitrios / University of Southern California, USA
Papadelis, Christos L. / Aristotle University of Thessaloniki, Greece
Papaloukas, Costas / University of Ioannina, Greece
Papastefanatos, Stefanos E. / University of Piraeus, Greece
Paraki, Jayanth G. / Telemedicine Research Laboratory, India
Penn, D. / University of New South Wales, Australia
Penn, Danielle / Queensland University of Technology, Australia
Petriu, Dorina / Carleton University, Canada
Petroudi, Dimitra / National and Kapodistrian University of Athens, Greece
Poravas, Efstratios / National and Kapodistrian University of Athens, Greece
Pristas, Ivan / Croatian Institute of Public Health, Croatia
Ramakrishnan, B. / M.I.T. Manipal, India
Reibnegger, Gilbert / Medical University Graz (MUG), Austria
Robolas, Pericles / National and Kapodistrian University of Athens, Greece
Sevdali, Maria / Scientific Collaborator of Technological Educational Institution Kalamata, Greece
Siassiakos, Konstantinos M. / University of Piraeus, Greece
Simpson, Lyn / Queensland University of Technology, Australia
Sintchenko, Vitali / University of New South Wales, Sydney, Australia
Smolle, Josef / Medical University Graz (MUG), Austria

Sriraam, N. / Multimedia University, Malaysia
Stanic, Arsen / Orthopaedic Clinic, Croatia
Starida, K. / University of Ioannina, Greece
Stevanovic, Ranko / Croatian Institute of Public Health, Croatia
Stockwell, Melinda / Queensland University of Technology, Australia
Taggart, J. / University of New South Wales, Australia
Tsipouras, Markos G. / University of Ioannina, Greece
Uhernik, Ana Ivicevic / Croatian Institute of Public Health, Croatia
Varanasi, Gomata L. / Samskruti, India
Wodny, Michael / Ernst-Moritz-Arndt-University, Germany
Wood, Leanne / Queensland University of Technology, Australia
Zarras, Georgios / National and Kapodistrian University of Athens, Greece
Zekios, Athanasios / National and Kapodistrian University of Athens, Greece
Zouridakis, George / University of Houston, USA
vanut, Boštjan / College of Healthcare Izola, University of Primorska, Slovenia
The premier reference source for information science & technology research

1
Chapter I
Electronic Health Records
Olga Galani
National and Kapodistrian University of Athens, Greece
Ageliki Nikiforou
ABSTRACT
The Electronic Health Record is a means of organizing patient data making profound use of
the advances in the field of information technology. Its purpose is to fulfill the various needs
for information not only of patients and healthcare providers but also of other beneficiaries.
The implementation of EHR systems in healthcare organizations is very complex and involves
many parameters. This article is about the challenges faced by those undertaking such a task
and about the potential benefits from a successful implementation.
INTRODUCTION have a number of significant disadvantages.

First of all, paper is a very fragile medium that
Advances in computer technology have the requires large storage facilities. In addition,
potential to solve some of the most persistent paper-based health records require large hu-
problems in healthcare. There is a consensus in man effort to keep the files and archives orga-
healthcare that there is a need for creating nized and updated, they are available only in
information and communication systems that one place at a time, and the aggregation of data
reduce cumbersome and outdated paperwork. for research is difficult. Furthermore, we are
Paper-based health records have been in use moving to a model where the patient is no longer
for centuries since they require relatively little a passive recipient of the services provided by
investment to use and produce compared to nurses, physicians, and others, but is an active
more sophisticated supports, and they are fa- partner (a consumer or a client) with the
miliar to users who do not have to acquire new healthcare practitioners. A successful partner-
skills or behaviors to use them properly. De- ship requires open access to the healthcare
spite all these, paper-based health records also information. The creation and implementation
Copyright © 2006, Idea Group Inc., distributing in print or electronic forms without written permission of IGI is prohibited.
Figure 1. Creation of an electronic health record
of electronic health record (EHR) systems is at • Outpatient and emergency depart-

the heart of addressing these needs (Daskalaki, ments
Lazakidou, Philipp, Jacob, & Berlien, 2001; • Community and allied health
Mantas, 2002; Medical Records Institute, http:/ • General practice
/www.medrecinst.com/index.asp; Wang et al., • Dental clinics
2003). • Diagnostic results (pathology and ra-
diology)
• Documents such as assessments, dis-
COMPONENT PARTS OF AN charge referrals, and letters
ELECTRONIC HEALTH RECORD • Multidisciplinary and multiservice
care plans
Figure 1 depicts an oversimplified view of how
the EHR is created.
To gain a more accurate appreciation of the USE AND BENEFITS OF
EHR’s complexity and breadth of information, THE EHR
one must recognize the wide range of health
information sources. Each time an individual The primary purpose of the EHR is to provide
visits a healthcare provider, data are gener- a documented record of care that supports
ated. Figure 2 identifies some of the sources of present and future care by the same or other
data for an EHR as listed by the Institute of
Medicine.
The component parts of utmost importance Figure 2. Sources of health-related data
of an EHR are the following.
• Patient’s demographic details

• Family history
• Allergies and alerts
• Medical history
• Procedures
• A summary of services provided to an
individual by
• Hospitals during admissions for treat-
ment
2
clinicians. This documentation provides a means Patients

of communication among clinicians contribut-
ing to the patient’s care. The primary benefi- • integrated health services
ciaries are the patient and the clinicians. • improved healthcare and decreased risks
Any other purpose for which the medical (e.g., adverse drug reactions)
record is used may be considered secondary as • not having to repeat basic information,
are any other beneficiaries. Much of the con- such as name and address
tent of EHRs is currently defined by secondary • increased confidence knowing that all
users as the information collected for primary healthcare professionals have access to
purposes was insufficient for purposes such as all relevant parts of one’s medical history
billing, policy and planning, statistical analysis, • access to their own health records helps
accreditation, and so forth. patients to make informed decisions about
Secondary uses of EHRs include the following: their health
• avoidance of duplicate, invasive, and/or
• Medico-Legal Uses: Evidence of care expensive tests
provided, indication of compliance with
legislation, reflection of the competence Public
of clinicians
• Quality Management: Continuous qual- • expanded reach of effective healthcare
ity-improvement studies, utilisation review, • more secure information
performance monitoring, benchmarking, • access to information about how the
accreditation healthcare system works
• Education
• Research: Development and evaluation Health Professionals
of new diagnostic modalities, disease-pre-
vention measures, epidemiological stud- • view of patient data
ies, population health analysis • access to other related and integrated
• Public and Population Health: Access patient information
to quality information enables the effec- • access through a portal to related health
tive management of real and potential services
public health risks • seamless care through the coordination of
• Policy Development: Health-statistics multiprofessional and multiagency care
analysis, trends analysis, casemix analysis • development of decision-support sys-
• Health-Service Management: Re- tems
source allocation and management, cost
management Health Administrators
• Billing, Finance, and Reimbursement:
Insurers, government agencies, funding • increased patient care time
bodies (Schloeffel & Jeselon, 2002) • access to data to support clinical gover-
nance and local planning
The use of EHR can yield to a number of • reduced healthcare costs
benefits, which can be described in terms of the
following attributes.
3
Policy Makers Figure 3. Clinical data repositories in use

and planned for use within 1 to 4 years
• improved and effective health mainte-
nance and education
• support for medical and administrative
decision-making processes
• improved long-term planning
Researchers
• access to timely, high-quality data for

research
• access to up-to-date research findings,
and treatment and medication options
• data aggregation
• improved trend analysis
(which requires a period of time to learn), or
Governments searching for the appropriate location to record
most pieces of information (which requires
• improved accountability time to search). Training is an important issue
• improved health-resource allocation as this is required by a large amount of the
(Upham, 2004) population. In addition, one should always keep
in mind the fact that the population has different
levels of computer literacy (Upham, 2004). All
THE EHR TODAY the above characteristics of the EHR have
contributed to a slow increase in the adoption of
The barrier to the adoption of the EHR that is such systems in many hospitals. One example
probably the most difficult to overcome is the of an important shift in the EHR market is seen
lack of easily apparent returns on investment in Figure 3, which includes responses to the
(ROI). Many writers on the subject have noted survey question, “What functions or compo-
that healthcare decision makers find it difficult nents of an EHR system do you have in use or
to readily demonstrate ROI or justify the ex- are planned for implementation?” Of the 436
penditure of dollars and time to undertake a who responded to the question, 35.6% said they
comprehensive EHR system within their orga- have already implemented the basic repository
nizations, particularly while healthcare costs capabilities of storing data, text, and reimburse-
continue to spiral out of control and taking into ment codes. Combined with those planning to
consideration the fact that the initial investment add these basic capabilities, the percentage is
on equipment can be quite expensive. Another expected to grow to 67% in the next 4 years. On
factor that complicates the adoption of the top of the fundamentals are plans to add storage
EHR is the fact that data are heavily structured, for clinical codes, voice or sound, and clinical
being recorded in their allocated space. This images.
implies a deep knowledge of the system to The EHR survey also included an analysis
know where to record any piece of information by market segment with some interesting find-
4
Figure 4. Some barriers, by market segment, to implementing EHRs
ings. Figure 4 shows some of the results from such records, we may also expect to find data
the 477 respondents to the question, “What are regarding populations of patients, integrated
the major barriers to your plans for implement- access to the biomedical literature, and interac-
ing an EHR?” tive environments for offering clinical guide-
For example, respondents from ambulatory- lines or frank consultative advice. We can
care facilities indicated that they face more envision a world in which the enterprise LAN
barriers in implementing EHRs than respon- (local area network) is seamlessly connected to
dents from hospitals or integrated delivery net- the full Internet, with integrated access to a
works. In particular, ambulatory-care respon- wide variety of information sources that are
dents reported challenges in finding an afford- geographically distributed well beyond local
able solution, creating a migration plan, finding institutions. Although such a concept raises
a solution that is not fragmented among ven- important issues related to patient privacy and
dors, evaluating solutions, and finding a solution confidentiality, there are technical and policy
that meets their technical requirements (Blair, measures that can be taken to help to assure
2003). that such virtual records are kept secure but
also available at times of medical need.
Realizing the vision described above will
THE EVOLUTION OF THE EHR depend on at least four factors.
Although we should always expect a medical • An Enhanced Internet: An Internet with

record to be populated with data about a spe- much higher bandwidth and reliability, in-
cific patient, in the electronic implementation of creased response time, and financial mod-
5
els that make the applications cost effec- oping greater intelligence within their sys-
tive and practical is required. Major re- tems to serve this dynamic industry.
search efforts are underway to address
some of these concerns, including the Since the actual realization of EHRs re-
federal Next Generation Internet activity quires full interoperability within and among
in the United States Exploratory efforts healthcare enterprises, total industry adoption
that continue to push the state of the art in is not likely to be achieved within the near
Internet technology, and all have signifi- future (most predict it will be 5 to 10 years).
cant implications for the future of Unlike EHRs, the continuity-of-care record
healthcare delivery in general and for (CCR) is designed to provide a snapshot of
computer-based health records in particu- essential patient information that will enable a
lar. physician to understand a patient’s context and
• Better Education and Training for provide appropriate care. Some of the core
Healthcare Providers: There is a dif- elements of the CCR are document-identifying
ference between computer literacy (fa- information (to and from fields, date sent, pur-
miliarity with computers and their routine pose), patient-identifying information, patient
uses in our society) and the knowledge of insurance and financial information, advance
the role that computing and communica- directives, patient health-status information (may
tions technology can and should play in our include conditions, diagnoses, problems, family
healthcare system. More medical- medical history, adverse reactions and aller-
informatics training programs and the ex- gies, social history and health-risk factors, medi-
pansion of existing programs are needed. cations, immunizations, vital signs and physi-
In addition, junior faculty in health-sci- ological measurements, laboratory results and
ence schools who may wish to seek addi- observations, and procedures and imaging),
tional training in this area should be sup- care documentations, care-plan recommenda-
ported. tions, and a list of healthcare practitioners. The
• Changes in the Management and Or- CCR is gaining momentum across the industry
ganization of Healthcare Institutions: because it will achieve many of the immediate
Healthcare provides some of the most short-term goals and benefits envisioned by
complex organizational structures in soci- HL7 as we continue to define and develop the
ety, and it is simplistic to assume that off- EHR: It may even provide the impetus needed
the-shelf products will be smoothly intro- to stimulate more rapid EHR development. It is
duced into a new institution without major increasingly being viewed as a practical, more
analysis, redesign, and cooperative joint- immediately achievable solution while the in-
development efforts. dustry continues to wait for a defined EHR
• Just Like the Healthcare Industry, (Carpenito, 2004; Golden, 2004; Shortliffe,
Technology Cannot Stand Still: Sys- 1999).
tems must continue to evolve to meet the
industry’s changing needs. While today’s
imaging and work-flow applications are CONCLUSION
excellent for viewing and accessing infor-
mation, healthcare institutions continue to The implementation of EHR systems is not
push for more. Suppliers must keep devel- simple; it concerns not a single system but,
6
rather, a collection of interlocking systems that health informatics: A nursing perspective

are tied to a series of complex clinical and (pp. 250-257). Amsterdam: IOS Press.
administrative work flows. This implementa-
Office of Health & Information Highway Health
tion will involve a long-term, highly coordinated
Canada. (2001). Toward electronic health
commitment from a large number of stakehold-
records. Retrieved from http://www.hc-
ers and a significant financial investment. The
sc.gc.ca
changes required are massive since the EHR-
related initiative is trying to implement a total Schloeffel, P., & Jeselon, P. (2002). ISO/TC
healthcare solution. To this end, projects are 215 ad hoc group report: Standards re-
under way that address part of the puzzle (i.e., quirements for the electronic health record
pharmaceutical systems), others are develop- & discharge/referral plans. Retrieved from
ing application models (emergency health http://www.iso.org
records) on which to base further development,
Shortliffe, E. (1999). The evolution of elec-
and still other projects are using new technol-
tronic medical records. Academic Medicine,
ogy models such as smart cards (Office of
74(4), 414-419.
Health & Information Highway Health Canada,
2001). Upham, R. (2004). The electronic health
record: Will it become a reality? Phoenix
Health Systems. Retrieved from http://
REFERENCES www.hipaadvisory.com
Wang, S., Middleton, B., Prosser, L., Bardon,
Blair, J. (2003). The EHR today. Health care
C., Spurr, C., Carchidi, P., et al. (2003). Aca-
informatics online. Retrieved from http://
demic publication proves that EMRs are cost
www.healthcare-informatics.com/index.htm
effective: A cost-benefit analysis of electronic
Carpenito, L. (2004). A report on the CCR. medical records in primary care. American
Symantec enterprise solutions. Retrieved from Journal of Medicine, 114(5), 397-403.
http://www.symantec.com
Committee of European Normalization, Tech- URL REFERENCES
nical Committee on Medical Informatics (CEN/
TC 251). (n.d.). Medical informatics vocabu- Center for Health Information Technology: http:/
lary working document (PT 011/N 300 /www.centerforhit.org
V.1.00).
Medical Records Institute: http://www.
Daskalaki, A., Lazakidou, A., Philipp, C., Jacob, medrecinst.com/index.asp
C., & Berlien, H. P. (2001). Introducing elec-
NSW Electronic Health Record: http://
tronic health record into laser medicine. Med.
www.nsw.com
Inform., 5, 85-86.
Golden, R. (2004). The evolving electronic
patient record system. Retrieved from http://
KEY TERMS
www.infotivity.com
CCR: The continuity-of-care record is an
Mantas, J. (2002). Electronic health record. In emerging standard for communicating patient
J. Mantas & A. Hasman (Eds.), Textbook in information electronically among providers. The
7
CCR is intended to provide a snapshot of es- Healthcare Record: Systematic record of

sential patient information, rather than a com- the history of the health of a patient kept by a
plete patient record, that will enable a physi- physician or other healthcare practitioner.
cian to understand a patient’s context and
HL7: A specification for a health-data in-
provide appropriate care. The format of the
terchange standard designed to facilitate the
CCR allows it to be used universally to help
transfer of health data resident on different and
to bridge the gaps between EHR systems and
disparate computer systems in a healthcare
improve the portability of patient information.
setting.
Decision-Support System: Any computer-
LAN (Local Area Network): A system of
based support of medical, managerial, adminis-
connecting computers and computer equipment
trative, and financial decisions in health using
together with physical links that do not use a
knowledge bases and/or reference material.
telecommunications network.
EHR: A healthcare record in computer-
Smart Card: An integrated circuit card that
readable format.
incorporates a processor unit. The processor
EHR System: Information system that may be used for security algorithms, data ac-
manages and operates on the EHR. cess, or for other functions according to the
nature and purpose of the card.
8
9
Chapter II
Security Health
Information Systems
Christina Ilioudi
University of Piraeus, Greece
Athina Lazakidou
ABSTRACT
Please provide an abstract.
INTRODUCTION mation. Nowadays medical data are being kept

in a computer, so we talk about an electronic
Before the computer age, healthcare informa- patient record (EPR) and not about a printed
tion was typically stored in the physician’s medical record. Thus, health information sys-
office. All information specific to a patient was tems rely upon a computerised infrastructure.
generally kept in a medical record and filed in The development of Internet technology and
the office filing cabinet. The introduction of Web-based applications made health informa-
technology changed how physicians and other tion more accessible than ever before from
health organizations keep personal health infor- many locations by multiple health providers and
Security in Health Information Systems
health plans. In the near future, the Internet will INFORMATION-FLOW CONTROL
probably be the platform of choice for process-
ing health transactions and communicating in- In security engineering, there are two ap-
formation and data. But along with this acces- proaches to information-flow control. The first
sibility come increased threats to the security of approach (Figure 1) is multilevel security (top-
health information, and those who would steal, down approach), in which lower level informa-
divert, alter, or misuse your information are tion may move up in the hierarchy, but higher
becoming even more skilled at finding what level information may not move down. The
they want and covering their tracks. main representative of this approach is the
In healthcare, it is very important to develop Bell-LaPadula policy. The second approach
secure systems because we want to ensure that (Figure 2) is multilateral security, in which
the medical data that contain the personal infor- information is prevented from flowing across
mation of patients will not be violated by any- departments. One representative of this ap-
one. Thus, what is under question is the avail- proach is the BMA model developed by the
ability of the right data to the right user at the British Medical Association to describe the
right time (availability). Information technology information flows permitted by medical ethics.
deeply affects the confidential relationship be- This kind of security applys very well in
tween patient and doctor since it increasingly healthcare systems as it covers the use of
surrounds and mediates it. Hence, the protec- techniques such as anonymity. However, we
tion of personal medical data (confidentiality) is will not make further analysis of it here because
a necessity without which medical treatment this is not our goal.
will hardly be successful. This will sometimes
include the anonymity of a patient. In addition to
the protection of data from unauthorized read- THREATS TO HEALTH
ing, data also have to be protected against INFORMATION SYSTEMS
unauthorized modification (integrity). Also,
healthcare professionals are personally respon- One question of vital importance for health
sible and mostly liable for their decisions in information systems could be “What are the
favour of a particular action or against it. This threats that could have a major impact on health
raises the need for health information systems information?” If we try to answer the question,
that are capable of providing an individual with we would count some of the potential threats. A
undoubted proof that he or she took a certain threat can be an agent who either accidentally
action or did not (accountability). Below we or intentionally gains unauthorized access to
will explain how exactly the information flow the protected IT systems. Threats may be
takes place in medical systems, which security physical or logical. Physical threats include the
model we used in the past, what the threats are, following.
which mechanisms are usually being used to
prevent the violence of those systems, and • employees (common threat)
finally we will present some examples of secu- • ex-employees
rity deficiency and what impacts have been • hackers
recorded. • terrorists
• criminals
• customers
10
• visitors tive password-protection mechanisms and

• the destruction of physical storage de- policies, not allowing critical data to walk
vices out of the door unprotected on laptops and
• natural disasters PDAs (personal digital assistants).
• Protect against hackers and computer
Protection against these threats comprises malice. Individuals (often ex-employees)
a variety of different methods and techniques. can wreak havoc to a computer system by
Logical threats involve unauthorized logical means of unauthorized access. We need
access to information. They can result in the authentication and authorization controls
disclosure of confidential information, the ille- that ensure that only authorized users gain
gal modification of data, or the destruction of access to a system and to only those parts
stored data. The threats can be classified as of the system necessary to perform his or
follows. her responsibilities. Appropriate authenti-
cation and access controls not only pro-
• The disclosure of confidential information tect against unauthorized access, but also
includes direct or indirect access to pro- reduce the risk of systems being infected
tected information. by malicious software.
• Illegal modification of data can be caused • Avoid public-relations issues. Breaches in
by improper, possibly accidental, data han- information security can be a source of
dling or intentional modifications by an embarrassment to an organization, espe-
illegal user (these threats are related to all cially one that relies on public trust.
attacks to data integrity). • Avoid productivity issues. Lapses in infor-
• Denial of service can be caused by mo- mation security can cause downtime and
nopolizing system resources in such a way the corruption of data and systems, which
that other users cannot access them. This in turn can cripple the operations of a
involves all attacks on availability. company.
• It is important to develop strong crypto-
graphic technologies end to end, where
SECURING HEALTH end points will range from patients’ homes
INFORMATION SYSTEMS to large hospitals, and often may terminate
in a mobile device such as a PDA or
Healthcare information systems are secure if a Internet-enabled cellular telephone. In
protection mechanism enforces the security addition, digital signatures should be used
policy. A security policy must have appropriate to verify that data, whether in transit or in
security features. These features should be a database, have not been modified by
supported by a security mechanism. In the unauthorized parties. Digital signatures
following, we specify those mechanisms. ensure that the accompanying data are
tamperproof and that signers cannot later
• Protect health information. Healthcare deny access or use.
systems will lose medical data unless they
take reasonable steps to protect them. Apart from security measures and policies
Operating computer networks should have through computer technologies, there is an-
firewalls, security monitoring, and effec- other way to ensure security in health informa-
11
tion systems. In particular, the Health Insur- is very important to develop mechanisms that
ance Portability and Accountability Act of 1996 minimize this risk. Therefore, we cannot allow
(HIPAA) has issued a detailed proposed regu- an insecure Internet connection in the internal
lation regarding security requirements for network of the healthcare system. Also,
healthcare information. HIPAA includes re- firewalls are another solution to minimize the
quirements to protect the security, integrity, impact of the threat.
and confidentiality of this health-related infor-
mation. To be HIPAA compliant, departments
(especially hospitals that work with health in- DISCUSSION
formation) must develop, implement, and en-
force a comprehensive security program in- Because of the rapid development of Internet
cluding administrative, technical, and physical technology and Web-based applications, it is
safeguards as determined appropriate for the obvious that in the near future, if we want to be
institution and data. In addition to developing more straightforward, our electronic medical
their own safeguards, departments are respon- records will contain personal health information
sible for taking steps to ensure that their affili- about our parents or genetic information. For
ates and service providers safeguard customer this reason, it is easy for everyone to under-
information in their care. stand how meaningfulness it is to find out
security models or extend those that already
exist. This is desirable because the most impor-
EXAMPLES OF SECURITY-RISK tant thing for healthcare systems is to ensure
IMPACTS our medical data.
The privacy, confidentiality, and security of
Below we mention the impact of threats in health information will be achieved only when
healthcare systems. We believe that this is a the following happen.
good way for those who are involved in
healthcare to take the enforcement of security • Privacy and confidentiality protections are
in such systems seriously. uniform and set the high standard through-
The most common security problem within out the country through federal laws that
healthcare systems is the access of employees establish fair, reasonable, and uniform
(threat from inside). Specifically, people who health-information practices, across all
work in a hospital have the ability to view the states. These laws should respect the
protected health information (PHI) of anyone. rights of the individual and the public, and
This raises the probability for legal action, apply to the medium in which such infor-
which causes major impacts. It is conceivable mation is stored, transferred, or accessed.
how important it is to enforce security policies. • An individual will have the right to access
It is important to the introduction of security his or her health information in any setting
policies that the people who have the authority (with minimal limits), to have an under-
to make decisions set boundaries under which standing of his or her privacy rights and
the staff can operate. options, to be notified about all information
Besides inside threats, it is possible for practices concerning his or her informa-
damage to occur in a healthcare system due to tion, to appropriately challenge the accu-
outside threats such as hackers. In this case, it racy of his or her health information, and
12
to opt in or authorize the collection or use REFERENCES

of information beyond what is originally
authorized by the individual or the law in Aderson, R. (n.d.). Security engineering: A
certain electronic or Internet situations. guide to building dependable distributed
• The collection and use of health informa- systems.
tion will be permitted only for legitimate
Baraani-Dastjerdi, A., Pieprzyk, J., & Safavi-
purposes and only as provided by law, and
Naini, R. (n.d.). Security in database: A sur-
will be uniform across all jurisdictions and
vey study. University of Wollongong, Depart-
entities and for all individuals.
ment of Computer Science.
• Credentialed health-information manage-
ment professionals, given their training Bleumer, G. (n.d.). Guidelines for a secure
and education in privacy and information- environment for information systems in medi-
release and health-information manage- cine: SEISMED. Hildesheim, Germany:
ment, are considered the primary custodi- Universität Hildesheim, Institut für Informatik
ans of health information and principal Samelsonplatz.
experts in maintaining the privacy, confi-
Castano, S., Fugini, M., Martella, G., & Samarati,
dentiality, and security of information in
P. (1995). Database security. Addison-Wesley,
the healthcare industry.
ACM Press.
• Laws, practices, and technologies are put
in place to provide protections required to Clark, D. D., & Wilson, D. R. (1987). A
maintain the appropriate privacy, confi- comparison of commercial and military com-
dentiality, security, and integrity of health puter security policies. Proceedings of IEEE
information. Computer Society Symposium on Security
and Privacy (pp. 184-194).
Cyber Security Industry Alliance (CSIA). (n.d.).
CONCLUSION
Ten steps for securing electronic health care
systems (CSIA security briefing).
The information being transferred between
hospitals or between departments that consti- Iowa State University. (n.d.). Health informa-
tute a hospital is referred to as sensitive data tion privacy and security policy.
because it concerns patients’ private data. It is
Kang, M. H., Froscher, J. N., & Costich, O.
easy for everyone to understand how important
(1992). A practical transaction model and
it is to ensure security for healthcare environ-
untrusted transaction manager for multilevel-
ments. As we use computer networks and the
secure database systems. Proceedings of 6th
Internet more often now than ever, when ex-
Annual IFIP WG11.3 Working Conference
changing health information, the threats will
on Database Security (pp. 289-310).
become bigger. The only way to protect our
sensitive information is a systematic approach Kang, M. H., Froscher, J. N., & Moskowitz, I.
to security. The people and organizations that S. (n.d.). An architecture for multilevel secu-
are involved with health information should rity interoperability. Center for High Assur-
document the steps they have taken to minimize ance Computer Systems, Naval Research Labo-
the risk to security. ratory.
13
National Health Policy Forum. (n.d.). Protect- Confidentiality: The property that infor-
ing the confidentiality of health informa- mation is not made available or disclosed to
tion. unauthorized individuals, entities, or processes
(ISO 7498-2 as cited in HISB Draft Glossary
Workgroup for Electronic Data Interchange
of Terms Related to Information Security in
(WEDI). (2003-2004). Strategic national
Health Care Information Systems).
implementation process (SNIP): Background
to information security. Cryptography: The art and science of us-
ing mathematics to secure information and
Workgroup for Electronic Data Interchange
create a high degree of trust in the electronic
(WEDI). (2004). Strategic national imple-
realm. See also public key, secret key, sym-
mentation process (SNIP): Background to
metric key, and threshold cryptography.
information security (Final draft).
Digital Signature: The encryption of a
message with a private key.
URL REFERENCES
ePHI (Electronic Protected Health In-
http://www.hipaadvisory.com/ formation): ePHI is any information specifi-
cally identifying a person that is stored elec-
http://www.healthcareedu.org/phs_page3c.cfm
tronically, or sent or shared electronically.
http://www.hipaabasics.com/glossary.htm
Hacker: A person who tries and/or suc-
http://www.public.iastate.edu/~health/docu- ceeds at defeating computer security mea-
ments/ sures.
http://www.hhs.gov/ocr/combinedregtext.pdf Healthcare: Care, services, or supplies
related to the health of an individual. Healthcare
http://www.hutchlaw.com/resources/docs/
includes, but is not limited to, the following: (a)
104047-v3.doc
preventive, diagnostic, therapeutic, rehabilita-
http://security.uwmedicine.org/ tive, maintenance, or palliative care, and coun-
securitypolicies.asp seling, service, assessment, or procedure with
respect to the physical or mental condition or
http://www.hhs.gov/ocr/combinedregtext.pdf
functional status of an individual that affects
http://www.ncipher.com/investors/glossary.php the structure or function of the body, and (b) the
sale or dispensing of a drug, device, equipment,
http://www.w3.org/TR/2002/WD-ws-gloss- or other item in accordance with a prescription.
20021114/#securitypolicy
Health Information: The term health infor-
http://library.ahima.org/xpedio/groups/public/ mation means any information, whether oral or
documents/ahima recorded in any form or medium, that is created
or received by a healthcare provider, health
KEY TERMS plan, public health authority, employer, life in-
surer, school or university, or healthcare clear-
Availability: Ensuring that authorized us- inghouse. It can relate to the past, present, or
ers have access to information and associated future physical or mental health or condition of
assets when required. an individual, the provision of healthcare to an
14
individual, or the past, present, or future pay- information values, protection responsibilities,
ment for the provision of healthcare to an and organization commitment for a system
individual. (OTA, 1993). The American Health Informa-
tion Management Association (AHIMA) rec-
HIPAA: The Health Insurance Portability and
ommends that security policies apply to all
Accountability Act of 1996.
employees, medical staff members, volunteers,
Integrity: Safeguarding the accuracy, com- students, faculty, independent contractors, and
pleteness, and control of information and pro- agents (AHIMA as cited in HISB Draft Glos-
cessing methods. sary of Terms Related to Information Secu-
rity in Health Care Information Systems). It
Password: A character string used as a key to
is part of the security-management process on
control access to files or to encrypt them.
the matrix.
Protected Health Information (PHI): Indi-
Trojan (or Trojan Horse): Trojans are pro-
vidually identifiable health information that is
grams (often malicious) that install themselves
transmitted by, or maintained in, electronic
or run surreptitiously on a victim’s machine.
media or any other form or medium. Protected
They do not install or run automatically, but may
health information excludes individually identi-
entice users into installing or executing them by
fiable health information in (a) education records
masquerading as another program altogether
covered by the Family Educational Rights and
(such as a game or a patch), or they may be
Privacy Act (FERPA), and (b) employment
packaged with hacked legitimate programs that
records held by the University of New England
install the trojan when the host program is
in its role as employer.
executed. Hence, a user thinks he or she is
Security Mechanism: A process (or a device installing a new game found for free, but will
incorporating such a process) that can be used actually be installing a nasty piece of software.
in a system to implement a security service that While trojans are not capable of spreading by
is provided by or within the system. themselves, there have been several reports of
worms that carry trojans, dropping them onto
Security Model: A schematic description of a
infected machines as they spread.
set of entities and relationships by which a
specified set of security services are provided Worm: A worm is a self-contained program
by or within a system. that spreads by creating multiple copies of
itself. Unlike viruses, worms do not require a
Security Policy: The framework within which
host file. Common modes of transmission used
an organization establishes needed levels of
by worms include e-mail, IRC, peer-to-peer file
information security to achieve the desired
sharing, and network drives.
confidentiality goals. A policy is a statement of
15
16
Chapter III
Development of a Health
Information System in a
Post-Communist Country
Ranko Stevanovic
Croatian Institute of Public Health, Croatia
Ivan Pristas
Ana Ivicevic Uhernik

Arsen Stanic
Orthopaedic Clinic, Croatia
ABSTRACT
Between 2001 and 2003, a project conducted in Croatia aimed to establish and develop a
health information system based on the latest technologies. Extraordinary results in the trial
run give the authors the ground to recommend such an approach to all transitional post
communist countries. The development of such systems is feasible in transitional countries
because most of them are still having one main insurer.
INTRODUCTION idea applied in the Croatian project was based

on the concepts developed by Professor Andrija
Between 2001 and 2003, a project conducted in Stampar. According to these, primary healthcare
Croatia aimed to establish and develop a health is a venue where the major health problems of
information system based on the latest tech- a population are resolved, and a point at which
nologies. The most important experience and outcomes of changes in the system are most
Development of a Health Information System in a Post-Communist Country
significantly reflected. A central health infor- and save enough money for “computerization to
mation system should be established and devel- happen spontaneously.” One should do exactly
oped in parallel with the primary healthcare the opposite: Computerization should be in-
activity. Extraordinary results in the trial run stalled in the system as a money keeping tool
give us the ground to recommend such an and implemented with the aim of exerting total
approach to all transitional post-communist control over the consumption as well as the
countries. The development of such systems is rationalizing of it in order to save substantially
feasible in transitional countries because most more money than the cost of installing the
of them still have one main insurer. In devel- information system.
oped countries, however, developing these could In fact, the share of investment in informa-
be difficult for they have a number of insurance tion systems is directly proportional to the fi-
companies that do not find their business inter- nancial effectiveness of the system (e.g., banks
est in the full integration of health information and insurance systems spend 5 to 6% of their
and data. Indeed, for countries with a single total budgets on the computerization of busi-
dominant, basic insurance company, the above ness operations). Whereas EU member states
is the only positive alternative. Subsequent spend at least 2 to 3% of the health budget on
linkups of supplemental, auxiliary, and other computerization, in the United States, this share
future insurers with the single information sys- varies around 4 to 10% (Bates, Ebell, Gotlieb,
tem on the national level will be much simpler to Zapp, & Mullins, 2003). Among transitional
make. countries, it is difficult to find one investing
At the beginning of 2004, that is, 6 months more than 0.2 to 0.5 % of its health budget. This
after software for primary healthcare was tested raises the question of whether this might be the
in Croatia, the European Public Health Alliance starting point for the vicious circle of ineffec-
published a document titled Communication tiveness in transitional countries’ health sys-
from the Commission to the Council, the tems.
European Parliament, the European Eco- The elements of new organizations (new
nomic and Social Committee and the Com- business rules) cannot be set, nor can the
mittee of the Regions. E-Health: Making foundation of a redesign be made without in-
Healthcare Better for European Citizens. An vesting in the computerization aligned with
Action Plan for a European E-Health Area modern concepts (Stevanovic, 2002a, 2003),
(text with EEA relevance). As made clear in under which information (processed data) be-
this document, European Union (EU) member comes a business resource (Krcmar,
states will start implementing in 2007 the solu- Stevanovic, Kovacic, & Merzel, 2001). At the
tion Croatia reached in 2003. Establishing and same time, the information communication sys-
developing this information system in transi- tem for primary healthcare warrants the confi-
tional countries is one of the strategic projects dentiality of data on patients and the standard-
for the coming years. Without rapid, reliable, ization of good practice for most common acute
and comprehensive information availability, and chronic mass diseases (Stevanovic & Erceg,
developing, implementing, and monitoring any 2003). The system should provide the basic
healthcare development strategy and system contents for the establishment of effective
reform would be difficult. One would be unrea- management.
sonable to expect a poor health system to keep
17
GOALS AND IMPROVEMENTS formance monitoring accurately measures the

effectiveness and outcomes, as well as team-
Ample help offered by computerization relates work coverage and contents.
to the rapid retrieval of documents and insurees
(Stevanovic, 2002b; Varga & Stevanovic, 2003), Benefits for the Health Ministry and
the replacement of manual data input, typing on Public Health
typewriters, and the writing of recipes, referral
notes, invoices, individual forms, and reports. The benefits include PHI (public health intelli-
The information system should give real time gence), the use of accurate and comprehensive
insight into the data and information, as well as data and information needed to run the public
enable prompt interventions within the system. health system and public health initiatives. MI
Improving the overall care of patients and (management intelligence), the utilization of
insurees, more rapid diagnosis, and accuracy in accurate and comprehensive data and informa-
prescribing a therapy are the major goals of tion needed by the ministry of health, health
such an approach. The system should enable managers, and decision makers for efficient
better utilization of the capacities, shorter wait- guidance and running of the system, is another
ing times, and shorter stays in health institutions benefit. An integration tool will enable the full
(Booth, 2003). integration of the health system’s information
The project was the first to introduce elec- and data. Currently, data on more than 500
tronic smart cards here for all physicians and parameters are collected from the primary
nurses. Thus, at each medical checkup or hos- health service with the aim of extracting just a
pital admission, the new information system few pieces of information, that is, qualitative
also verifies both the insuree’s and the indicators (exclusively on the level of ratios or
physician’s status and rights. Equally, as in trends). The future system with a few stan-
banking or other card business, only a linkup dardizes characteristics will offer the possibil-
between the magnetic card and the doctor’s ity of analyzing a much greater number of
card makes the transaction possible. indicators. Regarding the reporting of statistics,
registering with public health registries (obliga-
Insurance-Provider Benefits tory notifications of immunizations, infectious
diseases, melanomas, psychotics, the disabled,
BI (business intelligence), or the utilization of etc.) will be carried out automatically, interac-
accurate and comprehensive data and informa- tively, and proactively. Linkage makes the car-
tion required in health-insurance management, rying out of coordinated and joint preventive
is one benefit. Data are standardized, and their and curative interventions feasible. The system
input is independent of the wishes, ambitions, guarantees the confidentiality and safety of
and interests of teams and institutions. This sets personal information and of data on health and
the insurer and physician free of worries about diseases (VPN, PKI [public-key infrastruc-
data collection and primary data processing ture], encryption, smart cards, separate au-
because the information system takes care of thentication servers, data servers, and data
these with standardized applications. Thus, repositories). With an accident and catastrophe
doctors and insurers can devote all their energy early-alert system, one can boot the whole
to the execution of managerial and professional system with all data and information in 24
activities. Within the information system, per- hours.
18
Benefits for Physicians and Teams, Data Standards

and the Medical Chamber
For software applications and the central infor-
The utilization of guidelines, instructions, and mation system, use has been made of EU and
tools needed by physicians and insurees in certain other engineering standards in order for
order for standardized, quality care to be pro- this system to be open from the outset and built
vided and for more rapid diagnosis setting and for Croatia as an EU member state. It should
the selection of good and rational therapy in the support international and EU data standards
treatment of disease is one benefit (healthcare and classifications, for example, HL7 (Health
intelligence). The dissemination and utilization Level 7) Version 3, ICPC (International Clas-
of the knowledge necessary to provide good sification of Primary Care, 2nd edition), ICD-10
medical care is another (knowledge intelli- (International Classification of Diseases, 10 th
gence). Another benefit is still a linkage with revision), CEN TC 251 (European Committee
other participants in the process of treatment for Standardization/TC251 Work Groups), and
and prevention of disease (online linkage of all others.
care providers). Equally beneficial are the regu-
lated and safe utilization of all data on the health Technical Elements and Standards
and disease of an insuree in care, online consul- for the System and Applications for
tations with networked experts, e consultations Transitional Countries
with specialists, telemedicine, and the direct
engagement in scientific and technical public- Technical elements and standards should in-
health projects and programs. There is also the clude the following: one common system with
benefit of time and money savings in office, several licensed applications; an Internet infra-
postal, and other expenditures. structure (to enable paperless operation); XML
(extensible markup language) and HL7 stan-
Benefits for the Citizen, Patient, dards; system safety standards like PKI; mark-
and Insuree ing clinical and other documents with bar codes;
electronic medical files; global registration and
First, there is EI (equity intelligence), the equity an insuree database; a global database on
or enabling of all who are in the same medical codebooks (Hofmans-Okkes & Lamberts, 1996;
insurance (and market) position to avail them- Lamberts & Hofmans-Okkes, 1996; Lamberts
selves of the same conditions and quality. Then & Wood, 1987; Okkes, Oskam, & Lamberts,
there is the continuity, irrevocability, and trans- 1998; WONCA, 1998; World Health Organiza-
ferability of the safe storage of data on the care tion [WHO], 1994); access to external data-
of an individual insuree. Next are the benefits bases; data integration into the primary health
of guaranteed confidentiality and the security service’s information system; direct linkage
of personal data. The final benefit is the com- with the insurance administration, public health,
munication option, that is, two way (or multiple) and the ministry of health, e-recipes, e-refer-
communication between a physician and the ral-notes, e-business, and so forth; and the
citizen, patient, or insuree (e-information, e- standardized, equitable, and measurable use of
active-calling, e-instruction-sending). guidelines (Brage, Bensten, Bjerkedal, Nygard,
& Tellnes, 1996; Lamberts, Wood, & Hofmans-
Okkes, 1993).
19
CONCLUSION • The system should permit association with

interest groups for special research
• In transitional countries, an information projects, and business- and problem re-
system for primary healthcare is a strate- lated linking and networking.
gic component of health reform. Its goal is
to improve the quality of primary
healthcare and rationalize its consump- REFERENCES
tion.
• Investments in a primary-healthcare in- Bates, D. W., Ebell, M., Gotlieb, E., Zapp, J., &
formation system have no alternative. Mullins, H. C. (2003). A proposal for electronic
They are strategic and would yield a re- medical records in U.S. primary care. JAMA,
turn of investment in 2 years. 10, 1-10.
• In transitional countries, health computer-
Booth, N. (2003). Sharing patient information
ization projects should be defined (through
electronically throughout the NHS, BMJ, 327,
measurable targets) in phases by priorities
114-115.
and conducted all the way down to their
execution phase. Brage, S., Bensten, B. G., Bjerkedal, T., Nygard,
• New concepts and proven methodologies J. F., & Tellnes, G. (1996). ICPC as a standard
of project management should be used to classification in Norway. Fam Pract, 13, 391-
guarantee the effectiveness of the project 396.
(investment).
Hofmans-Okkes, I. M., & Lamberts, H. (1996).
• In countries in transition, the primary
The international classification of primary care
health service’s information system should
(ICPC): New applications in research and com-
allow physicians and nurses to switch to a
puter based patient records in family practice.
new, facilitated mode of operation that
Fam Pract, 42, 294-302.
gives them more time for patients and
practice-management improvement. Krcmar, N., Stevanovic, R., Kovacic, L., &
• A primary-healthcare information system Merzel, M. (2001). Health center, family medi-
grants access to data to all authorized cine and community health care reform. Sev-
staff needing it, as well as the proactive enth Congress on Family Medicine, 219-227.
use of knowledge, standards, guidelines,
Lamberts, H., & Hofmans-Okkes, I. (1996).
procedures, and algorithms. It should en-
Episode of care: A core concept in family
able direct IT communication between the
practice. J Fam Pract, 42.
citizen, wherever he or she may be, and
hospitals, specialists, home-care services, Lamberts, H., & Wood, M. (Eds.). (1987).
home-visiting services, and diagnostic units, ICPC: International classification of pri-
as well as all later linkages in the process mary care. Oxford: Oxford University Press.
of treatment.
Lamberts, H., Wood, M., & Hofmans-Okkes, I.
• The system should permit the utilization of
(Eds.). (1993). The international classifica-
diagnostic and therapeutic guidelines,
tion of primary care in the European com-
warrant equality in approaching the pa-
munity. Oxford: Oxford University Press.
tients, and make the necessary knowledge
available to the physician.
20
Okkes, I. M., Oskam, S. K., & Lamberts, H. KEY TERMS

(1998). Van klacht naar diagnose [From
complaint to diagnosis]. Bossum, Coutinho. Business Intelligence (BI): The process
of gathering information about a business or
Stevanovic, R. (2002a). Computer software for
industry matter, or a broad range of applica-
family medicine offices. Med Fam Croat, 9,
tions and technologies for gathering, storing,
30-33.
analyzing, and providing access to data to help
Stevanovic, R. (2002b). ICT of the primary make business decisions.
health care. First Croatian Congress on
European Committee for Standardiza-
Telemedicine with International Participa-
tion/TC251 Work Groups (CEN/TC251):
tion, 86-87.
Standardization in the field of health ICT to
Stevanovic, R. (2003). Collection and manag- achieve compatibility and interoperability be-
ing of health data, the linkage of informatics tween independent systems and to enable modu-
systems and quality of health data. In D. larity. This includes requirements on the health
Cvoriscec & V. Madaric (Eds.), Quality man- information structure to support clinical and
agement standards in hospital care (pp. 100- administrative procedures, and technical meth-
109). Koprivnica, MOH. ods to support interoperable systems, as well as
requirements regarding safety, security, and
Stevanovic, R., & Erceg, M. (2003). A pro-
quality.
posal for national preventive medicine informatic
& communication system. First Croatian Con- Extensible Markup Language (XML):
gress on Preventive Medicine and Health A simple, very flexible text format derived from
Promotion with International Participation, SGML (ISO 8879). Originally designed to meet
308. the challenges of large-scale electronic pub-
lishing, XML is also playing an increasingly
Varga, S., & Stevanovic, R. (2003). Health
important role in the exchange of a wide variety
informatics system: Preventive benefits. First
of data on the Web and elsewhere.
Croatian Congress on Preventive Medicine
and Health Promotion with International Health Level Seven (HL7): One of sev-
Participation, 295. eral departments of the American National
Standards Institute (ANSI), whose mission is to
WONCA. (1998). International Classifica-
provide standards for the exchange, manage-
tion Committee: The international classifi-
ment, and integration of data that support clini-
cation of primary care (2 nd ed.). Oxford:
cal patient care and the management, delivery,
Oxford University Press.
and evaluation of healthcare services.
World Health Organization (WHO). (1994).
International Classification of Diseases,
International statistical classification of dis-
10 th Revision (ICD-10): The international
eases and related health problems: Vol. 3.
standard diagnostic classification for general
Alphabetical index (10th revision). Geneva,
epidemiology and many management purposes.
Switzerland: Author.
21
International Classification of Primary Public-Key Infrastructure (PKI): A sys-

Care (ICPC): Instrument necessary to re- tem of digital certificates, certificate authori-
search general practice that was developed by ties, and other registration authorities that verify
the WONCA Classification Committee. It is and authenticate the validity of each party
available in more than 20 languages and has involved in an Internet transaction.
reached its second edition.
22
23
Chapter IV
Computerized Systems
Validation in the
Pharmaceutical Industry
Kashif Hussain
University of Valenciennes, France
Shazia Yasin Mughal

Sylvie Leleu-Merviel
ABSTRACT
This chapter has reviewed the regulatory requirements definitions of FDA, AFSSAPS, EC and
how computer systems can be validated with a relevant approach. By directing validation
activities into a clear action plan, first, we need to analyze the specific organizational need
according to the priorities and impacts to construct the required validation plan which can
fulfill the requirements. The aim of the authors is to focus on a general approach to validation
that can support several needs without addressing the technical aspects.
INTRODUCTION regulatory requirements. In the recent past, an

increase in the use of computerized-system
The term validation appeared in the ’60s, has validation (CSV) compelled quality and valida-
become a topic of business concern today, and tion standards to be more precise to meet the
is considered to be one of the core issues in the industrial need. The utility of CSV has been an
pharmaceutical industries in terms of meeting important catalyst in audits. Today, the phar-
Computerized Systems Validation in the Pharmaceutical Industry
maceutical industry makes use of CSV for Computerized System

several purposes (services, computers, equip-
ment, process verification, change manage- Several definitions are available to define a
ment [CM], etc.), exploring the ways how computerized system. We start with the prin-
validation ensures the system in use is compli- ciple described by the European Parliament in
ant and how to decrease failure rates, risks, Annex-11 (Computerized System), relating to
long-term costs, and so forth. Currently, official medicinal products for human use and investi-
inspections concentrate more and more on the gational medicinal products for human use:
validation of computerized systems due to good
manufacturing practices (GMPs; Hoffmann, The introduction of computerised systems
Kähny-Simonius, Plattner, Schmidi-Vckovski, into systems of manufacturing, including
& Kronseder, 1998). storage, distribution and quality control does
Worldwide regulatory authorities have is- not alter the need to observe the relevant
sued rules, regulations, and guidelines that are principles given elsewhere in the Guide.
aimed to ensure true and real practices in the Where a computerised system replaces a
pharmaceutical organizations concerning pub- manual operation, there should be no
lic health. These requirements to maintain qual- resultant decrease in product quality or
ity processes are the way to ensure that final quality assurance. Consideration should be
consumers receive only safe and effective given to the risk of losing aspects of the
medical products. These authorities verify the previous system which could result from
required compliance before approving the li- reducing the involvement of operators. (EC,
cense to the manufacturer. These require- 2003)
ments are available in several forms like GxP
(good [clinical, laboratory, manufacturing] prac- The Food and Drug Administration (FDA,
tice) or BPF (bonne pratique de fabrication). 1987) defines a computerized system as “com-
Each country has its own interpretation of puter hardware, software, and associated docu-
quality and compliance requirements for the ments (e.g., user manual) that create, modify,
development of systems and procedures to maintain, archive, retrieve, or transmit in digital
achieve quality; however, basic rules and prin- form information related to the conduct of a
ciples remain the same and are universally clinical trial.”
applicable. These regulatory requirements are A computerized system may include data
established to define uniform standards that input, electronic processing, and the output of
emphasize public health and safety as a first information to be used either for reporting or
concern. In all cases, these regulations give an automatic control. It may include automated
overview of the minimum requirements and manufacturing equipment, process-control sys-
dictate what must be done and by whom with- tems, automated laboratory equipment, labora-
out specifying how it is to be done. It is the tory-data capture systems, clinical or manufac-
responsibility of management with respective turing database systems, and so forth. Our
validation teams for identifying schedules, pri- study provides an overview of CSV accep-
orities, and resources required for the prepara- tance criteria in the pharmaceutical industry.
tion to meet these standards. To carry out our study, we used the regulations
applicable via the European Union parliament
24
for France, and for America, we use the FDA tion is defined by AFSSAPS as the establish-
as the reference authority. ment of the proof that the implementation or the
Validation is a process that requires dedi- use of every process, procedure, material, raw
cated attention, and we summarize a clear material, product, activity, or system really
approach to validation that can be incorporated makes it possible to reach the anticipated re-
with quality to implement validation principles. sults and the fixed specifications (Ministry of
We start with some general definitions of vali- Health & Social Protection, France, 2004).
dation used by AFSSAPS (French Health Prod- The directives of 2003/94/EC describe the
ucts Safety Agency), EC, FDA, and ISO (In- validation process:
ternational Organization for Standardization). The extent of validation necessary will de-
We go on to discuss a validation planning pend on a number of factors including the use to
method that may be suitable in general in high- which the system is to be put, whether the
assurance disciplines, and then we link valida- validation is to be prospective or retrospective
tion with quality. We conclude with some pos- and whether or not novel elements are incorpo-
sible future research opportunities in computer- rated. Validation should be considered as part
ized-system validation. of the complete life cycle of a computer sys-
tem. This cycle includes the stages of planning,
specification, programming, testing, commis-
VALIDATION APPROACH sioning, documentation, operation, monitoring
and modifying. (EC, 2003)
Validation remains one of the important issues The FDA (1987) defines validation as
in the pharmaceutical industry as it has for “[e]stablishing documented evidence that pro-
many years. Validation is an important aspect vides a high degree of assurance that a specific
in the age of global competition and the rapidly process will consistently produce a product
changing technology environment not because meeting its predetermined specifications and
of the regulatory requirements only, but also for quality attributes.” The organization addresses
an effective, true approach in the long run for the development of a proper validation plan that
the computer systems used in the industry. performs and documents validation tasks to
Validation is vital for organizational global com- assess software development.
petitiveness, growth, and compliance require- Validation should be considered a part of the
ments. The basic aim is always to assure quality complete life cycle of a computerized system
and that the process is capable to develop a (Stokes, 1998). Validation basically concerns
product meeting the predefined conditions and all steps in the production process that could
respecting all the applicable norms. Validation affect the final quality of the product (Hoffmann
teams are focusing on their role in the develop- et al., 1998). Thus, it is important to specify the
ment of quality orientation validation plans by compatibility and usefulness of the computer
providing valuable insight and guidance in order system, and to define specifications and quality
to keep pace with the latest changing technol- dimensions of the computerized system under
ogy. The major objective behind a validation consideration before developing or purchasing.
strategy is always to assure quality for all Therefore, the documentation of user require-
deliverables of a computer system and to verify ments and the documentation of system speci-
compatibility with all compliance (functional, fications are equally essential (Ermer, 2001).
security, integrity, etc.) requirements. Valida- The FDA in its guidance indicated the expecta-
25
tion of a risk-based approach to determining assumption that an organized approach, while

which systems should undergo validation. Most using one, many, or a combination of these
pharmaceutical and biotechnology profession- models, can result in fewer defects, thus, ulti-
als have experienced the qualitative difference mately providing better results (Uzzaman, 2003).
between making high-risk and low-risk deci- The documented collection of policies, pro-
sions. In low-risk situations, one is often willing cesses, and procedures is equally important.
to gamble and reach hasty, if not faulty, conclu- We will not go into the details of these models
sions. In high-risk situations, one usually takes as they are not in the primary objective of this
more time to gather and analyze all relevant chapter.
information (Woodrum, 1998).
As defined in the ISO/IEC 17025 interna- Validation of a Computerized
tional standard, validation is the “confirmation System
by examination and the provision of objective
evidence that the particular requirements for a The principles of good practices for the use of
specific intended use are fulfilled.” So, for a computerized system via validation are now
validation, both the procedures and systems well established as the means by which quality
components (hardware, software, etc.) that is built into the development of these systems.
can affect the results of measurements need to Major phases of the validation of a computer-
be validated. ized system producing documented evidence
As the validation of computerized systems are described as follows.
is a regulatory requirement for the FDA be-
cause of 21 CFR Part 11, validation can be • Network and infrastructure
treated as a proof of suitability that is linked to • Validation master plan
the whole system life cycle to ensure that the • User-requirements specification (URS)
system meets the user requirements, ensures • Risk analysis
compliance expectations, and will be properly • Supplier audit and vendor selection
maintained to guarantee a secure environment • Legal requirements
with accurate, reliable, and traceable informa- • Functional specification (FS)
tion throughout its life cycle. Validation is a • Validation strategy planning
process that begins with defining the require- • Design specification (DS)
ments and ends with ensuring that the needs are • Implementation and testing
being fulfilled consistently (Carter, 2005). • Qualification phase
Numerous life-cycle models (waterfall • Installation qualification (IQ)
method, enhanced waterfall method, spiral • Operational qualification (OQ)
model, V model, CMM [capability maturity • Performance qualification (PQ)
model], etc.) have been introduced in order to • Standard operating procedures (SOPs)
accomplish software validation. In general, • Training
software validation involves a series of activi- • Final validation report
ties and tasks that must be planned and ex- • Change management
ecuted at required stages depending on the life- • Maintenance and regular review
cycle model used, the software used, the asso- • Decommissioning
ciated risks, and the scope of changes made as
the project advances. This is based on the
26
Network and Infrastructure Tools to be defined and certain numbers of test runs
have to be done (FDA, 1987). Risk analysis
Infrastructure and network services including includes a strategy and methodology, the risk
LAN (local area network), WAN (wide area associated with different parts of the system,
network), intranet, security access, and so forth and an action plan to minimize these risks. The
must provide reliable, trustworthy information, analysis should result in solutions for all risks
data, and delivery, and must be assured and identified. What has been agreed upon during
maintained by detailed specification documen- the supplier or vendor selection and specified
tation (approved procedures) and change man- during the continuous contact with the supplier
agement. might give additional advice to define critical
functions and modules (Friedli, Kappeler, &
Validation Master Plan Zimmermann, 1998). For validation of the soft-
ware, risk analysis is affected module-wise
The master plan provides a complete documen- (Arnold, 1998).
tation list for the computerized system to be
validated according to the business and process Supplier Audit and
priorities. Vendor Selection
User-Requirements Specification Pharmaceutical companies are held respon-

sible for any inadequacy observed by the regu-
URS is a definition of the system requirements latory teams of their selected suppliers. A
in terms of how the user perceives the system supplier audit is recommended and required to
to operate in the intended use and environment. ensure the supplier is able to provide the ser-
The URS should define what one wants to do or vices required with all necessary documenta-
what the system is required to do for its users. tion. This audit includes the verification of a
It may identify integration with other existing quality-management system, standards, tools,
systems, functional needs (e.g., data presenta- programming standards, working conditions,
tion, records, reports), regulatory requirements, market repute, development methodology, cus-
working-environment constraints, system per- tomer support, change management, and so
formance, the maintenance required, the docu- forth. The audit-report findings should be con-
mentation required, and so forth. The final user firmed and verified. Depending on the audit
should be involved in this phase to ensure a results, the supplier or vendor may be asked to
good understanding of the system requirements reply to the audit observations and findings.
and the functionalities required of the installed
system. All URSs should be verified and as- Legal Requirements
sured before final confirmation.
Whenever external service providers are in-
Risk Analysis volved, legal contracts and service-level agree-
ments (SLAs) should be incorporated and as-
One of the most important steps in validation is sured so that the regulatory requirements of the
defining the risk areas. In order to guarantee as company are confirmed.
much security as possible during the routine
utilisation of the system, critical functions have
27
Functional Specification Testing is a sensitive issue, provoking failures

and maximizing confidence, that involves cer-
The FS provides details about what the system tain limitations depending on the goals. Testing
should do to satisfy the user requirements. The helps reliability and involves source-code re-
functional specification, also termed prelimi- view or peer review to identify any errors and
nary functional design, identifies any constraints to verify the codes with applicable coding stan-
or potential deviations from the planned re- dards. To evaluate the system performance
quirements. It should describe what the system against expectations, end users are required to
should do and how it should do it. It explains perform user acceptance tests. A review of the
how each feature of the system functions. user acceptance test results before system
installation at the operation site should be done
Validation Strategy Planning in order to satisfy the requirements. If executed
test plans and test results are verified, orga-
Validation strategy planning is used to define nized, and documented, significant qualification
the validation responsibilities, plan, strategy, workload can be reduced. Traceable records
and approach, and the acceptance criteria. among URS, FS, DS, and testing should be
However, for smaller systems, this phase can maintained at this stage in order to move for
be replaced by the validation master plan. The effective qualification.
final validation report is actually based on vali-
dation strategy planning. Qualification
Design Specification Qualification covers several steps including

installation qualification, operational qualifica-
The DS translates the requirements of the URS tion, and performance qualification; however,
into a technical solution. Both developer and depending upon the size and complexity of the
user involvement is required to facilitate this system, these steps can be carried out in single
phase, and a detailed review during the project stage. Installation and operational qualification
development phase is highly recommended to are usually based on system functionality.
maintain the design methodology. The DS in-
cludes the functional, structural, and technical Installation Qualification
design depending upon the system size and
complexity with detailed documentation. It ex- IQ ensures that the system has been installed
plains how the system is intended to function or against the predetermined design criteria and
operate. has been specified and set up in compliance
with the requirements of both parties (manu-
Implementation and Testing facturer and the user), and that complete docu-
mentation is available to prove this for the
The purpose of implementation and testing is to users. Usually, installation involves checking
assure the suitable implementation of the sys- that all major system hardware, software, tools,
tem components (hardware, software, etc.) peripherals, user manuals, and reference docu-
depending upon the size and complexity of the ments are readily accessible, and that a regular
system. Proper documentation including that of maintenance program along with backup is in
development tools is required at this stage. place.
28
Operational Qualification Training
The IQ of the test system should be completed Training is one of the essential elements for any
with complete SOPs and personal training be- successful validation. All concerned members
fore launching OQ. OQ is carried out as a (users, developers, and persons involved in
documented verification to ensure that the sys- development, validation, maintenance, and so
tem is in optimal working conditions and in- forth) of the validation project must be trained.
volves all functionalities and process opera- Training records must be documented and main-
tions, security access, and backup facilities tained. Training should be performed in a test
according to the predefined specifications. OQ environment in order to avoid any effects on the
should be performed in a test environment system. Typically, training initiates within the
when possible. After test execution, a review validation group. It is essential that the lead
report with expected and achieved results, de- validation resource for a given validation project
viations found, and a conclusion must be ap- initiate, facilitate, coordinate, and/or communi-
proved to commence PQ. cate the need for resource training (Neal, 2003).
Performance Qualification Final Validation Report
PQ is documented verification to assure that A review of all activities is done after PQ to

the system operates throughout all operating perform the final system release verification
ranges according to the requirements (stable and data transfer plan, if any, to assure all
system, operational documents in place, trained deliverables identified in the validation plan and
personnel) over a predefined period of time, to sum up the completion of the validation cycle
and that the results are in accordance with the via a final validation report. Based on positive
expectations in the system’s actual use envi- conclusions, the computer system can be re-
ronment. System stability through continuous leased for use (Friedli, 1998).
operation for a significant amount of time should
be confirmed. This allows encountering a wide Change Management
range of conditions to verify any error or fault
in system that does not appear in normal activi- Any changes to the system must be thoroughly
ties. documented and managed by a change-man-
agement process. CM verifies all impacts of a
Standard Operating Procedures proposed or planned change on the system
before that change is done. CM defines the
Approved SOPs must capture the details to complete set of processes employed for track-
standardize the validation-procedure perfor- ing all changes on systems, applications, and
mance. Departments charged with abiding by environments defined by a respective pharma-
or following these SOPs must first be trained ceutical regulatory conformance standard, and
against them. SOPs may involve system opera- manages systematically the effects of change.
tion, backup, restoration, security, data han- The changes are evaluated in order to decide
dling, change management, disaster recovery, the extent of validation required by the imple-
training, and so forth. mentation of the change.
29
Maintenance and Regular Review achieve customer satisfaction while allowing

for the system to be considered safe and effi-
The purpose of this phase is to assure the cient by the legal authorities of the company.
optimal use of the system by properly maintain-
ing the validation until the system is decommis-
sioned. Periodic reviews should be performed DISCUSSION
and documented including reviews of SOPs,
security, access, change management, backup Validation is highly dependent upon the effec-
files, training records, and so forth. tiveness of the planning that goes into it. In
many cases, validation may require consider-
Decommissioning able resources in terms of time, money, and
specialized personnel, so getting it right and
Decommissioning is done to retire a system getting it done on time is crucial. Organized,
after the completion of its intended use and well-planned validation increases the reliability
when it is no more required, using proper proof a computerized system, thereby decreasing
cedural methods and keeping in mind the im- compliance problems and increasing the
pact on the rest of the system. company’s confidence to fulfill the require-
The CM process should be used to track all ments of regulatory agencies. A better under-
changes related to the system decommissioning standing of the validation of a computerized
to avoid any impact on other systems. An system is of high importance and will continue
update of the disaster recovery plan (DRP) and to play an important role in various aspects to
the deactivation of SOPs if applicable are equally minimize risk and enhance compliance, overall
recommended. It is estimated that only 15% of effectiveness, and efficiency. Obviously, achiev-
organizations have attested to having an effec- ing the validation of a computer system involves
tive DRP plan (Snelham & Wingate, 2000). developing, maintaining, and releasing a com-
puterized system throughout its life cycle and
Quality Approach assures that the system consistently meets it
specifications, thus, making it worthy for its
In general, quality can be specified as the ability intended use. The cost of validation is deter-
of a product or service to fulfill the require- mined by the time spent on documentation, the
ments and expectations defined by the custom- development of protocols, SOPs, actual field-
ers. Quality is a dynamic issue that should lead work, data collection, and analysis (Stocker,
toward continuous improvement. We believe 1994). It is interesting to note that a good rule
that validation is directly linked with the quality of thumb is that total validation costs may run
of a system, and the use of quality tools can from 4 to 8% of the total project cost for typical
enhance the validation plan and assist in suc- pharmaceutical-plant expansion projects
cessful validation. (Gloystein, 1997). Some researchers have ex-
Various studies have considered validation perienced system for which validation reduced
as a part of a quality system (Randsell, 1996), maintenance-support costs by 75% (Wingate,
quality assurance (Muller, Gempler, Schweie, 1997).
& Zeugin, 1996), and total quality management We are hopeful that advancement in tech-
(Christensen, Kristiansen, Hansen, & Nielsen, nologies, standardization, and the harmonizing
1995), involving all resources (manpower, ma- of the definitions of the required validation
chines, knowledge, etc.) of a company to regulatory requirements will, in general, help
30
any organization develop a standard, consis- outline here encompasses the entire pharma-
tent, global approach for computerized sys- ceutical business environment. Validation should
tems. In the future, the key to mastering smart be regarded as part of an integrated concept to
validation will lie in weighing the real risks that ensure the quality, safety, and efficacy of phar-
can affect the computerized system. maceuticals (Ermer, 2001).
CONCLUSION REFERENCES
This chapter has reviewed the regulatory re- Arnold, L. (1998). Software assessment under
quirement definitions of the FDA, AFSSAPS, consideration of validation aspects: PPS and
and EC, and how computer systems can be PMS systems. Pharmaceutica Acta Helvetiae,
validated with a relevant approach. In directing 72, 327-332.
validation activities into a clear action plan, we
Carter, E. R. (2005). Systems validation: Appli-
first need to analyze the specific organizational
cation to statistical programs. BMC Medical
needs according to priorities and impacts to
Research Methodology, 5, 3.
construct a required validation plan that can
fulfill the requirements. Our aim is to focus on Christensen, J. M., Kristiansen, J., Hansen, A.
a general approach to validation that can sup- M., & Nielsen, J. L. (1995). Method validation:
port several needs without addressing the tech- An essential tool in total quality management.
nical aspects. It must be recognized that the The Proceedings of Sixth International Sym-
validation of a computerized system is highly posium on the Harmonization of the Role of
important. A well-defined validation plan can Laboratory Quality Assurance in Relation to
formalize the validation process, the associated TQM (pp. 46-54).
strategy, the responsibilities, and the accep-
EC. (2003). Directive of 2003/94/EC of 8 Oc-
tance criteria applicable to the computerized
tober 2003: Laying down the principles and
system, including risk analysis. The approach
guidelines of good manufacturing practice in
presented here can help to avoid some of the
respect of medicinal products for human use
common pitfalls in the validation of computer-
and investigational medicinal products for hu-
ized systems. The real challenge is in how a
man use. In Annex 11: Computerised system:
company plans for and documents the valida-
Vol. 4. Medicinal products for human and
tion activities for the system in order to better
veterinary use: Good manufacturing prac-
implement these guidelines in the least possible
tice (p. 11). Retrieved April 13, 2005, from
time and with the least possible resources. We
http://pharmacos.eudra.org/F2/eudralex/vol-4/
believe that the guidelines for validation can
home.htm
lead to a common approach and to establishing
common procedures and vocabulary. Valida- Ermer, J. (2001). Validation in pharmaceutical
tion can further lead to the harmonization of analysis: Part 1. An integrated approach. Jour-
audit inspections and the improvement of in- nal of Pharmaceutical and Biomedical Analy-
spection systems, thus improving the chances sis, 24, 755-767.
for successful inspection. Together with the
Food and Drug Administration (FDA). (1987).
general rules of validation applicable world-
Guidelines on general principles of process
wide, it is clear that the validation process
31
validation. Retrieved April 14, 2005, from Stokes, T. (1998). The survive and thrive
http://www.fda.gov/cder/guidance/pv.htm guide to computer validation. Buffalo Grove,
II: InterPharm Press.
Friedli, D., Kappeler, W., & Zimmermann, S.
(1998). Validation of computer systems: Prac- Uzzaman, S. (2003). Computer systems valida-
tical testing of a standard LIMS. Pharmaceutica tion: A system engineering approach. Pharma-
Acta Helvetiae, 72, 343-348. ceutical Engineering, 23(3), 52-66.
Gloystein, L. (1997). Protocol structure and IQ/ Wingate, G. A. S. (1997). Validating auto-
OQ costs. Journal of Validation Technol- mated manufacturing and laboratory appli-
ogy, 3(2), 140. cation: Putting principles into practices.
InterPharm Press.
Hoffmann, A., Kähny-Simonius, J., Plattner,
M., Schmidi-Vckovski, V., & Kronseder, C. Woodrum, D. T. (1998). Computer system
(1998). Computer system validation: An over- validation: Value added activities meeting regu-
view of official requirements and standards. latory imperatives. Drug Information Jour-
Pharmaceutica Acta Helvetiae, 72, 317-325. nal, 32, 941-945.
Ministry of Health & Social Protection, France.
(2004). Décrets, arrêtés, circulaires. Journal KEY TERMS
Officiel de la République française, 23(95).
Retrieved May 10, 2005, from http:// 21 CFR Part 11: FDA code of U.S. Fed-
agmed.sante.gouv.fr/htm/3/pta/ eral Regulations, Title 21, Part 11: Electronic
ptaa1_190804.pdf Records. Electronic signatures that define pa-
rameters by which pharmaceutical companies
Muller, K. M., Gempler, M. R., Schweie, M.-
can author, approve, store, and distribute records
W., & Zeugin, B. T. (1996). Quality assurance
electronically.
for biopharmaceuticals: An overview about
regulations methods and problems. AFSSAPS: The French Health Products
Pharmaceutica Acta Helvetiae, 71, 421-438. Safety Agency is the French authority for all
safety decisions taken concerning health prod-
Neal, C. (2003). Prerequisites for successful
ucts from their manufacturing to their market-
validation. Journal of Validation Technol-
ing, that is, medicinal products; raw materials
ogy, 9(3), 240-245.
for pharmaceutical use; organs, tissues, and
Randsell, T. E. (1996). The cost of validation. cells, and products of human and animal origins,
Journal of Validation Technology, 3(2), 142- and so forth.
143.
Bonne Practique de Fabrication (BPF):
Snelham, M., & Wingate, G. (2000). Validation Stands for bonne pratique de fabrication and is
laboratory information management systems. the equivalent of GMP.
Journal of Validation Technology, 6(4), 740-
Change Management: The complete set
748.
of processes employed for tracking all changes
Stocker, A. C. (1994). Why does validation in existing and new systems, applications, and
cost so much and take so long? And what we environments defined by a respective pharma-
can do about it. Journal of Validation Tech- ceutical regulatory conformance to mange sys-
nology, 1(1), 5-8. tematically the effects of change.
32
Capability Maturity Model (CMM): The duct, monitoring, recording, auditing, analysis,
capability maturity model is a method for evalu- and reporting of studies applicable to clinical or
ating and measuring the maturity of the soft- human studies in the evaluation of drugs, medi-
ware-development process of organizations on cal devices, biological products, and so forth.
a scale of 1 to 5. The Software Engineering The FDA ensures product quality applicable to
Institute (SEI) at Carnegie Mellon University in GCP via 21 CFR Parts 50, 54, and 56.
Pittsburgh, USA, developed the CMM.
Good Laboratory Practice (GLP): Good
Computerized System: A system that in- laboratory practice embodies an international
cludes software, hardware, application soft- set of quality regulations and guidelines appli-
ware, operating-system software, supporting cable to nonclinical studies in the evaluation of
documentation, and so forth, for example, auto- drugs, medical devices, biological products, and
mated laboratory systems; control systems; so forth. GLP provides a framework within
manufacturing-, clinical-, or compliance-moni- which the FDA ensures laboratory studies are
toring database systems; and so forth. planned, performed, monitored, recorded, re-
ported, and archived under 21 CFR Part 58.
Current Good Manufacturing Practice
(cGMP): Current good manufacturing prac- GxP: GxP is the generalization of any qual-
tice is an international set of quality regulations ity guidelines used in the pharmaceutical indus-
and guidelines applicable to the manufacture, try that groups together the following compli-
testing, and distribution of drugs, medical de- ance practices: cGMP, GLP, and GCP.
vices, diagnostic products, biological products
International Organization for Standard-
for human or veterinary use, and so forth. The
ization (ISO): The International Organization
FDA ensures product quality applicable to GMP
for Standardization is responsible for a wide
via the 21 CFR Part 210 and 211 regulations.
range of standards for a large number of indus-
Disaster Recovery Plan (DRP): A disas- tries. ISO provides a means of verifying that a
ter recovery plan is the plan for business con- proposed standard has met certain require-
tinuity in the event of a disaster that destroys ments for due process, consensus, and other
business resources. The goal of DRP is to criteria by those developing the standard.
recover the technical infrastructure that sup-
Life-Cycle Model: A life-cycle model is
ports business continuity in the event of a
the order in which a series of processes are
serious incident in the least possible time.
performed to create or update a product or
Food and Drug Administration (FDA): service. Examples include the waterfall, incre-
The Food and Drug Administration is the branch mental-build, and evolutionary-build approaches.
of the U.S. federal government that approves
Service-Level Agreement (SLA): Pro-
new drugs for sale and is responsible for ensur-
vides the predefined and standard level of
ing the safety and effectiveness of all drugs,
services available in the organization.
biologics, vaccines, and medical devices, in-
cluding those used in diagnosis, treatment, regu- Spiral Model: Also known as the spiral
lation, and so forth for the USA. life-cycle model, it is a systems-development
method used in information technology. This
Good Clinical Practice (GCP): Good clini-
model of development combines the features of
cal practice is an international set of quality
the prototyping model and the waterfall model.
regulations and guidelines for the design, con-
33
The spiral model is favored for large, expen- overall validation approach including qualifica-
sive, and complicated projects. tion. Validation includes but is not limited to
manufacturing processes, equipment, comput-
Standard Operating Procedure (SOP):
erized systems, and so forth.
A standard operating procedure is a written
document that describes in detail how a particu- V Model: The V model is a software-
lar procedure or method is executed for repeti- development model that describes the activities
tive use. It is generally intended to standardize and results that have to be produced during
the procedure performance. software development.
Validation: Documented evidence provid- Waterfall Model: Introduced in 1970, the
ing a high degree of assurance that a specific waterfall model is a software-development
process under consideration does what it pur- model describing the theoretical approach to
poses to do. Validation deals with the entire software development in which development is
system life cycle to ensure a system satisfies seen as flowing steadily through the phases of
user requirements, meets compliance expecta- requirements analysis, design, implementation,
tions, and is adequately maintained to provide a testing (validation), integration, and mainte-
secure environment with accurate, reliable, and nance. It is the basis for the V model used in the
traceable information from conception to re- pharmaceutical industry.
tirement. The term is also used to describe the
34
35
Chapter V
Chronic Disease Registers in
Primary Healthcare
M. F. Harris
University of New South Wales, Australia
D. Penn
J. Taggart
Andrew Georgiou
J. Burns
G. Powell Davies
ABSTRACT
Systematic care of patients with chronic diseases needs to be underpinned by information

systems such as disease registers. Their primary function is to facilitate structured care of
patients attending services—supporting identification of patients at risk, structured preventive
care and provision of care according to guidelines, and supporting recall of patients for
planned visits. In Australia general practitioners using division-based diabetes registers are
more likely to provide patient care that adhered to evidence-based guidelines. Critical data
issues include privacy, ownership, compatability, and capture as part of normal clinical care
and quality.
Chronic Disease Registers in Primary Healthcare
INTRODUCTION patients at risk, structured preventive care, the

provision of care according to guidelines, and
Chronic diseases account for over 70% of the the recall of patients for planned visits.
burden of disease in countries such as Australia
(Mathers, Vos, & Stevenson, 1999), and more Location
than one in four problems managed by general
practitioners relate to one or more of these Registers may be located at the service (e.g.,
chronic conditions (Australian Institute of within a general practice) or at the healthcare
Health and Welfare [AIHW], 2002). System- organization level (e.g., a diabetes centre or the
atic team care of patients with chronic diseases Division of General Practice). Registers within
such as diabetes and cardiovascular disease is practices or services have the advantage of
associated with improved health outcomes ease of data capture, avoiding problems of data
(Dunn & Pickering, 1998; Wagner, 1998; World transfer. Registers at the district or regional
Health Organization [WHO], 2001). This needs level held by a specialized service or primary-
to be underpinned by information systems that healthcare organization have greater capacity
assist with recall and audit according to evi- for analysis and are able to monitor the care
dence-based guidelines. provided by a multidisciplinary team across
In many countries, organizations of primary- different services.
care services have been established at local or
district levels. In Australia, the Divisions of Data
General Practice have played a key role in
supporting general practices to provide more Registers contain individual-identifying infor-
systematic care through disseminating evidence- mation, such as basic demographic profiles for
based guidelines, educating general practitio- recall purposes, together with information about
ners and consumers, supporting shared care the process and outcomes of care. This infor-
and self-management education, providing al- mation needs to be standardized to allow com-
lied health services, and coordinating local reg- parison using nationally accepted units of mea-
isters for recall and audit. The authors have surement and frequencies for routine testing
been involved in one widely implemented com- and recall. In Australia, standardized minimum
puter-based register system that has demon- data sets have been developed based on evi-
strated improvements in the quality of care. dence-based guidelines for the management of
diabetes or cardiovascular disease (National
Health Data Committee, 2003a, 2003b), over-
BACKGROUND seen by committees representative of primary-
care providers, nongovernment organizations
Functions (NGOs), specialist providers, government agen-
cies, and consumers.
Disease registers have a variety of purposes
ranging from facilitating longitudinal research Data Capture
to providing epidemiological surveillance. Their
primary function in primary healthcare is to Patient data may be recorded and captured in a
facilitate the structured care of patients attend- variety of ways. Historically, data have been
ing services: supporting the identification of recorded on forms or copies of patient-held records,
36
Figure 1. Sources of data capture for chronic disease registers in Australian general practice
extraction remote access pathology

from medical to register providers
record extraction
special form or
patient held e-mail Web or intranet
record diabetes
e-mail or centers or
mail chronic intranet hospitals
disease
register
which are subsequently entered from paper for- such registers may be an integral part of the
mat into a computer database either at the prac- patient registration system. In the United King-
tice or division level. Increasingly, data are either dom, case ascertainment via an electronic-
extracted from electronic health records (EHRs) record linkage method showed high concor-
and sent to a register or entered directly into a dance between general-practice registers and
Web-based register. HL7 (Health Level 7) mes- data collated from various sources (Morris et
sage specifications and agreed standard EHRs al., 1997).
allow register and recall developers to design In Australia, registers may be held within
systems that feature seamless interoperability for general practices, at a Division of General
the communication of clinical data. Practice, at a specialist centre, or at a hospital.
A survey of 81 Divisions of General Practice in
Reporting 2002 revealed that 31 had an electronic-regis-
ter recall system for diabetes and 8 had one for
Reports facilitate the functions of the register, CVD. The number has steadily increased since
principally providing prompts and audit reports 1993 (Georgiou, Burns, Penn, Infante, & Har-
against standards or other peer services (see ris, 2004; Penn, Burns, Georgiou, Powell Davies,
Table 1). & Harris, 2004). The National Integrated Diabe-
tes Program established by the Australian gov-
Implementation ernment in 2001 introduced incentive payments
for practices having a disease register to help
In countries where patients are required to support best-practice care and for the comple-
register with general practices, establishing tion of an annual cycle of care for diabetes.
Table 1. Types of reports generated from chronic disease registers

Function Reports
Recall Lists or recall notices for patients who are overdue for aspects of
care
Audit Frequency of visits or assessments
Behavioural or physiological risk factors
Intermediate health outcomes
Complications, hospitalizations, and death
Management: prescribing and referral
Follow-Up Lists of patients with poor control for interventions (e.g., referral)
Accountability Lists of patients for financial reporting or claims
Service Health-service attendance, workload, and population coverage
Management
37
Effectiveness and workshops to facilitate quality and service

improvement (Burns, Zingarelli, & Harris, 2002).
Register-recall systems are important facilita-
tors to structured evidence-based care for the
improved quality of clinical care (Weingarten et CRITICAL ISSUES
al., 2002). Research in Australia has shown
that general practitioners using division-based Privacy and Consent
diabetes registers were more likely to provide
patient care that adhered to evidence-based Privacy and consent are important issues, es-
guidelines (Harris, Priddin, Ruscoe, Infante, & pecially in sending any identified data from the
O’Tool, 2002). This has also been demon- practice to a district or health-service-organi-
strated for the secondary prevention of coro- zation register. Patients need to be informed
nary artery disease (Moher et al., 2001). that information will be sent to the register and
Recall or reminders for patients have been consent to the transfer of any identified infor-
shown to improve adherence with planned or mation (Burns et al., 2000). Consumers and
preventive care (Pirkis et al., 1998). However, other stakeholders need to be involved in deci-
it is important to ensure that such reminders are sions about the data to be collected, who has
culturally appropriate, are personalized, and access to them, and their reporting (CGPIS,
address financial barriers to attending the ser- 2004).
vice (Hunt et al., 1998).
Data Ownership
Registers in Australian Divisions of
General Practice Data ownership is also an issue for data col-
lated from practices or local services to a
In Australia, the Divisions Data Quality Im- district or primary-care organizational level
provement Program is based at the Centre for (Cromwell et al., 2001). In Australia, there has
General Practice Integration Studies (CGPIS) been extensive debate on this issue. This has
at the University of New South Wales (UNSW). led to the development of policies for data
The program provides resources, training, and extraction to the various levels (Carter et al.,
support for division diabetes and cardiovascu- 2000).
lar shared-care programs, especially those with
diabetes and/or cardiovascular-disease regis- Data Compatibility
ters based on the diabetes (National Diabetes
Outcomes Quality Review Initiative Data compatibility is a critical issue for regis-
[NDOQRIN], 2002) and cardiovascular-dis- ters so that practices can exchange and com-
ease data sets (CVDATA, 2002). The program pare data. This is dependent on a standardized
has collated data from division registers in 1999 minimum data set and a common coding system
(Carter, Bonney, Powell Davies, & Harris, for clinical software (CGPIS, UNSW, 2004).
2000) and in 2002 (covering the period of 2000
to 2002). Coverage of the estimated number of Data Capture
people with diagnosed diabetes for division
registers was estimated to be 20.2% in 2002. While the technology to capture clinical data is
Collective results were fed back to divisions available, there is a need for a national
38
interoperable clinical messaging infrastructure, FUTURE DIRECTIONS

messaging software that is minimally intrusive
to the clinicians’ work flow, and a seamless In Australia, we are likely to see a continued
exchange of data between messaging systems move of some functions, such as recall, from
(GPCG, 2004). the regional or primary-care organizational level
to services or practices. This will make issues
Data Quality of standardization and compatibility even more
important in allowing the comparison of acces-
Data quality is critically important in any infor- sibility and the performance of services. At the
mation system. Medical records that are the primary-care-organization or district levels, we
source of data for registers have significant will see the development of at least three
limitations because of their quality (Del Mar, register models:
Lowe, Adkins, & Arnold, 1996).
• A centralized database at the health-orga-
Coverage nization level with Web or intranet input of
data from local services
Participation in chronic disease registers is • A practice register-recall system with the
usually voluntary both for patients and provid- periodic extraction of data in identified or
ers. Data on such registers are therefore unidentified form for audit and quality
rarely representative of all providers or the improvement to the local or regional orga-
population of people with chronic diseases. In nization
general, such registers tend to include those • A data warehouse with linked data held in
with better quality of care (Nilasena & Lin- services and organizations (e.g., pathol-
coln, 1995). ogy providers)
Sustainability Choices between such systems will be made

on the basis of the resources required to set and
The sustainability of regional registers is a maintain them over time both within individual
critical issue. In Australia, the growth of divi- practices and primary-care organizations. Both
sion-based register-recall systems peaked in practices and primary-care organizations will
2002 and may now be declining. Ironically, this require strong direction from government to
decline may partly be due to the development of develop the capacity to allow their wider imple-
practice computer systems because of the dif- mentation.
ficulty of getting data from the practice to the
division in electronic form. Other factors con-
tributing to problems of sustainability include REFERENCES
the lack of capacity within divisions, competing
demands, or the lack of incentives for audits, Australian Institute of Health and Welfare
which allow comparison between practices or (AIHW). (2002). Australia’s health 2002.
services. Canberra, Australia: Author.
39
Burns, J., Carter, S., Bonney, M., Powell Davies, mation to general practices within the
G., & Harris, M. F. (2000). National divisions Illawarra derived from routinely collected,
diabetes program data collation project: computerised data. A second report on data
Vol. 7. Policy framework for collation of analysis and GP reaction. Report for the
data involving divisions of general practice. Commonwealth Department of Health and Fam-
Sydney, Australia: University of New South ily Services (p. 45). Wollongong: Centre for
Wales, School of Community Medicine, Centre Health Services Development.
for General Practice Integration Studies.
CVDATA. (2002). A minimum clinical data
Burns, J., Zingarelli, G., & Harris, M. F. (2002). set for the management of cardiovascular
Quality improvement in diabetes at practice disease in primary care. Melbourne, Austra-
and division level. General Practice and Pri- lia: National Heart Foundation.
mary Health Care Research Conference. Re-
Del Mar, C., Lowe, J. B., Adkins, P., & Arnold,
search: Making a Difference to Health and
E. (1996). What is the quality of general prac-
Health Care, Melbourne, Australia.
titioner records in Australia? Australian Fam-
Carter, S., Bonney, M., Powell Davies, G., & ily Physician, 13, S21-S25.
Harris, M. F. (2000). NDDP data collation
Dunn, N., & Pickering, R. (1998). Does good
project: Summary volume (Diabetes pro-
practice organization improve the outcome of
files). Sydney, Australia: The Centre for Gen-
care for diabetic patients? British Journal of
eral Practice Integration Studies.
General Practice, 48(430), 1237-1240.
Centre for General Practice Integration Stud-
Georgiou, A., Burns, J., Penn, D., Infante, F., &
ies (CGPIS), University of New South Wales
Harris, M. F. (2004). Register-recall systems:
(UNSW); Chronic Illness Alliance; & Austra-
Tools for chronic disease management in gen-
lian Divisions of General Practice. (2004).
eral practice. Health Information Manage-
General practitioners’ information data ex-
ment Journal, 33, 31-35.
change divisions’ capacity project final re-
port (GPCG Project 4 RFT 72/0203). Re- GPCG. (2004). HL7 systems and services:
trieved from http://www.gpcg.org/projects/ GPs information and data exchange evalu-
CurrentProjects.html ating messaging project final report (GPCG
Project E RFT 72/0203). Retrieved from http:/
Centre for General Practice Integration Stud-
/www.gpcg.org/projects/CurrentProjects.html
ies (CGPIS), University of New South Wales
(UNSW); Department of General Practice, Harris, M. F., Priddin, D., Ruscoe, W., Infante,
University of Adelaide; Chronic Illness Alli- F. A., & O’Tool, B. I. (2002). Quality of care
ance; Australian Divisions of General Practice; provided by general practitioners using or not
Collaborative Centre for e-Health, University using division based registers. Medical Jour-
of Ballarat; & Centre for Health Informatics, nal of Australia, 177, 250-252.
UNSW. (2004). Chronic disease functional-
Hunt, J. M., Gless, G. L., et al. (1998). Pap
ity project final report. Retrieved from http:/
smear screening at an urban aboriginal health
/www.gpcg.org/projects/CurrentProjects.html
service: Report of a practice audit and an
Cromwell, D., Arora, M., Dalley, A., Kreis, I. evaluation of recruitment strategies. Australia
(2001). The Epi-Beacon: A feasibility study and New Zealand Journal of Public Health,
on providing local epidemiological infor- 22, 720-725.
40
Mathers, C., Vos, T., & Stevenson, C. (1999). Orstein, S. M., Garr, D. R., et al. (1991).
The burden of disease and injury in Austra- Computer-generated physician and patient re-
lia (AIHW Cat. No. PHE 17). Canberra, Aus- minders: Tools to improve population adher-
tralia: Australian Institute of Health and Wel- ence to selected preventive services. Journal
fare. of Family Practice, 32, 82-90.
Moher, M., Yudkin, P., Wright, L., Turner, R., Penn, D. L, Burns, J. R, Georgiou, A., Powell
Fuller, A., Schofield, T., et al. (2001). Cluster Davies, G., & Harris, M. F. (2004). Evolution of
randomised controlled trial to compare three a register recall system to enable the delivery of
methods of promoting secondary prevention of better quality of care in general practice. Health
coronary heart disease in primary care. British Informatics Journal, 10, 163-174.
Medical Journal, 322, 1338.
Pirkis, J., Jolley, D., et al. (1998). Recruitment
Morris, A. D., Boyle, D. I. R., MacAlpine, R., of women by GPS for paper tests: A meta-
Emslie-Smith, A., Jung, R. T., Newton, R. W., et analysis. British Journal of General Prac-
al. (1997). The diabetes audit and research in tice, 48, 1603-1607.
Tayside Scotland (DARTS) study: Electronic
Wagner, E. H. (1998). Chronic disease manage-
record linkage to create a diabetes register.
ment: What will it take to improve care for chronic
British Medical Journal, 315(7107), 524-528.
disease? Effective Clinical Practice, 1, 2-4.
National Diabetes Outcomes Quality Review
Weingarten, S. R., Henning, J. M.,
Initiative (NDOQRIN). (2002). NDOQRIN
Badamagarau, E., Knight, K., Hasselblad, V.,
national minimum data set for the manage-
Gano, A., et al. (2002). Interventions used in
ment of diabetes. Australia: Australian Diabe-
disease management programmes for patients
tes Society.
with chronic illness: Which ones work? Meta-
National Health Data Committee. (2003a). analysis of published reports. BMJ, 325(7370),
Other data set specification: Cardiovascular 925-932.
disease (clinical) (AIHW Cat. No. HWI-46).
World Health Organization (WHO). (2001). In-
In National health data dictionary (pp. 62-
novative care for chronic conditions: Build-
63). Canberra, Australia: Australian Institute
ing blocks for action. Retrieved from http://
of Health and Welfare.
www.who.int/chronic_conditions/framework/en/
National Health Data Committee. (2003b).
Other data set specification: Diabetes (clinical)
KEY TERMS
(AIHW Cat. No. HWI-47). In National health
data dictionary (pp. 64-66). Canberra, Aus-
Audit: An analysis of data on patient care
tralia: Australian Institute of Health and Wel-
or the outcomes used to compare performance
fare.
against standards or other services.
Nilasena, D. S., & Lincoln, M. J. (1995). A
Chronic Disease: A disease present for 3
computer-generated reminder system improves
months or more.
physician compliance with diabetes preventive
care guidelines. Proceedings of the Annual Data Linkage: Linkage of individual health
Symposium of Computer Applications in data from different databases using a common
Medical Care 1995 (pp. 640-645). identifier.
41
Health Data Dictionary: Standard data Recall: A database generates reminders

definitions used in healthcare. for patients to receive planned care.
Physiological or Behavioral Risk Fac- Register: A database of patients with
tors: They increase the risk of a chronic dis- chronic diseases used to identify patients for
ease developing. follow-up, reminders, or audit.
42
44
Chapter VI
Standardization in Health and
Medical Informatics
Josipa Kern
Andrija Stampar School of Public Health, Zagreb University Medical School, Croatia
ABSTRACT
Standard is a thing or quality or specification by which something may be tested or measured.

The development of standards is organized on a global, international level, existing also on
a national level, well harmonized with an international one. International developers are
organizations working on this matter, like the International Organization for Standardization
(ISO) or the European Committee for Standardisation (Comité Européen de Normalisation—
CEN). Standards in health and medical informatics enable access to patient health records to
read or to add some new data relevant to other healthcare providers taking care of a patient.
Bad medical informatics can lead to patient deaths, and standardization in the field can
prevent this from happening.
INTRODUCTION according to strictly defined specification. Con-

sequently, it does not matter who produces the
When things go well, often it is because they components or where they are produced.
conform to standards (International Organiza- Industry put in the first demand for stan-
tion for Standardization [ISO], 2005). In the dards. Standardization is especially important
Oxford Dictionary of Modern English, there is for electronics, and for ICT and its application in
a lot of explanation of what the word standard different areas. Nowadays, the developing of
means, but, in the context of the first sentence, standards is organized on a global, international
the best meaning is the following: “standard is level, but it exists also on the national level, which
a thing or quality or specification by which is well harmonized with the international one.
something may be tested or measured.” A Developers of standards are organizations
personal computer is a standardized computer. and groups working on this matter. The leading
This means that all of its components are made standard developer in the world is the Interna-
Standardization in Health and Medical Informatics
tional Organization for Standardization. ISO is ISO and CEN have technical committees
a nongovernmental organization that was es- working in specific areas. ISO/TC215, estab-
tablished February 23, 1947. Its mission is to lished in 1998, and CEN/TC251, established in
promote the development of standardization 1991, are corresponding technical committees
and related activities in the world with a view to working on standardization in health and medical
facilitating the international exchange of goods informatics. Both standardization bodies coop-
and services, and to developing cooperation in erate and mutually exchange their standards.
the spheres of intellectual, scientific, techno- There are also a variety of other organiza-
logical, and economic activity (International tions and groups developing standards, either
Organization for Standardizaton, 2005). ISO cooperating with ISO and CEN or acting as
collaborates with its partner in international stan- administrators and coordinators in standardiza-
dardization, the International Electrotechnical tion. For example, there are Health Level 7
Commission (IEC), a nongovernmental body (HL7); Digital Imaging and Communications in
whose scope of activities complements ISO’s. Medicine (DICOM); the American National
The ISO and the IEC cooperate on a joint basis Standards Institute (ANSI), a nonprofit organi-
with the International Telecommunication Union zation that administers and coordinates the
(ITU), part of the United Nations organization U.S. voluntary standardization and conformity
whose members are governments. The ISO assessment system, and so forth.
standard can be recognized by the ISO logo,
ISO prefix, and the designation “International
Standard.” BACKGROUND
The European developer of standards is the
European Committee for Standardisation Definition
(Comité Européen de Normalisation, CEN). It
was founded in 1961 by the national standards A standard is a set of rules and definitions that
bodies in the European Economic Community specify how to carry out a process or produce
and EFTA countries. CEN promotes voluntary a product, or more precisely, a standard is a
technical harmonization in Europe in conjunc- document established by consensus and ap-
tion with worldwide bodies and its partners in proved by a recognized body that provides, for
Europe, and the conformity assessment of prod- common and repeated use, guidelines or char-
ucts and their certification (Comité Européen acteristics for activities or their results, aimed
de Normalisation, 2005). CEN cooperates with at the achievement of the optimum degree of
the European Committee for Electrotechnical order in a given context.
Standardization (CENELEC) and the European The main role of a standard is raising the
Telecommunications Standards Institute (ETSI). levels of quality, safety, reliability, efficiency,
A product of this cooperation is the European and interchangeability, and consequently low-
standard, which can be recognized by the prefix ering costs (International Organization for Stan-
EN. Any added prefix to the existing one, for dardization, 2005).
both the ISO and CEN standards, means that
this standard is the result of cooperation with Standard Creation Process
another standardization group or organization.
The prefix ENV in European standardization There are several phases in the process of
means that this standard is not yet a full stan- standardization. The first phase of this process
dard (it is under development by CEN). is characterized by demand for a standard.
45
There must be someone who needs a standard. be summarized as the practical realization of
Most standards are prepared at the request of ontology (Rodrigues, Trombert Paviot, Martin,
industry. The European Commission can also Vercherin, & Samuel, 2002). The standard
request the standards bodies to prepare stan- CEN ENV 12924 contains a security categori-
dards in order to implement European legisla- zation model for information systems in
tion. This standardization is mandated by the healthcare, distinguishing six categories plus
Commission through the Standing Committee some refinements. For each category, it speci-
of the Directive in support of the legislation. fies the required protection measures
Groups of users can also ask for a standard in (Louwerse, 2002). Standards support
the field of their interest. The second phase of interoperability and electronic-health-record
standardization is the developing of standards (EHR) communication. They support coopera-
by following specifications based on the needs tive work among health agents when it is nec-
defined in the first phase. Experts of a specific essary to share health-care information about
field work on related standards. After the stan- patients in a meaningful way. Examples of
dard has been approved by the standardization requirements for EHR are provided in four
body, it becomes a prototype and goes through themes: EHR functional requirements; ethical,
testing and evaluation. Positive test results legal, and security requirements; clinical re-
imply the dissemination of standards, and they quirements; and technical requirements. The
start to take effect. It should be highlighted that main logical building blocks of an EHR use the
a standard is dynamic, and it changes time after terminology of CEN/TC251 ENV13606 (Lloyd
time. Most standards require periodic revision. & Kalra, 2003; Maldonado, Crespo, Sanchis, &
Several factors combine to render a standard Robles, 2004; Marley, 2002). Many specific
out of date: technological evolution, new meth- medical records, like medical records of pa-
ods and materials, and new quality and safety tients suffering from beta-thalassaemia, which
requirements. To take account of these factors, are inevitably complex and grow in size very
the general rule has been established that all fast, are based also on ENV 13606 (Deftereos,
standards should be reviewed at intervals of not Lambrinoudakis, Andriopoulos, Farmakis, &
more than a predefined time period. On occa- Aessopos, 2001). The wider electronic exchange
sion, it is necessary to revise a standard earlier of clinical information between heterogeneous
(Hammond & Cimino, 2001). information systems in the delivery of diabetes
care demands a common structure in the form of
Standardization in Health and a message standard and close cooperation with
Medical Informatics: Why and What CEN/TC251 (Vaughan et al., 2000).
Standardization has been a major factor in

companies’ financial and clinical success, en- CURRENT STATUS OF
abling faster implementation, greater quality STANDARDIZATION IN HEALTH
control, and significant cost savings (Ball, AND MEDICAL INFORMATICS
Cortes-Comerer, Costin, Hudson, & August-
ine, 2004). The contribution of the standardiza- All the standards developers work through their
tion process in healthcare terminology initiated working technical committees in health and
by CEN/TC251 and supported now by the work medical informatics and a number of working
of CEN/TC215/WG3 to this new approach can groups specialized in a specific area.
46
What are Specific Areas of Work Table 1. Working groups of ISO/TC215

and Results of the ISO/TC215?
ISO/TC215 WG 1 Health records and modeling coordination
Table 1 shows working groups acting in the ISO/TC215 WG 2 Messaging and communication
ISO/TC215 WG 3 Health concept representation
ISO/TC215. Table 2 shows a list of standards ISO/TC215 WG 4 Security
ISO/TC215 WG 5 Health cards
given by this technical committee. ISO/TC215 WG 6 Pharmacy and medicines business
Table 2. Standards in health and medical informatics given by ISO/TC215
ISO/IEEE 11073-10101:2004 Health informatics—Point-of-care medical device communication—Part 10101:

Nomenclature
ISO/IEEE 11073-10201:2004 Health informatics—Point-of-care medical device communication—Part 10201: Domain
information model
ISO/IEEE 11073-20101:2004 Health informatics—Point-of-care medical device communication—Part 20101: Application
profiles—Base standard
ISO/IEEE 11073-30201:2004 Health informatics—Point-of-care medical device communication—Part 30200: Transport
profile -- Cable connected
ISO/IEEE 11073-30300:2004 Health informatics—Point-of-care medical device communication—Part 30300: Transport
profile -- Infrared wireless
ISO/TR 16056-1:2004 Health informatics—Interoperability of telehealth systems and networks—Part 1:
Introduction and definitions
ISO/TR 16056-2:2004 Health informatics—Interoperability of telehealth systems and networks—Part 2: Real-time
systems
ISO/TS 16058:2004 Health informatics—Interoperability of telelearning systems
ISO/TS 17090-1:2002 Health informatics—Public key infrastructure—Part 1: Framework and overview
ISO/TS 17090-2:2002 Health informatics—Public key infrastructure—Part 2: Certificate profile
ISO/TS 17090-3:2002 Health informatics—Public key infrastructure—Part 3: Policy management of certification
authority
ISO/TS 17117:2002 Health informatics—Controlled health terminology—Structure and high-level indicators
ISO/TR 17119:2005 Health informatics—Health informatics profiling framework
ISO/TS 17120:2004 Health informatics—Country identifier standards
ISO 17432:2004 Health informatics—Messages and communication—Web access to DICOM persistent
objects
ISO 18104:2003 Health informatics—Integration of a reference terminology model for nursing
ISO/TR 18307:2001 Health informatics—Interoperability and compatibility in messaging and communication
standards -- Key characteristics
ISO/TS 18308:2004 Health informatics—Requirements for an electronic health record architecture
ISO 18812:2003 Health informatics—Clinical analyser interfaces to laboratory information systems -- Use
profiles
ISO/TR 21089:2004 Health informatics—Trusted end-to-end information flows
ISO 21549-1:2004 Health informatics—Patient healthcard data—Part 1: General structure
ISO 21549-2:2004 Health informatics—Patient healthcard data—Part 2: Common objects
ISO 21549-3:2004 Health informatics—Patient healthcard data—Part 3: Limited clinical data
ISO/TS 21667:2004 Health informatics—Health indicators conceptual framework
ISO/TR 21730:2005 Health informatics—Use of mobile wireless communication and computing technology in
healthcare facilities—Recommendations for the management of unintentional electromagnetic
interference with medical devices
ISO 22857:2004 Health informatics—Guidelines on data protection to facilitate trans-border flows of personal
health information
ISO/TS is a normative document representing technical consensus within an ISO committee.

ISO/TR is an informative document containing information of a different kind from that normally published in a normative
document (Beolchi, 2003).
IEEE is the Institute of Electrical and Electronic Engineers, USA.
47
Table 3. CEN/TC251 working groups
CEN/TC251 WG 1 Communications: Information models, messaging, and smart cards

CEN/TC251 WG 2 Terminology
CEN/TC251 WG 3 Security, safety, and quality
CEN/TC251 WG 4 Technology for interoperability (devices)
Table 4. Standards in health and medical informatics given by CEN/TC251
EN 14484:2003 International transfer of personal health data covered by the EU data protection directive—High
level security policy
EN 14485:2003 Guidance for handling personal health data in international applications in the context of the EU
data protection directive
EN 1828:2002 Categorial structure for classifications and coding systems of surgical procedures
EN ISO 18104:2003 Integration of a reference terminology model for nursing (ISO 18104:2003)
EN ISO 18812:2003 Clinical analyser interfaces to laboratry information systems—Use profiles (ISO 18812:2003)
EN ISO 21549-1:2004 Patient healthcard data—Part 1: General structure (ISO 21549-1:2004)
EN ISO 21549-2:2004 Patient healthcard data—Part 2: Common objects (ISO 21549-2:2004)
EN ISO 21549-3:2004 Patient healthcard data—Part 3: Limited clinical data (ISO 21549-3:2004)
EN 12251:2004 Secure user identification for healthcare - Identification and authentication by passwords—
Management and security
EN 12252:2004 Digital imaging—Communication, workflow and data management (which endorses all of
DICOM as a European Standard)
What are Specific Areas of Work healthcare system are also candidates for inter-
and Results of the CEN/TC251? national standards. Standard formats need to
be defined for special kinds of data like images,
The work carried out by CEN/TC251 is men- signals and waveforms, sound and voice, and
tioned in Table 3. video, including motion video. Data security,
CEN/TC251 has been operating for 10 years. the security of objects and communication chan-
By October 2004, it had created 10 full stan- nels, and data archiving, especially in case of a
dards or EN (Table 4). catastrophe of any kind, should be standard-
ized. Some of these standards exist or have
been under way, but standardization is a con-
FUTURE TRENDS IN tinuous process, depending on the development
STANDARDIZATION IN HEALTH of information and communication technology,
AND MEDICAL INFORMATICS and therefore all standards need to be improved
and adapted to new technology coming day
There is no doubt that candidates for standard- after day.
ization in medical informatics will be core data
sets for healthcare speciality groups, decision-
support algorithms and clinical guidelines, and CONCLUSION
vocabulary. The identification of patients, con-
tent, and structure of electronic patient records, Standards in health and medical informatics are
and messages being communicating in the means to enabling better healthcare. Healthcare
48
is supposed to be better if healthcare providers Hammond, W. E., & Cimino, J. J. (2001).

can access data in a patient’s health record, Standards in medical informatics. In E. H.
and can read it or add some new data relevant Shortliffe & L. E. Perreault (Eds.), Medical
for other healthcare providers taking care of informatics: Computer applications in health
the patient. Information technology used in care and biomedicine (pp. 212-255). New
diagnostics produces data about a patient as York: Springer.
images, signals, or waves, as well as classic
International Organization for Standardization.
alphanumeric data. The format of such data
(ISO). (n.d.). Why standards matter? Re-
should be standardized, usable, and readable on
trieved February 11, 2005, from http://
any instrument of the same kind wherever it is
www.iso.org/iso/en/aboutiso/introduction/
in the world. Medical language should be stan-
index.html#one
dardized, and coding systems should be univer-
sal. The transfer of medical data should be Lloyd, D., & Kalra, D. (2003). EHR require-
secure, as well as data storage and communi- ments. Studies in Health Technology and
cation. Some patient data should be portable by Informatics, 96, 231-237.
health cards, especially for patients suffering
Louwerse, K. (2002). Demonstration results
from chronic diseases. Bad medical informatics
for the standard ENV 12924. Studies in Health
can kill a patient, and standardization in the field
Technology and Informatics, 69, 111-139
can help make this not happen.
(pp. 229-237).
Maldonado, J. A., Crespo, P., Sanchis, A., &
REFERENCES Robles, M. (2004). Pangaea: A mediator for
the integration of distributed electronic
Ball, M. J., Cortes-Comerer, N., Costin, M., healthcare records. In M. Fieschi, E. Coiera, &
Hudson, K., & Augustine, B. (2004). HCA Y. C. J. Li (Eds.), Medinfo 2004 (p. 1738).
Inc.: Standardization in action. Journal of Amsterdam: IOS Press.
Healthcare Information Management, 18(2),
Marley, T. (2002). Standards supporting
59-63.
interoperability and EHCR communication: A
Beolchi, L. (Ed.). (2003). European CEN TC251 perspective. Studies in Health
telemedicine glossary of concepts, technolo- Technology and Informatics, 87, 72-77.
gies, standards and users (5th ed.). Brussels,
Rodrigues, J. M., Trombert Paviot, B., Martin,
Belgium: European Commission.
C., Vercherin, P., & Samuel, O. (2002). Co-
Comité Européen de Normalisation (CEN, Eu- ordination between clinical coding systems and
ropean Committee for Standardisation). (n.d.). pragmatic clinical terminologies based on a
Retrieved February 11, 2005, from http:// core open system: The role of ISO/TC215/
www.cenorm.be/cenorm/index.htm WG3 and CEN/TC2511/WG2 standardisation?
Studies in Health Technology and
Deftereos, S., Lambrinoudakis, C.,
Informatics, 90, 401-405.
Andriopoulos, P., Farmakis, D., & Aessopos,
A. (2001). A Java-based electronic healthcare Vaughan, N. J., Cashman, S. J., Cavan, D. A.,
record software for beta-thalassaemia. Jour- Gallego, M. R., Kohner, E., Benedetti, M. M.,
nal of Medical Internet Research, 3(4), E33. et al. (2000). A detailed examination of the
49
clinical terms and concepts required for com- EN: European standard made by CEN.
munication by electronic messages in diabetes
European Standardization: Activity of the
care. Diabetes, Nutrition and Metabolism,
European authority for standards.
13(4), 201-209.
International Standardization: Activity
of the world authority for standards.
KEY TERMS
ISO (International Organization for Stan-
ANSI (American National Standards dardization): World authority for standards. It
Institute): The private, nonprofit membership is also the international standard made by the
organization responsible for approving official organization.
American national standards.
ISO/TC215: ISO technical committee for
CEN (Comité Européen de standardization in health and medical
Normalisation, European Committee for informatics.
Standardisation): The European authority for
Standard: A set of rules and definitions that
standards.
specify how to carry out a process or produce
CEN/TC251: CEN technical committee a product.
for standardization in health and medical
Standardization: Process of producing of
informatics.
standards.
50
51
Chapter VII
Basic Principles and Benefits
of Various Classification
Systems in Health
Dimitra Petroudi
Athanasios Zekios
ABSTRACT
The introduction of information systems in health progressively led tï coding systems. The
purposes of these systems are: recording causes of death, coding diseases and procedures,
etc. The most important medical coding system in our days is ICD (International Classification
of Diseases). Other coding systems that health professionals use are: SNOMED, LOINC,
MeSH, UMLS, DSM, DRG and HCPCS. There are also many Nursing Classification Systems,
such as: NANDA, NIC, NOC, ICNP, Omaha and HHCC. This chapter describes these coding
systems and their advantages.
INTRODUCTION clarity, stability, granularity, and the fact that it

can be developed. Coding systems usually aim
Since the introduction of information systems in to be accurate, have unambiguous expressions,
health, the demand for coding systems has been and are complete.
major. Medical coding systems are fundamen- In classification systems, groups of words
tal to medical record keeping as well as to or terms are collected together and organised.
gathering and communicating public health sta- Each of these terms will be associated with a
tistics. They are used for a variety of purposes: particular concept. Systems of classification
recording causes of death, coding diseases and have typically been hierarchical, meaning that
procedures, and physician billing and reim- more detail is obtained the further down the
bursement. The elements that make a coding hierarchy one proceeds, although ideas are still
system good are completeness, nonredundancy, linked and organised around common attributes.
Basic Principles and Benefits of Various Classification Systems in Health
Each concept within a classification system a worldwide standard for comparison of birth,
can also be given a numeric or alphanumeric death, and disease data.
code. The more extensive the coding system, ICD-9 includes diagnosis codes consisting
the more detail it can represent. of three to five numeric characters represent-
ing illnesses and conditions; alphanumeric E
codes describing external causes of injuries,
INTERNATIONAL poisonings, and adverse effects; and V codes
CLASSIFICATION OF DISEASES describing factors influencing health status and
contact with health services.
The most important medical coding system in Today, there is also the 10th version of ICD.
our days is the ICD (International Classifica- The ICD-10 consists of:
tion of Diseases). The purpose of the ICD is to
promote international comparability in the col- • Tabular lists containing cause-of-death
lection, classification, processing, and presen- titles and codes (Volume 1),
tation of health statistics. Since the beginning of • Inclusion and exclusion terms for cause-
the 20 th century, the ICD has been modified of-death titles (Volume 1),
about once every 10 years, except for the 20- • An alphabetical index to diseases, the
year interval between ICD-9 and ICD-10 (see nature of injuries, and external causes of
Table 1). The purpose of the revisions is to stay injury, and a table of drugs and chemicals
abreast with advances in medical science. (Volume 3), and
The ICD is copyrighted by the World Health • Descriptions, guidelines, and coding rules
Organization (WHO), which owns and pub- (Volume 2).
lishes the classification. Annual updates are
published by the Health Care Financing Admin- One benefit of ICD-10 is a more compre-
istration, now called the Centers for Medicare hensive scope. Table 2 gives examples of some
and Medicaid Services (CMS). of the subcategories provided in ICD-10 for the
ICD-9 is the ninth version of ICD. ICD-9 is capture of risk factors to health, such as lifestyle,
a classification system of diseases, health con- life management, psychosocial circumstances,
ditions, and procedures that represents the in- and the occupational or physical environment.
ternational standard for the labeling and nu- Another benefit is improved specificity and
meric coding of diseases. These codes provide currency. The results of a mapping from ICD-
Table 1. Revisions of the ICD according to the year of the conference in which they were
adopted and the years they were in use in the USA
Revision of the ICD Year of Conference in which Adopted Years in Use in USA
First 1900 1900-1909

Second 1909 1910-1920
Third 1920 1921-1929
Fourth 1929 1930-1938
Fifth 1938 1939-1948
Sixth 1948 1949-1957
Seventh 1955 1958-1967
Eighth 1965 1968-1978
Ninth 1975 1979-1998
Tenth 1992 1999-present
52
Table 2. Examples of some of the subcategories provided in ICD-10

ICD-10 Code Code Title
Z56.3 Stressful work schedule
Z58.1 Exposure to air pollution
Z63.0 Problems in relationship with spouse or partner
Z72.4 Inappropriate diet and eating habits
9-CM (Clinical Modification) to ICD-10 car- The ICD-9-CM consists of three volumes:
ried out in Australia showed that of a total of
13,600 ICD-10 codes, 50.8% were more spe- • A tabular list containing a numerical list of
cific than the ICD-9-CM codes, 31.5% were as the disease code numbers (Volume 1),
specific, and only 11.5% either were less spe- • An alphabetical index to the disease en-
cific or could not be compared. Other benefits tries (Volume 2), and
of ICD-10 include ongoing maintenance and • A classification system for surgical, diag-
updating, a single set of national standards, nostic, and therapeutic procedures (al-
international compatibility, a more effective phabetic index and tabular list; Volume 3).
structure, and better presentations and guide-
lines. Significant enhancements to the system’s Volumes 1 and 2 are used by physicians only
structure and presentation include an enlarged to assign diagnosis codes. Physicians also use
coding frame, hierarchic and logical presenta- Current Procedural Terminology (CPT) to re-
tion of codes, increased use of combination port medical and surgical procedures, and phy-
codes, and an improved format of the classifi- sician service codes rather than Volume 3 of
cation. the ICD-9-CM codes. The third volume of
The estimated costs for adopting the new ICD-9-CM is used by hospitals for reporting
ICD-10 coding systems are given as follows. inpatient procedures. The National Center for
Providers will incur costs for computer re- Health Statistics (NCHS) and the Centers for
programming; the training of coders, physi- Medicare and Medicaid Services are the U.S.
cians, and code users; and for the initial and governmental agencies responsible for over-
long-term loss of productivity among coders seeing all changes and modifications to the
and physicians. The cost of sequential conver- ICD-9-CM. The aim of the diagnosis codes
sion (10-CM then 10-PCS [Procedure Coding was epidemiological and not for billing func-
System]) is estimated to run from $425 million tions, but now, in the United States, the codes
to $1.15 billion in one-time costs plus some- are mostly used for billing and reimbursement
where between $5 million and $40 million a year purposes.
in lost productivity. A clinical modification of ICD-10 (ICD-10-
A related classification, the International CM) has been developed by the NCHS. Revi-
Classification of Diseases, Clinical Modifica- sions have been made to the draft of ICD-10-
tion (ICD-9-CM), is used in assigning codes to CM based on the comments received. An up-
diagnoses associated with inpatients, outpa- dated draft version of ICD-10-CM from June
tients, and physician-office utilization in the 2003 is now available for public viewing. How-
United States. The ICD-9-CM is based on the ever, the codes in ICD-10-CM are not cur-
ICD but provides for additional morbidity detail rently valid for any purpose or use.
and is annually updated.
53
DIFFERENCES BETWEEN ICD-9 Table 3. Examples of differences between

AND ICD-10 ICD-9-CM and ICD-10-CM
The tenth revision (ICD-10) differs from the ICD-9-CM ICD-10-CM

Precordial Chest Pain 786.51 R07.2
ninth revision (ICD-9) in several ways despite Asthma, Acute Exacerbation 493.92 J45.21
Thumb Laceration 883.0 S61.011a
the fact that the content is similar. First, ICD-
10 is printed in a three-volume set compared
to ICD-9’s two-volume set. Second, ICD-10
has alphanumeric categories (A00-Z99) rather
than numeric categories (001-999). Third, some for the identification of physician work. Its only
chapters have been rearranged, some titles intention is to identify inpatient facility services
have changed, and conditions have been rein a way not directly related to physician work,
grouped. Fourth, ICD-10 has almost twice as but directed toward the allocation of hospital
many categories as ICD-9. Fifth, some fairly services.
minor changes have been made in the coding
rules for mortality.
OTHER CODING SYSTEMS
DIFFERENCES BETWEEN SNOMED (Systematized Nomenclature of

ICD-9-CM AND ICD-10-CM Medicine) is a classification that is maintained
by SNOMED International. SNOMED’s de-
The current draft of ICD-10-CM contains a sign is based on detailed and specific nomencla-
significant increase in codes over ICD-10 and ture and has been successfully implemented
ICD-9-CM. Notable improvements in the con- internationally. The National Health Service in
tent and format of ICD-10-CM include the the United Kingdom has adopted a clinical
addition of information relevant to ambulatory version (SNOMED CT) as its preferred clinical
and managed-care encounters, expanded in- terminology.
jury codes, the creation of combination diagno- The LOINC (Logical Observation Identi-
sis-symptom codes to reduce the number of ties, Names and Codes) database provides a
codes needed to fully describe a condition, the set of universal names and ID codes for iden-
addition of a sixth character, the incorporation tifying laboratory and clinical observations. The
of common fourth- and fifth-digit subclassifica- aim of LOINC is to facilitate the exchange and
tions, laterality, and greater specificity in code pooling of clinical laboratory results, such as
assignment. blood hemoglobin or serum potassium, for clini-
ICD-10 is used to code and classify mortal- cal care, outcomes management, and research.
ity data from death certificates, having re- MeSH was originally developed by the
placed ICD-9 for this purpose as of January 1, United States’ National Library of Medicine
1999. ICD-10-CM is planned as the replace- (NLM) to index the world medical literature in
ment for ICD-9-CM, Volumes 1 and 2. MEDLINE (MeSH provides bibliographic head-
ICD-10-PCS is currently designated to re- ings for indexing). It also forms an essential
place Volume 3 of ICD-9-CM for hospital part of the NLM’s Unified Medical Language
inpatient use only. There is no intention for System (UMLS). MeSH is not an efficient
ICD-10-PCS to shape, form, or replace CPT indexing language for tasks such as classifying
episodes of patient care.
54
The UMLS project is a long-term research 511 categories. The purpose of DRGs is to
and development project at the NLM whose reduce hospital costs and reimbursements.
goal is to help health professionals and re- HCPCS (Healthcare Common Procedure
searchers to intelligently retrieve and integrate Coding System) was originally created for use
information from a wide range of disparate under the Medicare program. Today, HCPCS
electronic biomedical-information sources. This is used by virtually every payer in the United
makes it easier for users to link information States. HCPCS is comprised of three levels:
from patient record systems, bibliographic da-
tabases, factual databases, expert systems, and • Level I: CPT codes
so forth. The UMLS Knowledge Services can • Level II: National codes (i.e., J codes, A
also assist in data creation and indexing appli- codes, Q codes, C codes for OPPS only)
cations. UMLS is not itself a standard; it is a • Level III: Local codes
cross-referenced collection of standards and
other data and knowledge sources. It gives the There are many nursing classification sys-
chance to exchange health-care information tems, such as, NANDA (North American Nurs-
despite the multiplicity of coding systems in use ing Diagnosis Association), NIC, NOC, ICNP
today. (International Classification of Nursing Prac-
The UMLS metathesaurus contains map- tice), Omaha, and HHCC. The one that is most
pings to MeSH, ICD-9-CM, SNOMED, CPT, used is NANDA. It is a set of nursing diagnoses
and a number of other coding systems. The introduced by the North American Nursing
metathesaurus is organised by concepts, which Diagnosis Association in 1973. According to
means that alternate names (synonyms, lexical NANDA, the classification is based on “a
variants, and translations) for the same mean- selection of nursing interventions to achieve
ing are all linked together as one concept. The outcomes for which the nurse is accountable.”
metathesaurus adds information to the con- ICNP first appeared in 1989 and was created
cepts, including semantic types, definitions, and by the International Council of Nursing (ICN).
interconcept relationships. The purpose of this system was to place the
DSM (Diagnostic and Statistical Manual of common language of all nursing classification
Mental Disorders) provides numeric codes, systems into an international frame.
diagnostic criteria, and comprehensive defini-
tions of mental disorders. The codes and terms
in DSM-IV are fully compatible with ICD-9- BENEFITS OF CODING SYSTEMS
CM and ICD-10.
CPT is a listing of descriptive terms and A coding system is important because, first of
five-digit identifying codes for reporting medi- all, it groups, separates, abbreviates, and facili-
cal, surgical, and diagnostic services performed tates automated data processing and transmit-
by physicians and other health-care profession- tal. Codes are used to help participants in the
als. It first appeared in 1966 and is published health-care system digitize information so that
annually by the American Medical Association. billing, record-keeping, and practice-manage-
DRGs (Diagnosis-Related Groups) are cat- ment processes become more cost effective
egories of clinically similar illnesses that re- and reliable. Another important benefit is the
quire the same types of hospital resources to fact that health professionals gain precious
treat. Every patient can be classified into one of time during data recording and decision mak-
55
ing. Classification systems improve the com- CPRI Work Group on Codes and Structures,
munication among health professionals and Phase II: Evaluation of clinical coding schemes:
among health organisations. Coding makes it Completeness, taxonomy, mapping, definitions,
easier to handle medical insurance claims and and clarity. J Am Med Inform Assoc., 4(3),
to identify the provider on a predetermined 238-250.
basis. It also makes it easier for professionals to
Chute, C. G., Cohn, S. P., & Campbell, J. R.
compare medical data internationally and do
(1998). A framework for comprehensive health
research. Plus, coding and classifications help
terminology systems in the United States.
to standardize a clinical language. Classifica-
JAMIA, 5(6), 503-510.
tion can therefore be used as another way of
organising information and can act as a com- Chute, C. G., Elkin, P. L., Sheretz, D. D., &
mon language between health professionals, Tuttle, M. S. (1999). Desiderata for a clinical
enhancing the quality and usefulness of the terminology server. Proceedings of AMIA’99
communication. Annual Symposium.
Cimino, J. (1998). Desiderata for controlled
medical vocabularies in the twenty-first cen-
CONCLUSION
tury. Methods Inf Med., 37(4-5), 394-403.
Automated coding systems hold the potential ISO 1087-1:2000: Terminology work-vocabu-
for increased coding speed and accuracy com- lary, Part 1: Theory and application.
pared to unaided human coders. Coding and
Rector, A. L. (1999). Clinical terminology: Why
classification systems are developed rapidly
is it so hard? Methods Inf Med., 38(4-5), 239-
today. Specialists tend to use more and more
252.
codes and classifications in order to create
common international languages in all kinds of
sciences so that research will become easier. URL REFERENCES
http://apt.rcpsych.org/cgi/content/full/8/3/165
REFERENCES
http://citeseer.nj.nec.com/354766.html
Bakken, S., Campbell, K. E., Cimino, J. J., http://faculty.washington.edu/momus/
Huff, S. M., & Hammond, W. E. (2000). To-
http://rrc.gsk.com/ccp_issues/coding_systems.
ward vocabulary domain specifications for
htm
health level 7-coded data elements. JAMIA, 7,
333-342. http://secure.cihi.ca/cihiweb/dispPage.jsp?
cw_page=codingclass_icd10bene_e
Bechhofer, S. K., Goble, C. A., Rector, A. L.,
Solomon, W. D, & Nowlan, W. A. (1997). http://umls.nlm.nih.gov
Terminologies and terminology servers for in-
formation environments. Proceedings of STEP http://www.acep.org/1,33890,0.html
’97 Software Technology and Engineering http://www.amia.org/pubs/symposia/
Practice. D005782.PDF
Campbell, J. R., Carpenter, P., Sneiderman, C., http://www.bacts.org.uk
Cohn, S., Chute, C. G., & Warren, J. (1997).
56
http://www.biohealthmatics.com/health KEY TERMS

informatics/mlcls.aspx
Classification: Systematic representation
http://www.cams.co.uk
of terms and concepts and the relationship
http://www.cdc.gov/nchs/about/major/dvs/ between them.
icd10des.htm
Clinical Classification: A method of
http://www.cdc.gov/nchs/about/otheract/icd9/ grouping clinical concepts in order to represent
abticd9.htm classes that support the generation of indica-
tors of health status and health statistics.
http://www.cdc.gov/nchs/about/otheract/icd9/
abticd10.htm Codes: Numeric or alphanumeric abbre-
viations that can expand into some meaning.
http://www.cdc.gov/nchs/about/major/dvs/
icd10des.htm Coding System: A terminology and con-
text-free symbolic codes for each term.
http://www.cdc.gov/nchs/datawh/nchsdefs/
icd.htm Concept: An idea encompassing a class of
objects (“A concept is a unit of knowledge
http://www.cdc.gov/phin/data_models
created by a unique combination of character-
http://www.cms.hhs.gov/medicare/hcpcs/. istics”; ISO).
http://www.defoam.net/hubris/hubris01.htm Health Information: Information about an
identifiable individual that relates to his or her
http://www.hip.on.ca/search/160.html
previous, current, and future health. It is also
http://www.hl7.org/standards/icd10.htm knowledge derived from statistics or data de-
scribing and enumerating attributes, events,
http://www.hsl.unc.edu/services/guides/
behaviours, services, resources, outcomes, or
focusonmedcoding.cfm
costs related to health, disease, and health
http://www.medicalcodingandbilling.com/ services.
med_coding.htm
Modification: A slight change or alteration
http://www.mwsearch.com/ made to improve something or make it more
suitable.
http://www.opengalen.org
Nomenclature: An agreed system of as-
http://www.reimbursementcodes.com/
signed names.
medical_coding_d.html
Terminology: A set of words or expres-
http://www.snomed.org/products/content/
sions together with a definition used within a
mappings.html
certain field.
https://wwws.soi.city.ac.uk/intranet/students/
courses/mim/mi/lect2_2.htm
http://www.visualread.com
http://www.who.int/classifications/en
57
59
Chapter VIII
Virtual Reality in Medicine
Theodoros N. Arvanitis
University of Birmingham, UK
ABSTRACT
This chapter explores the technological quest of virtual reality within the field of medicine.
Although the author does not intend to provide an exhaustive review of the various health
informatics applications of VR over the past 15 years of its development, he presents some of
the major technological breakthroughs and their impact in the provision of healthcare
services to the point-of-need, (i.e. the patient).
INTRODUCTION ease. Today’s innovations in science and engi-

neering raise the potential for medical technol-
The continuing technological achievements of ogy to expand the frontier of healthcare deliv-
the modern era are changing dramatically the ery to unimaginable accomplishments. In this
ways in which we conduct our daily activities context, virtual-reality (VR) technologies have
and life. The medical field, through the provi- played an important role in revolutionising the
sion of high-quality healthcare to the patient, is practical provision of patient care. In recent
not an exception. The technological advances years, VR technology and its application to
that we have witnessed during the past few medicine are not a research curiosity anymore;
decades have had an enormous impact on the in several areas of clinical disciplines, the tech-
manner in which we diagnose and treat dis- nology and innovation are developing in such a
way that they can be adopted in routine prac- This chapter will explore the technological
tice, providing powerful tools in diagnostics, quest of virtual reality within the field of medi-
therapeutic planning, and interventions. cine. Although we do not intend to provide an
The enhancement of human health and, as a exhaustive review of the various health-
consequence, the improvement of the quality of informatics applications of VR over the past 15
human life, is one of the main objectives of years of its development, we aspire to present
scientific endeavour in the field of medicine. and discuss some of the major technological
The provision of high-quality patient care is the breakthroughs and their impact in the provision
ultimate outcome of any medical research and of healthcare services to the point of need, that
its related clinical implementation. The scien- is, the patient. Our analysis will focus mainly on
tific advancements in medicine have always the research motivations and challenges in the
benefited from simultaneous developments in routine use of the technology, while our argu-
engineering and technology. Virtual reality is ment will be about the socioeconomic effects
one of the important and recent technological and medicoethical concerns that relate to its
innovations that have made a significant impact implementation within the clinical practice.
in medicine, more specifically in its quest for the
high-quality provision of healthcare.
The medical applications of virtual reality THE CONCEPT OF VR IN THE
originated in the early 1990s from the need of CONTEXT OF MEDICINE
healthcare practitioners to visualise large
amounts of complex medical data, particularly Traditionally, virtual reality is defined as a form
in surgical planning, preoperative training, and of human-machine interaction technology, in
image-guided navigation during surgical proce- which human users are fully immersed within a
dures (Chinnock, 1994). Since the first surgical synthetic 3-D virtual environment (Ellis, 1994).
abdominal VR simulator in 1991 by Satava Users can interact, through all their senses,
(1993), the scope of virtual reality in medicine with any virtual objects and scenes of such an
has broadened, with applications ranging from environment as they are immersed in it with the
3-D (three-dimensional) immersive visualisation assistance of appropriate graphical displays
and manipulation of the cellular environment (usually, in the form of head-mounted display
(Guan et al., 2004) to clinically applied diagnos- technologies, HMD) and other nonvisual tech-
tic tools (e.g., virtual endoscopy technologies; nological modalities (e.g., auditory, haptic, etc.;
Lorensen, Jolesz, & Kikinis, 1995), advanced Pratt, Zyda, & Kelleher, 1995). Such an inter-
medical education and training (Zajtchuck & action gives a full sense of virtual presence to
Satava, 1997), augmented or enhanced surgery the user.
(Shuhaiber, 2004), medical therapy (Vincelli, This technology-oriented definition of VR,
Molinari, & Riva, 2001), and the virtual design although valid for many scientific, industrial,
of healthcare processes and environments and entertainment applications, is restrictive
(Kaplan, Hunter, Durlach, Schodek, & Rattner, when considered within the context of the
1995). Biomedical VR “has changed from [be- observed technological evolution of medical
ing] a research curiosity to a commercially and VR. Since the early years, many researchers
clinically important area of medical informatics and clinical practitioners have embraced the
technology” (Székely & Satava, 1999). concept and technology of medical VR by
adopting a broader definition and scope for its
60
application in the field. Depending on the re- health, etc.) and their corresponding impact to
quirements of the healthcare application at such a knowledge domain. In this chapter, we
hand, the implementation of immersion, in terms aim to ascertain the revolutionary character of
of technological devices used and sensory mo- medical VR technologies in improving the ap-
dalities involved, may vary. For instance, “in propriate provision of patient care. Thus, in the
some applications, real and virtual objects need discussion that follows, we explore the relevant
to be integrated making it necessary to present applications in terms of their innovation impact
and manipulate them simultaneously in a single and clinical significance within the broad themes
scene, leading to the development of hybrid of medical diagnostics and visualisation, therapy
systems referred to as augmented reality sys- interventions and planning, and medical educa-
tems” (Székely & Satava, 1999). McCloy and tion and training.
Stone (2001) argue that “although this so called
immersive technology is still evident today, only
10% of virtual reality applications warrant its VR IN MEDICAL DIAGNOSTICS
use,” while Riva (2003) states that only 20% of AND VISUALISATION
healthcare-related applications use immersive
equipment. On the other hand, particular reha- Medicine is an information-intensive field, where
bilitation (Rizzo & Buckwalter, 1997) and clini- a large volume of real data is collected experi-
cal neuropsychology applications (Rizzo, mentally for the purposes of interpretation in
Wiederhold, Riva, & Van Der Zaag, 1998) order to understand the effects of disease in
might benefit from an advanced immersive humans. To achieve such an interpretation of
interactive environment, where users are ac- disparate and complex data, we commonly
tive participants in the virtual world, interacting summarise all available and clinically related
in real-time with the functionality of virtual information through some form of meaningful
objects and/or scenes (Schultheis & Rizzo, visualisation. Traditional biomedical imaging
2001). In such a scenario, the therapist can techniques achieve the noninvasive visual rep-
change, in a controlled manner, the conditions resentation of human anatomy and the func-
of the immersive interaction of a patient with tional mapping of human physiology from the
the environment in order to study and manipu- cellular to the organism level. The advanced
late any parameters of the patient’s condition. interactive 3-D-graphics technology of virtual
Nevertheless, most biomedical VR practitio- reality (Haubner, Krapichler, Lösch, Englmeir,
ners advocate that the “key strength of virtual van Eimeren, & Kelleher, 1997) has further
reality…is that it supports and enhances real enhanced these visualisation possibilities by
time interaction on the part of the users” offering novel data-fusion approaches of hu-
(McCloy & Stone). man structural imaging with its corresponding
This user-centric stance makes it futile to functional mapping of human physiology and
classify and study the bulk of medical VR function (Soferman, Blythe, & John, 1998;
applications in terms of their technological char- Zajtchuck & Satava, 1997).
acteristics. Most authors in the current litera- On the level of microcellular visualisation,
ture investigate such applications on the basis we are currently experiencing a revolution in
of their subject-matter categorisation within a the 3-D representation of structural relation-
specific subdiscipline of medical and clinical ships within cells and human tissues, both in
science (e.g., surgery, rehabilitation, metal vivo and in vitro. Guan et al. (2004) recently
61
developed novel VR visualisation that provides of an internal organ of interest, followed by the
an intuitive, interactive way of viewing and execution of a visualisation “fly through” (Satava
manipulating 3-D cellular structures within an & Jones, 2002). The literature contains a vast
immersive synthetic environment. This break- amount of articles that successfully report the
through marks the beginning of further innova- clinical implementation and use of virtual
tive visualisation for many other micromolecular endocscopy in the organs of the colon (Vining,
imaging techniques, including electron micros- 1997), small bowel (Rogalla, Werner-Rustner,
copy (Frank, 2002) and NMR-based cellular Huitema, van Est, Meiri, & Hamm, 1998),
imaging (Blackband, Buckley, Bui, & Phillips, stomach (Springer et al., 1997), tracheo-bron-
1999). chial tree (Jones & Athtanasiou, 2005), and so
Novel VR visualisation of human anatomi- on.
cal structure has been rapidly realised since the The clinical significance of virtual endos-
completion of the Human Visible Project copy is currently being studied (e.g., Bhandari
(Ackerman, 1991). The rich anatomical data et al., 2004; Rapp-Bernhardt, Welte, Budinger,
sets have been acquired to serve as “a common & Bernhardt, 1998) as it is becoming more and
reference point for the study of human anatomy, more urgent to use it in clinical diagnostic
as a set of common public-domain data for practice. Virtual endoscopy lacks any known
testing medical imaging algorithms, and as a complications (in contrast to known issues of
testbed and model for the construction of image perforation, bleeding, etc. found in traditional
libraries that can be accessed through net- endoscopy), it is a totally noninvasive tech-
works” (Ackerman, 1998). A plethora of vir- nique, and its implementation is far more cost
tual-reality applications have been based on effective in terms of materials and personnel
these data, many of which have been regularly when compared with traditional invasive screen-
reported in the conference series Medicine ing methods (Dunkin, 2003).
Meets Virtual Reality (e.g., Westwood, Haluck,
Hoffman, Mogel, Phillips, & Robb, 2004;
Westwood, Hoffmann, Robb, & Stredney, VR IN CLINICAL THERAPY
1999). INTERVENTIONS AND PLANNING
The above-mentioned successes in human-
anatomy visualisation have been more notable The majority of current near-term therapeutic
in a particular medical application of VR that applications are identified in the areas of VR-
has pushed the technology into the frontier of technology-assisted surgical interventions (re-
medical diagnostics. Virtual endoscopy mote surgery and augmented- or enhanced-
(Lorensen et al., 1995; Rubino, Soler, reality surgery; Marescaux & Rubino, 2003;
Marescaux, & Maisonneuve, 2002; Wood & Marmulla, Hoppe, Mühling, & Eggers, 2005),
Razavi, 2002), one of the earlier endeavours in surgical planning, and surgical simulation (Satava
biomedical VR research, is now becoming a & Jones, 1998). Remote surgery applications
clinically acceptable new form of noninvasive link research in robotics and virtual reality.
screening for the diagnosis of structural abnor- Surgeons can manipulate equipment from a
malities in internal organs. The concept of remote site while having full haptic-sensory
virtual endoscopy is based on the 3-D recon- feedback through the use of telerobotic equip-
struction of CT (computed tomography) or ment. VR-enhanced remote surgery applica-
MRI (magnetic resonance imaging) data sets tions have been beneficial for defence medi-
62
cine and for solving the problem of providing progress of minimally invasive surgical train-
emergency treatment at a distance, where there ers, such as the MIST system, “a product for
is no availability of specialists locally. Satava’s training and assessment of surgical laparoscopic
(1995) original vision and concept of surgeons psychomotor skills” of trainee surgeons
in the future being equipped with virtual-reality (McCloy & Stone, 2001) that is now commer-
headsets and rehearsing real or robotic proce- cially available (Mentice Corporation, n.d.).
dures, using advanced computer-generated Currently, systems like MIST are undergoing
images, has recently become a reality with the clinical trials and validation and are slowly
successful completion of a transatlantic robot- becoming useful tools for surgical skills certifi-
assisted surgical intervention. On September 7, cation. Nonetheless, McCloy and Stone are
2001, surgeons in New York performed a arguing that despite the technological achieve-
laparoscopic cholecystectomy on a patient in ment in the area of surgical training VR simu-
France (Marescaux et al., 2001). Furthermore, lators, there is still a great need for objectively
many front-running concepts in image-guided validating the acquisition of skills of individual
surgery (augmented-reality concepts) are cur- surgeons during training: “The crucial factor
rently “undergoing consolidation through clini- that will determine the uptake of virtual reality
cal validation” (McCloy & Stone, 2001). Most technology by surgeons will be the demonstra-
of these applications are used both as image- tion that virtual reality is capable of delivering
guided and planning surgical tools. reliable and valid training and assessment sys-
For the above-mentioned applications, there tems.” Various researchers are currently work-
are still a few challenges to overcome. In order ing toward the identification of objective metrics
to achieve reliable real-time implementations of for the evaluation of surgical skills in real and
both remote and augmented-reality surgical virtual environments (Moody, Baber, &
procedures, further work is needed in ensuring Arvanitis, 2002; Moody, Baber, Arvanitis, &
that the technology will cope with the demands Elliott, 2003).
of the requirements for the procedures: The VR technologies are now starting to play a
improvement of high-performance 3-D graphi- crucial role as a clinical tool in the field of
cal visualisation and real-time image data-fu- neuropsychological assessment and rehabilita-
sion algorithms, together with the robust and tion (Riva, 2003; Vincelli et al., 2001). A new
resilient distributed transmission of large and era of clinical therapy interventions for clinical
complex data sets, are still important require- psychologists is at its beginning, in which the
ments to be met (Satava & Jones, 1998). properties of immersive interactive environ-
In the area of surgical simulation, we can ments are used by therapists to stimulate the
observe 15 years of continuous effort and real world of patients in a variety of flexible
development of surgical simulation technolo- interventions on psychological-distress cases.
gies for planning and training (Chinnock, 1994; Riva and Davide (1995) argue that the thera-
McCloy & Stone, 2001). Such systems can peutic effectiveness of VR clinical psychology
provide different levels of photo-realism in interventions is most likely to be increased
anatomical representations, different levels of compared to traditional methods, while sup-
fidelity in physical properties of tissue, and porting their arguments by reporting current
physiologic parameters (Satava & Jones, 1998). studies of the verification of the clinical effec-
Of the notable successes in VR surgical tiveness for six psychological disorders, includ-
simulation applications, we should mention the ing acrophobia, arachnophobia, panic disorders
63
with agoraphobia, body-image disturbances, radiology has been combined with virtual reality
binge-eating disorders, and fear of flight (Riva, (Hoffman, 1992; Hoffman, Irwin, Ligon, Murray,
2003). Even though the clinical potential is & Tohsaku, 1995). There is a great potential of
evident in this field, there are still some using VR with other established technologies,
medicoethical issues to be resolved. Whalley such as the Internet, in order to maximise the
(1995) argues that as this effort is still at a outputs of high-quality delivery of medical edu-
research stage, there is still cation (Samothrakis, Arvanitis, Plataniotis,
the possibility that some researchers place McNeill, & Lister, 1997).
their own advancement above the interests of a
particular patient. Some other researchers may
be unduly paternalistic especially when making CONCLUDING REMARKS
decisions about patients who because of mental
impairment or illness are unable to give their Our brief exploration of medical VR technolo-
informed consent. Potentially, VR machines gies and their application in both basic medical
may be prone to errors such that they introduce research and clinical practice has shown that
into the mental life of susceptible individuals the impact of the technology can be simply
specific distortions that serve to exacerbate the characterised as revolutionary in the context of
symptoms of mental illness or induce such enhancing the quality of patient care. The
symptoms when none were previously present. informatics tools that VR has offered to the
Therefore, there is a need for an open medical field are now widely accepted and
ethical debate within the research community, validated for everyday use in clinical practice.
while clearly medicoethical governance bodies The field of medical VR is not anymore a
have to regulate experimentation and applica- technological curiosity, but provides an appro-
tion of the technologies. priate and exemplar way of applying informa-
tion and communication technologies in
healthcare. However, despite the intensive re-
VR IN MEDICAL EDUCATION search and development over the past 15 years,
AND TRAINING there is still great potential for the expansion of
the capabilities that VR technology offers to
It is evident from the above discussion that the medicine. The current successful clinical imple-
visualisation, planning, and simulation aspects mentations show that VR has still the potential
of VR technology have important implications to offer a real value to further improving the
for medical education and training (Riva, 2003; appliance of informatics in healthcare and, as a
Székely & Satava, 1999). We could present a consequence, the quality of human life.
plethora of examples of successful applications
of VR in medical education and training, relat-
ing to the study of anatomy, surgical training, REFERENCES
and so forth. Many medical schools have suc-
cessfully combined the use of VR technologies Ackerman, M. J. (1991). The visible human
in established courses, such as the effort at the project. The Journal of Biocommunication,
University of California at San Diego, where an 18(2), 14.
established multimedia, computer-based edu-
Ackerman, M. J. (1998). The visible human project.
cation program on anatomy, pathology, and
Proceedings of the IEEE, 86(3), 504-511.
64
Bhandari, S., Shim, C. S., Kim, J. H., Jung, I. S., Hoffman, H. (1992). Developing network-com-
Cho, J. Y., Lee, J. S., et al. (2004). Usefulness patible instructional resources for UCSD’s core
of three-dimensional multidetector row CT (vir- curriculum. In J. D. Westwood, H. M. Hoffman,
tual gastroscopy and multiplanar reconstruc- R. A. Robb, & D. Stredney (Eds.), Medicine
tion) in the evaluation of gastric cancer: A meets virtual reality (pp. 90-92). Washington,
comparison with conventional endoscopy, EUS, DC: IOS Press.
and histopathology. Gastrointestinal Endos-
Hoffman, H., Irwin, A., Ligon R., Murray, M.,
copy, 59(6), 619-626.
& Tohsaku, C. (1995). Virtual reality-multime-
Blackband, S. J., Buckley, D. L., Bui, J. D., & dia synthesis: Next-generation learning envi-
Phillips, M. I. (1999). NMR microscopy: Be- ronments for medical education. The Journal
ginnings and new directions. Magnetic Reso- of Biocommunication, 22(3), 2-7.
nance Materials in Biology, Physics and
Jones, C. M., & Athanasiou, T. (2005). Is
Medicine, 9(3), 112-116.
virtual bronchoscopy an efficient diagnostic
Chinnock, C. (1994). Virtual reality in surgery tool for the thoracic surgeon? The Annals of
and medicine. Hospital Technology Series, Thoracic Surgery, 79(1), 365-374.
13(18), 1-48.
Kaplan, K., Hunter, I., Durlach, N. I., Schodek,
Dunkin, B. J. (2003). Flexible endoscopy simu- D. L., & Rattner, D. (1995). A virtual environ-
lators. Seminars in Laparoscopic Surgery, ment for a surgical room of the future. In R. M.
10(1), 29-35. Satava, K. S. Morgan, H. B. Sieburg, R.
Mattheus, & J. P. Christensen (Eds.), Interac-
Ellis, S. R. (1994). What are virtual environ-
tive technology and the new paradigm for
ments? IEEE Computer Graphics and Appli-
healthcare (pp. 161-167). Washington, DC:
cations, 14(1), 17-22.
IOS Press.
Frank, J. (2002). Single-particle imaging of
Lorensen, W. E., Jolesz, F. A., & Kikinis, R.
macromolecules by cryo-electron microscopy.
(1995). The exploration of cross-sectional data
Annual Review of Biophysics and
with a virtual endoscope. In R. M. Satava, K. S.
Biomolecular Structure, 31, 303-319.
Morgan, H. B. Sieburg, R. Mattheus, & J. P.
Guan, Y. Q., Opas, M., Cai, Y. Y., Zhang, X., Christensen (Eds.), Interactive technology
Xiong, Z. W., Wong, S., et al. (2004). Applica- and the new paradigm for healthcare (pp.
tion of virtual reality in volumetric cellular 221-230). Washington, DC: IOS Press.
visualization. Proceedings of the 2004 ACM
Marescaux, J., Leroy, J., Gagner, M., Rubino,
SIGGRAPH International Conference on
F., Mutter, D., Vix, M., et al. (2001).
Virtual Reality Continuum and its Applica-
Transantlantic robot-assisted telesurgery. Na-
tions in Industry (pp. 65-71).
ture, 413(6854), 379-380.
Haubner, M., Krapichler, C., Lösch, A.,
Marescaux, J., & Rubino, F. (2003).
Englmeir, K.-H., van Eimeren, W., & Kelleher,
Telesurgery, telemonitoring, virtual surgery and
K. (1997). Virtual reality in medicine: Com-
telerobotics. Current Urology Reports, 4, 109-
puter graphics and interaction techniques. IEEE
113.
Transactions on Information Technology in
Biomedicine, 1(1), 61-72. Marmulla, R., Hoppe, H., Mühling, J., & Eggers,
G. (2005). An augmented reality system for
65
image-guided surgery. International Journal rehabilitation: The state of the art. In G. Riva
of Oral and Maxillofacial Surgery, 34(6), (Ed.), Virtual reality in neuro-psycho-physi-
594-596. ology (pp. 123-146). Amsterdam: IOS Press.
McCloy, R., & Stone, R. (2001). Science, medi- Rizzo, A. A., Wiederhold, B., Riva, G., & Van
cine and the future: Virtual reality in surgery. Der Zaag, C. (1998). A bibliography of articles
British Medical Journal, 323(7318), 912- relevant to the application of virtual reality in
915. the mental field. Cyberpsychology and
Behaviour, 1(4), 411-425.
Mentice Corporation. (n.d.). Procedicus
MISTTM. Retrieved August 15, 2005, from http:/ Rogalla, P., Werner-Rustner, M., Huitema, A.,
/www.mentice.com/ van Est, A., Meiri, N., & Hamm, B. (1998).
Virtual endoscopy of the small bowel: Phantom
Moody, C. L., Baber, C., & Arvanitis, T. N.
study and preliminary clinical results. Euro-
(2002). Objective surgical performance evalu-
pean Radiology, 8(4), 563-567.
ation based on haptic feedback. In J. D.
Westwood, H. M. Hoffman, R. A. Robb, & D. Rubino, F., Soler, L., Marescaux, J., &
Stredney (Eds.), Medicine meets virtual real- Maisonneuve, H. (2002). Advances in virtual
ity 10 (pp. 304-310). Amsterdam: IOS Press. reality are wide ranging. British Medical Jour-
nal, 324(7337), 612.
Moody, C. L., Baber, C., Arvanitis, T. N., &
Elliott, M. (2003). Objective metrics for the Samothrakis, S., Arvanitis, T. N., Plataniotis,
evaluation of simple surgical skills in real and A., McNeill, M. D. J., & Lister, P. F. (1997).
virtual domains. Presence, 12(2), 207-221. WWW creates new interactive 3D graphics
and collaborative environments for medical re-
Pratt, D. R., Zyda, M., & Kelleher, K. (1995).
search and education. International Journal
Virtual reality: In the mind of the beholder.
of Medical Informatics, 47, 69-73.
IEEE Computer, 28(7), 17-19.
Satava, R. M. (1993). Virtual reality surgical
Rapp-Bernhardt, U., Welte, T., Budinger, M.,
simulator: The first steps. Surgical Endos-
& Bernhardt, T. M. (1998). Comparison of
copy, 7, 203-205.
three-dimensional virtual endoscopy with bron-
choscopy in patients with oesophageal carci- Satava, R. M. (1995). Medicine 2001: The king
noma infiltrating the trancheobronchial tree. is dead. In R. Morgan, R. M. Satava, H. B.
British Journal of Radiology, 71(852), 1271- Sieburg, R. Matthews, & J. P. Christensen
1278. (Eds.), Interactive technology and the new
paradigm for healthcare (pp. 334-339).
Riva, G. (2003). Applications of virtual envi-
Amsterdam: IOS Press.
ronments in medicine. Methods of informa-
tion in medicine, 42(5), 524-534. Satava, R. M., & Jones, S. B. (1998). Current
and future applications of virtual reality for
Riva, G., & Davide, F. (2001). Communica-
medicine. Proceedings of the IEEE, 86(3),
tions through virtual technologies: Identity,
484-489.
community and technology in the communi-
cation age. Amsterdam: IOS Press. Satava, R. M., & Jones, S. B. (2002). Medical
applications of virtual reality. In K. M. Stanney
Rizzo, A. A., & Buckwalter, J. G. (1997).
(Ed.), Handbook of virtual environments:
Virtual reality and cognitive assessment and
66
Design, implementation and applications (pp. three dimensionality in the relationship between
368-391). Mahwah, NJ: Lawrence Erlbaum patient and therapist. Studies in Health Tech-
Associates, Inc. nology and Informatics, 81, 551-553.
Schultheis, M. T., & Rizzo, A. A. (2001). The Vining, D. J. (1997). Virtual colonoscopy. Gas-
application of virtual reality technology in reha- trointestinal Endoscopy Clinics of North
bilitation. Rehabilitation Psychology, 46(3), America, 7(2), 285-291.
296-311.
Westwood, J. D., Haluck, R. S., Hoffman, H.
Shuhaiber, J. H. (2004). Augmented reality in M., Mogel, G. T., Phillips, R., & Robb, R. A.
surgery. Archives of Surgery, 139(2), 170- (Eds.). (2004). Medicine meets virtual reality
174. 12: Building a better you. The next tools for
medical education, diagnosis, and care.
Soferman, Z., Blythe, D., & John, N. W. (1998).
Amsterdam: IOS Press.
Advanced graphics behind medical virtual real-
ity: Evolution of algorithms, hardware, and soft- Westwood, J. D., Hoffman, H. M., Robb, R. A.,
ware interfaces. Proceedings of the IEEE, & Stredney, D. (Eds.). (1999). Medicine meets
86(3), 531-554. virtual reality 7. Amsterdam: IOS Press.
Springer, P., Dessl, A., Giacomuzzi, S. M., Whalley, L. H. (1995). Ethical issues in the
Buchberger, W., Stoger, A., Oberwalder, M., application of virtual reality in medicine. Com-
et al. (1997). Virtual computed tomography putational Biology in Medicine, 2, 107-114.
gastroscopy: A new technique. Endoscopy,
Wood, B. J., & Razavi, P. (2002). Virtual
29(7), 632-634.
endoscopy: A promising new technology. Ameri-
Székely, G., & Satava, R. M. (1999). Virtual can Family Physician, 66(1), 107-112.
reality in medicine. British Medical Journal,
Zajtchuck, R., & Satava, R. M. (1997). Medical
319(7220), 1305-1309.
applications of virtual reality. Communications
Vincelli, F., Molinari, E., & Riva, G. (2001). of the ACM, 40(9), 63-64.
Virtual reality as clinical tool: Immersion and
67
68
Chapter IX
Modelling and Simulation of
Biological Systems
George I. Mihalas
“Victor Babes” University of Medicine and Pharmacy, Romania
ABSTRACT
It is unanimously accepted that a theoretical approach of a system or phenomenon reveals new

features and offers a deeper insight into the intimate mechanisms. In life science, this
approach is mainly based on mathematical modeling, followed naturally by computer
simulation. The chapter presented here tries to give the reader a comprehensive view over the
main issues arising when attempting to build models of biological systems. A series of
applications is shortly presented. The second half of the chapter is dedicated to one of the most
interesting models: protein synthesis regulation. The example follows the classical steps: the
scheme of the processes to be described, the set of differential equations and the results,
including their possible interpretation.
INTRODUCTION viewpoints (Mihalas, Lungeanu, Kigyosi, &

Vemic, 1995):
Mathematical modelling proved to be a useful
research tool, offering an elegant and simple a. A system’s structure
description of a system (Ingram & Bloch, 1984). • Continuous models
The computer’s advent stimulated its use also • Discrete models
in complex systems, like biomedical systems b. An input-output relationship
(Garfinkel, 1965). There are several ways to • Deterministic models (subdivided into
approach a formal description of biological analytical models and models based
systems, yielding several types of models on differential equations)
(Brown & Rothery, 1993, Fishwick & Luker, • Stochastic models
1991). We can classify the models from various
Modelling and Simulation of Biological Systems
c. A modeled feature the experimental conditions that yield

• Models for structure measurable results.
• Models for function • Testing Hypotheses: Sometimes there
are no direct experimental procedures to
test two different hypotheses concerning
BUILDING MATHEMATICAL a certain phenomenon or process; in this
MODELS OF BIOLOGICAL case, we can simulate a system’s
SYSTEMS behaviour under the two conditions and
compare the simulation output with the
There are several typical steps followed for real behaviour.
building mathematical models (Keen & Spain, • Determining Parameters: For the pa-
1992). rameters that are not (easily) accessible
by direct measurements, we can simulate
• Defining and delimiting the system and the the process for a wide range of values of
process to be modelled the studied parameter. By comparing the
• Selecting the variables, both independent results with the real experimental behaviour,
and dependent variables we can estimate the parameter.
• Establishing the relations between the vari- • Prognosis: It is one of the largest appli-
ables cations and lets us analyze possible sys-
• Defining the input and output tem evolution.
• Setting initial conditions and the values of • Feeling the Phenomenon: A better un-
parameters and constants derstanding of the process is achieved
• Choosing the results representation when a system’s behaviour is analyzed
• Establishing the validation mode over several conditions.
• Didactic Applications: These are used
for educational purposes.
APPLICATIONS IN BIOMEDICAL
RESEARCH
AN EXAMPLE: COMPUTER
The results of a simulation are usually pre- SIMULATION OF PROTEIN-
sented as a graphical plot representing the SYNTHESIS REGULATION
system evolution over time; the experimental
conditions are represented by a set of input Delimiting the System and
parameters. the Process
There are several possible applications of
mathematical models and their corresponding The system is represented by a reduced (sim-
simulation programs in biomedical research plified) cell (Mihalas, Niculescu-Duvaz, &
(Mihalas, 1998). Simon, 1985). We take into account only three
components (Figure 1).
• Experimental Design: The system
behaviour for various values of input pa- • A DNA sequence on which we can
rameters is analyzed, letting us choose the distinguish two genes: the synthesis gene
best set of parameters corresponding to (SG) and the control gene (C)
69
Figure 1. Scheme of a protein synthesis • The activation degree, A, is also a depen-

control system. The variables and parameters dent variable.
are also denoted here. • Parameters include proportionality con-
stants for synthesis rates (ϕ for mRNA
and ψ for protein) and for decay rates (λ
for mRNA and η for protein), and the
coupling process between the protein P
and the control gene C for the
autoregulated process.
To establish the relations between vari-

ables, the processes described can be formally
written as:
 dx
• the messenger ribonucleic acid (mRNA)  = ϕA − λx
 dt
• the synthesized protein (P)  dy
 = ψx − ηy
 dt
The process is the protein synthesis, which  1
 A = 1 + y/k
comprises the following major steps: 
• mRNA is produced in the nucleus when Defining the Input and the Output
the SG is activated with a rate dependent
on the activation degree. • The input represents the starting condi-
• mRNA migrates into the cytoplasm and tions, that is, all parameters (ϕ, ψ, λ, η, k)
bounds to a ribosome; we should also con- and initial values for the variables (x0, y 0).
sider the mRNA degradation, with a decay • The output is represented by the time
rate proportional to its concentration. evolution of the two major components
• The protein P is synthesized on the ribo- (variables), that is, x(t) and y(t).
some with a rate proportional to the mRNA
concentration; again, we will also con- Setting Initial Conditions, (i.e.,
sider the protein decay. Associating values to ϕ , ψψ, λλ, ηη, k,
• The activation degree of the SG is regu- x0, and y0)
lated by the control gene C, whose state is
controlled either by the synthesized pro- We also set now in the iteration step a value for
tein P (self-regulation), or by another pro- dt, which is usually very small. It is important to
tein P’ (interregulation). note that for qualitative simulations, we do not
need exact values for the parameters and initial
Selecting the Variables and concentrations; it would suffice to work with
Defining the Parameters relative values (to consider one parameter as a
unit). However, when absolute values are avail-
• Concentrations are dependent variables; able, we will use them.
let [mRNA] = x and [P] = y.
70
Figure 2. Time evolution of a protein (y) on Figure 3. Phase diagram (Oscillatory

an arbitrary time scale behaviour is visible)
Choosing the Results Tomita et al., 1999) to more advanced software

Representation (Loew & Schaff, 2001; Tomita et al.); some
developers also pay special attention to the user
Usually, the time evolution is preferred: x(t) and interface (Ichikawa, 2001).
y(t) (Figure 2). An important step was taken when dedi-
However, we can also draw a phase dia- cated programming languages were proposed
gram y = f(x) (Figure 3), which is convenient (Cuellar, Lloyd, Nielsen, Bullivant, Nickerson,
for oscillatory systems or systems with multiple & Hunter, 2003; Hucka et al., 2003), which
steady states. bring the formal descriptions of biological pro-
cesses to a higher level.
Stating the Validation Mode Our example was taken from molecular
biology, but the use of modeling and simulation
When experimental data are available, a com- in life sciences covers a much wider range,
parison of simulated data with real data is from physiology (Randall, 1987) to integrated
recommended. However, we often consider approaches (Moolgavkar, 1986) and even health-
also the type of behaviour as a validation mode: care systems.
evolution toward steady state(s), damped or
sustained oscillations, and so forth (Mihalas,
Simon, Balea, & Popa, 2000). CONCLUSION
Computer simulations of biological processes

TRENDS FOR FUTURE are convenient tools for biomedical research,
DEVELOPMENTS and are easy to handle with various applica-
tions. The complexity of biological systems
The specific requirements for a convenient imposes limits in their use; most often the
simulation program have encouraged software results are limited to a small number of vari-
development, from simple forms (Mendes, 1993; ables and cover just certain relations, always
71
neglecting others. However, the programs will Loew, L. M., & Schaff, J. C. (2001). The
become more and more complex, covering virtual cell: A software environment for com-
more and more variables, trying to get closer to putational cell biology. Trends in Biotechnol-
the great challenge in this field in the (near) ogy, 19(10), 401-406.
future: to simulate an entire cell.
Mendes, P. (1993). Gepasi: A software pack-
age for modeling the dynamics, steady states,
and control of biochemical and other systems.
REFERENCES
Computer Applications in Biosciences
(CABIOS), 9(5), 563-571.
Brown, D. B., & Rothery, P. (1993). Models in
biology: Mathematics, statistics and com- Mihalas, G. I. (1998). Modelling and simulation
puting. Chichester, UK: John Wiley & Sons. in medicine and life sciences. Medical
Informatics, 23(2), 93-96.
Cuellar, A. A., Lloyd, C. M., Nielsen, P. F.,
Bullivant, D. P., Nickerson, D. P., & Hunter, P. Mihalas, G. I., Kigyosi, A., Lungeanu, D., &
J. (2003). An overview of CellML 1.1: A Vernic, C. (1998). Modeling and simulation in
biological model description language. SIMU- molecular pharmacology. MEDINFO’98, 372-375.
LATION: Transactions of the Society for
Mihalas, G. I., Lungeanu, D., Kigyosi, A., &
Modeling and Simulation International,
Vernic, C. (1995). Classification criteria for
79(12), 740-747.
simulation programs used in medical education.
Fishwick, P. A., & Luker, P. A. (Eds.). (1991). MEDINFO’95, 2, 1209-1213.
Qualitative simulation, modeling and analy-
Mihalas, G. I., Macovievici, G., Simon, Z., &
sis. New York: Springer Verlag.
Lungeanu, D. (1999). MOBISIM: Package for
Garfinkel, D. (1965). Simulation of biochemical simulation in molecular biology. Medical
systems. In R. W. Stacy & B. D. Waxman Informatics Europe ’99, 617-620.
(Eds.), Computers in biomedical research
Mihalas, G. I., Niculescu-Duvaz, I., & Simon,
(Vol. 1, pp. 111-134). New York: Academic
Z. (1985). Trigger with positive control for
Press.
gene activity regulation: Computer simulation.
Hucka, M., et al. (2003). The systems biology Studia Biophysica, 107(3), 223-229.
markup language (SBML): A medium for rep-
Mihalas, G. I., Simon, Z., Balea, G., & Popa, E.
resentation and exchange of biochemical net-
(2000). Possible oscillatory behaviour in P53-
work models. Bioinformatics, 19(4), 524-531.
MDM2 interaction: Computer simulation. Jour-
Ichikawa, K. (2001). A-Cell: Graphical user nal of Biological Systems, 8(1), 21-29.
interface for the construction of biochemical
Mihalas, G. I., Zaharie, D., & Kigyosi, A.
reaction models. Bioinformatics, 17(5), 483-
(2004). Step-wise or continous activation in cell
484.
regulation? A comparative study of p53-mdm2
Ingram, D., & Bloch, R. (1984). Mathematical interaction by computer simulation.
methods in medicine. Chichester, UK: John MEDINFO2004, 1758.
Wiley & Sons.
Moolgavkar, S. H. (1986). Carcinogenesis
Keen, R. E., & Spain, J. D. (1992). Computer modeling: From molecular biology to epidemiol-
simulation in biology. New York: Wiley-Liss.
72
ogy. Annuual Review of Public Health, 7, Deterministic Model: A mathematical

151-169. model based on a relation between the depen-
dent and independent variables; for each set of
Randall, J. E. (1987). Microcomputers and
values of the independent variables, there is
physiological simulation. New York: Raven
only one well-defined possible value of the
Press.
dependent variable.
Tomita, M., Hashimoto, K., Takahashi, K.,
Mathematical Modelling: A method to
Shimizu T. S., Matsuzaki, Y., Miyoshi, F. et al.
describe system behaviour by a set of relations
(1999). E-CELL: Software environment for
between the system’s characteristic variables
whole cell simulation. Bioinformatics, 15(1),
and parameters.
72-84.
Model: A simplified representation of a
system (a conventionally defined part of the
KEY TERMS
universe), keeping only a set of features (vari-
ables, parameters, and relations) considered
Computer Simulation: A computer pro-
relevant and neglecting irrelevant features.
gram that mimics the behaviour of the model
variables for a set of values of the model Stochastic Model: A mathematical model
parameters, and initial values of the indepen- based on probabilities for which there are sev-
dent variables. eral possible values for a set of values of the
independent variables; the model computes the
probabilities of various output values.
73
74
Chapter X
Virtual Reality Simulation in
Human Applied Kinetics and
Ergo Physiology
Bill Ag. Drougas
ATEI Education Institute of Epirus, Greece
ABSTRACT
Virtual reality is today an excellent tool for a full simulated experience in a modern
environment where any researcher or any individual scientist may work with vital experimental
environments or use parameters that sometimes does not really exist. It is already a vital step
for the future of science and for the modern experiment. Ergo physiology today has many
applications for research. We can find new unknown parameters for the human body
searching biokinetics and ergo physiology, and it is time to use modern technologies and
applications. The vital issues discussed in this chapter may offer many applications for human
kinetics and movement and may also discuss biokinetics research using the physical laws and
parameters in various biokinetics and physiology fields.
INTRODUCTION perimentation, and we can recognize that dur-

ing the past years, it has continued with great
Virtual reality is today an important part of success. An important field using this new way
modern scientific methodology and research, of research is the simulation of human move-
using modern high-speed computers of lately ment in ergo physiology and applied biokinetics
designed technologies for research and simula- science. Virtual reality is very useful for re-
tion in various scientific fields such as ergo searchers in these fields because they can have
physiology and biokinetics. This is a new field in simulations of the physical human body at any
contemporary science, methodology, and ex- time they want for study or experimentation.
Virtual Reality Simulation in Human Applied Kinetics and Ergo Physiology
BACKGROUND ISSUES
One of the first authors who wrote about virtual In the beginning, many other scientists worked
reality was Howard Rheingold (1991), who with computer data for virtual reality applica-
wrote about data visualization and 3-D CAD tions in various fields with very big success.
(computer-aided design) in which someone may Today, there are many different fields and
use his or her hands and fingers. Many applica- applications of virtual reality technology. Table
tions can be found from the middle war period 1 summarizes some of the virtual reality appli-
by the U.S. Air Force to create flying simula- cations similar to those of ergo physiology.
tions. Especially in the fields of ergo physiology
Myron Krueger, during the ’60s, worked on and biokinetics, virtual reality is used in many
the affiliation of human and computer with applications. Some of the characteristics of
special research in computer-controlled re- human movement, the human body, and param-
sponsive environments, which were named by eters such as space, geometry, color, and sound
him artificial reality (Krueger, 1993). He also may help virtual reality programs become more
designed the video place, a system that contains effective in various methodologies of research
a projection screen and a video camera that is and virtual applications.
controlled by a computer. By this method, The importance of this is to find a method-
human movements in each activity are trans- ology for using virtual reality and a way to
ferred to computer graphics in software recognize the results, such as some of the
(Boudouridis, 1994). official physiology results that can give to re-
So, there can be a connection between searchers many new discoveries in existing
human and technical things in space with com- science and theory, or future science research
puter graphics. This was one of the first meth- in finding new signals from the human body
odologies in human-kinetics research and appli- during simulations.
cations. Today we have all the modern technology to
Tom Furness was another scientist who make better simulations for the human body
designed the Super Cockpit for the U.S. Air
Force after many years of research. In a small
place, a human could use computers and a
HMD (head-mounted display) to understand Table 1. Summary of the virtual reality
vital secrets of the flight without any danger applications similar to those in ergo
(Furness, 1991). physiology fields
But the man who is the father of the terms
virtual reality and reality engine is Jaron Lanier, • Human behavior in flying simulations
• Human behavior in space simulations
an informatics scientist who, with another young • Learning and human-movement programs in
kinetics
man named Tom Zimmerman, established the • Neurodisease science
Visual Programming Language Research Inc • Rehabilitation
• Physical behavior in space
in 1980 (Boudouridis, 1994). This company was • Step-correction learning and research
the first to make important tools for virtual • Adapted methodologies in kinetics
• Special gymnastics programs
reality programs and applications, such as data • Study of the human senses and their characteristics
gloves and HMDs. • Pain confrontment
• Continuing education
75
Table 2. Important parameters in virtual kinetics today use many important laws of
reality simulations for ergo physiology and physics, physiology, mathematics, statistics, and,
human movement of course, medical data, and it is important to
continue to grow using new technologies like
• Length
• Height virtual reality. Table 2 summarizes some of the
• Time parameters that are important in virtual reality
• Sound frequency
• Color frequency simulation for the human body and movement.
• Wave length In some experiments, all of the parameters
• Velocity
• Pressure in Table 2 can be used, and in some others, only
• Area a few of them can be used. However, the
• Energy
• Power parameters in the simulations are very near to
those of physics, and the philosophy is about the
same in similar experiments.
Table 3 summarizes some of the fields in
and to see new fields that had not previously which virtual reality can help to recognize and
existed. This is, of course, the future of re- study the physical parameters of Table 2.
search. In every subject of the study of human
For example, it is not possible for anyone to abilities, there are various techniques and meth-
fly at high speeds without any danger so that odologies. However, the philosophies of all
scientists can see the behavior of the human methodologies are about the same and have
body or parameters such as blood pressure, the similar directions.
behavior of the heart, muscle energy, signals or So it is important to understand that with
other problems in the eyes, and so forth. But virtual reality simulations in ergo physiology,
with virtual reality simulations, we can today we may simulate human models in an environ-
register many of these parameters and stan- ment that does not really exist physically. In this
dards, see how they change after the experi- way we can find new and unknown parameters
mentation, and see it is used in our theories. of the basics of human biokinetics and ergo
The very best aspect of this method is that physiology. In the same way, a researcher can
we can stop the experiment when we must or establish new models of experiments and fields
when we want to begin the experiment again for research using many parameters of the
from the last step, or, if we want, we can design modern population and all the modern kinetic
another model with new parameters. So, we problems that appear in people. Researchers
can always put something new into our experi-
mentation and theory, and this is important Table 3. Summary of the fields of the physical
because every person is different from every parameters in virtual reality
other. By this methodology, we can register
statistical effects or make a very good math- • Kinetics conception
• Space-creation parameters
ematical analysis of the problem to continue • Relative-movement creation
with advanced research and measurements or • Color-conception studies
• Sound-conception studies
to make other project experiments. • Kinetics-parameters registration
So many past methodologies or past tech- • Study of velocity
• Study of frequency
nologies will be changed after these experi- • Study of space parameters
ments. Ergo physiology and applied human • Study and registration of time
76
can include these problems in the data of the Table 4. Summary of some of the simulated
simulation and study the methodology or the and virtual applications
effects within the simulation. For example, the
influence of a brain problem on human kinetics • Simulation of ergonomic environments
ability can be registered from various kinetics • Simulation-based training
• Rehabilitation and modern orthopedics
parameters and/or kinetics behavior and move- • Virtual patients and human organs for surgery
• Psychotherapy and applied physiology
ment in various daily skills. Many years before, • Sports simulations
this was not so clear, and many of these prob- • Physiological behavior with various phobias
• Biomechanical walking
lems seemed like they didn’t exist. In simula- • Virtual training for balance
tions, we have two bodies: the virtual and the
real.
It is important to understand that two human
bodies in the same experiment are the same but behavior, and, of course, for education pro-
also different. grams for researchers or students at the gradu-
We have two human bodies during the ex- ate or postgraduate levels.
periment that we can use, or many more if it is Table 4 summarizes some of the simulated
possible with the program. Virtual reality bioki- and virtual applications to human abilities.
netics research programs must have both hu- Many of the methodologies that exist today
man bodies: the real and the virtual one. The began first in front of a computer screen and,
parameters and simulations must be very close with the help of modern programs of simulation
between these two categories. These two bod- in physiology, biology, and mathematics, they
ies are different persons, but closely exhibit the continue to establish some of the new theories.
same movements. It is very important to recog- Today, electronic methods give us the op-
nize the abilities of the bodies. The information portunity to continue our research in a safer
about kinetics behavior or any other informa- work environment; in the past, various experi-
tion from any data structure that defines the mental problems in the physical environment
behaviors of the movement and kinetic param- were impossible to avoid. Virtual reality and its
eters of one virtual human can be passed di- development are currently underway in a num-
rectly to all the other virtual humans if there is ber of organizations such as NASA, IBM,
more than one human in the program of the Boeing, and so forth, and also in many official
experiment. In this way, we can have param- research programs of famous university labo-
eters and measurements from various kinetic ratories and technological educational insti-
models at the same time and study many differ- tutes. In many of these institutes, there exist
ent behaviors. This is the philosophy of a mul- many new international programs of human
tiple experiment. study and ergo physiology research.
Two or more students or researchers can
work with two or more models at the same time
and give different conclusions and different CONCLUSION
theories. Virtual humans are real models that
can be used many times for the same work and Virtual reality is today an excellent tool for a
for the same or a different research. This is fully simulated experience in a modern environ-
very good for the physiology studies and medi- ment where the researcher or simple scientist
cine experiments for human biokinetics and may work with vital experimental environments
77
or use any parameter that sometimes does not teleinformatics and applications in medi-
really exist. This is important for people who cine. ATEI of Epirus, Department of
are researching the abilities of the human body. Teleinformatics.
The vital issues discussed in this article may
Drougas, B. Ag. (2004a). Home care and the
offer many applications for human kinetics and
modern technologies (1st ed.). D-publications
movement, and biokinetics research using physi-
Arta Greece.
cal laws and parameters. In the same philoso-
phy, any researcher may propose or design new Drougas, B. Ag. (2004b). Telemedicine appli-
models for informatics scientists. Ergo physiol- cations in a contemporary environment. D-
ogy today has many applications for research, publications Arta Greece.
and the modern computer systems for simula-
Drougas, B. Ag. (2004c). Virtual reality and
tions and virtual reality give us many applica-
health: Manual of the Laboratory of the
tions and the opportunity to work in easier
Teleinformatics and Applications in Medi-
experimental environments to make and design
cine. ATEI of Epirus, Department of
experiments using real and unreal parameters
Teleinformatics.
without any danger for us or any other human.
This is already a step toward the future of Furness. (1991). Harnessing virtual space. So-
science and the experiments. ciety for Information Display Digest of Tech-
nical Papers (1st ed.).
Hoffman, H. G. (2004). Therapy applications
REFERENCES
of virtual reality. Scientific American (Greek
ed.), 28-36.
Boudouridis, M. (1994). The technology of
virtual reality. Presentation to the Interna- Koutsuris, D., Nikita, K., & Pavlopoulos, S.,
tional Summer School of Communications, Ath- (2004). Medical represented systems.
ens, Greece. Thesaloniki, Greece: Tziolas Publications.
Computers: Artificial intelligence. (n.d.). Time Koutsuris, D., Pavlopoulos, S., & Prentza, A.
Life, 73-77. (2003). Introduction into biomedical tech-
nology and analysis of medical signals.
Deisinger, J., Breining, R., & Robler, A. (n.d.).
Thesaloniki, Greece: Tziolas Publications.
ERGONAUT: A tool for ergonomic analysis in
virtual environments. In J. D. Mulder & R. van Krueger, M. (1993). Artificial reality. New
Liere (Eds.), Virtual Environments 2000, York: Addison-Wesley.
Proceedings of the 6th Eurographics Work-
Rheingold, H. (1991). Virtual reality. New
shop on Virtual Environments, Amsterdam,
York: Simon & Schuster.
The Netherlands. New York: Springer Wien.
Thalmann, D. (n.d.). Animating autonomous
Drougas, B. Ag. (n.d.). Rehabilitation by new
virtual humans in virtual reality. Swiss Fed-
technologies. Research and Theory Journal,
eral Institute of Technology, Computer Graph-
12-15.
ics Lab.
Drougas, B. Ag. (2003). Theory and labora-
Zeimbekakis, G., (2003). Telematics applica-
tory applications for the lesson of
tions. Athens: Modern Season Publications.
78
KEY TERMS Rehabilitation: A scientific methodology

using physical methods, especially applied move-
Biokinetics: A scientific division of ergo ments, kinetics, and modern psychology appli-
physiology for the research and recognition of cations, to change the human body and mind
physical laws, standards, and existing param- behavior, helping people to recreate their en-
eters of human movement and kinetics, or the ergy levels and their physical ability, or to use
changes from any external or internal effect the near environment in the best way to live
that can change any physiological movement. with their problems for a better life.
The creation of a physical model helps re-
Simulation: A number of presentations of
searchers to understand the problems of move-
different parameters for things, models, or en-
ment and kinetics and their relation with various
vironments designed from physical standards
pathological problems.
and laws. In this virtual presentation, one can
Ergo Physiology: A division of physiology use his or her experience to have an interactive
that searches physiological and physical pa- participation and use this virtual environment
rameters, constants, and standards for the cre- for personal training, education, rehabilitation,
ation of energy, power, balance, velocity, and and so forth.
other physical changes in the human body. This
Virtual Human: A virtual human body in a
division includes a vital search for the physical
3-D immersive, interactive simulation, created
and physiological laws related to various kinet-
by computer software, for the parallel presen-
ics problems for humans.
tation of reality in the virtual using any influence
Kinetics Behavior: A number of charac- from the real environment.
teristics of human movement including the physi-
Virtual Reality: An immersive, interactive
cal behavior of the human body in movement
simulation for real and unreal environments.
and kinetics. These characteristics draw a be-
We can also use this term for simple interactive
havior model for the physical body and offer the
models that cannot offer to a user any immersive
opportunity to researchers in the area of physi-
presentation in a virtual environment. Virtual
ology and biokinetics to search the behavior of
reality is the most useful tool in medicine today,
humans and to create scientific programs for
with many applications in various fields, espe-
rehabilitation.
cially in rehabilitation, education, and so forth.
Physical Biokinetics Parameters: A num-
Visualization: A medicine application us-
ber of different parameters, laws, and mea-
ing a 3-D human body model that has been
surements that are recognized from a physical
made from a number of different 3-D scans
model from physiology and human movement.
from a medical computed tomography.
We can use this model in various applications
recognizing human behavior and movement.
79
81
Chapter XI
Care2x in Medical
Informatics Education
Andreas Holzinger
Medical University Graz (MUG), Austria
Harald Burgsteiner
Graz University of Applied Sciences, Austria
Helfrid Maresch
Graz University of Applied Sciences, Austria
ABSTRACT
In this chapter the authors report about their experiences in education of both students of
healthcare engineering at Graz University of Applied Sciences, and students of medicine at
the Medical University Graz, gained during the winter term 2004. Care2x is an open source
Web-based integrated healthcare environment (IHE). It allows the integration of data,
information, functions, and workflows in one environment. The system is currently consisting
of four major components, which can also function independently: hospital information
system (HIS), practice management (PM), a central data server (CDS) and a health exchange
protocol (HXP). Although the components are under heavy development, the HIS has reached
a degree of stability, where one can use it at least for educational purposes. Various groups
also report the usage of enhanced versions of Care2x in real life settings. Our experiences in
both—very different—student groups have been very promising. In both groups the acceptance
was high and Care2x provided good insights into the principles of a hospital information
system. The medical students learned the principal handling of a HIS, whereas the engineering
students had the possibility to go deeper into technical details.
Care2x in Medical Informatics Education
INTRODUCTION tions, and common tasks of an HIS, but to also

let them work with a fully functional HIS during
In this chapter, the authors report about their lectures. This is essential, particularly if stu-
experiences in the education of students of dents are required to be able to work with
healthcare engineering (HCE) at Graz Univer- possibly any HIS in practice after only a short
sity of Applied Sciences, and students of medi- period of vocational adjustment. However, it
cine at the Medical University Graz, gained depends on many different factors regarding
during the winter term of 2004. Care2x is an which HIS to choose. One of the most impor-
open-source Web-based integrated healthcare tant is whether it is necessary to teach (with) a
environment (IHE). It allows the integration of particular HIS of a certain vendor, for example,
data, information, functions, and work flows in if this system is deployed in a network of local
one environment. The system currently con- hospitals. Another key factor, especially for
sists of four major components, which can also noncommercial educational institutions, is the
function independently: the hospital information economic impact of the introduction of a com-
system (HIS), practice management (PM), a mercial HIS at the university. Third, for the
central data server (CDS), and a health ex- education of students of medical informatics, it
change protocol (HXP). Although the compo- might also be reasonable to teach the process of
nents are under heavy development, the HIS developing (parts of) a bigger software engi-
has reached a degree of stability so that one can neering project. Hence, the need for an open-
use it at least for educational purposes. Various source system arises if one does not want to
groups also report the usage of enhanced ver- start the development of his or her own HIS.
sions of Care2x in real-life settings. Our expe- Although there are many more factors to con-
riences with both—very different—student sider in general, we chose Care2x as our pri-
groups have been very promising. In both groups, mary educational HIS for the following rea-
the acceptance was high and Care2x provided sons.
good insights into the principles of a hospital
information system. The medical students
learned the principal handling of an HIS, whereas CARE2X
the engineering students had the possibility to
go deeper into technical details. Care2x is a generic multilanguage, open-source
How to prepare both medical and engineer- project that implements a modern hospital in-
ing students in the best possible way for their formation system (the Web page of Care2x is
later work with modern HISs is a common located at http://www.care2x.org/). The project
question. Whereas students of engineering are was started in May 2002 with the release of the
rather enthusiastic about IT, students of medi- first beta version of Care2x by a nurse who was
cine are skeptical in general about using it. dissatisfied with the HIS in the hospital where
However, HISs are not widely accepted by he was working. As of today, the development
healthcare professionals; that is, barriers to the team has grown to over 100 members from
use of HIS are primarily sociological, cultural, over 20 countries. Care2x is a Web-based HIS
and organizational rather than technological that is built upon other open-source projects:
(Moore, 1996). the Apache Web server from the Apache Foun-
It seems plausible to not only give students dation (http://www.apache.org/), the script lan-
theoretical background about the structure, func- guage PHP (http://www.php.org/), and the re-
82
Figure 1. Help page describing the clinical path for starting a new surgery-operation
document
lational database-management system (RDMS) for some clinical paths. See Figure 1 for an
mySQL (http://www.mysql.com/). There exist example.
several source-code branches that try to inte-
grate the option to choose from other RDBMSs
like Oracle and postgreSQL. The latter one is REVERSE ENGINEERING
already supported in the current version at the
time of this writing (Deployment 2.1). For our The reverse engineering of existing complex
investigations, we chose the most feature-rich software packages starting at the source-code
version that was available from the Care2x level has a higher value for practical education
Web page in early fall of 2004. This release had than a new development. Bothe (2001) argues
the version number 2.0.2. Some minor deficien- that groups of students will rarely be able to
cies that we report later may already be fixed develop a project further than to a prototype
in the current version, Deployment 2.1. stage during a single lecture. Access to the
Care2x is a very feature-rich HIS that is source code is not available for most commer-
fully configurable for any clinical structure. It is cial HISs, which is another advantage of using
built upon different modules, which include, for Care2x as an educational system. In our first
example, in and outpatient administration, ad- lecture, the students of HCE were asked to test
mission, pharmacy, radiology (including DICOM all functions and paths of Care2x. They had to
[Digital Imaging and Communications in Medi- set up a small virtual clinic and employ doctors,
cine] image uploads), laboratories, ambulatories, nurses, and technical stuff. Finally, patients had
nursing, medics, DRGs (diagnosis-related to be admitted, attended to, and dismissed at all
groups), and so forth. Online help is available stations. In a second lecture in the upcoming
83
semester, our students have the assignment to 3. Familiarization with Care2x in practical
analyze a fully functional HIS at the source- sessions
code level. Since Care2x is built upon a modular 4. Practical work, specific work flows
structure, small teams of programmers have 5. Applying reverse engineering (HCE group
tasks like finding and fixing bugs in the current only in the second part of the lecture)
version, adding simple modules for special func- 6. Examination (both theoretical and practi-
tions not included in the official version, or cal)
implementing interfaces to other existing infor-
mation systems or medical equipment. In the During the education, the students were
spirit of open-source projects, reasonable addi- faced with the following strengths and weak-
tions and modifications can and should be pub- nesses of Care2x.
lished to the Care2x community Web page. Strengths:
• Everyone can make his or her own tools

LESSONS LEARNED • Work does not have to be done in a strict
order
Approximately 100 students from medicine and • Very flexible
25 from HCE participated (Figure 2). The • Easy to handle
whole lecture was built in the following way: • Continuing design and development
• Open source
1. Theoretical foundations of HIS in tradi- • Lots of different languages
tional lectures • Bg community that takes care of Care2x
2. Principles of Care2x explained (HCE • Easy to select the different departments
group’s lecture was more technology ori- and stations
entated)
Figure 2. Students at work with Crae2x (We assigned groups of 2-3 students with different
tasks related to the administration of a virtual hospital)
84
Weaknesses: the program has included this function. That did

not solve the problem completely: Every now
• No real standard between the modules and then the integrated back button of Care2x
• Documentation is only rudimentary led to nowhere, too. In addition, some pages did
• A few tools are not really easy to interpret not include the Care2x back button (inconsis-
• Lack of security measures tency), resulting in a blank page. This required
• Not a state-of-the-art user interface the user to restart at the very beginning and
• There is no global list of patients from click through all the menus once again, which
which to select one was boring for the students.
The general software problems that did not
concern the running process were not severe.
OBSTACLES IDENTIFIED However, there is a serious problem when it is
possible to admit a patient to more than one
During our lectures and trainings, there emerged station, or when it is possible to alter a patient’s
several problems while using Care2x. There record after his or her death.
are a lot of small bugs that caused troubles. The A severe problem that has to be solved is
biggest problem was that sometimes the browser that patients have to be discharged and then
responded with an “inactivity error” and the hospitalized again when we just want them to
session would time out. Most of the time this be transferred from the ambulatory to a station.
error message was shown, the last click had not There are some translation errors and miss-
been made but one single minute ago. The next ing notes. For example, if a new patient record
problem with the handling was that sometimes is being applied, there are red stars above some
the back button on the Web browser would lead properties. Although this is an obvious sign for
to nowhere because Care2x does not manage experienced users, it is not noted anywhere
this. Much later, we found out that the back why these stars appear. The students found out
button of the browser is unnecessary because that these stars show the minimum amount of
Figure 3. Example of a graphically embedded complex form, the diagnostic test order
85
Figure 4. Nursing information about stationary occupancy for one of our virtual wards
data that is required to create a patient record, and the software is not very intuitive. However,
but how would the students of medicine with if one trains with Care2x, the work flows
little experience in IT know this fact? become clearer and more logical. The online
It is also sometimes annoying that bits of help of Care2x should be better and more
information are hidden behind a link. For ex- comprehensive. Working with the software
ample, if you want to hospitalize a patient, you was very fun because you really can play with
have to remember the social insurance number a virtual hospital. Care2x is a very good possi-
because it is not shown in the place it is needed. bility for training with work flows in a hospital.
This is due to the fact that Care2x works with Further improvement of Care2x will open new
only one window. Sometimes there might just areas to work with this software.
be too little space to provide all the information
needed, and then the user has to write this
information down or remember it; this cannot REFERENCES
be the aim of an HIS.
Alpay, L., & Murray, P. (1998). Challenges for
delivering healthcare education through
CONCLUSION telematics. International Journal of Medical
Informatics, 50(1-3), 267-271.
Care2x is flexible open-source software. Al-
Bemmel, J. H. v., & Musen, M. A. (2000).
though there are some bugs, it has the potential
Handbook of medical informatics. Berlin,
to become functional software to support work
Germany: Springer.
flows within a (real) hospital. We think the
biggest problems are the documentation and Bothe, K. (2001). Reverse engineering: The
the deduction of treatments. Working with challenge of large-scale real-world educational
Care2x as a beginner is not very comfortable,
86
projects. 14th Conference on Software Engi- ture and stable product that can be used at least
neering Education and Training (p. 115). for educational purposes for both students of
medicine and students of medical informatics.
Fieschi, M. (2002). Information technology is
Some groups report the deployment of en-
changing the way society sees health care
hanced and adopted versions in real hospitals.
delivery. International Journal of Medical
Informatics, 66(1-3), 85-93. Diagnosis-Related Group (DRG): The
DRG system is an inpatient classification sys-
Haux, R. (1998). Health and medical informatics
tem based on several factors: the principal
education: Perspectives for the next decade.
diagnosis, secondary diagnosis, surgical fac-
International Journal of Medical
tors, age, sex, and discharge status. Under the
Informatics, 50(1-3), 7-19.
Medicare prospective payment system, hospi-
Haux, R. (2002). Health care in the information tals are paid a set fee for treating patients in a
society: What should be the role of medical single DRG category, regardless of the actual
informatics? Methods of Information in Medi- cost of care for the individual.
cine, 41(1), 31-35.
Digital Imaging and Communications in
Haux, R., Hasman, A., Leven, F. J., Protti, D. Medicine (DICOM): The DICOM image
J., & Musen, M. A. (1997). Education and format is commonly used for the transfer and
training in medical informatics. In J. Bemmel, J. storage of medical images. Visit Chris Rorden’s
V. Bemmel, & M. A. Musen (Eds.), Hand- DICOM page for information about the format
book of medical informatics (p. 537ff). Heidel- and free software to view and manipulate it.
berg, Germany: Springer.
Hospital Information System (HIS): It is
Haux, R., & Knaup, P. (2000). Recommenda- the central medical information system in most
tions of the International Medical Informatics hospitals in which most healthcare-related data
Association (IMIA) on education in health and (e.g., personnel, stations, patients and their
medical informatics. Methods of Information medical history, etc.) are stored.
in Medicine, 39(3), 267-277.
Medical Informatics: The rapidly devel-
Hovenga, E. J. S. (2004). Globalisation of health oping scientific field that deals with biomedical
and medical informatics education: What are information, data, and knowledge: their storage,
the issues? International Journal of Medical retrieval, and optimal use for problem solving
Informatics, 73(2), 101-109. and decision making. The emergence of this
new discipline has been attributed to advances
Moore, M. B. (1996). Acceptance of informa-
in computing and communications technology,
tion technology by health care professionals.
to an increasing awareness that the knowledge
Symposium on Computers and the Quality of
base of medicine is essentially unmanageable
Life (pp. 57-60).
by traditional paper-based methods, and to a
growing conviction that the process of informed
KEY TERMS decision making is as important to modern
biomedicine as is the collection of facts on
Care2x: An open-source HIS available from which clinical decisions or research plans are
http://www.care2x.org/. Care2x is a quite ma- made (Shortliffe, 1995).
87
Open Source: The idea of sharing the RDBMS (Relational Database Manage-
source code of applications or tools for free. ment System): A software package that man-
Other people are invited to elaborate on future ages a relational database, optimized for the
extensions and improvements. Most open-source rapid and flexible retrieval of data. It is also
projects are committed to one of the Gnu public called a database engine.
licenses (see http://www.gnu.org/licenses/
Reverse Engineering: Taking apart an
licenses.html).
existing system to analyze smaller or single
parts. The reduced complexity simplifies the
process of enhancing or understanding its func-
tions.
88
89
Chapter XII
An Object-Oriented Approach to
Manage E-Learning Content
Using Learning Objects
Andreas Holzinger
Josef Smolle
Gilbert Reibnegger
ABSTRACT
Learning objects (LO) are theoretically based on granular, reusable chunks of information.
In this chapter the authors argue that LOs should consist of more than just content, that is, they
should include pre-knowledge questions on the basis of the concept of the advanced
organizer, of self-evaluation questions (assessment), and finally of appropriate metadata. The
used metadata concept must be based on accepted standards, such as learning object
metadata (LOM) and the shareable object reference model (SCORM). A best practice example
of the realization of these concepts is the Virtual Medical Campus Graz (VMC-Graz), which
actually is the realization of an information system to make a new curriculum digitally
accessible.
INTRODUCTION should include preknowledge questions on the

basis of the concept of the advanced organizer,
Learning objects (LOs) are theoretically based self-evaluation questions (assessment), and fi-
on granular, reusable chunks of information. In nally appropriate metadata. The used metadata
this article the authors argue that LOs should concept must be based on accepted standards,
consist of more than just content; that is, they such as learning object metadata (LOM) and
An Object-Oriented Approach to Manage E-Learning Content Using Learning Objects
the shareable object reference model learning objects (ELOs; e.g., Muzio, Heins, &
(SCORM). A best-practice example of the Mundell, 2002) or reusable learning objects
realization of these concepts is the Virtual (RLOs; e.g., Polsani, 2003).
Medical Campus Graz (VMC-Graz), which Within the VMC-Graz, we use LOs as a
actually is the realization of an information new way of considering and handling learning
system to make a new curriculum digitally content. They include at least the following four
accessible. characteristics (compare also with the Center
We regard LOs as having a historical foun- for International Education of the University of
dation in the object-oriented paradigm of com- Milwaukee; Beck, n.d.):
puter science. Object orientation basically val-
ues the creation of components (called objects) • They must be much shorter than tradi-
that can be reused (Booch, 1994; Dahl & tional learning units, typically ranging from
Nygaard, 1966). 2 minutes to 15 minutes (absolute maxi-
Cisco (n.d.) defines such a learning object mum within the VMC-Graz is 45 min-
as “a granular, reusable chunk of information utes).
that is media independent.” The term informa- • They must be self-contained: Each learn-
tion chunk reaches back to Miller (1956); in his ing object can be used independently.
sense, a chunk is an information unit that can be • They must be tagged with metadata, which
perceived at one time by the individual and contain descriptive information allowing
stored in the short-term memory (STM). Chunks them to be easily found.
are generally information units that can be • They can be aggregated: Learning objects
individually complex and intra-individually very can be grouped into larger collections of
different (Simon, 1974). content, including traditional course struc-
Generally, the term media object is also tures.
often used, and for the purpose of e-learning,
this type of object is further defined as “digital
media designed and/or used for instructional INSTRUCTIONAL DESIGN
purposes” (South & Monson, n.d.). Such ob- THEORY AND LEARNING
jects range from simple text to video demon- OBJECTS
strations and interactive simulations (Holzinger
& Ebner, 2003). Instructional design theories (IDTs) describe
According to Wiley (2001), however, the methods of instruction and the situations in
main idea of LOs is to break educational con- which these methods should be used. The meth-
tent down into small chunks so that they can be ods can be broken into simpler component
(re)used in various learning environments, in methods and are probabilistic (Reigeluth, 1999).
the spirit of object-oriented programming. The IDT, or instructional strategies and criteria for
Learning Object Metadata Working Group of their application, play an important role in the
the IEEE Learning Technology Standards Com- application of learning objects. Combination
mittee (LTSC) refers to an LO as “any entity, and granularity are two factors that we con-
digital or non-digital, which can be used, re- sider vital:
used or referenced during technology enhanced
learning” (Robson, n.d.). Some authors use • Combination: Whilst the LTSC promotes
other terms; for example, they speak of e- international discussion around the tech-
90
Figure 1. A learning object within the VMC-Graz typically consists of four parts
Pre-knowledge
Content
questions
Self-assessment
questions Metadata
The structure of a Learning Object in the VMC-Graz
nology standards necessary to support be considered as the core learning-object no-

learning-object-based instruction, and tion, this question must be answered cautiously.
many people are talking about the finan- Luckily, within the VMC-Graz, this problem
cial opportunities about to come into exist- was relatively easy to solve due to the modular
ence, there is astonishingly little conver- and strict logic of the curriculum.
sation concerning the instructional design Within the VMC-Graz, a LO can have any
implications of learning objects (Wiley, granularity with the maximum didactical dura-
2001). tion of a lecture unit of 45 minutes. In any case,
• Granularity: The discussion of the prob- the produced LO must fit into this lecture unit.
lem of combining learning objects in terms For example, this is in close accordance with
of sequencing leads to another connection Reigeluth’s (1999) elaboration theory. Wiley
between learning objects and IDT. The (2001) synthesized this and other IDTs into a
most difficult problem facing the design- learning-object-specific instructional design
ers of learning objects is that of granular- theory called the learning object design and
ity (Wiley, 2001). How big should a learn- sequencing theory.
ing object be?
The IEEE LTSC leaves room for an entire PRACTICAL APPLICATION OF

curriculum to be viewed as a single learning LOs IN THE VMC-GRAZ
object, but such a large object view diminishes
the possibility of learning-object reusability. The general objective of the Virtual Medical
Due to the fact that reusability should always Campus Graz (http://vmc.meduni-graz.at) is
91
Figure 2. The logical structure of the VMC-Graz. The atomic unit is an LO that is then
assembled in lecture hours. An LO can have a maximum didactical size of one lecture hour.
Previous
knowledge Content
assessment
Final knowledge Metadata

assessment
VMC Graz Structural Hierarchy of Content
not a new learning platform, but the realization custom-made system, which was specially de-
of a tailor-made information system to make signed and developed by using a user-centered
the curriculum digitally accessible and to sup- development process (Holzinger, 2003, 2004).
port the end users in the creation of individual The content is developed by medical domain
work flows. Consequently, it is not aiming at experts in close cooperation with media spe-
providing traditional distance learning courses, cialists. The project, as such, should not be
but it contains accompanying material that sup- regarded in isolation, but rather as a part of the
ports the students before (e.g., prereadings), development of an e-learning strategy for the
during (e.g., hands-on experiments or simula- whole medical faculty. The solution of didactic
tions), and after the real lectures with corre- problems is central to this type of software
sponding material. Thus, the system does not project, and multimedia is one of the many
replace any lecture but supports every lecture, possible elements of the solution (Holzinger,
and the system can be used in any learning 2002).
scenario. An LO within the VMC-Graz consists usu-
Technically, the departmental knowledge ally of four parts:
covering the different disciplines is stored in
LOs and can be accessed via teaching and 1. Preknowledge questions,
learning module catalogs. The target audience 2. Learning material (content),
of the Virtual Medical Campus Graz is about 3. Self-evaluation questions, and
4,500 students and 600 teachers of the medical 4. Metadata.
university. This high number of users justified a
92
LOs form atomic units. These are grouped new material, a connection is made cognitively
within the VMC-Graz in lecture units (45- and the information is processed into long-term
minute lecture blocks, called lecture hours or memory (Bruning, Ronning, & Schraw, 1999).
lessons), thematic groups (topics, themes), and Schema theory is a cognitive learning theory
modules (see Figure 3 and Figure 4). An LO is that was introduced by Bartlett (1932). Piaget
technically unrestricted in the amount of data. (1961) described schemas as the basic building
The only limitation is didactical; that is, an LO blocks of knowledge and intellectual develop-
must not exceed a maximum of 45 minutes in ment. Schemas are extremely interesting in the
didactical size. The longest LO fills one lecture field of human-computer interaction, for ex-
hour with material. However, this didactical ample, to include knowledge structures that
length is determined by the lecturers them- store concepts in human memory, including
selves. They must know which material they procedural knowledge of how to use concepts
wish to supply to the students as support for a (Chalmers, 2003; Satzinger, 1998; Shapiro,
45-minute lecture block. 1999).
Preknowledge Questions The Content of an LO
The preknowledge questions serve in our LOs Taxonomy of Content

as advance organizers that are used as frame-
works for helping students to understand what Similar to Bloom’s (1956) famous taxonomy of
is to be learned. The term advance organizer educational objectives, Wiley (2001) also de-
was originally used by Ausubel (1960) to developed a taxonomy of LOs and differentiated
scribe a process of linking the upcoming unfa- between five learning-object types, which we
miliar learning material to the learners’ previ- also used within the VMC-Graz.
ously acquired knowledge. Generally, advance
organizers are defined as a kind of appropri- • A fundamental LO can include as content
ately relevant and inclusive introductory mate- either an image (JPEG, GIF, or others; in
rial, introduced in advance of the learning ma- medical education, images play an impor-
terial itself and used to facilitate the establish- tant role), a document (DOC, PDF, PPT,
ment of a meaningful set of learning (cf also etc.), a movie (MPEG, AVI, etc.), or any
with Ausubel, 1968; Corkill, Buring, & Glover, other file, for example, a simple text entry
1988; Kralm & Blanchaer, 1986). (containing only a literature reference to a
Advance organizers are closely related to hard-copy library book).
the schema model of cognitive processing. The • A combined-closed LO can contain, for
schema theory suggests that students learn example, a video with accompanying au-
better when information is presented in an dio.
associative organization. Students build new • A combined-open LO contains, for ex-
information on information that is already mas- ample, an (external) link to a Web page
tered, thus scaffolding new knowledge on top dynamically combining JPEG and
of old. In other words, learning progresses from QuickTime files together with extrane-
what is already known to what is unfamiliar, ously supplied textual material.
and then finally to the relationship between the • A generative-presentation LO can con-
two. When the prior knowledge is linked to the tain, for example, a JAVA applet.
93
• A generative-instructional LO may in- ity content on the one hand and a professional,
clude, for example, an execute instruc- media-didactic, and technical realization on the
tional transaction shell (Merrill, 1999), other hand.
which both instructs and provides practice
for any type of procedure. Self-Evaluation Questions
The purpose of the taxonomy of Wiley (2001) Self-evaluation methods make it possible for
was to differentiate between possible types of learners to check their progress (Bloom,
learning objects available for use in instruc- Hastings, & Madaus, 1971). Due to the fact
tional design. This taxonomy is not all encom- that multiple choices have been used at our
passing in that it includes only those LO types medical facility for a long time, we also support
that facilitate high degrees of reusability. Types all questions in a multiple-choice test style
of learning objects that hamper or even prevent (Gathy, Denef, & Haumont, 1991; McDonald,
reusability (e.g., an entire digital textbook cre- 2002). According to Burton, Sudweeks, Merril,
ated in a format that prevents any of the and Wood (n.d.), the difficulty of multiple-
individual media from being reused outside of choice items can be controlled by changing the
the textbook context) have been purposefully alternatives since the more homogeneous the
excluded. alternatives, the finer the distinction the stu-
dents must make in order to identify the correct
The Content answer. Normally, it takes much longer to
respond to an essay test question than it does to
The main content contribution comes from each respond to a multiple-choice test item. Conse-
of the 600 teachers. Mostly, they use their quently, students are able to answer many
available material, which encompasses written multiple-choice items in the time it would take
scripts (PDF, DOC), transparencies (PPT), to answer a single essay question. Teachers
images (GIF, JPEG, etc.), videos (AVI, MOV, can use this feature to assess a broader sample
etc.), and any combination of these. For the of the course content in a shorter time. An
support of good content development, we pro- essential point is grading because multiple-
vide special training courses and written tutori- choice accelerates the reporting of test results
als, as well as a hotline and a FAQ (frequently to the student; thus, any follow-up clarification
asked question) section that is based on previ- of instruction may be done before the course
ous experience. We also make sure that the has proceeded much further (Burton et al.,
teachers include their preknowledge questions, n.d.).
self-evaluation questions, and the proper
metadata. Metadata
Of course, multimedia content must be de-
signed effectively in order to maximize the true Experience from other projects has generally
capabilities that multimedia has for enhancing shown that these are mostly technology driven
human learning (Holziner, 2002). Within the without enough commitment to content, content
VMC-Graz, the cooperation of the domain spe- management, and above all metadata strategies
cialists together with media experts ensures (Holzinger, Kleinberger, & Müller, 2001). On
appropriate content development. This func- the one hand, it is necessary to provide all users
tional separation secures qualitatively high-qual- (in our case, students and teachers) with the
94
possibility to find relevant material quickly; on facilitate their exchange. Only an LMS may
the other hand, we aim for interoperability of launch an SCO. An SCO itself is not allowed to
the learning material within an international launch other SCOs (Dodds, n.d.).
context. Consequently, such a project can only
be successful when it is fully committed to the
implementation of metadata activities. It is not CONCLUSION AND
just a project but a strategy, which raises aware- LESSONS LEARNED
ness of the possibilities of these metadata.
Correspondingly, our LOs are developed Generally, the auspicious theoretical concept of
according to accepted standards for interna- learning objects was not easy to carry out in
tional education as a basis for worldwide net- practice. It needs a lot of awareness rising
working in the form of RLOs. These LOs are amongst the teachers and the provision of
stored in the repository and are arranged in information to realize the advantages of these
lectures, themes, and modules by the VMC new concepts. As an incentive, we always
logic. pointed out the future advantages that the suc-
We consistently used the SCORM, Version cessful completion of learning material would
1.2. SCORM is a reference model that defines bring.
a Web-based-learning content model, which The handling of the LO editor proved to be
consists of a set of interrelated technical speci- successful, although we weakened our strict
fications. In November 1997, the U.S. Depart- concepts (originally, teachers were forced to
ment of Defense (DoD) and the White House fill in every part), providing prefilled sections
Office of Science and Technology Policy (with default settings) and allowing the post-
(OSTP) launched the Advanced Distributed poned production of the preknowledge and
Learning (ADL) initiative (ADLNet, n.d.). The self-evaluation questions (although we person-
metadata model of the LOM standard inte- ally recommend strictness).
grated in the SCORM supports the retrieval of We found that most of the teachers did not
learning objects in varying constellations. like the creation of the preknowledge ques-
SCORM denominates the smallest unit that can tions. Some even refused to provide any ques-
be administered by a learning-management sys- tions. Thus, we also had to weaken our previous
tem (LMS) as a sharable content object (SCO). concept wherein the creation of preknowledge
An SCO represents so-called assets, which use questions was obligatory. We advocate strongly
the SCORM run-time environment to commu- the advantages of the advance-organizer con-
nicate with different systems. cept and provide pay-off possibilities within
This SCO represents the lowest level of special VMC courses. The dislike of the
content granularity that can be tracked by any preknowledge-question section is easy to ex-
system. An SCO should be principally indepen- plain: The teachers, mainly medical doctors,
dent of the learning context and therefore be lack the exorbitant time required to construct
reusable in different learning situations. More- questions that reflect exactly the content and
over, several SCOs can be assembled to form the necessity for assessing the students’ under-
learning or exercise units on a superordinate standing of the material. However, once they
level. To make a potential reuse practicable, had created the questions, they gained a deeper
SCOs should be small units. They can be the understanding of their material and of the knowl-
basis for sharable content repositories that edge they expected from their students. Conse-
95
quently, the students benefit from this effort. summative evaluation of student learning.
Also, if the teachers get feedback about pos- San Francisco: McGraw Hill.
sible troubles of the students, eventually they
Booch, G. (1994). Object-oriented analysis
will also get a return on their investment. How-
and design with applications. Redwood City,
ever, the self-evaluation questions were re-
CA: Benjamin/Cummings.
garded as useful and important by every stu-
dent. The students like to see their own progress Bruning, R. H., Ronning, R. R., & Schraw, G.
and thus are able to reflect about the content. J. (1999). Cognitive psychology and instruc-
There is still scientific research to be carried tion (3rd ed.). Upper Saddle River, NJ: Prentice-
out, including extensive research in the ex- Hall.
change of LOs in an international context, in
Burton, S. J., Sudweeks, R. R., Merrill, P. F., &
measuring and benchmarking the quality of the
Wood, B. (n.d.). How to prepare better mul-
content, and in gaining understanding of the
tiple-choice test items: Guidelines for uni-
optimal granularity of such LOs with the aim to
versity faculty. Retrieved from http://
support maximum exchangeability and usability.
testing.byu.edu/faculty/handbooks.asp
Chalmers, P. A. (2003). The role of cognitive
REFERENCES theory in human-computer interface. Comput-
ers in Human Behavior, 19(5), 593-607.
ADLNet. (n.d.). Advanced distributed learn-
Cisco. (n.d.). E-learning glossary. Retrieved
ing. Retrieved from http://www.adlnet.org
from http://www.cisco.com/warp/public/10/
Ausubel, D. P. (1960). The use of advance wwtraining/elearning/pdf/elearn_glossary.pdf
organizers in the learning and retention of mean-
Corkill, A. J., Bruning, R. H., & Glover, J. A.
ingful verbal material. Journal of Educational
(1988). Advance organizers: Concrete versus
Psychology, 51, 267-272.
abstract. Journal of Educational Research,
Ausubel, D. P. (1968). Educational psychol- 82, 76-81.
ogy: A cognitive view. New York: Holt,
Dahl, O.-J., & Nygaard, K. (1966). SIMULA:
Rinehart & Winston.
An ALGOL-based simulation language. Com-
Bartlett, F. C. (1932). Remembering. London: munications of the ACM, 9(9), 671-678.
Cambridge University Press.
Dodds, P. (n.d.). Advanced distributed learn-
Beck, R. J. (n.d.). University of Milwaukee, ing, sharable content object reference model
Center for International Education: Learn- (SCORM): ADL version 1.3 application pro-
ing objects. Retrieved from http:// file working draft 1.0, March 26, 2003,
www.uwm.edu/Dept/CIE/AOP/LO_what.html version 1.2, release 2001. Retrieved from
http://www.adlnet.org
Bloom, B. S. (1956). Taxonomy of educa-
tional objectives, handbook 1: Cognitive Gathy, P., Denef, J.-F., & Haumont, S. (1991).
domain. New York: Longmans Green. Computer-assisted self-assessment (CASA) in
histology. Computers & Education, 17(2),
Bloom, B. S., Hastings, J. T., & Madaus, G. F.
109-116.
(1971). Handbook on formative and
96
Holzinger, A. (2002). Multimedia basics: Vol. Miller, G. A. (1956). The magical number seven,
2. Learning. Cognitive fundamentals of plus or minus two: Some limits of our capacity
multimedial information systems. New Delhi, for processing information. Psychological
India: Laxmi. Review, 63, 81-97.
Holzinger, A. (2003). Experiences with user Muzio, J. A., Heins, T., & Mundell, R. (2002).
centered development (UCD) for the front end Experiences with reusable e-learning objects:
of the virtual medical campus Graz. In J. A. From theory to practice. The Internet and
Jacko & C. Stephanidis (Eds.), Human-com- Higher Education, 5(1), 21-34.
puter interaction, theory and practice (pp.
Piaget, J. (1961). On the development of
123-127). Mahwah, NJ: Lawrence Erlbaum.
memory and identity. Worchester, MA: Clark
Holzinger, A. (2004). Application of rapid University Press.
prototyping to the user interface development
Polsani, P. R. (2003). Use and abuse of reus-
for a virtual medical campus. IEEE Software,
able learning objects. Journal of Digital In-
21(1), 92-99.
formation, 3(4).
Holzinger, A., & Ebner, M. (2003). Interaction
Reigeluth, C. M. (1999). Instructional design
and usability of simulations & animations: A
theories and models: A new paradigm of
case study of the Flash technology. Interact
instructional theory. Hillsdale, NJ: Erlbaum.
2003, 777-780.
Robson, R. (n.d.). IEEE learning technology
Holzinger, A., Kleinberger, T., & Müller, P.
standards committee (LTSC). Retrieved from
(2001). Multimedia learning systems based on
http://ltsc.ieee.org
IEEE learning object metadata (LOM). ED-
Media World Conference on Educational Satzinger, J. W. (1998). The effects of concep-
Multimedia, Hypermedia and Telecommuni- tual consistency on the end user’s mental mod-
cations (pp. 772-777). els of multiple applications. Journal of End
User Computing, 10(3), 3-14.
Kralm, C., & Blanchaer, M. (1986). Using an
advance organizer to improve knowledge appli- Shapiro, A. M. (1999). The relationship be-
cation by medical students in computer-based tween prior knowledge and interactive over-
clinical simulations. Journal of Computer views during hypermedia-aided learning. Jour-
Based Instruction, 13, 71-74. nal of Educational Computing Research,
20(2), 143-167.
McDonald, M. E. (2002). Systematic assess-
ment of learning outcomes: Developing mul- Simon, H. A. (1974). How big is a chunk?
tiple-choice exams. Sudbury, MA: Jones & Science, 183, 482-488.
Bartlett.
South, J. B., & Monson, D. W. (n.d.). A
Merrill, M. D. (1999). Instructional transaction university-wide system for creating, capturing,
theory (ITT): Instructional design based on and delivering learning objects. Retrieved from
knowledge objects. In C. M. Reigeluth (Ed.), http://reusability.org/read/chapters/south.doc
Instructional design theories and models: A
Wiley, D. A. (n.d.). Connecting learning ob-
new paradigm of instructional theory (pp.
jects to instructional design theory: A definition,
397-424). Hillsdale, NJ: Erlbaum.
97
a metaphor, and a taxonomy. The instruc- Granularity: The breadth and depth of an
tional use of learning objects. Retrieved object’s content as it relates to reusability.
from http://reusability.org/read/chapters/
Learning Object: Any digital resource
wiley.doc
that can be reused to mediate learning.
Wiley, D. A. (Ed.). (2001). Agency for In-
Metadata: Descriptive information and
structional Technology and the Association
designed to help users and managers locate,
for Educational Communications and Tech-
organize, access, and use objects effectively.
nology. New York.
Module: A grouping of readings, activities,
tasks, and assignments that are organized around
KEY TERMS
a central topic or theme. Breaking content into
components supports the organization of knowl-
Aggregated Learning Object (ALO): A
edge and reduces the cognitive load of the
combination of learning objects that can be
learner. For example, a beginning algebra course
broken down into separate parts without losing
includes the following modules:
the integrity of each part.
Multimedia: A combination of text, graph-
Assessment: Any process used to system-
ics, audio, animation, video, and/or simulation.
atically evaluate the knowledge level of learn-
Typically, combinations of media can provide
ers.
deeper explanations or illustrations of content
Content Aggregation: Any process of than data presented in one medium.
building a new learning object from one or more
Reusable: To be placed in different situa-
existing objects or assets. For example, if a
tions, environments, or locations for different
pretest reveals that students are missing some
purposes or functions by different end users.
key skills or core knowledge, an instructor
might locate a series of objects and link them Shared Content Object Reference
using a common interface as a strategy for Model (SCORM): Defines a Web-based-
remediation. learning content-aggregation model and run-
time environment for learning objects. The
Curriculum: A set of courses, modules, or
SCORM is a collection of specifications adapted
other organized learning experiences that con-
from multiple sources to provide a comprehen-
stitute a complete, cohesive, and coherent pro-
sive suite of e-learning capabilities that enable
gram of study.
the interoperability, accessibility, and reusabil-
Digital Asset: Any audio, animation, ity of Web-based learning content.
graphic, photograph, text, or video that may
convey information, does not have a learning
objective attached, and is not multimedia.
98
99
Chapter XIII
Motivating Healthcare Students in
Using ICTs
Boštjan vanut
College of Healthcare Izola, University of Primorska, Slovenia
ABSTRACT
Motivating and teaching healthcare students to use information and communication technologies
represent a challenge. For the successful integration of healthcare and technology, there must
be an investment in the organization, but particularly in its people. Motivation and a lot of
practical work are mandatory for teaching informatics in healthcare. A practical knowledge
of informatics is an investment for healthcare students that can improve their quality of study,
work efficiency, and everyday life. In this article, four examples of connecting healthcare jobs
with informatics are presented. Connecting healthcare students’ work and everyday lives is
an efficient way of motivating them to use information and communication technologies.
INTRODUCTION majority of my students, particularly the older

ones, are afraid of new technologies and try to
The usage of modern information and commu- avoid them. The causes can be the following:
nication technologies in healthcare became
mandatory. Services and technologies like • Inappropriate education in information tech-
telemedicine, e-learning, medical and pharma- nologies
ceutical databases, robotized tele-operating • No opportunity to do practical work with
systems, computer-assisted radiology, and so computer applications and services like
forth can provide healthcare professionals with the World Wide Web and e-mail
access to the latest knowledge and help them in • The opinion that healthcare specialists are
their everyday work. not computer specialists
Working as a lecturer of informatics in • The belief that they are too old to learn and
healthcare at the College of Healthcare Izola, use information technologies
University of Primorska, I realized that the
Motivating Healthcare Students in Using ICTs
• Not knowing that the usage of information PRACTICAL EXAMPLES OF

technologies in healthcare can reduce MOTIVATING HEALTHCARE
costs, increase productivity, and help STUDENTS TO LEARN ABOUT ICT
healthcare professionals stay focused on
their patients Example 1: Von Neuman Model of
the Computer with a Screwdriver
The research in computer applications ap- and Medical Input-Output Devices
plied to healthcare is intensified. The proper
information system at the right time can be a When we have a class full of healthcare stu-
competitive advantage. On the other hand, dents and we try to explain to them the Von
inadequate information systems can be disas- Neuman model of the computer, we soon notice
trous to an institution. The development and that 90% of them are not interested in the topic
implementation of information technology into at all. First, we must motivate them by bringing
an organization is no easy matter and, particu- in a personal computer and a screwdriver.
larly in healthcare, where users of these sys- Physical contact with the computer and its
tems may be patients, the successful introduc- parts reduces the fear of the computer that
tion of systems may be difficult. Often this is some students have.
not because of technological problems; typi- Then we try to link the model of the com-
cally, the technologies employed are relatively puter to the everyday life of a healthcare pro-
mature as they have been in service for some fessional. After we explain what input and
time and have been tested in commerce or output devices are, we present them with the
industry (Fitch, 2004). For the successful inte- monitor, mouse, printer, keyboard, and other
gration of healthcare and technology, we have devices specific to healthcare and medicine,
to invest in people and the organization. As a such as the following:
lecturer of informatics in healthcare, I have the
opportunity to teach students to use information • Electrocardiograms (ECG or EKG)
and communication technologies. Connecting • Radiography devices
their work and everyday lives is an efficient • Ultrasound devices
way of motivating healthcare students to use • Laboratory devices
them.
In this chapter, some practical examples of
the motivation of students in learning informa-
tion and communication technologies, and using Figure 1. Input-output devices that are
them in everyday life, are described. Motiva- specific to healthcare and medicine
tion is an internal process that creates and
maintains the desire to move toward goals
(http://www.psychadvantage.com/glossary.html).
Our aim is to transform students into advanced
users, capable to use the computer and operat-
ing system, determine a simple bug, use the
Internet and so forth, and apply their knowledge
of information and communication technologies
in healthcare.
100
Students often have the wrong ideas about Figure 2. Searching the data by hand or
the computer. They see it as a box with a with the help of a computer
monitor, mouse, and keyboard. We must to
explain them that the computer can also be a
blood-pressure monitor, a pulse-meter watch, a
complex laboratory device, and so forth. In this
way, a student extends the term computer with
more familiar terms: We make the computer
more familiar to them.
Example 2: Software is a Tool that

Prevents the Worst Scenarios these scenarios as real as possible and motivate
students to use the computer efficiently.
Teaching healthcare students to use Excel,
Calc, SPSS, and similar tools can be very Example 3: The Internet and Its
difficult. For instance, when they open Excel, Services
they find several sheets, toolbars, and menus,
and they usually do not know how to use these When you explain to healthcare students the
tools and cannot imagine how this application definition of the Internet, you must not start out
can possibly facilitate their work. My tactic is to with using terms like protocols, packets, rout-
prepare a scenario and make them realize the ers, servers, gateways, DNS servers, DDNS,
power of these tools. I usually prepare a table DHCP servers, modems, optical fibers, satel-
with the data of 5,000 people, including their lite communication, TERENA, certificates,
names, surnames, personal identification num- IPSec, sources of information, and so forth. We
bers, sex, blood groups, birth dates, home ad- have to keep in mind that our aim is to make
dresses, telephone numbers, and so forth. I healthcare students advanced users and not
print the table and also save it somewhere on information and communication technology ex-
the server. perts. Too much technical details can de-moti-
After that, the students must pretend there vate them.
is a car accident and that in 30 minutes they My aim is to explain what the Internet is,
need 100 blood donators with blood type AB+ what services it offers to users, and how these
who are older than 18 years. Those students services can help healthcare professionals in
who are not familiar with the computer will their jobs and private lives.
probably try to find the donators manually and The World Wide Web, e-mail, file transfer
run out of time. Then I explain to them that they protocol, Telnet, and so forth—all these terms
are responsible for their patients and must find must be connected with students’ everyday
the right way to accomplish the task in time. In lives. When explaining the World Wide Web, we
this way, we can explain to them that using tools must show to the students the possibilities that
like filters can help them find a solution and this service offers. We can start with searching
solve the problem in someone’s life. pages like http://www.altavista.com, http://
With examples like this, students realize that www.google.com, http://www.yahoo.com, and
computer applications can solve problems faster then search healthcare, nursing, pharmaceuti-
and in an efficient way. The aim is to make cal, and medical home pages.
101
Real-life scenarios must be included in the • Find the address and telephone number of
study. For instance, a patient from the USA a friend in New Zealand
spends the holidays in Europe and accidentally • Find what the price is of a train ticket from
needs a doctor. After the visit, the patient asks Budapest to Moscow
a nurse to prepare the bill in U.S. dollars • Find the three main pharmaceutical com-
because of the formality of his or her insurance panies in China
company. The nurse can immediately find the • Find some important Web pages and
recent exchange rate and edit the bill. Another Internet portals about healthcare
example is when a patient from South Africa • Find databases of libraries
uses a special medicine and you are not able to • Find the most important healthcare con-
find a substitution for it in your documentation. ferences in Central and Eastern Europe
If we want to give our students the idea of
how big the Internet is, we can explain it by Students can be up to date with recent re-
telling them that Internet services are used by search in healthcare and medicine with the help
millions of people all over the world. We can of newsgroups in nursing (e.g., sci.med.nursing),
explain that there are hundred of thousands of pharmacy (e.g., sci.med.pharmacy), nutrition
servers, routers, switches, and so forth. But the (e.g., sci.med.nutirtion), and informatics in medi-
most effective way is to give them simple tasks. cine (e.g., sci.med.informatics). It is very im-
For example, with the help of the World Wide portant to show the students Web portals like
Web, we can have them do the following: Mediline, Web of Science, and so forth, where
students can find recent publications. We can
• Find the telephone numbers of three col- also present students with the possibilities of e-
leges of healthcare in Venezuela learning and include it in their study.
• Find the picture of the lighthouse Kereon
in France
Figure 3. Data encryption ( vanut, 2003)
Diabetic patients Wfjeiosdklwmegpovsdèlvj Diabetic patients:

1.Doe John fydslkawempolèvmiprlsvk 1.Doe John
2. John Doe njioplwekwejnvfo<ialkvnh 2.John Doe
… sxdcoljkwhnjvsdiolxckjwe …
…
Unencrypted document Encrypted document Unencrypted document
ENCRYPTION DECRYPTION
INTERNET
Sender Receiver
intruder
encryption key - E decription key -D

Jwoeisldfuj4opwaeifjeroil 5823'jefw023wefjo23r78f
dvjnlkrgvcolu90234587 u0osdiu738902r78or5z
102
Example 4: Data Encryption CONCLUSION
When you start to explain to students the prob- An information system is made of hardware,
lem of encrypting data, they often ask, “Why do software, data, the organization, and people.
I have to encrypt and decrypt the data?” Fortu- People are a very important part of every
nately, there exists a good example for this: a information system. As a lecturer of informatics
list of diabetic patients that must be sent to the in healthcare, I have the goal to motivate my
clinic immediately. Pharmaceutical companies students to use information and communication
would pay millions for such lists and they will do technologies by connecting healthcare with
anything to find them. Sending these lists by e- informatics and computer science. The ex-
mail unencrypted can be very dangerous be- amples in this chapter are efficient ways to
cause the companies can get them. motivate students and are also good chances
In this way, the principles of encryption and for lecturers to learn about healthcare and
decryption, and public and private keys can be medicine.
explained to the students. We need three per-
sonal computers—a sender, a receiver, and an
intruder—a hub, a program like Network Ana- REFERENCES
lyzer, which runs on an intruder computer, and
a list of diabetic patients. First, we send the data Fitch, C. J. (2004). Information systems in
from the sender to the receiver unencrypted. healthcare: Mind the gap. Proceedings of the
With the help of a program like Network Ana- 37 th Hawaii International Conference on
lyzer, we can receive all the packets sent from System Sciences, Waikoloa, HI.
the sender computer and show to the students
Sriæa, V., Teven, S., & Pavliæ, M. (1995).
the unencrypted data captured from the in-
Informacijski sistemi: Gospodarski vestnik.
truder computer.
Ljubljana, Slovenia: Gospodarski vjestnik.
Then we repeat the scenario with encrypted
data and explain to the students the following: vanut, B. (2003). Informatics in healthcare.
Koper, Slovenia: Author.
• The sender can encrypt the data without
knowing the receiver’s private key.
• The intruder will spend many years de-
KEY TERMS
crypting the data, and the data will be
Data Encryption: Scrambling data by the
certainly be unusable because the patients
use of a key and a transformation technique,
may all be deceased by that time.
which protects the confidentiality of the infor-
• The receiver can easily decrypt the data
mation being transferred.
with the private key.
Information and Communication Tech-
This way, students can realize that sending nologies: The study of developing and using
unencrypted information over the Internet can technology to process information and aid com-
be a danger for data confidentiality, and that if munication.
we want to protect the data, we must take into
account the fact that intruders will do anything Motivation: Internal process that creates
to access them. and maintains the desire to move toward goals.
103
Organization: A formal group of people other documents as well as graphics, audio, and
with one or more shared goals. video files.
World Wide Web: A system of Internet Von Neuman Model: The most common
servers that support specially formatted docu- conceptual model of a computer, in which a
ments in a language called HTML (HyperText computer is considered to consist of linear
Markup Language), which supports links to memory and a central processor.
104
105
Chapter XIV
The User Agent Architecture and
E-Learning in Healthcare and
Social Care
Konstantinos M. Siassiakos
Stefanos E. Papastefanatos
Athina A. Lazakidou
ABSTRACT
E-learning is developing rapidly worldwide. The volume of the information that e-learning
systems render grows, too. Nowadays, the critical issue is to acquire more knowledge using
less time and effort. Contemporary e-learning systems, although they embody up-to-date
technology, artificial-intelligence techniques, and pedagogical methods, address too flat a
spectrum of users. The user agents deal with this weakness. E-learning has the potential to
transform learning for healthcare and social care, supporting the aims of the National Health
System Plan and raising standards of care for patients and service users across healthcare
and social care. The solution proposed is based upon the user-agent architecture and
confronts the individual issues of the health and social sector.
INTRODUCTION precious global resource that is the embodiment

of human intellectual capital and technology.
Society is entering into an era where the future As people begin to expand their understanding
essentially will be determined by people’s abil- of knowledge as an essential asset, they are
ity to use knowledge wisely. Knowledge is a realizing that in many ways the future is limited
The User Agent Architecture and E-Learning in Healthcare and Social Care
only by imagination and the ability to leverage HEALTH AND SOCIAL SECTOR
the human mind. As knowledge increasingly PARTIES
becomes a key strategic resource, the need to
develop comprehensive understanding of knowl- As the health and social sector holds great
edge processes for the creation, transfer, and significance in society, the spectrum of the
deployment of this unique asset is becoming parties involved is wide and manifold. A system
critical. willing to provide knowledge to this spectrum
The issue about the difference between must be flexible, accurate, effective, and direct,
knowledge and information is today the subject and at the same time it must be able to accept
of much literature, discussion, planning, and and elaborate information from this spectrum.
some action. Moreover, the involvement of Only by recording the involved parties and their
knowledge management in e-learning systems specifications can an e-learning system have
is a crucial matter: Contemporary e-learning success.
systems increasingly take advantage of knowl- The spectrum discussed above can be bro-
edge-management techniques to utilize the great ken down into the following parties (user groups):
volume of information that they render.
On the other hand, nowadays the academic • Individuals
community is addressing more and more the • Employers
rise of the online community that will be instru- • Managers
mental in the realization of advanced learning • Healthcare professionals
societies. Internet online environments enable • Providers
new and interesting designs for the support of • Organizations
traditional learning and for the development of • Patients and service users
new forms of learning. Ideally, users will be • All staff
able to access all forms of knowledge in any
combination, from any location at any time. These groups have different properties that
This, of course, implies considerable complex- induce the specifications of an integrated e-
ity in the software design and a substantial level learning environment. By developing a unified
of intelligence across the systems: from the set of internal mechanisms and an interface, an
servers to the networks, to the user interfaces. e-learning environment cannot have equal suc-
Although e-learning environments can be used cess throughout the whole spectrum of users.
widely either for educational or for training The answer to the problem is the appropri-
purposes, the problem that still exists is the ate formation of the learning-content structure,
efficient management of content and effective- but also the creation of respective procedures
ness for users. regarding the interaction between the content
E-learning has the potential to transform and the user.
learning for health and social care, supporting Before proceeding to the presentation of the
the aims of the NHS Plan and raising standards solution proposed, it is useful to introduce the
of care for patients and service users across Gagne model of learning and recall (Figure 1).
health and social care. The solution proposed is The model is based on the theory of information
based upon the user-agent architecture and elaboration. It represents the procedure of
confronts the individual issues of the health and knowledge generation within the human brain.
social sector. The interaction between the environment and
106
The User Agent Architecture in E-Learning in Healthcare and Social Care
Figure 1. Gagne model of learning and recall
the human is performed through receptors and The user agents will monitor the input and
effectors (input and output); the data are saved output data of the environment. So, they
in the temporary (short-term) memory, while will be able to record the users’ steps,
repeating and performing save them in the long- interests, and needs, and also the feed-
term memory. Then, the response generator back of the environment. After a period of
performs actions, using data from memory and training, the user agents will be able to
controlling them by executive control and ex- propose, filter, search, and present the
pectancies. The mechanisms of knowledge appropriate information to the user.
acquisition (or the learning procedure) can be • Rapid, reliable, and accurate content re-
managed by using the appropriate methods and store and presentation.
tools. • Compatibility with pedagogical and e-learn-
The architectural model proposed (Figure 2) ing models.
is a coordinated cooperation of user agents,
content management, learning-objects metadata, Before the information is saved in the con-
and e-learning models. The goal of this model tent database, the metadata assignment proce-
can be divided into the following axes. dure takes place. This procedure does not
identify the data directly, but assigns several
• Categorization of the information accord- property metadata tags. These tags are used by
ing to the user’s needs. The information the user agents to categorize the data. Then,
will have a different hierarchy and will be the data are saved in the content database along
sorted according to the user’s personal with their metadata tags.
needs. This will be achieved with metadata The user agents are the users’ representa-
tags, which will categorize the content tives: They forward their selections, record
with the support of the user agents. their steps, filter information, and collect the
• Users’ progress monitoring and support. environment’s feedback. This is achieved
The user agents will be the entrance point through the elaboration of the metadata tags.
of the users in the e-learning environment. The users’ agents are trained, for an initial
107
Figure 2. Architecture of the proposed model
LEARNING CONTENT
Information flow
Information flow
Information flow
Information Information Information

Elaboration Elaboration Elaboration
User Agent User Agent User Agent
Information flow Information flow Information flow
Individual Patients Managers
period of time, to identify the correct data by mechanisms. Since the user is authorized to use
searching through their metadata tags. the e-learning environment, a profile is created
After the training period, the users’ agents that matches his or her status. There will be a
have a complete view of their users’ needs, pool of profiles (templates) that will be deter-
selections, interests, and so forth. Then, they mined and created by experts. This pool will
are able to work online even if the user is off cover the whole spectrum of health and social
line. They can perform searches, filter the sector parties. Each profile will describe the
results, and present them to the user when he or party (user-group) properties (interests, needs,
she connects back to the system. job descriptions and specifications, skills, etc.).
According to the above, the users’ agents The profile templates will be the basic entities
are a part of Gagne’s learning and recall model. of the e-learning environment. When a user
They simulate the learning and facilitate the creates a profile (by selecting a template and
recall process. The technology used for the filling in the information required), a new in-
implementation of the users’ agents is based on stance of the template begins to live in the
artificial neural networks and associative memo- environment. The representative of the profile
ries. is the user agent. The user agent is always
online, whether the user is connected to the
environment or not. When the user is online, the
USER AGENTS’ CUSTOMIZATION agent monitors the user’s steps and paths,
AND ADVANTAGES filters the information flow, suggests useful
information, brings similar users together, and
The user-agent architecture has great benefits so forth. When the user is off line, it searches
regarding the customization of the learning for useful information, updates the user profile
108
according to the progress made, streamlines ware has become integral for both on-campus
the information flow in order to form the next e- and distance learning classes due to its ability to
learning course, and so forth. organize syllabi, lecture materials, handouts,
There are several advantages to the sector and assignments.
parties. First of all, the information stored in the Physicians and medical students on a clini-
environment (especially by the passing of time) cal rotation may soon be able to attend on-
is categorized and starts to form knowledge. As campus seminars or lectures from a laptop
the user proceeds to acquiring knowledge, he or across town, across the state, or across the
she helps the other users with similar profiles to country. Telehealth programs designed to im-
improve their learning procedure because the prove medical education for students in rural
user agents communicate with each other, ex- healthcare areas are developing rotations that
changing experience. Since there is feedback link several community hospitals with a spon-
to the user-agent mechanisms, by the passing soring medical school.
of time the results improve and the searches for Distance learning is still in its infancy with
information become more and more accurate little standardization, but most academic medi-
and successful. cal centers have or are planning some type of
Another great advantage is the automation distance learning programs. The availability of
of the formation of learning courses. By stream- technical support for both instructors and stu-
lining and elaborating the information, the user dents is a critical issue for all technology-
agents are able to form a learning course that mediated learning.
will cover a thematic unity. Moreover, the Distance learning and telehealth are still
course can be reviewed and adjusted by the new concepts. However, communications tech-
experts (e.g., system administrators, authors, nologies available today (though dependent on
etc.), and even be exported as a single course individual systems’ capacities) are able to con-
to a CD-ROM. Another benefit of the user nect healthcare professionals (with patients
agents is the transformation of information to and other healthcare professionals) and infor-
knowledge. Knowledge is not a set of informa- mation systems (for data and information ex-
tion. Although the definition of knowledge is change) located at distant sites. Emerging medi-
beyond the framework of this chapter, it is cal applications that can occur between geo-
widely acknowledged that the information needs graphically dispersed locations include initial
to be elaborated to conduct knowledge. Such and continuing medical education, clinical ser-
an elaboration is being performed by the user vices delivery and consultation, patient educa-
agents, producing knowledge. tion, and healthcare management and adminis-
tration.
With increasing demands on physicians with
DISCUSSION regards to their knowledge base and productiv-
ity, distance learning and telehealth may pro-
As computer and communications technologies vide an opportunity for busy clinicians to meet
are incorporated as teaching tools, traditional clinical responsibilities to a more dispersed and
materials and technologies are being adapted. diverse population while still participating in
Histology and pathology slides that previously educational activities. These teleprograms may
required a microscope are now digitized for take more time and be more expensive in the
study on a computer. Course-management soft- short term; however, as the technology be-
109
comes more prevalent and the number of users Meyen, E. L. (2000). Using technology to
increases, these programs may offer an effi- move research to practice: The online acad-
cient alternative to meet the escalating de- emy. Their world 2000. New York: National
mands of a rapidly changing healthcare envi- Center for Learning Disabilities.
ronment.
Pulido, P., & Requena, J. (2003). PAFAMS:
Panamerican Federation of Associations of
Medical Schools. Alternatives in “e-learn-
CONCLUSION
ing” for health professionals in Latin America
and the Caribbean.
E-learning is developing rapidly worldwide. The
volume of the information that e-learning sys- Rosenberg, H. (2003). The effectiveness of
tems render grows, too. Nowadays, the critical computer-aided, self-instructional programs in
issue is to acquire more knowledge in less time dental education: A systematic review of the
and effort. The chaotic volumes of information literature. J Dent Educ., 67(5), 524-532.
provided must be elaborated and transformed
Rosenberg, M. J. (2001). E-learning: Strate-
into knowledge. Contemporary e-learning sys-
gies for delivering knowledge in the digital
tems, although they embody up-to-date tech-
age. McGraw-Hill.
nology, artificial-intelligence techniques, and
pedagogical methods, address too flat a spec- Sakai, S., Mashita, N., Yoshimitsu, Y., Shingeno,
trum of users. The user agents deal with this H., Okada, K., & Matsushita, Y. (n.d.). An
weakness. Also, the wide and manifold spec- efficient method of supporting interactions
trum of the health and social sector parties will for an integrated distance learning system.
definitely benefit from the architecture pro-
Schreiber, D. A., & Berge, Z. L. (1998). Dis-
posed as most e-learning systems lack this
tance training: How innovative organiza-
feature.
tions are using technology to maximize learn-
ing and meet business objectives. San Fran-
cisco: Jossey-Bass.
REFERENCES
Zahm, S. (2000). No question about it e-learn-
Berridge, E., et al. (2000). Computer-aided ing is here to stay: A quick history of the e-
learning for the education of patients and family learning evolution. E-Learning, 1(1), 44-47.
practice professionals in the personal care of
diabetes. Comput Methods Programs Biomed.,
URL REFERENCES
62(3), 191-204.
Devitt, P., Smith, J. R., & Palmer, E. (2001). http://agelesslearner.com/intros/elearning.html
Improved student learning in ophthalmology
http://derekstockley.com.au/elearning-
with computer-aided instruction. Eye, 15(5),
definition.html
635-639.
http://www.brandonhall.com/public/glossary/
Hubbs, P. R., Rindfleisch, T. C., & Godin, P.
glossary.html
(1998). Medical information on the Internet.
JAMA, 280, 1363.
110
http://www.emeraldinsight.com/Insight/html/ Learning-Content Management System

Output/Published/EmeraldFullTextArticle/Ar- (LCMS): A learning-content management sys-
ticles/0460100403003.png tem is an environment where developers can
create, store, reuse, manage, and deliver learn-
http://www.iste.org/research/reports/tlcu/
ing content from a central object repository,
internet.html
usually a database. LCMS generally works with
http://www.ochsnerjournal.org/ochsonline/ content that is based on a learning-object model.
?request=get-document&issn=1524- These systems usually have good search capa-
5012&volume=003&issue=01&page=0022 bilities, allowing developers to find quickly the
text or media needed to build training content.
KEY TERMS Synchronous E-Learning: Synchronous
or live e-learning means that communication
Asynchronous E-Learning: Asynchro- occurs at the same time between individuals,
nous e-learning happens when communication and information is accessed instantly. Examples
between people does not occur simultaneously. of synchronous e-learning include real-time
Some examples of asynchronous e-learning chat, and video or audio conferencing.
include taking a self-paced course, exchanging
User Agent: A program or device that can
e-mail messages with a mentor, and posting
be used to access the Web. This includes brows-
messages to a discussion group.
ers, Web robots, intelligent agents, advanced
Distance Learning: Instruction provided hardware, specialized software, and so on.
by a human separated by place from the learner.
E-Learning: The delivery of a learning,
training, or education program by electronic
means. E-learning involves the use of a com-
puter or electronic device (e.g., a mobile phone)
in some way to provide training, educational, or
learning material.
111
112
Chapter XV
E-Learning in Healthcare
and Social Care
Maria Kalogeropoulou
Maria Bastaki
Polyxeni Magoulia
ABSTRACT
E-learning has the potential to transform learning for healthcare and social care, supporting
the aims of the NHS Plan and raising standards of care for patients and service users across
health and social care. This chapter sets out a vision of healthcare and social care services
in the 21st century, and a strategy for making it a reality. The authors present and discuss here
the basic principles and benefits of e-learning for healthcare professionals, medical students,
and patient education.
INTRODUCTION usually include some form of support, whether

face to face or electronic, and can often be
E-learning is the use of interactive technologies blended with classroom methods. It can offer
to support and improve learning. It is not just learners and tutors many services, including
about online courses. E-learning can include a access to resources, information, and advice. It
range of technologies from CD-ROMs (com- can reduce the time spent on administration,
pact disc read-only memory) to electronic and help with the planning, recording, and track-
whiteboards or online simulations. It should ing of learning and development. An e-learning
E-Learning in Healthcare and Social Care
strategy is therefore really an aspect of a • The rapidly changing workplace

strategy for effective learning. • A more competitive job market
E-learning has the potential to transform • Increased emphasis on teamwork
learning for health and social care, supporting • Informal learning
the aims of the NHS Plan and raising the • Technology
standards of care for patients and service users • The need for professionalism
across health and social care. This document
sets out a vision of health- and social-care E-learning is increasingly widely used by
services in the 21st century and a strategy for learners in schools, colleges, and universities. It
making it a reality. The vision is of a health and is also widely used in work-based learning and
social sector in which: corporate education, and in industry and the
public sector. E-learning is therefore a signifi-
• Patients and service users have the infor- cant factor in the personal and professional
mation they need to be involved in their development of the 1.2-million-plus people who
own care, and know that staff have the work in the health sector, and the 1.4 million
skills and expertise to give them the high- who work in social care. Clearly, e-learning is
est standards of care, important when it comes to acquiring job-re-
• All health- and social-care staff can ac- lated knowledge and skills.
cess the learning opportunities and sup-
port they need to develop personally and
professionally, CAPABILITIES AND BENEFITS
• Flexible learning is a central part of every- OF E-LEARNING
day work for everyone,
• The highest standards of professionalism E-learning strategy sets out generic e-learning
are found throughout all occupations and capabilities for the education sector. It is useful
communities, to look at how these could apply to health and
• People share knowledge, resources, ex- social care.
pertise, and good practice within and
across their communities, and • Individualized Learning: Meeting the
• Resources are used effectively to provide needs of all staff, including those working
lifelong learning and continuous develop- in remote locations, in the home, or in
ment opportunities for all staff now and in small organizations, or whose work re-
the long term. quires them to be mobile
• Personalized Learning Support: Ex-
In the past, involvement in learning has been ploring learning pathways and resources,
largely a matter of personal preference and finding the right courses and materials,
opportunity, governed by the individual’s own and tracking work-based learning
motivation, their seniority, the availability of • Collaborative Learning: Including col-
suitable learning, and the support of colleagues laboration between learners on work-
and supervisors. Increasingly, factors such as based projects or action research (on, for
the ones listed here mean that learning is be- example, national service frameworks),
coming a central part of everyone’s working and supporting health informatics commu-
life:
113
nities or health- and social-care • For organizations: Becoming partners

interprofessional groups in workforce-development functions and
• Tools for Educators and Employees: promoting knowledge management
Support for innovation by customizing or • For Patients and Service Users: Get-
creating learning resources or simulations ting individuals and communities involved
• Virtual Learning Worlds: Online in improving care outcomes
master’s classes and simulations, and ac- • For all Staff: Interlinking the technolo-
cess to virtual campuses or wider learning gies used for learning and for work
environments
• Flexible Study: On-demand learning,
which people can access when and where E-LEARNING FOR HEALTHCARE
they need it PROFESSIONALS
• Online Communities of Practice:
Bringing together specialist communities, In recent years, as the demand for lectures
practitioners, learners, community or vol- without the limitations of time and place by
untary workers, and service users and those who have jobs and require lifelong educa-
care givers tion grows, there are more and more expecta-
• Quality at Scale: Providing access to e- tions on the implementation of distance learning
learning resources and services right systems (Sakai, Mashita, Yoshimitsu, Shingeno,
across the sector, without variations in Okada, & Matsushita, n.d.). There is a double
standards, that are linked to information, source of demand for services: institutional and
and HR (human-resource) and manage- personal. Initially, the demand will basically
ment systems arise from the public education system of health
professionals and technicians. In this initial
E-learning can offer huge practical ben- phase, recipients of information and courses
efits. The value of e-learning for all parties is will be of two kinds: technical staff of health
clear: divisions or institutions in every state, and man-
aging, medical, and nursing staff of hospitals
• For Individuals: Freedom to develop, and ambulatories in individual states (Pulido &
both personally and professionally, through Requena, 2003).
accessible learning opportunities Until now, professionals were required to
• For Employers: Engaging staff and pro- take time away from their practices and per-
moting a sense of ownership and involve- sonal lives to attend in-person training sessions.
ment The introduction of an Internet-based training
• For Managers: Achieving business and curriculum provides physicians with the flex-
performance targets ibility and convenience of learning at their own
• For Health Professionals: Better col- pace and on their own time.
laboration and communication, and creat- In-depth, efficient training for physicians
ing more development opportunities enables them to deliver an even higher standard
• For Providers: Widening participation in of care to patients. E-Learning provides physi-
learning, at work, in the community, and at cians with the means to address their training
home needs immediately (Raz, 2002). The e-learning
114
program will help participants assimilate the ciously, can lead to information overload. Fur-
basic principles and practices of therapies, thermore, whilst they are valuable for develop-
including patient selection, where the treatment ing learning in the cognitive domain, they are
should be positioned in the care continuum for less useful for improving interpersonal skills
diseases, and the significant symptoms of these and changing attitudes (Marshall & Kirwan,
diseases. The e-learning curriculum consists of 2004).
modules, and some of them are free for physi- The best balance seems to be reached by
cians. The modules are interactive and include providing small-group sessions with a facilita-
animation and streaming video, as well as exer- tor to consolidate the subject, and to place it
cises and review questions. In addition, partici- within the current clinical context. Further-
pants receive a CD-ROM containing educa- more, the involvement of the students in the
tional surgical or other kinds of videos, and production of e-learning packages helps to im-
audio programs on patient-physician communi- part a sense of ownership to the students and
cation. Finally, physicians have the option of overcome any resistance to the use of alterna-
communicating with e-learning faculty mem- tive teaching methods. General Medical Coun-
bers. cil (GMC) guidance on reducing the amount of
information imparted to students (GMC &
Tomorrow’s Doctors, 2003) must be squared
E-LEARNING FOR with ever increasing amounts of information
MEDICAL STUDENTS freely available on the Internet—human con-
tact is vital for this.
It is important to note that a lot of people believe The University of Hull’s School of Nursing,
that traditional training combined with a strong Social Work, and Applied Health Studies has
implementation process and coaching can cre- developed an interdisciplinary, part-time mod-
ate measurable results for students. E-learning ule called Facilitating E-Learning in Health and
can increase the retention of learned informa- Social Care Education and Practice to provide
tion by an additional 35% to 45% when blended health- and social-care professionals involved
with other training approaches and technolo- in education with the opportunity of considering
gies. how they can integrate e-learning into their
Undergraduate medical students are a group practice and teaching. A major step in imple-
of individuals who often have a list for knowl- menting the university’s e-learning strategy
edge along with the skills and opportunities to within the school has been to implement a
use information technology. Furthermore, as virtual learning environment: an integrated set
the doctors of the future, they will be increas- of electronic teaching tools that are available to
ingly involved with technology in their profes- students and teachers online, allowing informa-
sional practice. tion transfer, learning-material delivery, and
Past experience in developing e-learning communication through e-mail, discussion, and
packages for students learning tutorials has chat to support student learning (Santy, 2003).
shown them to be powerful tools for learning as The module has the potential to attract stu-
there are exciting opportunities for the use of dents not only nationally, but also internationally
images, video, and links to external sites or because the module can be delivered at any
relevance. However, they are extremely labour time to any place with an Internet connection.
intensive to set up and, unless created judi- E-learning can make these students more inde-
115
pendent and lifelong learners. New technology ing™ engine technology to offer compelling,
has been transforming the way we think about intuitive, and effective e-learning to some of the
education in health and social care, and the most demanding users of information: everyday
increasing use of technology in practice set- patients trying to better understand their dis-
tings. eases and the therapeutics that may help them
(Patel, 2003).
Patient-education e-learning courses em-
E-LEARNING FOR power physicians, allied health professionals,
PATIENT EDUCATION and patients alike with the knowledge critical to
the understanding of diseases and the thera-
GeneEd develops new e-learning systems for peutic regimens used to treat them. E-learning
patient education for major pharmaceutical cli- courses are actively responding to the ever-
ents. As the leading provider of advanced e- changing training needs of the life-science in-
learning solutions for the life-science industry, dustry, and the introduction of this new patient-
it announced that it has extended its e-learning education curriculum reiterates the strength of
curriculum to span all subject areas, from drug our science-focused curriculum design.
discovery and molecular medicine to patient
education.
The GeneEd patient-education curriculum DISCUSSION
that was announced (in addition to the over-100
therapeutic courses released over the last two While there are clearly many benefits and
quarters) represents a significant expansion of advantages to e-learning, we also need to take
GeneEd’s product line. GeneEd products are into account the barriers, challenges, and disad-
extensively used in the research, development, vantages associated with it. These include high
and sales divisions of major global pharmaceu- development costs, barriers to access for dis-
tical organizations. In combination with the advantaged learners or those with disabilities,
soon-to-be-released regulatory and compliance and the misconception that online learning is a
curriculum, the patient-education program solitary and unsupported activity. It is particu-
greatly enhances the GeneEd product portfolio larly important to address any barriers relating
to address all of the mission-critical training to potential users so that e-learning really does
needs of modern life-science organizations. As benefit all target groups. To gain full benefit, we
the role of molecular medicine broadens its will need to take steps to guard against a
reach within pharmaceutical organizations, potential digital divide by addressing both ac-
employees, physicians, and patients alike are cess and skills. We also need to achieve the
required to absorb the complex scientific and right balance between e-learning and tradi-
medical issues this new science brings. GeneEd tional methods. While e-learning can make a
is the only organization providing validated e- powerful contribution to large-scale engage-
learning across the spectrum. These techniques ment in learning, as well as the tailoring of
range from genomics and proteomics, through learning to individual needs, it should not and
disease states, all the way to physician and cannot replace all other approaches to learning.
patient education. GeneEd is utilizing its award- An e-learning strategy should be one aspect of
winning, proprietary Repurposing™ and View- a wider learning strategy.
116
CONCLUSION General Medical Council (GMC) & Tomorrow’s

Doctors. (2003). Recommendations on un-
E-learning is developing rapidly worldwide. Any dergraduate medical education.
strategy will therefore need to have an interna-
Marshall, R. W., & Kirwan, J. R. (2004).
tional dimension so that e-learning for the NHS
Development of e-learning tutorials in rheu-
can be genuinely world class. We may need to
matology.
harness international knowledge and best prac-
tice, and use research to benchmark NHS e- Pulido, P., & Requena, J. (2003). Alternatives
learning against the very best globally. Further- in “e-learning” for health professionals in
more, with the increasing pace of globalization, Latin American and the Caribbean.
learning resources and opportunities that origi- Panamerican Federation of Associations of
nate abroad will increasingly be available, and Medical Schools (PAFAMS).
health and social care will require systems that
Sakai, S., Mashita, N., Yoshimitsu, Y., Shingeno,
are compatible and interoperable with those in
H., Okada, K., & Matsushita, Y. (n.d.). An
use in other fields at home and abroad.
efficient method of supporting interactions
E-learning is not an end in itself, nor is it a
for an integrated distance learning system.
marginal activity related only to online courses
or distance learning. It will increasingly em-
brace all aspects of learning and will therefore URL
form a fundamental part of how people will
learn in 10 or 20 years. Although we cannot http://www.cmwdc.nhs.uk/elearning/
predict exactly which technologies or which elearningstrategy.pdf
models of learner support will be most widely
http://www.dfes.gov.uk/consultations2/16/
used, existing examples of leading practice,
whether in the United Kingdom or elsewhere in http://www.dfes.gov.uk/elearningstrategy/
the world, provide some indications. These
http://www.edu.gov.mb.ca/ks4/tech/wbc/
examples, as well as alternative scenarios,
wbcgloss.html
should inform the emerging e-learning strategy
for health and social care. http://www.geneed.com
http://www.google.com
REFERENCES http://www.interstim.com
Department of Education and Skills (DfES). http://www.medtronic.com

(2003). Towards a unified learning strategy. http://www.nhsu.nhs.uk./webportal/learning/
Department of Health (DOH), Cheshire and elearning/StrategyDocument.pdf
Merseyside, Cumbria and Lancashire, and http://www.training.interstim.com
Greater Manchester WDCs. (2003). Deliver-
ing e-learning in the NHS. Getting the blend http://www.UniversityofHull.com
right: A strategic approach for the north
west.
117
KEY TERMS Learning Objects: Course materials de-

veloped according to a standard (e.g., IMS)
CD-ROM: High-capacity optical storage that allows the easy sharing of materials. Ma-
medium. terials are modular and can be used for a
variety of purposes and outputs.
Distance Learning: Learning in which
students and instructors are separated by dis- Online: Connected to a computer network.
tance or time.
Online Community: Learners who, al-
E-Learning: A wide set of applications and though separated by distance and time, share a
processes including Web-based training, virtual common experience.
classrooms, digital collaboration, and computer-
Online Conferencing: Communications
based training.
happening over a network. They can be syn-
Face to Face (F2F): Students and teach- chronous via chat or asynchronous via discus-
ers are in the same location at the same time. sion groups.
Internet-Based Learning or Training: Web-Based Course (WBC): Distance
Courses delivered via Internet technologies education course materials supported by com-
(these were text based before the Web). puter- mediated communications and delivered
asynchronously via the World Wide Web. An-
Learning Networks: Communities of
cillary materials such as print, videos, and CD-
learners connected via computer networks.
ROMs may be required.
118
119
Chapter XVI
Potential Benefits and
Challenges of Computer-Based
Learning in Health
Athina A. Lazakidou
Christina Ilioudi
Andriani Daskalaki
Max Planck Institute of Molecular Genetics, Germany
ABSTRACT
Computer-based learning has been developed for the beginning medical student and the
experienced practitioner, for the lay person and the medical expert. There are many
advantages to online and computer-based learning when compared to traditional face-to-face
courses and lectures. Information technologies are providing new opportunities for linking
medical schools around the world for sharing computer-based learning materials. In this
chapter, the authors present examples of actual programs that are being used to support
medical education for each of these categories of learners.
INTRODUCTION medical students to learn to use information

technology effectively. Computers can play a
Information technology is an increasingly im- direct role in the education process; students
portant tool for accessing and managing medi- may interact with educational computer pro-
cal information: both patient-specific knowl- grams to acquire factual information and to
edge and more general scientific knowledge. learn and practice medical problem-solving tech-
Medical educators are aware of the need for all niques. In addition, practicing physicians may
Potential Benefits and Challenges of Computer-Based Learning Health
use computers to expand and reinforce their puter, and complete the exam provided in
professional skills throughout their careers the program.
(Shortliffe & Perreault, 2001).
The application of computer technology to Levels I, II, III, and IV are the types of
education is often referred to as computer- computer-based training that would be most
assisted learning (CAL), computer-based edu- effective in addressing performance gaps among
cation (CBE), or computer-aided instruction international health workers. To qualify for
(CAI). these levels, a computer-based training pro-
Computer-based learning has been devel- gram must meet the following commonly ac-
oped for the beginning medical student and the cepted criteria (Dickelman, 1994).
experienced practitioner, for the lay person and
the medical expert. In this article, we present • Be easy to enter and exit
examples of actual programs that are being • Provide a simple way to move forward
used to support medical education for each of and backward (i.e., from screen to screen)
these categories of learners. • Be consistent in its key conventions
• Offer context-sensitive prompts and helps
• Provide tracking feedback (e.g., where
TYPES OF COMPUTER-BASED have I been? Where am I now? How
TRAINING much more is there to go?)
• Offer bookmarks (i.e., quit now, resume
There are four levels of computer-based train- later)
ing (CBT), each based on the application’s • Always offer a way out
complexity and its level of interactivity with the
user (Dulworth & Carney, 1996):
COMPUTER-BASED TRAINING IN
• Level I. Customized Linear Presen- HEALTHCARE
tation: Training similar to a standard
PowerPoint overhead presentation with In the health setting, CBT can be delivered in a
little interactivity preservice or in-service mode, as follows:
• Level II. Instructor-Led, Nonlinear
Presentation: Training by a facilitator • Preservice Training: Computerized
accompanied by navigation through the training delivered in health-education, nurs-
information on a computer without the use ing, and medical-school curricula through
of multimedia the use of software tutorials with or with-
• Level III. Facilitator-Led Training: A out professor facilitation, followed by ex-
multimedia presentation accompanied by aminations programmed in the computer
classroom-based training program or given by an instructor
• Level IV. Self-Paced Training: A mul- • In-Service Training: Health workers
timedia presentation that trainees use in- use CD-ROMs independently on their own
dependently with minimal assistance (also computers for stand-alone training, meet
known as stand-alone training). Individu- at a computer lab where facilitator-led
als can train at their own pace, either at an courses are coupled with the computer
outside lab or on their own desktop com- program, or attend the lab according to
120
Potential Benefits and Challenges of Computer-Based Learning in Health
their own schedules and review the mate- There are many advantages to online and
rials at their own pace computer-based learning when compared to
traditional face-to-face courses and lectures.
Research has shown that computer training There are a few disadvantages as well.
is particularly well suited to visually intensive,
detail-oriented subjects, such as anatomy and Main Advantages of Online or
kinesiology. This is because it allows text to be Computer-Based Learning
combined with still and moving graphics, with
the display of this information controlled by the • Class work can be scheduled around work
learner (Toth-Cohen, 1995). For example, com- and family.
puters can be particularly effective in present- • It reduces the travel time and travel costs
ing the following (Phillips, 1996): for off-campus students.
• Students may have the option to select
• Subjects that are difficult to conceptual- learning materials that meet their level of
ize, such as microscopic processes knowledge and interest.
• Material that is three dimensional and • Students can study anywhere they have
difficult to visualize on traditional two- access to a computer and an Internet
dimensional media such as books or connection.
whiteboards • Self-paced learning modules allow stu-
• Simulations of expensive or complex pro- dents to work at their own pace.
cesses, where the mechanical details of • There is the flexibility to join discussions
performing the process or the impossibil- on the bulletin-board threaded discussion
ity of using the real equipment may hinder areas at any hour, or visit with classmates
understanding and instructors remotely in chat rooms.
• Instructors and students both report that
e-learning fosters more interaction among
BENEFITS OF COMPUTER- students and instructors than in large lec-
BASED LEARNING ture courses.
• E-learning can accommodate different
Students may be learning more from using their learning styles and facilitate learning
computers than from attending lectures, ac- through a variety of activities.
cording to a study published in July 2004 at • It develops knowledge of the Internet and
studentbmj.com (http://www.studentbmj.com/ computers skills that will help learners
back_issues/0800/news/265a.html). throughout their lives and careers.
Research psychiatrists at the School of • Successfully completing online or com-
Medicine, University of Leeds, compared stu- puter-based courses builds self-knowledge
dents’ use of a computer-based multimedia and self-confidence, and encourages stu-
package with lecture-based teaching on the dents to take responsibility for their learn-
subject of anxiety. They found that even though ing.
students felt they learned more in the lecture • Learners can test out of or skim over
theatre, they gained more from using a com- materials already mastered and concen-
puter package. trate their efforts on mastering areas con-
taining new information and/or skills.
121
Main Disadvantages of Online or • A comprehensive medical-education re-

Computer-Based Learning source, the Medical Education Service,
available to teachers and learners world-
• Learners with low motivation or bad study wide. It will provide users with education-
habits may fall behind. ally and technologically state-of-the-art
• Without the routine structures of a tradi- medical-education resources and services
tional class, students may get lost or con- at low cost.
fused about course activities and dead- • A cost-effective alternative-track curricu-
lines. lum for undergraduate medical students
• Students may feel isolated from the infor the early years of the undergraduate
structor and classmates. program, and the right to complete the
• The instructor may not always be avail- clinical stages of their training in a partner
able when students are studying or need medical school.
help. • Customized postgraduate and continuing
• Slow Internet connections or older com- professional development (CPD) pro-
puters may make accessing course mate- grams that can be taken at the time and
rials frustrating. place of choosing of the learner, thereby
• Managing computer files and online learn- facilitating convenient and cost-effective
ing software can sometimes seem com- lifelong learning.
plex for students with beginner-level com- • Customised medical and multiprofessional
puter skills. health education through the IVIMEDS
• Hands-on or lab work is difficult to simu- Foundation appropriate to the needs and
late in a virtual classroom. circumstances of developing countries in
regard to curriculum, localization, language,
and mode of delivery.
THE INTERNATIONAL VIRTUAL
MEDICAL SCHOOL Benefits to Partner Institutions
The International Virtual Medical School IVIMEDS offers membership to an interna-

(IVIMEDS, http://www.ivimeds.org) is a ma- tional network of partner institutions sharing
jor international collaboration created to meet resources to enrich individual member curricu-
the challenge facing medical education through lums and to enhance the ability to deliver cost
innovative approaches that exploit develop- effectively high-quality medical-education pro-
ments in educational thinking and information grams. Benefits include the following:
and communication technologies. Currently, 37
leading medical schools located in 14 countries • Improved finances and assets by provid-
have committed financial and human resources, ing the means to develop an enhanced
and have agreed to share learning resources to curriculum with additional topics and ap-
make a reality of the IVIMEDS vision. proaches to learning, open access to medi-
The International Virtual Medical School cal training for students of different back-
provides the following: grounds, and a global market for home-
grown educational resources and strate-
gies.
122
• Rapid, effective execution of new ap- study), best suited to the financial, per-
proaches that may be beyond the budget sonal, and educational circumstances of
and scope of any one institution, drawing individual students.
upon a global body of expertise in subject
matter, educational theory, and technol- Benefits to Society
ogy.
• Reduced risk associated with curriculum IVIMEDS offers an approach to medical edu-
changes by sharing innovative thinking cation and training that is both adaptable and
and benefiting from other schools’ experi- cost effective. Benefits to society include the
ences in curriculum planning and their use following:
of learning technologies and learning-man-
agement systems. • Flexibility to expand and contract num-
• Quality resources and innovative ap- bers of learners to meet changing circum-
proaches to medical education, which can stances.
contribute to the creation of a curriculum • Wider access to medical education for
that is coherent, integrated, student cen- students, including disadvantaged and
tered, and authentic. mature students.
• Training doctors to focus on the needs of
Benefits to Students, Trainees, and particular communities, with the potential
Practicing Doctors that qualified doctors will return to work in
these communities.
IVIMEDS will provide an innovative curricu- • Training doctors with an appropriate high
lum and/or curriculum elements tailored to the level of competence in information han-
changing educational needs of students, train- dling and an aptitude for self-directed
ees, and medical professionals. Benefits in- learning and continuing professional de-
clude the following: velopment.
• Cost-effective training with schools work-
• A learner-centered approach with stu- ing together to blend e-learning with face-
dents exposed to just-in-time learning with to-face learning in a variety of educational
theory closely linked to practice. and clinical settings.
• Adaptive learning or “just-for-you” learn-
ing catering to individual learning styles
and interests. CONCLUSION
• Curriculum frameworks provided by es-
tablished learning outcomes, a broad cur- We recognize that technology impacts
riculum map, and a bank of virtual pa- healthcare-education, research, and science
tients. educators in the areas of research, classroom
• Blended learning including anytime, any- teaching, and distance education. While the
where electronic study guides, face-to- overall effect is not yet fully assessable, the
face and online tutors, and peer-to-peer presence of technology in so many different
learning. aspects of the profession makes it important to
• Flexible learning based in a variety of more clearly recognize and appreciate its cur-
settings (e.g., a university teaching hospi- rent and potential role.
tal, a local health centre, or home-based
123
Information technologies can be educators’ Henry, S. B., & Waltmire, D. (1992). Comput-
tools in finding creative ways that encourage erised clinical simulations: A strategy for staff
students to self-test, self-question, and self- development in critical care. American Jour-
regulate learning and in helping them to create nal of Critical Care Nursing, 1(2), 99-107.
solutions to complex problems. Information
Jaffe, C. C., & Lynch, P. J. (1993). Computers
technologies are providing new opportunities
for clinical practice and education in radiology.
for linking medical schools around the world for
Radiographics, 13(4), 931-937.
sharing computer-based learning materials.
They open a wide horizon for acquiring and Johannson, S. L., & Wertenberger, D. H. (1996).
expending medical knowledge originated in any Using simulation to test critical thinking skills of
part of the world without the limitations of time, nursing students. Nurse Education Today, 16,
space, or distance. 323-327.
The use of computers and information tech-
Lauri, S. (1992). Using a computer simulation
nology in medical education should be regarded
program to assess the decision making process
as an additional tool and must never be a goal in
in child health care. Computers in Nursing,
itself but part of flexible learning. On the con-
10(40), 171-177.
trary, clinical medical education should always
be centered on direct patient contact and bed- Lauriland, D. (1995). Multimedia and the chang-
side education. While we urge for direct patient ing experience of the learner. The British
contact, we believe that using stimulations would Journal of Educational Technology, 26(3),
also benefit the student in training. 179-189.
Phillips, R. (1996). Developers guide to inter-
active multimedia: A methodology for edu-
REFERENCES
cational applications. Perth, WA: Curtin
University.
Dickelman, G. J. (1994). Designing and man-
aging computer-based training for human re- Romiszowski, A. (1994). Individualization of
source development. In C. E. Schneier, C. J. teaching and learning: Where have we been?
Russell, R. W. Beatty, & L. S. Baird (Eds.), Where are we going? Journal of Special
The training and development sourcebook. Education Technology, 2, 182-194.
Amherst, MA: Human Resource Develop-
Shortliffe, E. H., & Perreault, L. E. (2001).
ment Press.
Medical informatics: Computer applications
Dulworth, M. R., & Carney, J. (1996). Improve in health care and biomedicine (2 nd ed.).
training with interactive multimedia: Info- New York: Springer.
line 9601. Alexandria, VA: American Society
Toth-Cohen, S. (1995). Computer-assisted in-
for Training and Development.
struction as a learning resource for applied
Evans, L. A., Brown, J. F., & Heestand, D. E. anatomy and kinesiology in the occupational
(1994). Incorporating computer-based learning therapy curriculum. American Journal of
in a medical school environment. Journal of Occupational Therapy, 49(8), 821-827.
Biocommunication, 21(1), 10-17.
124
FURTHER READING KEY TERMS
http://www.asme.org.uk/ Computer-Aided Instruction (CAI): The

application of computer technology to educa-
http://www.cdlhn.com
tion (also called computer-assisted learning and
http://www.chime.ucl.ac.uk/ computer-based education).
http://www.dso.iastate.edu/dept/asc/elearner/ Computer-Assisted Learning (CAL):
advantage.html The application of computer technology to edu-
cation (also called computer-aided instruction
http://www.emedicine.com
and computer-based education).
http://www.health.state.mn.us/divs/hrm/dl/
Computer-Based Education (CBE): The
compbased.html
application of computer technology to educa-
http://www.ifmsa.org/partners/wfme_he.htm tion (also called computer-assisted learning and
computer-aided instruction).
http://www.interactive-designs.com/cbl1.htm
Multimedia Content: Information sources
http://www.ivimeds.org
that encompass all common computer-based
http://www.lib.uiowa.edu/commons/cbl.html forms of information, including texts, graphics,
images, video, and sound.
http://www.med.cam.ac.uk/html/teaching/
DepMed/Phase1/computer.html Simulation: A system that behaves ac-
cording to a model of a process or another
http://www.personal.dundee.ac.uk/~cdvflore/
system; for example, a simulation of a patient’s
http://www.qaproject.org/pubs/PDFs/ response to therapeutic interventions allows a
researchcbtx.pdf medical or nursing student to learn which tech-
niques are effective without risking human life.
http://www.sph.umn.edu/publichealthplanet
Tutoring System: A computer program
designed to provide self-directed education to a
student or trainee.
125
127
Chapter XVII
Brain Mapping in
Functional Neurosurgery
George Zouridakis
University of Houston, USA
Javier Diaz
Farhan Baluch
ABSTRACT
Functional brain mapping is a procedure that can be used to identify cortical areas that
mediate sensorimotor and higher cognitive brain functions, such as language, attention,
memory, and cognition. Clinically, it is currently used for preoperative surgical planning in
patients suffering from intractable epilepsy and brain tumors, and may soon have significant
applications in brain injury, stroke, dementia, and developmental disorders. Functional brain
mapping is also a very powerful research tool in the area of cognitive neuroscience and,
lately, in psychiatry. Recent technological advances in neuroimaging techniques, the
development of large sensor arrays, the use of sophisticated computer systems and superior
graphics, gradually make more apparent the relevance of this technique in providing answers
to complex questions about the structural and functional connectivity of the brain, and the way
it represents and processes information.
INTRODUCTION rological diseases resistant to drug therapy,

such as refractory epilepsy, Parkinson’s dis-
Functional neurosurgery refers to those surgi- ease, essential tremor, and intractable chronic
cal interventions that are intended to improve pain. Ablative procedures entail the permanent
the function of the central or peripheral nervous disconnection of certain neural pathways, while
system. It is usually reserved for chronic neu- augmentative ones make use of implantable
Brain Mapping in Functional Neurosurgery
devices that modulate the function of dysfunc- population of cortical neurons whose activation
tional neuronal assemblies through the chronic gives rise to electromagnetic signals that can be
electrical stimulation of specific neuronal path- recorded outside the head. The electrical as-
ways. pects of brain activation can be recorded in an
In all cases, accurate localization of the electroencephalogram (EEG) by placing a set
intended surgical target and of the cortical of electrodes on the scalp, while the corre-
areas that are responsible for vital brain func- sponding magnetic aspects can be captured in
tions, such as sensation, movement, and speech, a magnetoencephalogram (MEG) by placing an
is of paramount importance because resection array of coils in close proximity to the head.
of such critical brain areas can have devastat- Brain-mapping procedures based on the combi-
ing results. The procedure employed is called nation of high-resolution MRI and EEG or MEG
functional brain mapping and aims at visualizing recordings are known as electrical source im-
the relationship between neural structures and aging and magnetic source imaging, respec-
their function. tively.
In general, MEG systems are expensive as
they require special cryogenic equipment, a
BACKGROUND magnetically shielded room, and daily monitor-
ing and maintenance. They are also only avail-
During the past several years, a number of able in a handful of places around the world, so
noninvasive functional imaging modalities, in- the clinical usefulness of MEG is limited. On
cluding functional magnetic resonance imaging the other hand, EEG equipment is portable,
(fMRI; Binder, Frost, Hammeke, Cox, Rao, & does not require any special maintenance, is
Prieto, 1997), positron-emission tomography readily available in practically all clinical set-
(PET; Peterson, Fox, Posner, Mintun, & Raichle, tings, and even the most sophisticated systems
1998), regional cerebral blood flow (rCBF; that incorporate dense-array sensors are at
Friberg, 1993), and single-photon-emission com- least one order of magnitude less expensive
puted tomography (SPECT; Gomez-Tortosa, than MEG. Moreover, recent advances in hard-
Martin, Sychra, & Dujovny, 1994), have been ware and the development of new mathemati-
used to map brain function with varying de- cal tools for modeling the intracranial sources
grees of success. However, the most reliable and the head make dense-array EEG (dEEG) a
approach to date still relies on direct electrical very attractive methodology because it can
stimulation of the exposed cortex (Ojemann, provide a temporal resolution of one millisecond
Ojemann, Lettich, & Berger, 1989), a proce- or less and a very high spatial sampling. There-
dure that is highly invasive and unpleasant, and fore, brain mapping based on dEEG and MRI
is performed mostly intraoperatively on an can have a significant impact on patient care. A
awake patient. state-of-the-art dEEG system (ActiveTwo,
More recently, however, completely BioSemi, The Netherlands) that is available in
noninvasive procedures have been success- our lab features 256 recording channels for
fully used for brain mapping (Ebersole & Wade, whole head coverage at a sampling rate of 5
1990; Peterson et al., 1998; Zouridakis, Simos, kHz per channel and uses active electrodes
Breier, & Papanicolaou, 1998). These proce- with built-in preamplifiers for noise cancella-
dures rely on the fact that the performance of tion.
certain brain functions involves only a small
128
Figure 1. Somatosensory (squares), language metabolic demand on the brain. Cancer-related

(circles), and epileptogenic (triangles) fatigue is the most common and debilitating side
cortical areas in an epileptic patient effect of cancer and its treatment.
Typically, brain tumors and medically in-
tractable epilepsy require surgical treatment,
but there is a narrow therapeutic window be-
tween treatment and debilitating side effects
due to surgery. Accurate preoperative brain
mapping is of paramount importance because
resection of functionally vital brain areas can
have devastating results.
BRAIN-MAPPING PROCEDURES
APPLICATION EXAMPLES Several brain functions can be studied using

recordings of spontaneous neurophysiological
Medically intractable epilepsy and brain cancer activity and of evoked responses (Zouridakis &
are two of the most debilitating diseases that Iyer, 2003). The primary auditory, somatosen-
severely affect the quality of life of those sory, and visual cortices can be identified by
suffering from the conditions, and they also analyzing responses to auditory, somatosen-
have a heavy impact on the lives of those caring sory, and visual stimuli, respectively. Epilepto-
for, and supporting, the patients. genic brain areas can be identified by analyzing
Refractory epilepsy is a neurological disor- spontaneous activity containing interictal spikes,
der characterized by recurrent seizures origi- while brain areas that contribute to language
nating from abnormal electrical activity in the function can be identified using evoked re-
brain. Patients experience a variety of symp- sponses obtained, for example, with a task for
toms, which can recur many times in a day, the continuous recognition of single words pre-
including loss of consciousness and possibly a sented either visually or aurally. The intracra-
fall, rigidity, jerking motions, incontinence of nial sources of the externally recorded re-
urine, weakness or paralysis, and even visual sponses can be modeled, for example, as a
and sensory hallucinations. Usually these symp- series of dipoles, whose locations can be visu-
toms are followed by a period of confusion and alized by projecting the estimated sources onto
deep sleep. Recurrent seizures severely limit a participant’s MRI.
school achievement, employment prospects, An example of a brain map (Zouridakis &
and normal daily activities, such as driving a Iyer, 2003) obtained using MEG recordings is
vehicle or operating machinery. shown in Figure 1, in which the somatosensory,
Brain cancer is a life-threatening disease language, and epileptogenic cortical areas are
that typically has a debilitating unremitting course shown as squares, circles, and triangles, re-
until the patient eventually expires: The physi- spectively. Figure 2 shows an example of map-
cal mass of the tumor is disruptive, the tumor ping the primary auditory cortex based on dEEG
cells cannot carry out normal cell functions, and (Zouridakis & Iyer, 2003). Shown are the evoked
the fast-growing cells place an extraordinary responses obtained from 256 recording elec-
129
Figure 2. N1 component (left), surface potentials around the N1 peak (middle), and cortical
areas activated (right) after auditory stimulation with tones of 1 kHz
trodes and the N1 component used to fit the been recognized (Papanicolaou et al., 1999;
single-dipole model (left), the 3-D distribution Zouridakis, Simos, Breier, et al., 1998;
of surface potentials at the N1 peak (middle), Zouridakis, Simos, & Papanicolaou, 1998).
and the estimated intracranial sources (right). A very important part of the mapping proce-
After mapping, surgical treatment may con- dure is the processing of the signals that repre-
sist of the implantation of a Vagus nerve stimu- sent brain activation, namely, the ongoing EEG
lator or the surgical resection of the epileptoge- and the evoked potentials out of which the
nic cortical areas. In the latter case, the func- intracranial sources are estimated. Thus, the
tional maps estimated from the surface neuro- development of advanced mathematical tools is
physiological recordings can be verified intra- of paramount importance. We have recently
operatively through direct electrical stimulation developed procedures (Zouridakis & Iyer, 2004)
of the exposed cortex to avoid postoperative that can reliably separate the scalp recordings
deficits associated with inadvertent resection into the neuronal activation that results from the
of eloquent cortex. experimental task and other activity that in-
cludes the ongoing background rhythm and
artifacts. Our methodology is based on inde-
FUTURE TRENDS pendent component analysis (ICA) and em-
ploys single-trial recordings in an iterative fash-
The previous paragraphs give some examples ion to obtain improved estimates of the true
of how functional source imaging can be used evoked response embedded in each single trial.
for clinical purposes. Both MEG and EEG This procedure, termed iterative ICA (iICA),
large-array systems have received Food and was used with single-trial auditory responses
Drug Administration (FDA) approval for clini- obtained from normal participants to demon-
cal and research use, and they are gradually strate the existence of aberrant (positive-go-
becoming available in many centers around the ing) N100 components that were intermixed
world. In particular, in the case of epilepsy with the expected normal (negative-going) ones
surgery, the usefulness of source imaging as a (Zouridakis & Iyer, 2005b). When the same
noninvasive tool to preoperatively delineate the procedure was applied to recordings from
extent of a lesion (to be resected) and of the schizophrenic participants, it was found that the
eloquent cortex (to be preserved) has already percentage of aberrant responses was much
130
Figure 3. Series of expected (value 0) and aberrant (value 1) responses obtained from a
normal (left) and a schizophrenic (right) participant after auditory stimulation with tones of
1 kHz
higher in the schizophrenia patients than in the Figure 4. Spatiotemporal pattern of brain
normal controls (Iyer & Zouridakis, 2004; activation quantified by channel coherence.
Zouridakis & Iyer, 2005a). Thus, we believe Average coherence across all channels (top),
that this measure may be used as a noninvasive the N100-P200 complex of the auditory
biological marker of schizophrenia. Figure 3 evoked responses recorded from 24 channels
shows an example of the classification patterns (middle), and the cortical areas that are
obtained from typical normal controls and a synchronized before stimulation (bottom left)
schizophrenic participant. and at the N100 peak (bottom right)
In an effort to understand information pro-
cessing and brain dynamics, and how the brain’s 1
connectivity influences its function, we are

Coherence
0.8
0.6
investigating the concept of functional connec- 0.4

0.2
tivity, in which we employ descriptive mea- 6

-200 -100 0 100 200
Time / ms
300 400 500
sures of similarity to study the various spa- 4
2
tiotemporal patterns of brain activation formed
0
between spatially distinct regions of the cere- -2
bral cortex during specific tasks. -4
-6
A simple example is shown in Figure 4. -8
-200 -100 0 100 200 300 400 500
Single-trial responses obtained from 24 chan-
nels around the head were processed with the
iICA procedure mentioned earlier and aver-
aged to produce clean evoked potentials, one
from each channel, which are shown superim-
posed in the middle panel of Figure 4. Then, the
coherence between all possible pairs of chan-
nels in a 20-millisecond sliding window was
computed as a function of time, and the average
131
computed across all channels is shown in the the development of new mathematical tools for
top panel of Figure 4. As it can be seen, the modeling the head and the intracranial sources
coherence function shows very high values underlying the externally recorded neuro-
during the evolution of the N100-P200 complex physiological signals gradually make more ap-
of the evoked response, and much lower values parent the relevance of this technique by pro-
outside this complex. The spatial extent of the viding answers to complex questions about the
functionally connected channels estimated be- structural and functional connectivity of the
fore the onset of the stimuli (time 0) and at the brain, and the way it represents and processes
N100 peak (time 100) is shown in the left and information.
right bottom panels of Figure 4, respectively.
REFERENCES
CONCLUSION
Binder, J. R., Frost, J. A., Hammeke, T. A.,
Functional brain mapping is a procedure that Cox, R. W., Rao, S. M., & Prieto, T. (1997).
attempts to establish the relationship between Human brain areas identified by functional
brain structures and their functions. It has a magnetic resonance imaging. Journal of Neu-
wide range of potential applications as it can roscience, 17, 353-362.
reliably identify cortical areas that mediate
Ebersole, J. S., & Wade, P. B. (1990). Spike
sensorimotor and higher cognitive brain func-
voltage topography and equivalent dipole local-
tions, such as receptive and expressive lan-
ization in complex partial epilepsy. Brain To-
guage, on an individual-participant basis.
pography, 3, 21-34.
Clinically, it is currently used in the diagno-
sis of disease and the assessment of disease Fender, D. H. (1991). Models of the human
progression, and for preoperative surgical plan- brain and the surrounding media: Their influ-
ning in functional neurosurgery, which is a ence on the reliability of source localization.
rapidly advancing field that offers minimally Journal of Clinical Neurophysiology, 8(4),
invasive and highly effective treatment options 381-390.
for many difficult neurological disorders. Fu-
Friberg, L. (1993). Brain mapping in thinking
ture clinical applications of mapping can be
and language function. Acta Neurochirurgia,
extended to include brain injury and stroke
56(Suppl.), 34-39.
assessment, dementias, developmental disor-
ders, as well as further characterization of Gomez-Tortosa, E., Martin, E. M., Sychra, J. J.,
higher cortical areas involved in attention, & Dujovny, M. (1994). Language-activated
memory, and cognition. Functional brain map- single-photon emission tomography imaging in
ping is also an active area of research and has the evaluation of language lateralization-evi-
rapidly developed as a powerful tool in cogni- dence from a case of crossed aphasia: Case
tive neuroscience and, in recent years, has seen report. Neurosurgery, 35, 515-519.
widespread application in psychiatry.
Iyer, D., & Zouridakis, G. (2004). Single-trial
Recent technological advances in
EP analysis improves separation of normal and
neuroimaging techniques, the development of
schizophrenia subjects. Proceedings of the
larger sensor arrays, the use of sophisticated
First Joint Conference of the EEG & Clini-
computer hardware and superior graphics, and
132
cal Neuroscience Society (ECNS) & the In- ceedings of the 27 th IEEE EMBS Annual
ternational Society for NeuroImaging in Psy- International Conference, Shanghai, China.
chiatry (ISNP), Irvine, CA.
Zouridakis, G., Simos, P. G., Breier, J. I., &
Ojemann, G., Ojemann, J., Lettich, E., & Burger, Papanicolaou, A. C. (1998). Functional hemi-
M. (1989). Cortical language localization in left, spheric asymmetry assessment in a visual lan-
dominant hemisphere: An electrical stimulation guage task using MEG. Brain Topography,
mapping investigation in 117 patients. Journal 11, 57-65.
of Neurosurgery, 71, 316-326.
Zouridakis, G., Simos, P. G., & Papanicolaou,
Papanicolaou, A. C., Simos, P. G., Breier, J. I., A. C. (1998). Multiple bilaterally asymmetric
Zouridakis, G., Willmore, L. J., Wheless, J. W., cortical sources account for the auditory N1m
et al. (1999). Magnetoencephalographic map- component. Brain Topography, 10, 183-189.
ping of the language-specific cortex. Journal
of Neurosurgery, 90, 85-93.
KEY TERMS
Peterson, S. E., Fox, P. T., Posner, M. I.,
Mintun, M., & Raichle, M. E. (1988). Positron Coherence: A measure of the dependence
emission tomographic studies of cortical anatomy (or similarity) of two random processes.
of single-word processing. Nature, 331, 585-
Correlation: A measure of the strength of
589.
the linear relationship between two random
Zouridakis, G., & Iyer, D. (2003). Functional processes.
brain mapping through intracranial source im-
Electroencephalogram (EEG): Tracings
aging. In J. E. Moore, Jr. & G. Zouridakis
representing the spontaneous electrical activity
(Eds.), Biomedical technology and devices
of the brain (brain waves).
handbook. Boca Raton: CRC Press.
Electroencephalography (EEG): The
Zouridakis, G., & Iyer, D. (2004). Improved
measurement of the electrical activity of the
estimation of evoked potentials using an itera-
brain by placing electrodes on the scalp.
tive independent component analysis proce-
dure. WSEAS Transactions on Signal Pro- Epilepsy: A chronic neurological condition
cessing, Robotics and Automation, 2(1), 288- characterized by recurrent seizures.
291.
Evoked Potential: Electrical potential re-
Zouridakis, G., & Iyer, D. (2005a). Phase corded on the scalp following the presentation
aspects and localization analysis of the auditory of a sensory stimulus.
N100 component. Proceedings of the Joint
Functional Imaging: The use of imaging tech-
Meeting of the Fifth International Confer-
niques to localize brain areas that mediate specific
ence on Bioelectromagnetism and the Fifth
sensorimotor and cognitive brain functions.
International Symposium on Noninvasive
Functional Source Imaging within the Hu- Functional Magnetic Resonance Imag-
man Brain and Heart, Minneapolis, MN. ing (fMRI): The use of MRI to measure the
hemodynamic response related to neural activ-
Zouridakis, G., & Iyer, D. (2005b). Single-trial
ity in the brain.
analysis of the auditory N100 component. Pro-
133
Magnetoencephalography (MEG): The Positron-Emission Tomography (PET):

measurement of the magnetic activity of the Measures emissions from radioactive chemi-
brain by placing cryogenic equipment with coils cals injected into the bloodstream to produce
close to the scalp. functional images of the brain.
134
135
Chapter XVIII
ECG Diagnosis Using Decision
Support Systems
Themis P. Exarchos
University of Ioannina, Greece
Costas Papaloukas
Markos G. Tsipouras
Yorgos Goletsis
Dimitrios I. Fotiadis
Biomedical Research Institute—FORTH, Greece
Lampros K. Michalis
ABSTRACT
ECG is one of the most common signals used in medical practice due to its noninvasive nature
and the information it contains. Several systems and various automated approaches have been
developed that use computer technology to provide ECG diagnosis. These systems detect
abnormalities and other features in the ECG signal and produce a decision which helps the
physician when performing diagnosis. ECG decision support systems can serve as a diagnostic
tool for specific cardiac anomalies such as myocardial ischaemia and arrhythmia.
ECG Diagnosis Using Decision Support Systems
ELECTROCARDIOGRAM duration and amplitude to the more complex like

slopes, intervals, frequencies, or other discrimi-
The electrocardiogram (ECG) is a clinical test nating indices. These are used in the diagnosis
that records the electrical activity of the heart. stage since the values of certain features are
ECG is used to measure the rate and regularity indicators of the existence of an underlying
of heartbeats as well as the size and position of disease. Apparently, the measurement accu-
the chambers, the effects of drugs or devices racy (Acc) is vital at this stage, and computer-
used to regulate the heart, and the presence of ized methods are used to address it efficiently.
any damage to the heart. An ECG is useful in The last stage in the ECG analysis is the
determining whether a person suffers from a diagnosis, where explicit medical knowledge is
heart disease. If a person has chest pain or utilized. Collaboration with medical experts is
palpitations, an ECG will determine if the heart necessary, and the individual characteristics of
is beating normally. If a person is under medi- each patient complicate the decision-making
cations that affect the heart or if the patient is task. Various automated approaches have been
on a pacemaker, an ECG can provide informa- proposed. These systems detect the abnormali-
tion on the immediate effects of changes in ties in the ECG and some of them can also
activity or medication levels. An ECG may be produce interpretations for the decisions made.
included as part of a routine examination in ECG analysis can help diagnose specific car-
patients over 40 years old. diac anomalies such as myocardial ischaemia
and arrhythmia.
ECG ANALYSIS
MYOCARDIAL ISCHAEMIA
Automated ECG analysis consists of a series of DIAGNOSIS
procedures that can be utilized in order to
produce useful clinical information to help the Myocardial ischaemia is the condition in which
physician to reach a diagnosis concerning the oxygen deprivation to the heart muscle is ac-
pathophysiological condition of the patient’s companied by the inadequate removal of me-
heart faster and safer. ECG analysis consists of tabolites due to reduced blood flow or perfu-
four stages: (a) signal acquisition, (b) process- sion. This reduced blood supply to the myocar-
ing, (c) feature extraction, and (d) diagnosis. dium causes alterations in the ECG signal, such
Signal acquisition should fulfill certain specifi- as deviations in the ST segment and changes in
cations concerning the sampling frequency the T wave (Goldman, 1982). The detection and
(100Hz to 1 KHz), the resolution (number of assessment of those alterations in long-duration
bits for each sample, 6 to 16), and the sensitiv- ECGs is a simple and noninvasive method to
ity, which expresses the signal’s amplitude diagnose ischaemia. In Figure 1, some of the
range (usually 5 mV or 6 mV). The digital ECG typical ECG features employed for the diagno-
signal is then processed and filtered to suppress sis of myocardial ischaemia are shown.
noise and enhance the relevant ECG character- Myocardial-ischaemia diagnosis using the
istics. ECG signal can be described as a sequence of
In the feature-extraction stage, all the rel- two tasks: ischaemic beat detection and
evant ECG characteristics are recognized and ischaemic episode definition. The first is re-
some of their features are computed. The ex- lated to the classification of beats as normal or
tracted features vary from simple ones like the ischaemic. Several techniques have been pro-
136
Figure 1. Typical ECG features extracted for myocardial-ischaemia diagnosis: (a) ST-segment
deviation, (b) ST-segment slope, (c) ST-segment area, and (d) T-wave amplitude
(a) (b)
R
R
S T slo p e
T
P T
P
isoelectric ST deviation isoelectric

line
line
J80 point
S
J point J80 point
S
J point
(c) (d)
R
R
ST area
T
T peak
T P
P
isoelectric T a m p litu d e
isoelectric line
line
J80 point S
S
J point
posed for ischaemic beat classification, which Michalis, 2004), and classification using asso-
evaluate the ST-segment changes and the T- ciation rules (Exarchos, Papaloukas, Fotiadis,
wave alterations using different methodologi- & Michalis, in press).
cal approaches. More specifically, they use Cardiac beat detection and classification is
parametric modeling (Papaloukas, Fotiadis, a key process in the definition of the ischaemic
Likas, & Michalis, 2002a; Pitas, Strintzis, episodes in the ECG signal. The accuracy of
Grippas, & Xerostylides, 1983), wavelet theory the beat classification influences ischaemic
(Senhadji, Carrault, Bellanger, & Passariello, episode definition, in which sequences of
1995), a set of rules (Papaloukas, Fotiadis, ischaemic beats need to be identified. Various
Liavas, Likas, & Michalis, 2001; Papaloukas, methods have been proposed for ischaemic
Fotiadis, Likas, Stroumbis, & Michalis, 2002), episode detection based on a set of rules
artificial neural networks (Maglaveras, (Papaloukas, Fotiadis, Liavas, et al., 2001; Silipo,
Stamkopoulos, Pappas, & Strintzis, 1998; Taddei, & Marchesi, 1994), artificial neural
Papadimitriou, Mavroudi, Vladutu, & networks (Papadimitriou et al., 2001;
Bezerianos, 2001; Papaloukas, Fotiadis, Likas, Papaloukas, Fotiadis, Likas, & Michalis, 2002b;
& Michalis, 2002b; Stamkopoulos, Diamantaras, Silipo & Marchesi, 1998), fuzzy logic (Vila,
Maglaveras, & Stintzis, 1998), multicriteria Presedo, Delgado, Barro, Ruiz, & Palacios,
decision analysis (Goletsis, Papaloukas, Fotiadis, 1997), and other signal-analysis techniques
Likas, & Michalis, 2003), genetic algorithms (Jager, Mark, Moody, & Divjak, 1992; Lemire,
(Goletsis, Papaloukas, Fotiadis, Likas, & Pharand, Rajaonah, Dube, & LeBlanc, 2000;
137
Taddei, Costantino, Silipo, Edmin, & Marchesi, ing the ST segment and the T wave of each beat
1995). is the input to the beat-classification system,
The most common techniques for the beat- and the output is the classification of the beat.
classification problem are neural- and rule- The input to the network is produced using
based approaches. Neural-based approaches principal component analysis to reduce dimen-
have resulted in high performance but exhibit sionality. The network performance in beat
an important drawback due to their inability to classification was tested on a subset of the
provide explanations for the classification deci- European Society of Cardiology ST-T Data-
sions. Rule-based approaches exhibit the highly base (European ST-T Database Directory,
desirable feature of interpreting the decisions, 1991), providing 90% sensitivity (Se) and 90%
but their performance is not equally satisfac- specificity (Sp). The neural beat classifier is
tory in terms of accuracy. integrated in a four-stage procedure for
A recent knowledge-based system analyzes ischaemic episode detection.
the ECG signal using a four-stage algorithm Another technique employs a similar ap-
(Papaloukas, Fotiadis, Liavas, et al., 2001). The proach as fuzzy logic, multicriteria decision
four stages correspond to noise handling and analysis, for ischaemic beat recognition (Goletsis
ECG feature extraction, beat classification, et al., 2003). It deals with assigning objects,
window classification, and the identification of namely the cardiac beats, into predefined cat-
ischaemic episode duration. In the first stage, egories. In order to characterize each beat as
the preprocessing of the ECG recording is ischaemic or not, the beat is compared to
performed to achieve noise removal and the already assigned category prototypes. Similar-
extraction of the signal features to be used for ity between each beat and the prototype is
beat characterization. In the second stage, each computed, and each beat is assigned to the
beat is classified as normal, abnormal category to which the most similar prototype
(ischaemic), or artefact. This information is belongs. The fuzzy pairwise comparison is made
used in the third stage (the window-character- for a number of criteria that employ the ST-
ization stage), where each 30-second ECG segment deviation and slope, the T-wave am-
window is classified as ischaemic or not. In the plitude and polarity, and the patient’s age. For
fourth stage, the identification of start and end each criterion, two parameters are estimated,
points of each ischaemic episode is performed. the similarity and the dissimilarity, while the
The above system is used to detect the overall comparison outcome is aggregated into an in-
episodes of ischaemia, but to distinguish also difference index with the use of criterion
the ST episodes from the T episodes weights. All the parameters of the method,
(Papaloukas, Fotiadis, Likas, Stroumbis, et al., thresholds and weights, were adjusted using
2002). The latter is of great clinical importance medical experience. A more recent approach
since the prognosis of ischaemic episodes with uses genetic algorithms for the automatic cal-
ST-segment changes is worse than those with culation of the thresholds and the weights
T-wave alterations only. (Goletsis et al., 2004).
Another methodology, which is based on The latest developed approach for ECG
ANNs, has been developed for the detection of beat classification employs data-mining tech-
ischaemic episodes in long-duration ECG re- niques and especially algorithms that use asso-
cordings (Papaloukas, Fotiadis, Likas, & ciation rules for the classification (Exarchos et
Michalis, 2002b). The raw ECG signal contain- al., in press). A methodology based on a three-
138
Table 1. Comparison of the performance of several methods for myocardial-Ischaemia

diagnosis
Method Se1 (%) Sp2 (%) Acc3 (%)

ANN & parametric modelling (Papaloukas et. al., 2002) 81 84
Rule-based (Papaloukas et. al., 2001) 70 63
Feed forward ANN and nonlinear principal components analysis 79 75
(Stamkopoulos et. al., 1998).
Bidirectional associative memories ANN (Maglaveras et. 56
al.,1998)
ANN (Classification partitioning-Self organising map) 74
(Papadimitriou et. al., 2001)
ANN (Classification partitioning-Self organising map & radial 77
basis function) (Papadimitriou et. al., 2001)
ANN (Classification partitioning-Self organising map & support 80
vector machine) (Papadimitriou et. al., 2001)
ANN & principal components analysis (Papaloukas et. al., 2002a) 90 90
Multicriteria decision analysis (Goletsis et. al., 2003) 90 89
Genetic algorithms & multicriteria decision analysis (Goletsis et. 91 91
al., 2004)
Association rule mining (Exarchos et. al., 2005) 87 93 90
1
Se: Sensitivity
2
Sp: Specificity
3
Acc: Accuracy
stage schema was developed. The three stages parametric modeling (Papaloukas, Fotiadis,
correspond to noise handling and ECG feature Likas, & Michalis, 2002a), bidirectional asso-
extraction, feature discretization, rule mining, ciative memories (Maglaveras et al., 1998), and
and beat classification. In the first stage, the the Kohonen self-organizing map algorithm
preprocessing of the ECG recording is per- combined with radial basis functions or support
formed to achieve noise removal and the ex- vector machines (Papadimitriou et al., 2001).
traction of the signal features that are used for In order to evaluate the performance of
beat characterization. In the second stage, ev- automated systems for myocardial-ischaemia
ery continuous valued feature is discretized (it diagnosis, a standard reference database has
is transformed to categorical) in order to be been developed: the European Society of Car-
utilized in the next stage. In the third stage, diology ST-T Database (European ST-T Da-
association-rule mining algorithms are applied tabase Directory, 1991).
to generate association rules, which are used to In Table 1, we can see the performance of
establish the beat-classification model. The several systems designed for beat classification.
methodology was evaluated using data from the
European Society of Cardiology ST-T Data-
base, and the obtained sensitivity and specific- ARRHYTHMIA DIAGNOSIS
ity were 87% and 93%, respectively.
Other approaches for beat classification are Arrhythmia can be defined as either an irregu-
based on the combination of an auto-associa- lar single heartbeat (arrhythmic beat), or as an
tive, nonlinear ANN and a radial basis function irregular group of heartbeats (arrhythmic epi-
ANN (Stamkopoulos et al., 1998), ANNs and sode). Arrhythmias can affect the heart rate,
139
causing irregular rhythms, such as slow or fast namical modeling (Owis, Abou-Zied, Youssef,
heartbeat. Arrhythmias can take place in a & Kadah, 2002). It is noticeable that all meth-
healthy heart and be of minimal consequence ods address the detection of only a few types of
(e.g., respiratory sinus arrhythmia, which is a arrhythmia (atrial tachycardia, ventricular ta-
natural periodic variation in heart rate, corre- chycardia, atrial fibrillation, and ventricular fi-
sponding to respiratory activity), but they may brillation). ECG beat-by-beat classification is
also indicate a serious problem that may lead to another field of interest, where each beat is
stroke or sudden cardiac death (Sandoe & classified into several different rhythm types
Sigurd, 1991). utilizing artificial neural networks (Dokur &
Several researchers have addressed the Olmez, 2001), fuzzy neural networks (Osowski
problem of the automatic detection and classi- & Linh, 2001), the “mixture of experts ap-
fication of cardiac rhythms. Some techniques proach” (Hu, Palreddy, & Tompkins, 1997),
are based on the detection of a single arrhyth- hermite functions combined with self-organiz-
mia type and its discrimination from normal ing maps (Lagerholm, Peterson, Braccini,
sinus rhythm, or the discrimination between Ebendrandt, & Sornmo, 2000), and time-fre-
two different types of arrhythmia utilizing time- quency analysis combined with knowledge-
domain analysis (Throne, Jenkins, & DiCarlo, based systems (Tsipouras, Fotiadis, & Sideris,
1991), the sequential hypothesis-testing algo- 2002). These methods classify more arrhyth-
rithm (Thakor, Zhu, & Pan, 1990), threshold- mic beat types, but they focus on single-beat
crossing intervals (Clayton, Murray, & Campbell, classification and not arrhythmic episode de-
1993), artificial neural networks (Clayton, tection.
Murray, & Campbell, 1994; Yang, Device, & Most of the studies are based on the analysis
Macarlane, 1994), time-frequency analysis of the ECG signal. In these methods, ECG
(Afonso & Tompkins, 1995; Tsipouras & features are extracted and used for the detec-
Fotiadis, 2004), fuzzy adaptive resonance theory tion and/or classification of arrhythmias. How-
mapping (Ham & Han, 1996), and the sequen- ever, the presence of noise makes feature
tial detection algorithm (Chen, Clarkson, & extraction difficult and in some cases impos-
Fan, 1996). Another category of methods for sible. Also, most of the methods are time con-
arrhythmia detection and classification is based suming and ineffective for real-time analysis.
on the detection of different heart rhythms and An alternative would be to use only the RR-
their classification in two or three arrhythmia interval signal. In this case, it is expected that
types, and the normal sinus rhythm. Techniques certain types of arrhythmias can be detected
belonging to this category include multiway and classified.
sequential-hypothesis testing (Thakor, A recent work for an arrhythmia-detection
Natarajan, & Tomaselli, 1994), wavelet analy- method based on time and time-frequency analy-
sis (Khadra, Al-Fahoum, & Al-Nashash, 1997), sis (Tsipouras & Fotiadis, 2004) utilizes only the
artificial neural networks (Minami, Nakajima, RR-interval signal and heart-rate features. Ini-
& Toyoshima, 1999), complexity measure tially, the RR-interval duration signal is ex-
(Zhang, Zhu, Thakor, & Wang, 1999), tracted from ECG recordings and segmented
multifractal analysis (Wang, Zhu, Thakor, & into small intervals. The analysis is based on
Xu, 2001), wavelet analysis combined with both time and time-frequency features. Time-
radial basis function neural networks (Al- domain measurements are extracted and sev-
Fahoum & Howitt, 1999), and nonlinear dy- eral combinations between the obtained fea-
140
Table 2. Comparison of several research attempts for arrhythmic beat classification
Acc1
Method Signal Dataset
(%)
Feature extraction: 7,185 beats from MIT-BIH
cumulants of the second, 4,035 training – 3,150 testing
third and fourth order ECG Í : 2,250 A : 658 96.06%
Classification: L : 1,200 V : 1,500
fuzzy hybrid neural network R : 1,000 I : 472
(Osowski, & Linh, 2001) E : 105
Feature extraction:
discrete wavelet transform 3,000 beats from MIT-BIH
Classification: ECG N, L, R, P, p, a, E, V, F, f: 300 from each category 95.7%
intersecting Spheres network 1,500 training – 1,500 testing
(Dokur, & Olmez, 2001)
Feature extraction: 25 min from each record in MIT-BIH 200 series
PCA in 29 points from QRS, excluding records 212, 217, 220, 222 and 232
instantaneous and average
RR-interval, QRS complex
width ECG 95.52%
Classification: N : 43897 V : 5363
mixture of experts (SOM,
LVQ)
(Hu et. al., 1997).
108,963 beats from MIT-BIH
Feature extraction: N : 74053 F : 803

hermite functions, L : 8074 b : 472
RR-interval R : 7259 e : 16
ECG 98.49%
Clustering: A : 2544 j : 229
self organizing maps a : 150 E : 106
(Lagerholm et. al., 2000) J : 83 P : 7028
S : 2 f : 982
V : 7129 Q : 33
Feature extraction: 30,000 beats from MIT-BIH
RR-interval RR-
N, P, f, p, V, F : 2,950
Classification: interval : 25,188 95.85%
Q, L, R e,j,n,E : 265
knowledge-based system signal
(Tsipouras et. al., 2002) A, a, J, S : 1,213 [, !, ] : 384
93,349 beats from MIT-BIH

N, P, f, p, 98.20%
: 86,262 V : 6,183
Feature extraction: L, R, Q
RR-interval RR- [, !, ] : 484 BII : 420
Classification: interval
knowledge-based system signal 109,880 beats from MIT-BIH
(Tsipouras et. al., 2005) N, P, f, p, 94.26%
: 102,793 V : 6,183
L, R, Q
[, !, ] : 484 BII : 420
1
Acc: Accuracy
tures are used to train a set of neural networks. ences and Technology, 1997), and satisfactory
Short-time Fourier transform, and several time- results are obtained for both sensitivity and
frequency distributions are used in the time- specificity for arrhythmic segment detection
frequency analysis. The final decision is made (87.5% and 89.5% respectively for time-do-
using a set of rules. The proposed approach is main analysis, and 90% and 93% respectively
tested using the MIT-BIH Arrhythmia Data- for time-frequency-domain analysis).
base (Harvard-MIT Division of Health Sci-
141
Another approach for arrhythmic beat clas- REFERENCES

sification and arrhythmic episode detection and
classification (Tsipouras, Fotiadis, & Sideris, in Afonso, V. X., & Tompkins, W. J. (1995).
press) is also based only on the RR-interval Detecting ventricular fibrillation. IEEE Engi-
signal. A three RR-interval sliding window is neering in Medicine and Biology, 14, 152-
used in the arrhythmic beat classification algo- 159.
rithm. Classification is performed for four cat-
Al-Fahoum, A. S., & Howitt, I. (1999). Com-
egories of beats: normal, premature ventricular
bined wavelet transformation and radial basis
contractions, ventricular flutter and fibrillation,
neural networks for classifying life-threatening
and 2o heart block. The beat classification is
cardiac arrhythmias. Medical & Biological
used as the input of a knowledge-based, deter-
Engineering and Computing, 37, 566-573.
ministic automaton to achieve arrhythmic epi-
sode detection and classification. Six rhythm Chen, S. W., Clarkson, P. M., & Fan, Q.
types are classified: ventricular bigeminy, ven- (1996). A robust sequential detection algorithm
tricular trigeminy, ventricular couplet, ventricu- for cardiac arrhythmia classification. IEEE
lar tachycardia, ventricular flutter and fibrilla- Transactions on Biomedical Engineering,
tion, and 2o heart block. The achieved scores 43, 1120-1125.
for the MIT-BIH Arrhythmia Database indi-
Clayton, R. H., Murray, A., & Campbell, R. W.
cate high performance: 98% accuracy for ar-
F. (1993). Comparison of four techniques for
rhythmic beat classification and 94% accuracy
recognition of ventricular fibrillation of the sur-
for arrhythmic episode detection and classifi-
face ECG. Medical & Biological Engineer-
cation.
ing and Computing, 31, 111-117.
In Table 2, several systems and methodolo-
gies that address the problem of arrhythmic Clayton, R. H., Murray, A., & Campbell, R. W.
beat classification are presented. F. (1994). Recognition of ventricular fibrillation
using neural networks. Medical & Biological
Engineering and Computing, 32, 217-220.
CONCLUSION
Dokur, Z., & Olmez, T. (2001). ECG beat
classification by a hybrid neural network. Com-
We presented a general review of ECG diagno-
puter Methods and Programs in Biomedi-
sis using decision-support systems. Computer-
cine, 66, 167-181.
ized ECG processing can be divided into sev-
eral stages. The most important ones are the European Society of Cardiology. (1991).Euro-
analysis and the diagnosis. The variety of the pean ST-T database directory. Pisa: S.T.A.R.
systems developed in order to address the
Exarchos, T. P., Papaloukas, C., Fotiadis, D. I.,
automation of these two tasks shows the great
& Michalis, L. K. (in press). An association
interest for this scientific area and its clinical
rule mining based methodology for automatic
importance. The developed methods differ in
diagnosis of myocardial ischemia using ECGs.
terms of accuracy, but each one of them has
IEEE Transactions on Biomedical Engineer-
unique advantages. A proper combination of
ing.
the above techniques can improve the results of
ECG analysis and diagnosis. The physician, Goldman, M. J. (1982). Principles of clinical
however, remains responsible to evaluate the electrocardiography (11 th ed.). Los Altos,
systems’ output and make the final diagnosis. CA: LANGE Medical Publications.
142
Goletsis, Y., Papaloukas, C., Fotiadis, D. I., Lemire, D., Pharand, C., Rajaonah, J.-C., Dube,
Likas, A., & Michalis, L. K. (2003). A B., & LeBlanc, A. R. (2000). Wavelet time
multicriteria decision based approach for is- entropy, T wave morphology and myocardial
chemia detection in long duration ECGs. Pro- ischemia. IEEE Transactions on Biomedical
ceedings of IEEE EMBS Fourth Interna- Engineering, 47, 967-970.
tional Conference of Information Technol-
Maglaveras, N., Stamkopoulos, T., Pappas, C.,
ogy and Applications in Biomedicine (ITAB
& Strintzis, M. (1998). ECG processing tech-
2003) (pp. 230-233).
niques based on neural networks and bidirec-
Goletsis, Y., Papaloukas, C., Fotiadis, D. I., tional associative memories. Journal of Medi-
Likas, A., & Michalis, L. K. (2004). Automatic cal Engineering in Technology, 22, 106-111.
ischemic beat classification using genetic algo-
Minami, K., Nakajima, H., & Toyoshima, T.
rithms and multicriteria decision analysis. IEEE
(1999). Real-time discrimination of ventricular
Transactions on Biomedical Engineering,
tachyarrhythmia with Fourier-transform neural
51, 1717- 1725.
network. IEEE Transactions on Biomedical
Ham, F. M., & Han, S. (1996). Classification of Engineering, 46, 179-185.
cardiac arrhythmias using fuzzy ARTMAP.
Harvard-MIT Division of Health Sciences and
Technology. (1997). MIT-BIH arrhythmia
ing, 43, 425-430.
database (3 rd ed.) [CD-ROM]. Cambridge,
Hu, Y. Z., Palreddy, S., & Tompkins, W. J. MA.
(1997). A patient-adaptable ECG beat classi-
Osowski, S., & Linh, T. H. (2001). ECG beat
fier using a mixture of experts approach. IEEE
recognition using fuzzy hybrid neural network.
44, 891-900.
ing, 48, 1265-1271.
Jager, F., Mark, R. G., Moody, G. B., & Divjak,
Owis, M. I., Abou-Zied, A. H., Youssef, A. M.,
S. (1992). Analysis of transient ST segment
& Kadah, Y. M. (2002). Study of features
changes during ambulatory monitoring using
based on nonlinear dynamical modelling in ECG
the Karhunen-Loève transform. Proceedings
arrhythmia detection and classification. IEEE
of the IEEE Computers in Cardiology (pp.
691-694).
49, 733-736.
Khadra, L., Al-Fahoum, A. S., & Al-Nashash,
Papadimitriou, S., Mavroudi, S., Vladutu, L., &
H. (1997). Detection of life-threatening car-
Bezerianos, A. (2001). Ischemia detection with
diac arrhythmias using wavelet transformation.
a self-organizing map supplemented by super-
Medical & Biological Engineering and Com-
vised learning. IEEE Transactions on Neural
puting, 35, 626-632.
Networks, 12, 503-515.
Lagerholm, M., Peterson, C., Braccini, G.,
Papaloukas, C., Fotiadis, D. I., Liavas, A. P.,
Ebendrandt, L., & Sornmo, L. (2000). Cluster-
Likas, A., & Michalis, L. K. (2001). A knowl-
ing ECG complexes using hermite functions
edge-based technique for automated detection
and self-organizing maps. IEEE Transactions
of ischemic episodes in long duration electro-
on Biomedical Engineering, 47, 838-848.
143
cardiograms. Medical and Biological Engi- changes in ischemic patients. Proceedings of

neering and Computing, 39, 105-112. the IEEE Computers in Cardiology (pp. 225-
228).
Papaloukas, C., Fotiadis, D. I., Likas, A., &
Michalis, L. K. (2002a). An expert system for Stamkopoulos, T., Diamantaras, K.,
ischemia detection based on parametric model- Maglaveras, N., & Strintzis, M. (1998). ECG
ing and artificial neural networks. Proceedings analysis using nonlinear PCA neural networks
of the European Medical Biological Engi- for ischemia detection. IEEE Transactions on
neering Conference (pp. 742-743). Signal Processing, 46, 3058-3067.
Papaloukas, C., Fotiadis, D. I., Likas, A., & Taddei, A., Costantino, G., Silipo, R., Edmin,
Michalis, L. K. (2002b). An ischemia detection M., & Marchesi, C. (1995). A system for the
method based on artificial neural networks. detection of ischemic episodes in ambulatory
Artificial Intelligence in Medicine, 24, 167- ECG. Proceedings of the IEEE Computers in
178. Cardiology (pp. 705-708).
Papaloukas, C., Fotiadis, D. I., Likas, A., Thakor, N. V., Natarajan, A., & Tomaselli, G.
Stroumbis, C. S., & Michalis, L. K. (2002). Use (1994). Multiway sequential hypothesis testing
of a novel rule-based expert system in the for tachyarrhythmia discrimination. IEEE
detection of changes in the ST segment and the Transactions on Biomedical Engineering,
T wave in long duration ECGs. Journal of 41, 480-487.
Electrocardiology, 35, 27-34.
Thakor, N. V., Zhu, Y. S., & Pan, K. Y. (1990).
Pitas, I., Strintzis, M. G., Grippas, S., & Ventricular tachycardia and fibrillation detec-
Xerostylides, C. (1983). Machine classification by a sequential hypothesis testing algo-
tion of ischemic electrocardiograms. Pro- rithm. IEEE Transactions on Biomedical En-
ceedings of the IEEE Mediterranean gineering, 37, 837-843.
Electrotechnical Conference (MELECON),
Throne, R. D., Jenkins, J. M., & DiCarlo, L. A.
Athens, Greece.
(1991). A comparison of four new time-domain
Sandoe, E., & Sigurd, B. (1991). Arrhythmia: techniques for discriminating monomorphic
A guide to clinical electrocardiology. Bingen, ventricular tachycardia from sinus rhythm us-
Germany: Publishing Partners Verlags GmbH. ing ventricular waveform morphology. IEEE
Senhadji, L., Carrault, G., Bellanger, J. J., &
38, 561-570.
Passariello, G. (1995). Comparing wavelet
transforms for recognizing cardiac patterns. Tsipouras, M. G., & Fotiadis, D. I. (2004).
IEEE Engineering in Medicine and Biology Automatic arrhythmia detection based on time
Magazine, 14, 167-173. and time-frequency analysis of heart rate vari-
ability. Computer Methods and Programs in
Silipo, R., & Marchesi, C. (1998). Artificial
Biomedicine, 74, 95-108.
neural networks for automatic ECG analysis.
IEEE Transactions on Signal Processing, Tsipouras, M. G., Fotiadis, D. I., & Sideris, D.
46, 1417-1425. (2002). Arrhythmia classification using the RR-
interval duration signal. In A. Murray (Ed.),
Silipo, R., Taddei, A., & Marchesi, C. (1994).
Computers in cardiology (pp. 485-488).
Continuous monitoring and detection of ST-T
Piscataway: IEEE.
144
Tsipouras, M. G., Fotiadis, D. I., & Sideris, D. which is expressed through interunit connec-
(in press). An arrhythmia classification system tion strengths, and can make this knowledge
based on the RR-interval signal. Artificial In- available for use.
telligence in Medicine.
Data-Mining Association Rules: The
Vila, J., Presedo, J., Delgado, M., Barro, S., process of discovering valuable information
Ruiz, R., & Palacios, F. (1997). SUTIL: Intel- from large amounts of data stored in databases,
ligent ischemia monitoring system. Interna- data warehouses, or other information reposi-
tional Journal of Medical Informatics, 47, tories is called data mining. Association-rule
193-214. mining is the discovery of association relation-
ships or correlations among a set of items.
Wang, Y., Zhu, Y. S., Thakor, N. V., & Xu, Y.
H. (2001). A short-time multifractal approach Genetic Algorithm: Genetic algorithms
for arrhythmia detection based on fuzzy neural are derivative-free, stochastic optimization
network. IEEE Transactions on Biomedical methods based loosely on the concepts of natu-
Engineering, 48, 989-995. ral selection and evolutionary processes.
Yang, T. F. Device, B., & Macarlane, P. W. Heart-Rate Variability: The alterations
(1994). Artificial neural networks for the diag- of the heart rate between consecutive heart-
nosis of atrial fibrillation. Medical & Biologi- beats.
cal Engineering and Computing, 32, 615-
Ischaemic Episode: An ischaemic epi-
619.
sode is defined as a time period of no less than
Zhang, X. S., Zhu, Y. S., Thakor, N. V., & 30 seconds containing ischaemic beats.
Wang, Z. Z. (1999). Detecting ventricular ta-
RR-Interval Signal: The signal produced
chycardia and fibrillation by complexity mea-
from the durations between consecutive R
sure. IEEE Transactions on Biomedical En-
waves.
gineering, 45, 548-555.
Sensitivity: The sensitivity of a test defines
the probability that the test will be positive
KEY TERMS
(pathologic) when the outcome is positive
(pathologic).
Artificial Neural Network: An artificial
neural network is a massive parallel, distributed Specificity: The specificity of a test defines
processor made up of simple processing units. the probability that the test will be negative (nor-
It has the ability to learn from knowledge, mal) when the outcome is negative (normal).
145
147
Chapter XIX
Information Processing in
Clinical Decision Making
Vitali Sintchenko
University of New South Wales, Sydney, Australia
ABSTRACT
This chapter outlines an information-processing model of clinical decision-making which is

described as a function of the task, the decision maker, and the context. Attributes of the task,
the decision maker, and the decision environment are highly interrelated and often
interdependent. They directly affect the use of clinical evidence. We argue that information
processing is modified significantly by the decision-making context and decision task
characteristics. Knowledge of clinical decision-making is therefore becoming increasingly
important when designing an intervention that will produce sustained behavioural change. An
exploration of the context and information seeking aspects of prescribing is emerging as a
first step towards building the concept of task-specific decision support design.
BACKGROUND First, the decision maker must seek information

cues from the environment, such as clinical
The information-processing approach focuses signs and symptoms, medical history, or results
on information seeking and use as the key of diagnostic investigations. Selective attention
attribute in any decision making. It describes based on beliefs of the individual clinician plays
decision making as a process of information a critical role in the filtering of which cues to
inputs and outputs, and identifies parts of deci- process and which to ignore (Wickens &
sion making that may benefit from decision Hollands, 1999).
support (Elson, Faughnan, & Connelly, 1997). The information cues selected form the
An information-processing model of clinical basis for the situation assessment, which in-
decision making explaining the information flow cludes the identification of decision goals, an
during a clinical encounter is shown in Figure 1. assessment of how critical the problem is, and
Information Processing in Clinical Decision Making
Figure 1. An information-processing model of decision making (Modified from Wickens &

Hollands, 1999)
Decision context
Clinical
cues
Action Cue Selective

filtering attention
Information
Decision seeking Clinical Memory
assessment
Response
selection
the risks associated with possible outcomes, as tant possible outcome and their implicit utilities
well as a comparison of the result of this is more likely to occur for nonurgent healthcare
assessment with previous experience. decisions, but resource-saving reasoning heu-
It is believed that clinical assessment or ristics are often applied by a clinician who is
problem solving is based on external informa- examining an acutely ill person in an emergency
tion cues and knowledge stored in working and department where there is time pressure on the
long-term memories, and that medical exper- staff (Kushniruk, 2001; Patel & Kaufman,
tise is based more on knowledge than on expert 1998).
reasoning. The concept of the mental schema, During clinical problem solving, a number of
or clusters of related information that can be diagnostic hypotheses and management op-
accessed rapidly and then utilised for decision tions are generated. This number is limited by a
making, has been introduced to explain how person’s short-term memory. Relevant mental
humans acquire and store information. The schemas or clusters of related information avail-
number, size, and range of mental schemas or able are reviewed and matched. The matching
illness scripts, and the ability to retrieve and of schemas seems to be more rapid than
apply this information correctly form the foun- hypothetico-deductive reasoning as it imposes
dation on which expertise resides. less of a cognitive load than reasoning (Schmidt,
Research from a number of domains has Norman, & Boshuizen, 1990). However, when
demonstrated the importance of situation as- there is no acceptable match between selected
sessment as it assists in matching the type of cues and mental schemas, more cognitively
reasoning used by a decision maker to the demanding reasoning strategies are applied. If
characteristics of the task. A medical practitio- data gathering and processing steps of clinical
ner, for example, uses different decision-mak- assessment provide insufficient information for
ing strategies to manage a patient depending on selecting a response, then the decision maker
the clinical problem and the urgency of the task. may seek additional information cues and use
For instance, the assessment of every impor- decision-support tools. The alternative options
148
are then considered and assessed according to also been identified as potential contributors to
their values and cost, and the one with the suboptimal decision making (Elstein, 1999). The
highest value-cost ratio is usually selected. information-processing model describes deci-
sion making as a function of the task, the
decision maker, and the context. Each of these
CLINICAL DECISION MAKING can be characterised by a range of variables
that potentially can modify their relative impact
Information seeking is one of the most impor- on the decision making. Specifically:
tant aspects of information processing. It con-
sists of uncertainty-resolution stages as the • Problem or task attributes include differ-
decision maker moves from problem recogni- ences in decision variables (e.g., number
tion and assessment to the generation of alter- of decision alternatives) and information
natives and action (Wilson, 1997). Evidence cues,
suggests that information seeking can be • Decision maker attributes represent dif-
characterised by personal information-seeking ferences in skills associated with different
styles reflecting personal attitudes and beliefs. levels of task knowledge and cognitive
For example, it has been shown that clinicians abilities, and
actively pursue only about a third of the clinical • Context attributes recognise variations in
questions they have (Wyatt, 2000). These styles the decision environment that could influ-
are embedded in the activity that generates ence decision making.
information seeking and are responsive to in-
formation needs (Wilson, 1997). The informa- The interaction of these attributes influ-
tion-seeking style may be viewed as an inter- ences the choice of the strategy used for mak-
mediary variable that integrates different ating decisions. This interaction can also affect
tributes of a decision maker and determines his the information-seeking behaviour employed
or her information-seeking behaviour. for the task. Therefore, these attributes and
Decision makers often opt to seek additional their potential impact on clinical decisions need
information as a part of their decision making. to be reviewed in more detail.
Evidence suggests that information seeking is
an integral part of a particular decision strat- Patient- and Disease-Specific Task
egy. Tasks that require the processing of large Attributes
amounts of information in a very short period of
time tend to induce less analytical processing Clinical decisions are shaped by multiple fac-
(Kushniruk, 2001). The main objective of any tors and can be associated with variables re-
decision strategy chosen is to ensure the opti- lated to a patient and his or her disease. Several
mal choice and solution of the problem. Several patient-specific task attributes determine the
factors may open the problem-solving process selection of a particular management strategy,
to error, such as the uncertain nature of a task, such as the patient’s age, the duration and
lack of information to match correct schema, progression of symptoms, findings at examina-
insufficient knowledge, or cognitive biases tions, concerns about adverse outcomes if treat-
(Elstein, 1999). Cognitive errors (i.e., decisions ment is withheld, the prognosis, and the patient’s
that are inconsistent with people’s own prefer- and/or relatives’ expectations.
ences) and communication interruptions have
149
Disease-specific attributes also have an 1999). For example, two clinicians with differ-
impact on clinical decisions. For example, clini- ent risk-taking attitudes may manage the same
cians adjust the intensity of care in response to patient with a suspected infection with or with-
a patient’s condition and prognosis to assure out antibiotics. Importantly, an individual’s prac-
the longest possible survival time even when tice style shapes the content, sequence, and
the prognosis is poor. As Connors (1999, p. 5) format of information usage and persists over
has put it, “we are so afraid of not doing enough time (Davis, Gribben, Scott, & Lay-Yee, 2000;
that we often do too much.” In this context, the Long, 2002).
opportunity for errors in decision making is A large and systematic variation in clinical
high. decision making between different countries
Diagnostic uncertainty caused by the non- and healthcare systems, between and within
specific presentations of many life-threatening different clinical specialties, and between dif-
but treatable conditions is another significant ferent decision makers has been documented
modifier of clinical decisions. Second, the risk (Metlay, Shea, & Asch, 2002). This may reflect
of potential adverse drug events, malpractice variations in clinical training and decision mak-
litigation concerns, and sociocultural and eco- ing (Long, 2002; Metlay et al.), continuous
nomic pressures are usually considered (Brad- subspecialisation in healthcare, and differences
ley, 1991). All these factors may potentially in information sources used and/or beliefs (Gos-
affect the associated information seeking. ling, Westbrook, & Coiera, 2003). Differences
in an individual clinician’s practice style, risk
Important Attributes of a Clinician handling, and the intensity of care, as well as a
lack of consensus on the best practice and
Personal attributes of clinicians also shape their definitive studies identifying truly beneficial
decisions. Two phenomena that are of impor- interventions contribute to this problem. For
tance to any intervention that aims to optimise example, risk-averse physicians are more likely
clinicians’ decision making are the variability of to seek additional information before deciding
clinicians’ practice styles and information over- to treat or to do nothing, which increases the
load. use of resources. Evidence also suggests that
thresholds for adopting new therapies differ
Practice Styles Contribute to the between clinical specialties (Metlay et al.).
Variation in Decision Making Existing variations also create challenges for
evaluators of healthcare improvement inter-
Clinicians do not always behave in accordance ventions.
with normative principles of decision analysis. There have been suggestions that these
They do not make their decisions on a purely variations are due to the differences in clini-
rational basis but often deviate from the ideal cians’ knowledge and experience in evidence
decisions in a number of ways and show distinct evaluation (Kushniruk, 2001), and, therefore,
decision-making styles (McKinlay, Lin, Freund, access to specific knowledge may improve
& Moskowitz, 2002). Evidence suggests that clinical decision making (Elstein, 1999; Long,
clinical decisions depend more on a clinician’s 2002). However, this view has been challenged
behaviour than on the clinical picture because by research observations that better knowl-
of personal differences in the evidence-evalu- edge in a specific domain did not ensure better
ation stage of the decision process (Connors, decision making.
150
Information Overload Challenges Context Attributes

Decision Quality
Background context (e.g., time pressure, body
As in many other areas of modern medicine, of knowledge available about different clinical
evidence and knowledge are constantly being syndromes, diagnostic procedures, etc.) makes
reexamined, and the half-life of the truth of each clinical episode unique. Contextual fac-
evidence is limited. A mismatch between the tors also include specific clinical roles of
increasing amount of information available to healthcare decision makers, the number of
clinicians and human cognitive resources has decisions they are working on simultaneously,
led to information overload. the perceived importance of each decision, the
Information overload is one of the most established interaction patterns in a healthcare
pronounced phenomena of current clinical prac- team, and the existing information environ-
tice in general and prescribing in particular. For ment.
an individual, information overload means a Decision context includes both the physical
perception that the information associated with and the social context of decision making. Each
a task is greater than that which can be man- decision setting presents a particular mix of
aged effectively. This overload creates a de- resources such as people, tools, and events that
gree of stress for which the individual’s coping can be used to help to solve a problem and
strategies are ineffective (Coiera, 2000; Yu & determine the way in which such resources are
Chiang, 2002). For the healthcare system, in- used. It has been suggested, based on simula-
formation overload reduces the overall effec- tion analyses, that context variables may have
tiveness of routine healthcare procedures. Taken a larger effect on decision making than task
together, information overload leads to decision variables (Chu & Spires, 2000). There is a
makers’ thought patterns becoming more com- perception among many clinicians that they
plex, diverse, and less predictable (Yu & have less time available than in the past be-
Chiang). Importantly, decision quality suffers cause of increased patient volumes, a greater
under situations of information overload as it demand for documentation, and the increasing
makes doctors’ reasoning vulnerable to a num- complexity of modern practice. For example, in
ber of cognitive biases (Ayton & Pascoe, 1995; the critical-care setting, the prescribing task is
Elstein, 1999); clinicians may have problems performed under significant time pressure be-
with the effective integration of complex infor- cause of the severity of patients’ conditions and
mation, even if all the necessary information is the possible impact of delayed actions (Elstein,
available upon which to make a valid decision. 1999).
Context of Decision Context as a Decision Modifier
While the different types of information-pro- There are three important aspects of decision
cessing strategies used by people when making context that may affect clinical decision mak-
decisions have been extensively studied ing. First, every clinical decision occurs in a
(Wickens & Hollands, 1999), the role of the patient-specific context. Essentially, decision
decision context has only recently received making is a judgment task in which a clinician
some attention. makes judgments about the value of different
151
criteria (e.g., positive pathology report) based terns and the association of nonclinical factors
on the value of information sources in a particu- with decision choice suggest opportunities to
lar environment. The clinician must make judg- improve clinical decision making.
ments and take actions in a fast-paced, dy- We argue that information processing is
namic environment with information that is modified significantly by the decision-making
often incomplete or uncertain. context and decision task characteristics.
Lastly, evidence suggests that the social Knowledge of clinical decision making is there-
structures of healthcare teams are important fore becoming increasingly important when
contributors to the uptake of information tech- designing an intervention that will produce sus-
nology (Coiera, 2004; Gosling et al., 2003). It tained behavioural change. An exploration of
seems that participants playing different roles the context and information-seeking aspects of
in an organization have different information prescribing is emerging as a first step toward
requirements. Within a functional group such building the concept of task-specific decision-
as consultant physicians, or across roles within support design.
a subspecialty, different teams such as inten-
sive-care, cardiology, or infectious-disease com-
munities may develop, each having its own REFERENCES
shared understanding of events and ideas, shared
vocabulary, and shared knowledge. Each group Ayton, P., & Pascoe, E. (1995). Bias in human
looks at a process from a different perspective, judgment under uncertainty? The Knowledge
has different skill requirements, and defines the Engineering Review, 10(1), 21-41.
context by their own interest in the process. For
Bradley, C. P. (1991). Decision making and
example, a consultant may be interested in
prescribing patterns: A literature review. Fam-
high-level information like the clinical unit’s
ily Practice, 8(3), 276-287.
performance indicators while a registrar or
resident in training needs information support Chu, P. C., & Spires, E. E. (2000). The joint
for his or her basic prescribing decisions. Such effect of effort and quality on decision strategy
contextual specifications should be included in choice with computerised decision aids. Deci-
system requirements. sion Sciences, 31, 259-292.
Coiera, E. (2000). When conversation is better
than computation. Journal of American Medi-
CONCLUSION
cal Informatics Association, 7(2), 277-286.
Attributes of the task, the decision maker, and Coiera, E. (2004). Four rules for the reinvention
the decision environment are highly interrelated of health care. British Journal of Medicine,
and often interdependent. Their interaction can 328, 1197-1199.
only be assessed in the context of a specific
Connors, A. F. (1999). The influence of prog-
decision. The decision maker’s concepts and
nosis on care decisions in the critically ill.
use of evidence are very dynamic and can also
Critical Care Medicine, 27(1), 5-6.
depend on the user’s knowledge and personal
beliefs. Observed practice variation may also Davis, P., Gribben, B., Scott, A., & Lay-Yee,
be indicative of undesirable treatment prefer- R. (2000). Do physician practice styles persist
ences. Furthermore, variation in practice pat- over time? Continuities in patterns of clinical
152
decision-making among general practitioners. Schmidt, H. G., Norman, G. R., & Boshuizen,
Journal of Health Service Research Poli- H. P. (1990). A cognitive perspective on medi-
cies, 5(4), 200-207. cal expertise: Theory and implications. Aca-
demic Medicine, 65, 611-621.
Elson, R. B., Faughnan, J. G., & Connelly, D. P.
(1997). An information process view of infor- Wickens, C. D., & Hollands, J. G. (1999).
mation delivery to support clinical decision Engineering psychology and human perfor-
making: Implications for system design and mance (3 rd ed.). Upper Saddle River, NJ:
process measures. Journal of American Medi- Prentice Hall.
cal Informatics Association, 4(3), 266-278.
Wilson, T. (1997). Information behaviour: An
Elstein, A. S. (1999). Heuristics and biases: interdisciplinary perspective. In P. Vakkari, R.
Selected errors in clinical reasoning. Academic Savolainen, & B. Dervin (Eds.), Information
Medicine, 74, 791-794. seeking in context (pp. 39-52). London: Gra-
ham Taylor.
Gosling, A. S., Westbrook, J. I., & Coiera, E.
W. (2003). Variation in the use of online clinical Wyatt, J. C. (2000). Clinical questions and
evidence: A qualitative analysis. International information needs. Journal of Royal Society
Journal of Medical Informatics, 69(1), 1-16. of Medicine, 93(2), 168-171.
Kushniruk, A. W. (2001). Analysis of complex Yu, P. L., & Chiang, C. I. (2002). Decision
decision-making processes in health care: Cog- making, habitual domains and information tech-
nitive approaches to health informatics. Jour- nology. International Journal of Informa-
nal of Biomedical Informatics, 34(5), 365- tion Technology and Decision Making, 1(1),
376. 5-26.
Long, M. J. (2002). An explanatory model of
medical practice variation: A physician re- KEY TERMS
source demand perspective. Journal of Evalu-
ation in Clinical Practice, 8(2), 167-174. Context of a Decision: The context of any
decision task can be defined as the collection of
McKinlay, J. B., Lin, T., Freund, K., &
surrounding influences that make a situation
Moskowitz, M. (2002). The unexpected influ-
unique and comprehensible.
ence of physician attributes on clinical deci-
sions: Results of an experiment. Journal of Information Impact: Modifications in a
Health and Social Behavior, 43, 92-106. practice or protocol.
Metlay, J. P., Shea, J. A., & Asch, D. A. Information Utilisation: Applying infor-
(2002). Antibiotic prescribing decisions of gen- mation to make a difference in the thoughts and
eralists and infectious disease specialists: actions of users.
Thresholds for adopting new drug therapies.
Information Utility: The relevance of in-
Medical Decision Making, 22(6), 498-505.
formation to a policy or protocol.
Patel, V., & Kaufman, D. (1998). Medical
Situation Assessment: The process of
informatics and the science of cognition. Jour-
the identification and clarification, by the deci-
nal of American Medical Informatics Asso-
sion maker, of the decision problem.
ciation, 5(6), 493-502.
153
154
Chapter XX
Data Mining Techniques and
Medical Decision Making for
Urological Dysfunction
N. Sriraam
Multimedia University, Malaysia
V. Natasha
H. Kaur
ABSTRACT
Data mining has been emerging recently as a viable computational tool for autonomous
decision making especially in the field of medical applications. It has provided diagnostic
solutions for skin and breast cancer detection, brain tumor detection, and also for other
classification problems. In this chapter, we explore two data mining techniques, namely,
association mining and decision tree mining, for predicting the life span of the kidney failure
patients who have undergone routine dialysis. The total parameters used for this study were
28 attributes. The optimal prioritized parameters that decide the survival rate are reported and
it can be concluded from the experimental results that the decision tree approach yields
promising results.
INTRODUCTION in medical fields such as kidney dialysis, skin

cancer and breast cancer detection, and also
The process of extracting useful information biological sequence classification (Fernando,
from a set of data is called data mining. Data Juan, & Angel, 2002; Krzysztof & William,
mining techniques have been used as a recent 2002; Kusiak, Dixon, & Shah, 2005; Linhua et
trend for gaining diagnostics results, especially al., 2004). It is well known that the primary
Data Mining Techniques and Medical Decision Making for Urological Dysfunction
problem for urological dysfunction is acute and of water and electrolytes to balance the changes
chronic renal failure, which can be treated (Sherwood, 1993). For a person with kidney
through dialysis. This chapter describes the failure, waste products that are produced from
data mining techniques for predicting the life food tissue, namely, urea and creatinine, start
span of a kidney dialysis patient. Since we are accumulating in the body, and hence the fluid
interested in predicting the output for each level and water homeostasis will be imbalanced.
individual, data mining tools have been opted as This eventually leads to endocrine failure and
the decision making tools due to their individual- results in death.
based functionality compared to other analysis Dialysis refers to any medical treatment
tools, which are population based. They offer a that aims at replacing normal kidney function by
valid, novel, and helpful solution for the identi- artificial means. The treatment is prescribed
fication of patterns of data, and also develop for patients with end-stage renal failure (ESRF).
high-confidence predictions for each individual. Dialysis is an effective life-saving treatment.
Given the predicted outcome, more and better Without dialysis, the life expectancy for a pa-
treatments can be made available for every tient with ESRF is less than a year. Life expect-
individual. Furthermore, the resulting output ancy on treatment may be as long as 16 years,
serves as a major contribution to medical-care depending on the age and health status of the
centres in providing enhanced treatment to patients. There are about 370,000 kidney pa-
kidney-dialysis patients to prolong their life tients who are undergoing treatment for dialy-
spans. Therefore, it is hoped that the applica- sis, and its annual cost is $11.1 billion in the
tion of these techniques could provide us a United States of America (Kusiak, Dixon, et
rough estimation for the survival prediction of al., 2005; U.S. Renal Data System [USRDS],
dialysis patients based on certain weighed pa- 2002). Even though the number of kidney pa-
rameters. About 29 parameters were consid- tients has been rising yearly with a growth rate
ered as data from the database of kidney pa- of 6%, little attention is given to kidney health.
tients to provide an effective estimate of one’s Out of the figure mentioned above, the total
survival length after kidney failure. number of patients suffering from chronic renal
failure is 260,000, and about 50,000 patients die
yearly (Cooper, 1999; Kusiak, 2004; USRDS).
KIDNEY FAILURE AND DIALYSIS When the kidney is functioning at less than 50%
of its normal capacity, it eventually leads to
The kidney plays an important role in the body chronic renal failure. In the end stage of renal
due to its basic functionality of processing all failure, kidney function is at less than 10 to 15%
toxic waste together with excess water and of normal capacity. In Malaysia at the end of
salt. Generally, it is an organ that filters about 2002, a total of 2,223 patients were accepted
189 liters of liquid from the blood (1% original for dialysis compared to 43 patients in 1980, and
filtrate, which appears in the final urine as prevalent dialysis patients increased rapidly
waste product and water). These waste prod- from a total 59 in 1980 to 8,954 in 2002 (Zaki et
ucts that are produced from food tissue are al., 2003). The acceptance rate for dialysis
urea and creatinine. However, its main ability is increased very rapidly from 3 per 1 million to 91
to retain a proper stability of extra cellular fluid per 1 million in 2002. Death rates for
(ECF) and electrolyte homeostasis. This pro- haemodialysis have remained at 10% or lower
cess can be done by maintaining the secretion per year throughout the years from 1980 to
155
Figure 1. Schematics diagram for data-mining techniques for predicting the survival period
of kidney-failure patients
2003; CAPD death rates were higher at 10 to trees (DTs), association rules, neural networks,
20% (Zaki et al.). Based on research, it has classification rules, and the k-nearest algorithm
been found that Malaysian patients who have for handling medical and nonmedical applica-
undergone kidney transplants have a better tions. The main objective of this research chap-
chance of survival compared to patients who ter is to develop an automatic decision making
just undergo dialysis. tool through data mining to predict the survival
period of the kidney failure patient undergoing
dialysis. Survival analysis that uses standard
DATA MINING TECHNIQUES statistical tools, for example, logistic regres-
sion, Cox’s model, and factorial designs, were
Data mining, also called knowledge discovery found to be population-based models whereas
in databases, is defined as the nontrivial extrac- data mining tools provide decision making for
tion of implicit, previously unknown, and poten- individual patient predictions, which are ob-
tially useful information from data. It involves tained from the probability of or the distance
the identification of patterns or relationships in from the population estimates (Kusiak, Dixon,
data. There are several data mining techniques et al., 2005). For survival prediction problems,
available in the literature (Antonie, Zaïane, & decision trees and rough-set trees have been
Coman, 2001; DeClaris, Shalvi, Duong, & Luu, used with limited parameters (Kusiak, Dixon, et
1996; Hogl et al., 2001; Inanda & Terano, 2002; al.; Shah, Kusiak, & Dixon, 2003). In our
Kusiak, Dixon, et al., 2005; Kusiak, Kern, experimental work, we have used more param-
Kernstine, & Tseng, 2000) such as decision eters for evaluation purposes using association
156
Table 1. Parameter priority loss, and blood type as they have less influence
on the survival length prediction. Due to incon-
No Parameter
1. Kt/V sistent readings, parameters such as type of
2. URR access, type of machine, data transfer in, data
3. Blood Flow Rate
4. Dialysate Flow rate transfer out, and glucose were excluded for this
5. Venous Pressure
6. Bicarbonate
experimental study.
7. Potassium Clustering is a method with the objective to
8. Total Blood Volume
9. Time Dialysis find natural groupings called clusters. The natu-
10. Diff_Pst_PreSupines ral groupings should be in a large dimensional
11. Diff_Pst_PreStands
12. Diff_PreSupine_Stand_S set. Clustering plays a major role in data mining
13. Diff_Pst_PreSupine_D applications such as text mining, medical diag-
14. Diff_Pst_PreStand_D
15. Diff_PreSupineStand_D nosis, and computational biology (Aslandogan
16. Pulse_Stand_Pre
17. Pulse_Stand_Post et al., 2004; Dandekar et al., 2001; Krzysztof &
18. Pulse_Supine_Pre William, 2002). The objects are clustered to-
19. Pulse_Supine_Post
gether if they are similar to one another accord-
ing to certain measures, and further clustering
is done with dissimilar objects. The parameters
mining and decision tree mining. Figure 1 shows that were obtained from the medical centre had
the overview of our work. some loops in which some data were not filled
in as the data were missing in the centre itself.
Such parameters included standing blood pres-
DATA CLUSTERING sure, supine blood pressure, venous pressure,
blood type, date of transfer into the medical
In order to evaluate the effectiveness of the centre, and date of the transfer out. As only
data mining algorithm, 182 dialysis data sets certain data were left in a loop, clustering was
were used for training and 66 data sets for applied to these parameters to decide their
testing. The test data belonged to new patients values. The values that have been clustered fall
who had not undergone dialysis. The prepro- under three categories, namely, high, low, and
cessing steps start with the classification of medium (normal), which have the values 2, 0,
dialysis parameters into three modes based on and 1, respectively. Clustered data have these
clinical condition, namely, priority mode, significant values to represent them for deci-
nonpriority mode, and excluded parameter. Data sion making. High and low values are defined
collection is based on known as well as un- as:
known indicators (parameters). Dialysis pa-
tients received the treatment three times a Low< Normal < High, (1)
week, and readings were listed before and after
the dialysis. The priority mode contains the where a low value is below the average normal
parameter that had more influence on making value, and a high value is the above the average
an effective decision on the survival period of normal value. The range of values that are
the patient. Table 1 shows the details, in which categorized into low, medium, and high levels
the higher the number, the lower the parameter for the dialysis parameters is shown in Table 2.
priority (as defined by clinicians). Nonpriority These values have been derived based on medi-
parameters consist of age, sex, diagnosis, weight cal experts’ feedback.
157
Table 2. Clinical ranges of dialysis mate of the survival period for the kidney
parameters dialysis patient.
After all the data have been clustered to
Parameters Low Medium High
significant values of H, M, and L, or 2, 1, and 0,
Blood pressure <120 120 - 140 > 140
(systolic) the parameters are framed into a table. Once
Blood pressure < 70 70 – 90 > 90 the parameters are framed, the decision is
(diastolic)
calculated. Decisions are determined as such:
Pulse rate <60 60 – 90 > 90
Time of dialysis <4 4 ( 3 x 52) >4
(hour) Decision (D) = X and Y, (2)
Blood flow rate < 250 250 – 300 > 300
( ml )
where X and Y are prioritised dialysis param-
Venous Pressure < 100 100 – 150 > 150
Bicarbonate < 66.0 66.0 – 66.08 > 66.08 eters.
(g) The preprocessed data sets are further ap-
Potassium <5.20 5.20 – 5.228 > 5.228
(g) plied to the data mining algorithm.
Total blood < 70 % 70 % - 74% > 74%
volume
Kt/v < 1.4 1.4 – 1.6 > 1.6
URR < 72 72 – 73 > 73 EVALUATION
Dialysate flow < 500 500 > 500
rate
In order to determine the survival period of the
kidney failure patient using the training data
sets, the following processes are carried out.
It can be observed from Table 1 that certain
parameter (10- to 15) values are framed based 1. Association mining of the data sets
on certain measures. For example, consider the 2. Determination of classification accuracy
parameter Diff_Pst_PreSupine_S. 3. Prediction of survival length by decision
This parameter is based on the difference tree mining
between the postreading for systolic supine 4. Testing the mining algorithms with the
(sitting) and the prereading for systolic supine available test data (66 data sets)
(sitting). By using the simple operators, H – H
= H, Association rules are applied to the data
M- M = M, and L – L = L, the data can be sets belonging to Table 3 describing the events
clustered. After clustering, the prioritised pa- that tend to occur together. They provide infor-
rameters are combined as shown in Table 3 in mation about the if-then statement, which can
order to evaluate the effectiveness of the esti- be computed directly from the data that are
Table 3. Combinational prioritised dialysis parameters
Parameters (Total Parameters)
Case A Kt/V, Venous Pressure, Blood Flow Rate, Dialysate Flow Rate, URR (5)
Case B Bicarbonate, Potassium, Total Blood Volume, Time dialysis (4)
Case C Diff_Pst_PreSupine_S, Diff_Pst_PreStand_S, Diff_PreSupineStand_S,
Diff_PstSupineStand_S (4)
Case D Diff_Pst_PreSupine_D, Diff_Pst_PreStand_D, Diff_PreSupineStand_D,
Diff_Pst_SupineStand_D (4)
Case E Pulse_Supine_Pre, Pulse_Supine_Post (2)
158
probabilistic in nature. In association analysis, Rule 2

the if-then statements are sets of items that are
disjoint. The effectiveness of the association IF Kt/V, URR, Blood flow rate, Dialysate flow
rule is relatively based on two indicators, namely, rate, Venous pressure AND Diff_Pst_
support and confidence. Support is simply the PreSupine_D, Diff_Pst_PreStand_D,
number of transactions that include all param- Diff_Pre_Supine_Stand_D, Diff_Pst_
eters in the if-then statement of the rule (i.e., Supine_Stand_D IS (>=76.47%) THEN CON-
probability that a randomly selected transaction SEQUENCES IS H (support(a)17;
from the database will contain all the items in support(c)35; 13; 0.895798)
the if-then part). Confidence is the ratio of the
number of transactions that include all param- Case C_D_E
eters in the if-then statement to the number of
transactions that include all the parameters in Rule 1
the if statement. After the association mining of
all possible cases, it was found that only 10 IF Diff_Pst_PreSupine_D, Diff_Pst_
combinational cases yielded clinically useful PreStand_D, Diff_Pre_Supine_Stand_D,
information. Rules obtained from these cases Diff_Pst_Supine_Stand_D AND Diff_Pst_
are considered for further investigation. Some PreSupine_D, Diff_Pst_PreStand_D, Diff_
of the rules obtained from association mining Pre_Supine_Stand_D, Diff_Pst_Supine_
are shown in the following sections. Stand_D AND Pulse_Supine_Pre, Pulse_
Supine_Post IS (>=76.67%) THEN CONSE-
Case A_B QUENCES IS H (support(a)30; support(c)33;
23; 0.952525)
Rule 1
Rule 2
IF Kt/V, URR, Blood flow rate, Dialysate flow
rate, Venous pressure AND Diff_Pst_ IF Diff_Pst_PreSupine_D, Diff_Pst_
PreSupine_D, Diff_Pst_PreStand_D, Diff_ PreStand_D, Diff_Pre_Supine_Stand_D,
Pre_Supine_Stand_D, Diff_Pst_Supine_ Diff_Pst_Supine_Stand_D AND Diff_Pst_
Stand_D IS (>=77.78%) THEN CONSE- PreSupine_D, Diff_Pst_PreStand_D,
QUENCES IS H (support(a)9; support(c)38; 7; Diff_Pre_Supine_Stand_D,
0.839181) Diff_Pst_Supine_Stand_D AND Pulse_
Supine_Pre, Pulse_Supine_Post IS (>=66.67%)
Case A_C THEN CONSEQUENCES IS M
(support(a)12; support(c)30; 8; 0.911111)
Rule 1
Case A_B_C_D_E
IF Kt/V, URR, Blood flow rate, Dialysate flow
rate, Venous pressure AND Diff_Pst_ Rule 1
PreSupine_D, Diff_Pst_PreStand_D, Diff_Pre_
Supine_Stand_D, Diff_Pst_Supine_Stand_D IS IF Kt/V, URR, Blood flow rate, Dialysate
(>=81.25%) THEN CONSEQUENCES IS H flow rate, Venous pressure AND Bicarbon-
(support(a)16; support(c)35; 13; 0.951786) ate, Potassium, Total blood volume, time
159
Table 4. Classification accuracy obtained for dialysis-parameter combinations
Priority group Total Below Median Above Median

parameters
CASEA_B 9 65.22 50
CASEA_C 9 76.47 -
CASEA_D 9 - 76.47
CASEB_C 8 62.5 52.82
CASEB_D 8 - 58.33
CASEC_D 8 50 -
CASEC_D_E 10 58.33 66.67
CASEA_B_C_D 17 94.44 76.92
CASEA_B_C_D_E 19 97.44 88.89
dialysis AND Diff_Pst_PreSupine_S, Diff_ The mining process resulted in a list of rules
Pst_PreStand_S, Diff_Pre_Supine_Stand_S, whereby specific rules with good confidence
Diff_Pst_Supine_Stand_S AND Diff_Pst_ intervals were selected first according to the
PreSupine_D, Diff_Pst_PreStand_D, Diff_ maximum occurrence and the minimum occur-
Pre_Supine_Stand_D, Diff_Pst_Supine_ rence. The rules with the maximum and mini-
Stand_D AND Pulse_ Supine_Pre, mum occurrence are used to get the classifica-
Pulse_Supine_Post IS (>=94.44%) THEN tion accuracy. A good confidence interval is
CONSEQUENCES IS M (support(a)18; essential for this clinical study. Classification
support(c)39; 17; 0.992877) accuracy requires a good confidence interval to
obtain the number of correctly classified ob-
Rule 2 jects from the test set. The results of classifica-
tion accuracy are shown in Table 4
IF Kt/V, URR, Blood flow rate, Dialysate flow The DT is one of the most popular classifi-
rate, Venous pressure AND Bicarbonate, Po- cation algorithms in current use for data mining
tassium, Total blood volume, Time dialysis AND and machine learning. The DT algorithm cre-
Diff_Pst_PreSupine_S, Diff_Pst_PreStand_S, ates decision trees or sets of decision rules
Diff_Pre_Supine_Stand_S, Diff_Pst_Supine_ based on the concept of the information gained.
Stand_S AND Diff_Pst_PreSupine_D, A decision tree takes as input an object or
Diff_Pst_PreStand_D, Diff_Pre_Supine_ situation described by a set of properties, and as
Stand_D, Diff_Pst_Supine_Stand_D AND outputs a yes-no decision, thereby representing
Pulse_Supine_Pre, Pulse_Supine_Post IS Boolean functions. The basis of a decision tree
(>=94.12%) THEN CONSEQUENCES IS H is that the outcome will be 50-50. Predictions
(support(a)17; support(c)39; 16; 0.989442) for the survival length of kidney failure patient
are made using the decision tree algorithm.
160
Figure 2. Decision-tree mining for kidney dialysis
Thirty-two rules were generated by the DT Rule 2

algorithm based on two conditions: the survival
length above the median (H) and below the IF (Kt/V_URR>=1.6; 73) AND (Blood_flow_
median (M). Figure 2 shows the structure of a rate>= 300) AND (Dialysate_flow_rate>=500)
DT for kidney-dialysis parameters. AND (Venous_pressure>= 150) AND
H represents the low risk factor of survival, (Bi_Po<66.0 ; 5.20) AND (Total_blood_ vol-
and M indicates the high risk factor. The fol- ume<70.0%) AND (Time_dialysis<4) THEN
lowing are the exemplary rules generated by (Survival_length=Below_med); [83.33%]
DT mining (only a few are given).
Rule 3
Survival Rules Generated by the
Decision-Tree Algorithm IF (Kt/V_URR>=1.6; 73) AND (Blood_flow_
rate>=300) AND (Dialysate_flow_rate>=500)
Rule 1 AND (Venous_pressure>=150) AND
(Bi_Po>=66.08; 5.228) AND (Total_blood_
IF (Kt/V_URR>=1.6; 73) AND (Blood_flow_ volume>=74.0%) AND (Time_dialysis>=4)
rate>=300) AND (Dialysate_flow_rate>=500) AND (Diff_Pst_PreSupine_S>0.0) AND
AND (Venous_pressure>=150) AND (Bi_Po>= (Diff_Pst_PreStand_S>0.00) AND (Diff_Pre_
66.08; 5.228) AND (Total_blood_volume> Supine_Stand_S>0.0) AND (Diff_Pst_Supine_
=74.0%) AND (Time_dialysis>= 4 ) THEN Stand_S>0.0) AND (Diff_Pst_PreSupine_
(Survival_length=Below_med); [16.667%] D>0.0) AND (Diff_Pst_PreStand_D>0.00)
161
AND (Diff_Pre_Supine_Stand_D>0.0) AND rate>=500) AND (Venous_pressure>=150)

(Diff_Pst_Supine_Stand_D>0.0) THEN AND (Bi_Po<66.0; 5.20) AND (Total_
(Survival_length=Below_med); [3.5714%] blood_volume<70.0%) AND (Time_dialysis<4)
AND (Diff_Pst_PreSupine_S>0.0) AND
Rule 4 (Diff_Pst_PreStand_S>0.00) AND
(Diff_Pre_Supine_Stand_S>0.0) AND
IF (Kt/V_URR>=1.6; 73) AND (Blood_ (Diff_Pst_Supine_Stand_S>0.0) AND (Diff_
flow_rate>=300) AND (Dialysate_flow_ Pst_PreSupine_D<=0.0) AND (Diff_Pst_
rate>=500) AND (Venous_pressure>=150) PreStand_D<=0.00) AND (Diff_Pre_Supine_
AND (Bi_Po>=66.08; 5.228) AND Stand_D<=0.0) AND (Diff_Pst_Supine_
(Total_blood_volume>=74.0%) AND Stand_D<=0.0) AND (Pulse_Supine_Pre<=60)
(Time_dialysis>=4) AND (Diff_Pst_ AND (Pulse_Supine_Post<=60) THEN
PreSupine_S>0.0) AND (Diff_Pst_PreStand_ (Survival_length=Below_med); [4.1667%]
S>0.00) AND (Diff_Pre_Supine_Stand_S>0.0) The rules generated by the DT predict the
AND (Diff_Pst_Supine_Stand_S>0.0) AND survival length of the kidney failure patient in
(Diff_Pst_PreSupine_D<=0.0) AND (Diff_ two modes based on the combination of
Pst_PreStand_D<=0.00) AND (Diff_Pre_ prioritised dialysis parameters.
Supine_Stand_D<=0.0) AND (Diff_Pst_
Supine_Stand_D<=0.0) THEN (Survival_ • Survival length=Above_median. This
length=Below_med); [96.4286%] means the dialysis patient does not need
much attention from the clinician, and his
Rule 5 or her level of dialysis is sufficient for
survival.
IF (Kt/V_URR>=1.6; 73) AND (Blood_flow_ • Survival length=Below_median. This
rate>=300) AND (Dialysate_flow_rate>=500) indicates that the patient needs special
AND (Venous_pressure>=150) AND attention as he or she is at a high risk level.
(Bi_Po>=66.08; 5.228) AND (Total_blood_ Furthermore, the DT rules predict the
volume>=74.0%) AND (Time_dialysis>=4) survival period based on a threshold level
AND (Diff_Pst_PreSupine_S>0.0) AND of the parameters.
(Diff_Pst_PreStand_S>0.00) AND (Diff_Pre_
Supine_Stand_S>0.0) AND (Diff_Pst_Supine_ Rule 1 indicates that the survival length is
Stand_S>0.0) AND (Diff_Pst_PreSupine_ below median with a low chance of survival:
D<=0.0) AND (Diff_Pst_PreStand_D<=0.00) 16.6667%. Meanwhile, Rule 2 indicates that
AND (Diff_Pre_Supine_Stand_D<=0.0) AND the patient is also below median but does not
(Diff_Pst_Supine_Stand_D<=0.0) AND need much attention as his or her survival
(Pulse_Supine_Pre<=60) AND (Pulse_Supine_ chance is 83.33%, which is a fairly high per-
Post<=60) THEN (Survival_length=Below_ centage of survival. It can be concluded from
med); [96.6667%] the above two rules that the impact of param-
eters such as bicarbonate, potassium, total blood
Rule 6 volume, and time dialysis decides the survival
length and the level of risk. This holds good for
IF (Kt/V_URR>=1.6; 73) AND (Blood_ the rest of the rules obtained by DT mining
flow_rate>=300) AND (Dialysate_flow_
162
CLINICAL IMPORTANCE Example 1.
Priority – overall = 13 – 8
To further confirm the effectiveness of our Overall 8
evaluation scheme, we also used the test data to = 0.625 x 100 %
predict the survival length, and 88.5% of our = 62.5 %
results match with the clinician opinion. From a
clinical point of view, it can be concluded that
there are certain values that are significant in
deciding the survival prediction. The significant
• Type of access
parameters found from the prioritised param-
• Type of machine
eters are the following:
• Pulse rate
• Total blood volume
• Bicarbonate
• Total dialysis time
• Potassium
• Diagnosis
• Total blood volume
• Time of dialysis
The final trade-off of the significant param-
• Difference of post- and prereading in
eter is calculated as such (see Example 1).
supine systolic
standing systolic
CONCLUSION
• Difference of prereading in supine and
standing systolic
The most significant result obtained from this
• Difference of postreading in supine and
research was demonstrating that data mining,
standing systolic
data transformation, data partitioning, and deci-
sion making algorithms were useful for the
supine diastolic
survival prediction of dialysis patients. Analyz-
ing and comparing the data mining rules pro-
standing diastolic
duced a list of significant parameters such as
• Difference of prereading in supine and
bicarbonate, potassium, total blood volume, time
standing diastolic
of dialysis, difference of post- and prereading in
• Difference of postreading in supine and
supine systolic, difference of post- and
standing diastolic
prereading in standing systolic, difference of
• Pulse rate
prereading in supine and standing systolic, dif-
ference of post- and prereading in supine dias-
The survival length rule generated was mined
tolic, difference of prereading in supine and
using data mining algorithms on priority data.
standing diastolic, difference of postreading in
When all the data were mined using the data
supine and standing diastolic, and pulse rate. A
mining algorithm, the significance parameters
conservative approach was applied while han-
were found, as shown here:
dling these cases, whereby the outcome below
median was assigned to patients who had a
• Dialysate flow rate
shorter survival time compared to patients who
• Bicarbonate
were above median. Below median patients
163
were those who had been placed on the trans- Krzysztof, J. C., & William, M. G. (2002).
plant list, whose condition was deteriorating, Uniqueness of medical data mining. Artificial
and whose chances of surviving were small. H Intelligence in Medicine, 26, 1-24.
and M represented patients with a higher chance
Kusiak, A., Dixon, B., & Shah, S. (2005).
of survival and patients who had a lower chance
Predicting survival time for kidney dialysis pa-
of survival, respectively. The outcome resulted
tients: A data mining approach. Computers in
in the conclusion that DT-generated rules con-
Biology and Medicine, 35(4), 311-327.
tain more significant parameters than the rules
produced by association rules. Therefore, DTs Kusiak, A., Kern, J. A., Kernstine, K. H., &
produce higher trade-offs compared to the Tseng, B. T. L. (2000). Autonomous decision-
latter. The medical relevance of significant making: A data mining approach. IEEE Trans-
parameters was established. The final trade- actions on Information Technology Biomedi-
off of the significant parameter was 62.5%. cine, 4(4), 274-284.
Lihua, L., Tang, H., Wu, Z., Gong, J., Gruidl,
M., Zou, J., et al. (2004). Data mining tech-
REFERENCES
niques for cancer detection using serum
proteomic profiling. Artificial Intelligence in
Antonie, M. L., Zaïane, O. R., & Coman, A.
Medicine, 32, 71-83.
(2001). Application of data mining techniques
for medical image classification. Proceedings Shah, S., Kusiak, A., & Dixon, B. (2003). Data
of the Second International Workshop on mining in predicting survival of kidney dialysis
Multimedia Data Mining (pp. 94-101). patients. Proceedings of Photonics West -
Bios, Lasers in Surgery: Advanced Charac-
Cooper, J. (1999). US incidence of kidney
terization, Therapeutics, and Systems XIII,
failure is the highest in the world. The medical
4949 (pp. 1-8).
reporter. Retrieved from http://
medicalreporter.health.org/tmr0799/ Sherwood, L. (1993). Human physiology:
kidney.html From cells to systems (3rd ed.). Wadsworth
Publishing Company.
DeClaris, N., Shalvi, D., Duong, T., & Luu, T.
(1996). Computational intelligence-based meth- U.S. Renal Data System (USRDS). (2002).
odologies for population studies and laboratory USRD 2002 annual data report: Atlas of end
medicine decision aids. Proceedings of the stage renal disease in the United States.
International Neural Network Society. Bethesda, MA: National Institutes of Health,
National Institutes of Diabetes and Digestive
Fernando, A., Juan, P. C., & Angel, L. (2002).
and Kidney Diseases.
Combining expert knowledge and data mining
in a medical diagnosis domain. Expert Systems Zaki, M., Shaariah, W., Liu, W. j., Hooi, L. S.,
with Applications, 23, 367-375. Goh, B. L., Philip, N. J., et al. (2003). Eleventh
report of the Malaysian Dialysis & Trans-
Inada, M., & Terano, T. (2002). Interactive
plant Registry.
data mining from clinical inspection data. IEEE
Conf Sys Man and Cyber, 4, 6.
164
KEY TERMS Data Mining: The process of analyzing

data to identify patterns or relationships.
Association Mining: Association rule min-
Decision Tree: A decision tree partitions
ing determines the correlation relationships
data into smaller segments called terminal nodes
among large sets of data and shows attribute
or leaves that are homogeneous with respect to
value conditions that occur frequently together
a target variable.
in a given data set.
Dialysis: Dialysis is a method of removing
Clustering: Clustering is the classification
toxic substances (wastes) from the blood when
of similar objects into different groups, or dis-
the kidneys are unable to function properly.
tributing data sets into clusters or subsets.
Renal Failure: Sudden and often tempo-
rary loss of kidney function.
165
166
Chapter XXI
Spline Fitting
Michael Wodny
Ernst-Moritz-Arndt-University, Germany
ABSTRACT
Given are the m points (xi,yi), i=1,2,…,m. Spline functions are introduced, and it is noticed that
the interpolation task in the case of natural splines has a unique solution. The interpolating
natural cubic spline is constructed. For the construction of smoothing splines, different
optimization problems are formulated. A selected problem is looked at in detail. The
construction of the solution is carried out in two steps. In the first step the unknown Di=s(xi)
are calculated via a linear system of equations. The second step is the construction of the
interpolating natural cubic spline with respect to these (xi,D i), i=1,2,…,m. Every optimization
problem contains a smoothing parameter. A method of estimation of the smoothing parameter
from the given data is motivated briefly.
INTRODUCTION properties. Their application is widespread.

There are multidimensional splines and many
Model fitting in data mining requires the atten- mathematical generalizations, too, especially
tion of at least three aspects: the data, the model on Hilbert spaces.
to be fitted, and the optimization criterion. This Spline fitting involves the calculation of the
general situation is specified for simplicity as parameters of the chosen spline function from
follows. The data are a two-dimensional set of the given data. The data, the class of the
real numbers (xi,yi), i = 1, 2, ..., m, and the model functions, and the optimization criterion deter-
is a class of spline functions. Selected optimiza- mined the calculation method for the param-
tion criteria are subsequently explained. eters of the spline function.
Spline functions are a class of functions that Let (xi,yi), i = 1, 2, …, m be given. We
is characterized by general mathematical prop- assume that x1<x2<x3<...<xm.
erties, instead of data-driven or problem-driven
Spline Fitting
DEFINITION generalize the afterward-derived calculation

procedure of the parameters of a natural cubic
A spline function s(x) of degree n is a func- spline to arbitrary natural splines.
tion defined on Ñ. s(x) is given by some Looking at the definition, a cubic spline can
polynomial of degree n or less in each of the be written in the interval [x i,xi+1] as:
intervals (-∞,x 1], [x i,x i+1 ], i = 1, 2, …, m-1,
and [x m,+∞). All derivatives of s(x) up to s(x) = A i(x-xi)3 + Bi(x-xi)3 + Ci(x-xi) +Di.
order n-1 are supposed to be continuous every-
where. s(x) is called a natural spline if n = 2k- Obviously,
1 is odd and s(x) is given in each of the intervals
(-∞,x1] and [x m,+∞) by polynomials of degree s(x i) = Di = yi, i=1, 2, ..., m-1. (1)
k-1 or less.
For example, the class of all splines of Furthermore, it follows that s”(x i) = 2Bi, and
degree n withthe knots xi includes all polynomi- so:
als of degree n or less.
B i = s”(xi)/2. (2)
INTERPOLATION The second derivation s"(x) is a straight line

with the slope (s"(x i+1)-s"(xi))/(xi+1-xi). Conse-
A special task of curve fitting is the interpola- quently, one obtains:
tion problem. One looks for a function f(x)
satisfying the strong conditions f(xi) = y i for all A i = 1/3(Bi+1-Bi))/(xi+1-xi) (3)
i from 1 to m.
The interpolation problem has a unique solu- for all i = 1, 2, …, m-1.
tion for natural splines s(x). The main property The Ci can be represented with the data and
of natural splines is proved by deBoor (2001) the Bi as:
and Schoenberg and deBoor. In the case n = 3,
the result was given already by Holladay (1957). Ci=(yi+1-yi)/(xi+1-xi) - 1/3(xi+1-xi)(2 Bi+1+Bi)
(4)
THEOREM as seen in Equations (1), (2), and (3).

From these considerations it follows that an
Let s(x) be the interpolating natural spline of interpolating cubic spline is uniquely deter-
degree n = 2k-1, with respect to (xi, y i), i = 1, 2, mined and completely represented by (xi,s(x i))
..., m, and f(x) any interpolating function with and B i (especially Bm= s"(xm)/2). One uses the
continuous derivatives up to order k. Then continuity of s'(x) in the x i, i=1, 2, ..., m to
b b calculate the unknown Bi. Consequently:
∫ s ( x ) dx ≤ ∫ f ( x ) dx for all a ≤ x1 and b ≥ xm.
(k ) 2 (k ) 2
a a
s’(xi) = 3Ai-1(xi-xi-1)2 + 2Bi-1(xi-xi-1) + Ci-1 =
In the case where k > 1, the strong inequality Ci, i = 2, 3, ..., m-1
is valid.
We furthermore refer to natural cubic splines holds true. Short remodeling together with the
because they are widely used. It is possible to specifications ∆xi := xi+1-x i and ∆yi := yi+1-y i, i =
1, 2, ..., m-1 leads to:
167
Spline Fitting
Figure 1. An example of an interpolating natural cubic spline
 ∆ y i ∆ y i −1  (OP1).
∆ x i −1 B i −1 + 2 ( ∆ x i −1 + ∆ x i ) B i + ∆ x i B i +1 = 3  − 
 ∆ x i ∆ x i −1  xm m
Minimize µ ∫ f " ( x) dx + ∑ ( y i − f ( xi ) )
2 2
for all over the
(5) x1 i =1
interval [x1, xm] twice continuous differentiable

These are only m-2 equations (i = 2, 3, ..., m-1) functions f(x), f 0 C2[x1,xm] and fixed µ∃0.
for the determination of the m unknown Bi. Addi-
tionally, from the definition it follows that s"(x1) C2[x1,xm] denotes the set of all over [x1,xm]
= B1 = 0 and s"(xm) = Bm = 0 because s"(x) is twice continuous differentiable real funtions.
continuous everywhere. The parameter µ controls the trade-off be-
Summarizing, the calculation of an inter- tween the “roughness”, as measured by
xm
polating natural cubic spline requires the solu-
∫ f " ( x)
2
dx , and the coincidence of the data and
tion of a linear system of equations. x1
m
the function f(x) as measured by ∑ (( y − f ( x )) . It i =1

i i
2
must be chosen in advance.

SMOOTHING SPLINES The interpolating spline results when the
object function of (OP1) is reduced to the least
Interpolation methods are not suitable when the squares criterion.
data are influenced by measurement errors. Another problem was formulated and solved
Smoothing splines can be constructed in this by Reinsch (1967):
situation. For this, several optimization prob-
lems are explained subsequently: (OP2). xm
Minimize ∫ f " ( xm) dx for all f0C2[x1,xm] satisfying

2
x1
the condition ∑ (( yi − f ( xi )) ≤S.
2
i =1
168
Spline Fitting
A third criterion follows: s(x i) = Di, i=1, 2, …, m is solved. µ is the

smoothing parameter.
(OP3). In (OP2) and (OP3), smoothing parameters
m
are S and T, respectively. The interpolating
Find the minimum of ∑ (( y
i =1
i − f ( xi ) )
2
for
spline results for µ = 0. The straight line is the
xm
consequence of µ → ∞.
f0C 2[x1,xm] where ∫ f " ( x ) dx ≤ T .
2
dx
x1
Estimation of µ
Due to Holladay’s theorem (1957), each of
the problems (OP1), (OP2), and (OP3) is The estimation of µ requires statistical context.
uniquely solved by a natural cubic spline. The Suppose yi = g(x i) + ei. The ei are stochastically
problems OP1, OP2, and OP3 can be reformu- independent realizations of a random variable e
lated as quadratic optimization problems in the with expectation E[e] = 0 and variance V(e) =
Ñm with regard to Equations 1 to 5. σ 2.
The construction of the solution s(x) of OP1 Let µ ≥ 0 be arbitrary but fixed. Let sµ (x)
will be demonstrated now. denote the uniquely determined solution to
In the first step, the unknown Di=s(x i) will (OP1). Furthermore:
be calculated via the uniquely solvable linear- m
system of equations MSE(µ)= 1
m ∑ E[( s
i =1
µ ( xi ) − g ( xi )) 2 ]
[I + 2µQ TëZ-1ëQ]ëD = Y. defines the average mean-squared error, and

m
In Equation 6, I denote the m-dimensional PSE(µ)= 1
m ∑ E[( s
i =1
µ ( xi ) − yi* ) 2 ]
identity matrix:
defines the average predictive squared error
 2( ∆x1 + ∆x2 ) ∆x 2 0 ... 0  with new, unused y*i.
 
 ∆x 2 2( ∆x2 + ∆x3 ) ∆x3 
 0 ∆x 3 2( ∆x3 + ∆x4 )
...
...
0
0 
The parameter µ can be estimated by the
1 . . . ... . 
Z := 
3
 well-known cross-validation method. It works

.
.
. . ...
...
.
.


,
 0
.
... ∆
.
xm − 3 2( ∆ xm −3 + ∆ x m −2 ) ∆ x m −2 
as follows. Remove exactly one observation
 ∆xm − 2 2( ∆xm − 2 + ∆xm −1 ) 
 0 ... 0
(x i,yi) from the data. Calculate the smoothing
spline s µ− i (x ) with regard to the remaining m-1
observations with the given µ. The response of
 1  1 1  1 
 −  +  
 ∆x1  ∆x1 ∆x2  ∆x 2
0 ... 0
 the separated single point is measured by (yi -
 
 0

1
∆x 2
 1
−  +
1 
 ∆x 2 ∆x3 

1
∆x 3
... 0 

sµ−i ( xi ) )2. Do so for all observations and calcu-
Q :=  . . . . . .  m
∑( y
  ,
 . . . . . . 
late CV(µ) = 1
m i − s µ−i ( xi )) 2 .
 . . . . . . 
  i =1
 0 1  1 1  1 


0 ...
∆x m − 2
−  + 
 ∆x m − 2 ∆x m −1  ∆x m −1  It can be shown that PSE(µ) = MSE(µ) + σ
and E[CV(µ)] is approximately PSE(µ). This is
motivation to use the minimum µ* of CV(µ) as
∆xi := xi+1-x i, D = (D1, D 2,…Dm)T and Y =
an estimator for µ and sµ *(x) as an estimator for
(y1,y2,…,ym)T.
g(x).
At first glance, the calculation of CV(µ) is
In the second step, the interpolation problem
very expensive.
169
Spline Fitting
Figure 2. A smoothing spline and noisy data
Define H = (hij)i,j=1…m := [I + 2µQTëZ-1ëQ]-1. lating or smoothing spline functions. Numerische

Mathematik, 12, 66-82.
Hastie and Tibshirani (1990, pp. 46-48)
Craven, P., & Wahba, G. (1979). Smoothing
proved the equation
noisy data with spline functions: Estimating the
2
m
 yi − sµ ( xi )  correct degree of smoothing by the method of
CV(µ)= 1
m ∑ 
i =1  1 − hii 
 .
generalized cross-validation. Numerische
Mathematik, 31, 377-403.
The generalized cross-validation method
GCV(µ) of Craven and Wahba (1979) replaces De Boor, C. (2001). A practical guide to
each hii by the mean of the hii. In the past, it was splines. New York: Springer-Verlag.
easier to calculate the trace of H. Currently
DeVore, R. A., & Lorentz, G. G. (1993).
available algorithms compute hii in O(m) opera-
Constructive approximation: Polynomials
tions. So, the original motivation for GCV is no
and splines approximation. Berlin, Germany:
longer valid.
Springer.
Eubank, R. L. (1988). Spline smoothing and
REFERENCES nonparametric regression. New York:
Dekker.
Ansalone, P. M., & Laurent, P. J. (1968). A
Greville, T. N. E. (Ed.). (1969). Theory and
general method for the construction of interpo-
170
Spline Fitting
applications of spline functions. New York: observation is removed. The rest serve as a
Academic Press. training sample.
Hastie, T. J., & Tibshirani, R. J. (1990). Gener- Interpolating Spline: Let m points (xi,yi),
alized additive models. In Monographs on i=1, 2, …, m, x 1<x2<x3<...<xm be given. An
statistics and applied probability (No. 43). interpolating spline s(x) is a spline function
London: Chapman and Hall. satisfying the so-called interpolating conditions
s(x i)=yi for all i=1, 2, …, m.
Holladay, J. C. (1957). A smoothest curve
approximation. Mathmeatical Tables and other Natural Cubic Spline: This function is a
Aids to Computation, 11, 233-243. natural spline of degree n=3. This means k=1.
Hutchinson, M. F., & de Hoog, F. R. (1985). Natural Spline Function: Spline s(x) is
Smoothing noisy data with spline functions. called a natural spline if the degree of s(x) is
Numerische Mathematik, 47, 99-105. n=2k-1 (odd) and s(x) is given in each of the
intervals (-∞,x 1] and [xm,+∞) by polynomials of
Knott, G. D. (2000). Interpolating cubic
degree k-1 or less.
splines. Boston: Birkhäuser.
Smoothing Spline: A smoothing spline does
Reinsch, C. H. (1967). Smoothing by spline
not satisfy the interpolating conditions. The
functions. Numerische Mathematik, 10, 177-
starting point of its construction is an optimiza-
183.
tion problem. For example:
Wodny, M. (1998). Ausgewählte probleme der
kurvenanpassung. Greifswalder Seminarberichte
xm m
Heft 5, Gutzkow, GinkgoPark Mediengesellschaft. µ ∫ f " ( x) 2 dx + ∑ ( yi − f ( xi ) )
2
minimize (OP1)
x1 i =1
KEY TERMS
for all in the interval [x1,xm] twice continuous
Cross-Validation Method: Cross-valida- differentiable functions f(x). µ ≥ 0 is a given
tion is a model-evaluation method. Holdout smoothing parameter. The optimization prob-
cross-validation is the simplest kind. Observa- lem (OP1) has a unique solution, and this solu-
tions are randomly chosen from the given data tion is a natural cubic spline s(x). For µ=0 s(x)
set to form the validation data. The remaining it is the interpolating spline, and for µ → ∞ we
observations are retained as the training data get the straight line.
and allow for the determination of the model
Spline Function of Degree n: Let m real
parameters. So, it is possible to give predictions
numbers x1<x2<x3<...<x m be given. A spline
via the model and compare it with the validation
function s(x) of degree n is a function defined
data that have not already been seen. This
on Ñ. s(x) is given by some polynomial of degree
provides the background of model validation. A
n or less in each of the intervals (-∞,x 1], [xi,x i+1],
special case of holdout cross-validation is the
i = 1, 2, …, m-1, and [xm,+∞). All derivatives of
so-called leave-one-out method. Exactly one
s(x) up to order n-1 are continuous everywhere.
171
172
Chapter XXII
Parameter Estimation
Karl-Ernst Biebler
ABSTRACT
Parameters are numbers which characterize random variables. They make possible the
summarizing description of the observations, serve as the basis of statistical decisions and are
calculated from the data. Point estimations and confidence estimations are introduced.
Samples of the observed random variable are a starting point. The maximum-likelihood
method for the construction of parameter estimations is introduced here. Examples concern
the normal distributions and the binomial distributions. Approximate methods of the parameter
estimation also can be too inaccurate at large sample sizes. This is demonstrated in an example
from genetics.
INTRODUCTION parameters can be components of data models,

for example, the numbers α and β in the simple
Statistical parameter estimation is a standard model Y = α X + β.
task in most data-mining procedures. It is pre- The calculation of an approximate value for
supposed that the data are a sample of the the interesting parameter from the observed
interesting random variable. data is called point estimation. A confidence
Parameters, unknown values of character- estimation delivers a certain region in which the
izing numbers concerning the observed random interesting parameter is contained with given
variable, can be calculated from the data. Such (high) probability. It reflects the information
a calculation is called an estimate if it is carried content of the data with respect to the param-
out following statistical principles. eter and the sample size better than point
Examples of parameters are the expecta- estimation.
tion, the standard deviation, the median, quantiles, Parameter-estimation procedures are de-
and so forth of a random variable. Furthermore, veloped following special principles and consid-
ering the distribution of the random variables. The maximum-likelihood estimator MLE
They should fulfill quality requirements. α M LE calculates αMLE from the given data so
that L(x 1,...,x N, α MLE) is a maximum of the
likelihood function.
POINT ESTIMATION
Examples
Definition: Let X be a random variable, FX,a (x)
its distribution function, and (X1, ..., XN) a Normal Distribution
sample of size N. A point estimator α N of the
parameter α is a function α N = α N (x1,...,xN) of Let X be a normal distribution X » N (:, Φ2) with
the sample. A point estimation is the value of α N expectation E(X) = m and variance V(X) = σ2.
at given data (x1, ..., xN ). Then the parameter α = (:, Φ2) is an element of
The most essential methods for the con- a two-dimensional set. The MLE of the expec-
struction of point estimators are based on the tation is the sample mean:
maximum-likelihood principle, the Bayesian
principle, the moments principle, the minimum- 1 N
chi-square principle, or the least squares prin-

x =
N
∑x ,
i =1
i
ciple. The last named is not a statistical prin-

ciple but a geometric one. Its relation to the and the MLE of the variance is:
maximum-likelihood principle is the clue of the
Gausss-Markov theorem in mathematical sta- 1 N
tistics. Only one method is introduced here.

s M LE 2 =
N
∑(x
i =1
i − x )2.
Maximum-Likelihood Estimation Binomial Distribution
The idea of the maximum-likelihood principle For a binomial-distributed random variable, the
consists in the choice of the parameter value so relative frequency of the event in a sample of
that the observation gets maximal probability. size N is an MLE of its unknown probability.
Definition: Let the parameter be " 0 A f |m, The One-Locus-Two-Allele Model in

X be a continuous random variable with prob- Population Genetics
ability density fX," (x) depending on " and (x1, ...,
xN) be the sample data. Then: A two-allele model regarding the alleles A1 and
A2, and P(A1) = p and P(A2) = 1 – p, without any
N
dominance relations, is described by the proba-
L ( x 1 , ... x N , α) = ∏f
i =1
X ,α (xi )
bilistic model
is called the likelihood function of (x1, ..., xN ) {( A 1 A 1 ), ( A 1 A 2 ), ( A 2 A 2 ) } ; 

and ". M1 =  
 P ( A 1 A 1 ) = p , P ( A 1 A 2 ) = 2 p (1 − p ), P ( A 2 A 2 ) = (1 − p ) 
2 2
For a discrete random variable, the input is:

The phenotypes observed are exactly the
genotypes AiA j. The maximum-likelihood esti-
N
L ( x 1 , ... x N , α) = ∏ P( X
i =1
= xi ) . mator for p based on a sample of size N with
173
observed genotype frequencies N(A iA j) re- Examples

lated to M1 is the so-called gene-counting
method: Normal Distribution
p M 1 = ( 2 N ( A 1 A 1 ) + N ( A 1 A 2 )) / ( 2 N ) . The sample mean is an unbiased and effective

estimator for the expectation of a normally
If A1 is the dominating allele, you obtain: distributed random variable. The point estima-
tor sMLE2 is not unbiased for:
 {( A 1 A 1 , A 1 A 2 ), ( A 2 A 2 ) } ; 
M2 =   N −1
 P ( A 1 A 1 , A 1 A 2 ) = 1 − (1 − p ) , P ( A 2 A 2 ) = (1 − p ) 
2 2
E ( s M LE 2 ) = α.
N
and the MLE: Consequently, the usual sample variance:
N 1 N
p M 2 = 1 −
N (A 2 A 2 )
. s2 =
N −1
s M LE 2 = ∑(x − x)2
N − 1 i =1 i
N
If A2 is the dominating allele, the probabilis- is an unbiased estimator. It is also an effective

tic model: estimator of the variance.
Binomial Distribution
{( A 1 A 1 ), ( A 1 A 2 , A 2 A 2 ) } ; 
M3 =  
 P ( A 1 A 1 ) = p , P ( A 1 A 2 , A 2 A 2 ) = 1 − p 
2 2
The relative frequency of the event in a sample
of size N is an unbiased and effective estimator
gives the MLE: of its unknown probability.
N ( A1 A1 ) The One-Locus-Two-Allele Model in

p M 3 = .
N Population Genetics
Properties of Point Estimators The random variable 2N p M 1 describes the

number of alleles of type A1 in a sample of size
The characterization of the quality is also nec- 2N. It is binomially distributed with parameters
essary for the point estimators. Since estima- 2N and p. The estimator p M 1 is unbiased for:
tors are random variables, concepts are used
for this from the probability calculus. 1 1
E ( p M 1 ) = E ( 2 N p M 1 ) = 2 Np = p .
2N 2N
Definition: A point estimator α N is called
unbiased if its expectation is E (α N ) = α for all α Its variance:
∈ A. The bias BN (α N ) of the point estimator α N is
BN (α N ) = EN (α N ) – α. An unbiased point estima-
1 1 p (1 − p )
tor α N * is called effective if its variance is minimal V ( p M 1 ) = V ( 2 N p M 1 ) = 2 N p (1 − p ) =
(2 N ) 2 (2 N ) 2 2N
in the class of all unbiased estimators of α.
174
coincides with the inverse of the Fisher infor- X −µ

» N (0, 1).
mation. Due to the Rao-Cramer inequality, the σ/ N
maximum-likelihood estimator p M 1 is efficient.
Both p M 2 and p M 3 are neither unbiased nor X −µ
Set = u 1 −ε / 2
effective estimators of p. σ/ N
The random variables N(1- p M 2 ) 2 and
N( p M 3 )2 are binomially distributed with param- with u1 – ε / 2 the (1 – ε / 2) – quantile of the
eters N, (1-p)2 and N, p2, respectively. The standard normal distribution. Consequently:
estimators p M 2 and p M 3 are both asymptotically
efficient and asymptotically unbiased accord- P ( X − u 1 −ε / 2 ⋅ σ / N ≤ µ ≤ X + u 1 −ε / 2 ⋅ σ / N ) = 1 − ε.
ing to the theory of estimation. Their asymptotic
variances (Rao-Cramer inequalities) are cal- This means:
culated from the related Fisher information as:
[X − u 1 −ε / 2 ⋅σ / N ; X + u 1 −ε / 2 ⋅ σ / N ]
2p − p 2
V ( p M 2 ) = is a (1 – ε ) – confidence interval for the
asym p 4N
expectation.
and The quantile u1 – ε / 2 must be replaced by the
quantile tN – 1; 1 – ε / 2 of the t-distribution if the
variance of the random variable X is unknown
1 − p2 and estimated by the sample variance s2.
V ( pˆ M 3 ) = .
asymp 4N The (1 – ε) – confidence interval now reads:
(Biebler, Jäger, & Wodny, 2003)

[X − t N −1 ; 1 −ε / 2 ⋅s / N ; X + t N −1 ; 1 −ε / 2 ⋅ s / ]
N .
CONFIDENCE ESTIMATION Binomial Distribution
A point estimation from sample data gives only Let X be binomially distributed with the param-
an approximate value of the unknown parameters N and p, where X » B (N, p). There are
eter. It was the idea of Neyman (1935) to three possibilities for the calculation of a confi-
calculate an interval of parameter values con- dence interval CI of p.
sistent with the data. Method 1: Calculate CI exactly with re-
spect to the binomial distribution with the pa-
Definition: A set CI Ì A Í |m is called the rameters N and p.
confidence interval for the parameter α at the A (1 – ε) – confidence interval for p is CI =
confidence level ε > 0 if P(α ∈ CI) ≥ 1 – ε. [pl; pu], where the interval bounds are solutions
of the equations:
Examples
N
N ε
∑  m  p m
l ( 1 − p l ) N −m =
2
Normal Distribution m =k
Let X » N (:, Φ2). Then X » N (:, Φ2 / Ν) and and
175
Figure 1. Exact width of the 0.95 confidence Figure 2. Difference between the width of
interval as a function of p, where N = 50 the exact and the width of the asymptotic due
to maximum-likelihood-calculated 0.95
confidence intervals as a function of p,
where N = 50
N
N ε
∑  m  p
m =k
m
u (1 − p u ) N −m =
2
, respectively.
The solution of these equations is not simple.

This method is most frequently used.
Calculation problems occur for large values of
Method 3: Calculate CI asymptotically
N. The exact calculation of confidence inter-
according to the limit theorem from the theory
vals for the parameter p of a binomial distribu-
of maximum-likelihood estimations under ref-
tion for arbitrary N is possible with the help of
erence to the normal distribution with the ex-
incomplete Beta functions Ip (Johnson, Kotz, &
pectation value p and the variance Vasymp.
Kemp, 1992):
The One-Locus-Two-Allele Model in
p
I p ( k , N − k + 1) =
∫t
0
k
(1 − t ) N −k
dt
= P( X ≥ k) ,
Population Genetics
1
∫t
0
k
(1 − t ) N −k d t
It shall be demonstrated that approximate meth-
ods to noteworthy faults can lead to the simple
and example of the calculation of phenotype prob-
abilities for phenylketonuria (PKU) from popu-
N
N k lation data. The model of inheritance of PKU is
P( X ≥ k) = ∑ 
j =k
 p (1 − p ) N −k .
j M3 because the available tests of herozygocity
are applied only in special situations. The data-
Calculation tools are available in mining operation N ( p M 3 )2 is a binomially dis-
MATHEMATICA and in SAS (e.g., Daly, tributed random variable with parameters N
1992). and p2. This allows for the confidence estima-
Method 2: Calculate CI asymptotically action for p.
cording to the central-limit theorem of Laplace The calculation methods yield different con-
under reference to the normal distribution with fidence intervals for p. Figures 1, 2, and 3
the expectation Np and the variance Np(1-p). illustrate these differences for an example.
176
Figure 3. Difference between the width of Figure 4. Exact width B = p 2 of the 0.95
the exact and the width of the asymptotic due confidence interval of p 2 as a function of
to the Laplace central-limit-theorem- sample size N, calculated in the
calculated 0.95 confidence intervals as a neighbourhood of p 2 = 0.0001
function of p, where N = 50
0.00013
0.00012
0.00011
0.0001
0.00009
0.00008
0.00007
100000 150000 200000 250000 300000 350000
necessary sample size of N = 38,413 with

Method 3. The exact Method 1 yields N =
The differences between the methods of
45,865 here.
confidence-interval calculation will be illus-
The information content has to be carefully
trated also in the context of the sample-size
judged also for large-population genetic data
calculation.
sets. Already, for the very simple model of
The confidence level is 0.05 in the follow-
inheritance M 3 for PKU, there are considerable
ing; the width B of the confidence interval has
differences between the approximate and the
to be fixed. For given values of p, necessary
exact calculated sample sizes. Consequently,
sample sizes are calculable from the probability
the approximate calculation methods should not
distributions of p M 3 .
be used any longer.
PKU is one of the most frequent hereditary
diseases. For a certain population, the allele
probability of PKU is supposed as p = 0.01.
REFERENCES
This is a realistic order of magnitude. The
incidence (phenotype probability) for someone
Bickel, P. J., & Doksum, K. A. (1977). Math-
having the disease is p2 = 0.0001.
ematical statistics: Basic ideas and selected
Figure 4 shows the width B = p2 of the 0.95
topics. NJ: Prentice Hall.
confidence interval of the phenotype probabili-
ties p2 as a function of the sample size N. Biebler, K. E., Jäger, B., & Wodny, M. (2003).
For B = p2 = 0.0001, the necessary sample How exactly do we know inheritance param-
size is N = 173,146 (Method 1). The approxi- eters? In P. Perner, R. Brause, & H. G.
mate Method 2 yields N = 153,649. Holzhütter (Eds.), Medical data analysis (pp.
Consider now the estimation of the allele 9-14). Berlin, Germany: Springer.
probability p. For p = 0.01 and the same width
B of the 0.95 confidence interval, you get a Cox, D. R., & Hinkley, D. V. (1974). Theoreti-
cal statistics. London: Chapman & Hall.
177
Daly, L. (1992). Simple SAS macros for the ability of a sample may be expressed via the
calculation of exact binomial and Poisson con- conditional probability of the random variable
fidence limits. Comput. Biol. Med., 22, 351- given the parameter values of its distribution (a
361. priori distribution). Following Bayes, the a pos-
teriori distribution of the parameters given the
Efron, B., & Tibshirani, R. J. (1993). An intro-
sample data can be expressed. The parameters
duction to the bootstrap. London: Chapman
can be calculated so that the given sample data
& Hall.
becomes more probable according to the a
Encyclopedia of Biostatistics. (1998). posteriori distribution.
Chichester: Wiley.
Beta Functions: The function
1
Johnson, N. L., Kotz, S., & Kemp, A. W. B ( p , q ) = ∫ t (1 − t ) d t , p > 0, q > 0, is called
p −1 q −1
0
(1992). Univariate discrete distributions (2 nd a Beta function. An incomplete Beta function is
α
ed.). New York: John Wiley & Sons. defined as B ( p , q ) = ∫0 t p −1 ( 1 − t ) q −1 d t , 0 < " < 1.
Beta functions and their inverses are applied in
Lachin, J. M. (2000). Biostatistical methods:
statistics, for example, for tail probability calcu-
The assessment of relative risks. New York:
lations of binomial distributions.
John Wiley & Sons.
Binomial Distribution: The observation
Lehmann, E. L. (1983). Theory of point esti-
of the occurrence vs. the nonoccurrence of an
mation. New York: John Wiley & Sons.
event in a random experiment. When the ran-
Neyman, J. (1935). On the problem of confi- dom experiment is repeated, the number k of
dence intervals. Annals of Mathematical Sta- occurrences in n trials is a random variable. Its
tistics, 6, 111-116. probability distribution is the binomial distribu-
tion with the parameters n and the probability p
Neyman, J. (1937). Outline of a theory of
for the occurrence of the event.
statistical estimation based on the classical
theory of probability. Philosophical Transac- Confidence Interval: From sample data,
tion of Royal Society of London (Series A), an interval is calculated that contains the de-
236, 333-380. sired parameter of a random variable at a
predefined level of confidence. Confidence
Vogel, F., & Motulsky, A. G. (1979). Human
levels are probabilities, for example, 0.95 or
genetics. Berlin, Germany: Springer.
0.99.
Effective Point Estimator: A point esti-
KEY TERMS mator is the statistical method to calculate a
parameter value from sample data. Effective-
Asymptotically Efficient: The property of
ness is one of its most desired properties. It is
efficiency almost applies to large sample sizes.
the best exploitation of the information con-
Asymptotically Unbiased: The property tained in the data for the purpose of parameter
of being unbiased almost applies to large sample calculation.
sizes.
Fisher Information: Consider a sample of
Bayesian Principle: Statistical calculation size n of a random variable. The Fisher infor-
principle founded on Bayes’ theorem. The prob- mation is calculated via the probability distribu-
178
tion of the random variable and the sample size pected and observed values of that random
n. It can be understood as a measure for the variable becomes minimal.
information content of the sample and is closely
Moments Principle: Probability distribu-
related to the minimal variance of a statistical
tions are characterized by parameters and, on
point estimator due to the Rao-Cramer inequal-
the other hand, by moments. One can calculate
ity.
from the given sample data the moments, and
Gauss-Markov Theorem: Linear models from these the desired parameters.
of random variables are of importance in re-
Normal Distribution: The normal distri-
gression analysis, for example. The Gauss-
butions are a special type of probability distribu-
Markov theorem describes the special way to
tions for continuous random variables. They
get point estimates of the parameters of linear
are completely characterized by the two pa-
models.
rameters mean and variance. The standard
Least Squares Principle: This principle normal distribution has a mean of zero, a vari-
concerns curve fitting and is of geometric na- ance of one, and the well-known Gaussian bell-
ture. Neither random variables nor samples are shape density function. The normal distribu-
required. Curve parameters are calculated so tions play an important role in applied statistics.
that the sum of squared differences between
Parameter: A characterizing quantity, es-
function values and data values becomes mini-
pecially in statistics. Examples are the mean
mal.
and the variance of a random variable, or the
Likelihood Function: The likelihood func- coefficients in a linear regression model. It is a
tion expresses the likelihood (simply, the prob- standard task of statistics to calculate param-
ability) of a sample of the random variable in eters from data.
observation. It is derived from the related prob-
Point Estimation: When a point estimator
ability function (in case of a discrete random
is applied to given data, the calculation result is
variable) or the related probability density (in
called a point estimation.
case of a continuous random variable).
Point Estimator: This is a statistical method
Maximum-Likelihood Estimator: It is a
to calculate a parameter value from sample
calculation procedure and the result of the
data.
application of the maximum-likelihood principle
to a parameter-estimation problem. Population Genetics: Population genetics
studies processes of heredity in populations. It
Maximum-Likelihood Principle: It con-
concerns biology and medicine first, but also
cerns random variables and their parameters.
fields from history to ethnology. Most of its
These parameters are calculated from the data
methods come from biometry. The effects of
so that the likelihood (simply, the probability) of
evolution, migration, selection, and mutation on
the sample at hand becomes a maximum.
the genetic constitution of populations are es-
Minimum-Chi-Square Principle: The pecially investigated.
unknown parameters of a random variable are
Rao-Cramer Inequality: A statistical point
calculated so that the difference (measured by
estimator is a random variable. Its variance has
Pearson’s chi-square formula) between ex-
a lower bound (Rao-Cramer bound) under cer-
179
tain suppositions concerning the probability dis- Unbiased Point Estimator: A point esti-
tribution of the observed random variable. The mator is the statistical method to calculate a
Rao-Cramer bound depends only on this distri- parameter value from sample data. Being unbi-
bution and the sample size, but not on the ased is one of its most desired properties. This
sample data. means there is the absence of any systematic
failure in the sense of statistics.
180
181
Chapter XXIII
The Method of Least Squares
Bernd Jaeger
ABSTRACT
The method of least squares is a geometric principle of curve fitting. The unknown parameters
of a function are calculated in such a way that the sum of squared differences between
function values and measurements gets minimal. Examples are given for a linear and a
nonlinear curve fitting problem. Consequences of model linearizations are explained.
INTRODUCTION The method of least squares is explained at

its simplest examples in the following para-
Model fit is a general task in data mining. It is graphs. In addition, difficulties occurring in the
a basic component of general problems like model linearization are demonstrated.
optimization, statistical data evaluation, data The so-called method of least squares is a
imaging, and so forth. The method of least universal method for the calculation of the
squares (MLS) is a widely used principle of parameters α1, ..., αk of a model function y =
geometric character to fit a model to given fα1,...,α1 (x), which in the best possible way goes
data. The method goes back to the work of C. through a given set of points (xi, yi), i = 1, ..., n.
F. Gauss and A. M. Legendre. The basic idea is to minimize the sum of the
Suppositions concern merely the model to squared distances between the function
be adapted: The data must be numbers. Many f α1 ,...,αk ( xi ) and the measurement yi:
data-related optimization criteria are exten-
sions of the classic least squares method. Sta-
( )
n
g (α1 ,..., α k ) = ∑ f α1 ,...,α k ( xi ) − y i
2
tistical parameter estimation and the method of .
i =1
least squares are closely connected in linear
statistical models (e.g., the Gauss-Markov theo- For that, the equations
rem in mathematical statistics).
∂ g (α1 ,..., α k ) minant of the coefficients of the linear system

= 0 , j = 1, ..., k is different from zero.
∂α j
Unfortunately, there are only a few model
functions that lead to such an easily solvable
formulate the necessary conditions. These k system of equations. Nonlinear systems of equa-
equations are the system for the determination tions result mostly. In these cases, one must be
of the unknown parameters α1, ..., αk. The content with iterative approximate solutions.
solutions must still be examined to determine
whether they satisfy the sufficient conditions
for a minimum. EXAMPLE 2
The model function y = f(x) = c ⋅ exp(dx) leads

EXAMPLE 1 to nonlinear equations for the determination of
the parameters c and d. It is calculated from:
A linear function y = f(x) = ax + b shall be
fitted to the set of points. The function: n
g1 ( c, d ) = ∑ (( c ⋅ exp( dxi )) − yi ) (∗),
2
n i =1
g ( a, b) = ∑ ((axi + b) − y i )
2
i =1
∂ g1 (c, d ) n
= ∑ 2(( c ⋅ exp(dxi )) − y i )exp( dxi ) =0 , and
depends on the unknown a and b and shall be ∂c i =1
minimized.
The necessary conditions for the existence ∂ g1 ( c , d ) n
= ∑ 2(( c ⋅ exp( dxi )) − yi )(c ⋅ exp( dxi ) )xi =0 .
of a minimum are: ∂d i =1
∂ g ( a , b) n The resulting system of nonlinear equations

= ∑ 2(( axi + b) − y i )xi =0 and
∂a i =1 follows:
∂ g (a , b) n
 n 2
n
= ∑ 2(( axi + b) − y i ) =0 . ( I )  ∑ (exp(dxi ) ) c − ∑ yi exp(dxi ) = 0
∂b i =1  i =1  i =1
 n 2 2  n 
( II )  ∑ (exp(dxi ) ) c −  ∑ xi y i exp(dxi ) c = 0
The conditions lead to a linear system of  i =1   i =1 
equations:
One can get the solution using a numeric
 n 2  n   n  approximation method, for example, the Gauss-
( I )  ∑ xi a +  ∑ xi b =  ∑ xi y i  Newton method. Most iterative procedures need
 i =1   i =1   i =1 
 n
  n
 a start value (cs, ds). This start value shall be
( II )  ∑ xi  a + nb =  ∑ y i  .
 i =1   i =1  contained in a close neighbourhood of the solu-
tion. However, this is not a sufficient condition
for the convergence of the algorithm toward
The solution (a0, b0) of this system of equa-
the desired solution.
tions is uniquely determined in case the deter-
182
n
Figure 1. Measurements (xi , yi) and the MLS
g 2 ( h, d ) = ∑ (( dxi + h ) − Log ( y i ) ) (∗ ∗)
2
fitted function y = f1(x) = 3.3139.exp(–0.4263x) i =1
y
4
with an explicit solution.

The inverse transformation (exp(h0), d0) of
3 the solution (h0, d0) into the original space of
parameters does not yield any solution of the
original problem (*). This has been known
2
already for a long time (Wittstein, 1882). To-

day, this is nevertheless still frequently misun-
1 derstood, too.
The data in Table 1 are concentrations yi of
an active agent of a medication measured xi
0
0 1 2 3 4 times in the serum of a patient. The concentra-

x
tion-time function is supposed to be y = f1(x) =

c ⋅ exp(dx).
The parameters c = 3.3139 and d = - 0.4263
The following procedure seems seductive: are calculated via MLS. Figure 1 represents the
Transform the original problem into a problem measurements and the fitted function.
that is linear dependent on the searched param- The linear function Log(y) = 1.13948017 -
eters and fit it to the transformed data without 0.45372974*x is fitted to the log measurements
the use of an approximation method. You get in Figure 2. With the inverse transform of the
for the problem (*) of Example 2: parameters, f2(x) = 3.12514 exp(-0.45372974x)
results. The squares of differences between
Log(f1(x)) = Log(c) + dx = h + dx the measurements and the model functions are
indicated in Columns 4 and 6 of Table 1.
and, with (xi, Log(yi)), the new minimum prob-
lem:
Figure 2. Transformed model Log(y) =
1.13948017 – 0.45372974*x and the log
measurements
Table 1. Measurements f 1(x) and f 2(x) yt
2
xi yi exact (MLS) transformation

f1(x) = 3.3139 * exp(-0.4263x) f2(x) = 3.1251 *exp(-0.4537x)
f1(x) (f1(xi) - yi)2 f2(x) (f2(xi)- yi)2
0.00 3.74184 3.31300 0.18390 3.12514 0.38032 1
0.25 3.79231 2.97808 0.66296 2.79001 1.00459

0.50 0.65957 2.67702 4.07011 2.49083 3.35350
0.75 2.54683 2.40639 0.01972 2.22372 0.10440
1.00 2.99364 2.16313 0.68975 1.98526 1.01683
1.25 1.61417 1.94445 0.10909 1.77237 0.02503 0
1.50 1.03785 1.74788 0.50414 1.58231 0.29643
1.75 1.58489 1.57118 0.00019 1.41263 0.02967
2.00 1.41626 1.41235 0.00002 1.26114 0.02406
2.25 1.99844 1.26957 0.53124 1.12591 0.76131
2.50 1.60539 1.14123 0.21544 1.00517 0.36026 -1
2.75 1.20958 1.02586 0.03375 0.89738 0.09747
3.00 0.94700 0.92215 0.00062 0.80115 0.02127
3.25 0.57369 0.82893 0.06515 0.71524 0.02004
3.50 0.93629 0.74513 0.03654 0.63854 0.08866
3.75 0.14282 0.66980 0.27771 0.57006 0.18253
-2
4.00 0.68264 0.60209 0.00649 0.50893 0.03017 0 1 2 3 4
Ó 7.40682 Ó 7.79655 x
183
The function f2(x) does not solve the minimal Gauss-Newton Algorithm: It is an itera-
problem (*) as can be seen from the sums of the tive algorithm to solve a nonlinear system of
deviation squares in Column 4 (7.40682) and in equations or to find the minimum of a function.
Column 6 (7.79655). In a line-search version of the Gauss-Newton
algorithm, the search direction and the follow-
ing point of iteration (x1(n+1), …, x m(n+1)) is calcu-
REFERENCES lated by an approximated linear system on the
point (x1(n), …, xm(n)).
Armitage, P., & Berry, G. (1991). Statistical
Method of Least Squares (MLS): Least
methods in medical research. Oxford:
squares problems appear in data-fitting appli-
Blackwell Sc. Pub.
cations. Suppose that the function f(x) depends
Fisher, R. A. (1973). Statistical methods and on a parameter vector (p1, …, pm), and yi is the
scientific inference. New York: Hafner. actual observation (including errors) of the
system at xi. Then g = ∑ (f(xi)-yi)² is the sum of
Hartung, J. (1984). Statistik, lehr: Und
the squares of the residuals between the func-
handbuch der angewandten statistik.
tion and the observations. Methods for the
München: Oldenbourg Verlag.
calculation of the minimum of g are methods of
Lachin, J. M. (2000). Biostatistical methods. least squares.
New York: John Wiley & Sons.
System of Linear or Nonlinear Equa-
Rosner, B. (1989). Fundamentals of biosta- tions: A system with m linear equations (being
tistics. Boston: PWS-Kent Publishing Com- valid simultaneously), n unknown variables x1,
pany. x2, …, xn, fixed variables aij (i = 1, …, m; j = 1,
…, n) and bi (i = 1,…,m), and can be written as:
Wittstein, T. (1882). Ein zusatz zur methode der
kleinsten quadrate. Z. Math. Phys., 27, 315-
317. a11 x1 + a12 x2 + …. + a 1n xn = b1
a21 x1 + a22 x2 + …. + a 2n xn = b2
…….
KEY TERMS am1 x1 + am2 x2+ …. + amn xn = bm;
Gauss-Markov Theorem: The theorem

otherwise, it is a nonlinear system. A simple
says that in a linear model in which the errors
method to solve the system for a small number
are independent and identically normally dis-
of equations is the Gauss algorithm. Gauss-
tributed with expectation zero and with equal
Jordan elimination and Cholesky decomposi-
variances, the best linear unbiased estimators
tion are the methods for larger systems of linear
and the least squares estimators are the same.
equations.
Even with weaker conditions, the theorem holds
true: The errors need to have expectation zero,
and need to be uncorrelated with equal vari-
ances.
184
186
Chapter XXIV
The Data-Information-
Knowledge Model
Andrew Georgiou
ABSTRACT
The generation and transformation of data into information and knowledge is a basic formula
in health informatics. This process is often represented in a model that portrays each
component hierarchically with data at the bottom followed by an intermediary layer of
information and topped by the knowledge layer. This model is a simple way to conceptualize
important components of the informatics process, but it also has major limitations. The capture
of data does not lead seamlessly to information or knowledge. The process is much more
complex involving a multi-faceted web of interactions and issues.
INTRODUCTION nizational planning and management (Peel, 1994;

Shortliffe, 1991).
Health informatics is still an emerging and The heterogenous nature of the discipline
rapidly expanding academic discipline (Greenes means that it finds itself enmeshed in the many
& Shortliffe, 1990), located at the intersection methodological and epistemological issues in-
of ICT and the many areas of healthcare. Its volved in the practice of healthcare (Georgiou,
growth is a direct consequence of the dramatic 2002). Indeed, even health informatics’ most
expansion of ICT across health services over basic formula—the generation and transforma-
the last two decades. More than just the study tion of data into information and knowledge—
of computers within medicine and its related invites divergent opinions about the assump-
fields, health informatics embraces a number of tions that underpin its claims (Hirschheim et al.,
different fields and activities including patient 1995).
care, healthcare research, education, and orga-
The Data-Information-Knowledge Model
THE HEALTH-INFORMATICS Figure 1. The informatics model

MODEL
The process of distinguishing different forms

and objects, and naming and categorizing them Knowledge
is an essential part of our ability to interact and
communicate. This process is no less important Information
to medical science and its need for communica-

Data
tion channels to describe illnesses and share
treatment options (Coiera, 2003). Models form
an important element to the way we perceive
and interpret the world (NHS Executive, 1996).
Not only do they help us to describe and com-
municate aspects of the world, they also assist
in understanding what is going on, perhaps even (based on principles of logical implication, e.g.,
helping us to change and manage a part of statements of certainty), abduction (which aims
reality. to establish links between observations such as
The model often referenced within health cause and effect, e.g., statements of possibil-
informatics is the data-information-knowledge ity), and induction (whereby generalizations are
model. It is used to help identify and understand generated from specific examples to formulate
the different components and interrelationships general rules; Coiera, 2003; Degoulet & Fieschi,
within healthcare (Abdelhak et al., 1996; Coiera, 1997).
2003; Degoulet & Fieschi, 1997; Sheaff & This hierarchical model has underpinned
Peel, 1995). The origin of this basic model can many of the information infrastructures within
be traced back to the 19th century and the healthcare including the use of disease classifi-
development of the functions of taxonomy and cation data to help research and plan healthcare.
classification. Early statisticians used and de- The United Kingdom National Health Service’s
veloped classification systems as knowledge Information Management and Technology strat-
repositories developed from data and informa- egy in the 1990s described the process as a
tion (Desrosieres, 1998). Their model involved “language of health” (NHS Centre for Coding
three essential parts arranged hierarchically, and Classification, 1996). It posited healthcare
with data at the bottom and an intermediary information as based on clinical terms that
layer of information topped by the knowledge could then be transformed into classification
layer. systems such as the International Statistical
Within this model (see Figure 1), data take Classification of Diseases and Related
on the character of facts or observations, which, Health Problems (World Health Organization
in and of themselves, have little or no meaning. [WHO], 1993), and from there into casemix
They take on significance only when they are groups for cost and resource management.
provided with a contextual framework to man- Even the growth of evidence-based medi-
age and make sense of them. Information is cine (EBM) with its commitment to using sci-
assumed to be the product of processed data. ence, research, and evidence to guide decision
The generation of knowledge then proceeds making (Appleby et al., 1995) envisages infor-
through a complex process involving deduction mation flows that broadly replicate the
187
informatics model. The main tasks involved in Many of the problems associated with the
EBM can be described as: model are directly applicable to coding and
classification systems. These are tools de-
• Finding, appraising, and using research- signed for a specific purpose. There is no pure
based knowledge in decision making; set of codes or terms that can be universally
• Using systems for managing medical applied in healthcare. This is because they are
knowledge, and for obtaining, storing, and heavily context dependent and change over
promoting the use of evidence; and time with developments in medical science and
• Promoting and facilitating evidence-based understanding (Coiera, 2003). The choice of
decision making. (Muir Gray, 1997) which classification to use must be determined
by the area of investigation (WHO, 1993). In
It is relatively easy to conceptualize these the NHS in the ’90s, leading clinicians and
tasks using associated informatics functions health-service commentators questioned the
such as knowledge browsing to assess informa- reliance on ICD classification codes that were
tion from a knowledge base; messaging to geared to producing administrative contract
communicate records, assessments, and refer- data to the detriment of clinical data (Hopkins,
rals; and counting to generate and analyze data 1996; Wyatt, 1995). The capture of classifica-
(Benson, 1997). It is not surprising, therefore, tion codes and their transformation into larger
that the growth of EBM has been linked to aggregate groups was not producing the type of
organizational commitments to establishing ro- information or knowledge required.
bust clinical information systems alongside EBM has made dramatic inroads into many
greater attention to health informatics (Muir levels of healthcare, but it is not a perfect
Gray, 1997). decision-making tool (Cohen et al., 2004). Crit-
ics of EBM question the ability of large data-
bases to provide best guidance for clinical
LIMITATIONS OF THE practice. There is always the potential for the
INFORMATICS MODEL misapplication of EBM using irrelevant or out-
dated evidence from randomized and controlled
The data-information-knowledge informatics trials, systematic reviews, and expert guide-
model is an abstraction: a simple way to con- lines (Evans, 1995; Greenhalgh & Worrall,
ceptualize a complex process. Therein lies its 1997).
value but also its greatest weakness. The model The statistical technique of meta-analysis in
views the generation of knowledge as a linear EBM is based on the belief that trial results
process whereby the capture of data from one from apparently similar interventions (e.g.,
side of the spectrum can lead seamlessly to drugs) can be pooled. This process works best
information and knowledge on the other side with many trials of a single drug, as in the case
(Georgiou, 2002). The problem with this con- of aspirin for diseases in coronary, cerebral, or
ception is that the whole process of knowledge peripheral arteries (Antiplatelet Trialists Col-
management is viewed overwhelmingly as just laboration, 1994). But reality is not always
a matter of capturing, organizing, and retrieving straightforward. Trial registers and the follow-
information based on databases, mining, docu- up of nontrial patients demonstrate that inclu-
ments, and so forth (Thomas et al., 2001). It is sion and exclusion criteria lead to the recruit-
not so simple. ment of younger and more mobile patients who
188
can attend follow-up care, are free from other ism, which recognizes that reality exists inde-
diseases, and are not being treated with mul- pendently of our thoughts or beliefs (Robson)
tiple drugs. This leads to an implicit bias against and aims to explore the context and inherent
older people (Hampton, 2000). Other criticisms mechanisms that generate events (Danermark
of EBM contend that it has led to an obsession et al., 1997; Mingers, 2004; Sayer, 2004).
with measurement and accountability, which
fosters illusions that complexities in medicine
can be reduced to numbers to be manipulated CONCLUSION
(Charlton & Miles, 1998; Goodman, 1998). As
Cohen et al. (2004, p. 40) point out, “There is no The health-informatics model offers a simple
‘mean tendency’ for a single patient. A therapy and straightforward way to conceptualize im-
is beneficial for a person, or it is not.” portant components of the informatics process.
Information-system developments and data- Its value lies overwhelmingly in its ability to
modeling enterprises are always underpinned differentiate the concepts that underpin the
by certain general assumptions that guide their process. The model is not new; indeed, it has
approach. Examining these assumptions is a been associated with the development and use
prerequisite to better understanding the prac- of statistics to help understand, monitor, and
tice of informatics (Hirschheim et al., 1995). even govern society (Desrosieres, 1998;
The health-informatics model sits within a posi- Georgiou, 2001; Georgiou & Pearson, 2002).
tivist framework, which has traditionally domi- This probably explains why the model is more
nated information-systems research and sys- closely identified with statistically oriented ap-
tems development (Mingers, 2004). Positivism proaches involving disease classifications and
visualizes reality as the sum of sense impres- EBM. However, it is very important to under-
sions (Robson, 2002) with a focus on recording stand the limitations of the model. Knowledge
and measuring events and then using statistical generation is not a seamless operation. Under-
and mathematical models to capture the pat- standing the use of information systems and
terns that appear in the data (Mingers, 2004). their impact requires one to consider the com-
While positivist approaches dominate many plex web of interactions that are implicated in
fields of scientific enquiry, they have been the process (Coiera, 2004). It is clearly an
criticized for their emphasis on superficial facts involved course of action incorporating social
without providing an understanding of the un- and technical issues, scientific debate, and suc-
derlying mechanisms or meanings to individuals cesses and failures. The traditional health-
(Bowling, 1997). informatics model offers a start to understand-
There are other approaches to knowledge ing this process, but it is by no means the finish
generation that attempt to deal with the appar- of the story.
ent restraints of positivism. Phenomenological
approaches interpret the world in terms of
meanings that construct an individual social REFERENCES
reality (Robson, 2002). Knowledge generation
is not viewed as a logically ordered accumula- Abdelhak, M., Grostick, S., Hanken, M., &
tion of facts leading to scientific laws, but a Jacobs, E. (1996). Health information: Man-
process of conflicts and paradigm shifts (Kuhn, agement of a strategic resource. Philadel-
1970). Other approaches include critical real- phia: WB Saunders.
189
Antiplatelet Trialists Collaboration. (1994). Desrosieres, A. (1998). The politics of large

Collaborative overview of randomised trials of numbers: A history of statistical reasoning.
antiplatelet therapy. I: Prevention of death, Boston: Harvard University Press.
myocardial infarction, and stroke by prolonged
Evans, J. G. (1995). Evidence-based and evi-
antiplatelet therapy in various categories of
dence-biased medicine. Age & Ageing, 24(6),
patients. British Medical Journal, 308(6921),
461-463.
81-106.
Georgiou, A. (2001). Health informatics and
Appleby, J., Walshe, K., & Ham, C. (1995).
evidence based medicine: More than a mar-
Acting on the evidence of a review of clini-
riage of convenience? Health Informatics
cal effectiveness: Sources of information,
Journal, 7(3-4), 127-130.
dissemination and implementation. Birming-
ham, UK: HSMC. Georgiou, A. (2002). Data information and
knowledge: The health informatics model and
Benson, T. (1997). The message is the medium.
its role in evidence-based medicine. Journal of
Health Service Journal, 107(5538), 4-5.
Evaluation in Clinical Practice, 8(2), 127-
Bowling, A. (1997). Research methods in 130.
health. Buckingham, UK: Open University
Georgiou, A., & Pearson, M. (2002). The role
Press.
of health informatics in clinical audit: Part of the
Charlton, B. G., & Miles, A. (1998). The rise problem or key to the solution? Journal of
and fall of EBM. Quarterly Journal of Medi- Evaluation in Clinical Practice, 8(2), 183-
cine, 91(5), 371-374. 188.
Cohen, A. M., Stavri, P. Z., & Hersh, W. R. Goodman, N. W. (1998). Clinical governance.
(2004). A categorization and analysis of the British Medical Journal, 317(7174), 1725-
criticisms of evidence-based medicine. Inter- 1727.
national Journal of Medical Informatics,
Greenes, R., & Shortliffe, E. (1990). Medical
73(1), 35-43.
informatics: An emerging academic disci-
Coiera, E. (2003). Guide to health informatics pline and institutional priority. Journal of the
(2 nd ed.). London: Oxford University Press. American Medical Association, 263, 1114-
1120.
Coiera, E. (2004). Four rules for the reinven-
tion of health care. British Medical Journal, Greenhalgh, T., & Worrall, J. G. (1997). From
328(7449), 1197-1199. EBM to CSM: The evolution of context-sensi-
tive medicine. Journal of Evaluation in Clini-
Danermark, B., Ekstrom, M., Jakobsen, L., &
cal Practice, 3(2), 105-108.
Karlsson, J. C. (1997). Explaining society:
Critical realism in the social sciences. Lon- Hampton, J. R. (2000). Evidence, guidelines,
don: Routledge. audit and cardiology: Principles and problems in
secondary and tertiary care. In A. Miles, J. R.
Degoulet, P., & Fieschi, M. (1997). Introduc-
Hampton, & H. B. London (Eds.), NICE, CHI
tion to clinical informatics. New York:
and the NHS reforms (pp. 65-78). London:
Springer.
Aesculapius Medical Press.
190
Hirschheim, R., Klein, H. K., & Lyyvtinen, K. Sheaff, R., & Peel, V. (1995). Managing
(1995). Information systems development and health service information systems: An in-
data modelling: Conceptual and philosophi- troduction. Buckingham: Open University
cal foundations. Cambridge, UK: Cambridge Press.
University Press.
Shortliffe, E. H. (1991). Medical informatics
Hopkins, A. (1996). Clinical audit: Time for a and clinical decision making: The science and
reappraisal? Journal of the Royal College of the pragmatics. Medical Decision Making,
Physicians of London, 30(5), 415-425. 11(Suppl. 4), S2-S14.
Kuhn, T. (1970). The structure of scientific Thomas, J., Kellogg, W., & Erickson, T. (2001).
revolutions. Chicago: University of Chicago The knowledge management puzzle: Human
Press. and social factors in knowledge management.
IBM Systems Journal, 40(4), 863-884.
Mingers, J. (2004). Re-establishing the real:
Critical realism and information systems. In J. World Health Organization (WHO). (1993).
Mingers & L. Willcocks (Eds.), Social theory International statistical classification of
and philosophy for information systems (pp. diseases and related health problems (10 th
372-406). Chichester, UK: John Wiley & Sons. revision, Vol. 2). Geneva, Switzerland: Au-
thor.
Muir Gray, J. A. (1997). Evidence-based
healthcare. Edinburgh, UK: Churchill Wyatt, J. C. (1995). Hospital information man-
Livingstone. agement: The need for clinical leadership. Brit-
ish Medical Journal, 311(6998), 175-178.
NHS Centre for Coding and Classification.
(1996). An introduction to the NHS Centre
for Coding and Classification Version 4.0 KEY TERMS
IMG F6163. Loughborough, UK: Information
Management Group, NHS. Classifications: The action of classifying
or arranging in classes according to common
NHS Executive. (1996). Building a bigger
characteristics or affinities.
picture. UK.
Code: The unique numerical identifier as-
Peel, V. (1994). Management-focused health
sociated with a concept, which may be associ-
informatics research and education at the Uni-
ated with a variety of terms all with the same
versity of Manchester. Methods of Informa-
meaning.
tion in Medicine, 33, 273-277.
Critical Realism: The view that reality
Robson, C. (2002). Real world research.
exists independently of our thoughts or beliefs.
Oxford: Blackwell Publishers Ltd.
Research refers to this reality rather than con-
Sayer, A. (2004). Foreword: Why critical real- structing it.
ism? In S. Fleetwood & S. Ackroyd (Eds.),
Epistemology: The theory or science of
Critical realist applications in organisation
the method or grounds of knowledge.
and management studies (pp. 6-20). London:
Routledge. Models: Representations of real objects or
phenomena, or templates for the creation of
objects or phenomena.
191
Paradigm: A set of ideas about the phe- Positivism: A philosophical view that aims
nomena under enquiry. to discover laws using quantitative methods and
emphasizes positive facts.
Phenomenology: A philosophical belief
that interprets and experiences the world in Term: In medical terminology, it is a recog-
terms of meanings and actively constructs an nized name for a medical condition or treat-
individual social reality. ment.
192
193
Chapter XXV
Goals and Benefits of Knowledge
Management in Healthcare
Odysseas Hirakis
Spyros Karakounos
ABSTRACT
The aim of this chapter is to explain the role of knowledge management and how it can be
successfully applied in the area of healthcare in order to improve health services and to
increase patients’ satisfaction. The first part of this chapter is about explaining the theories
beyond knowledge management as “what is knowledge” and how it can be transformed and
captured across people and organizations. The second part consists of the theory of
knowledge management and the benefits of it in the area of healthcare in comparison with the
old traditional systems. Knowledge management systems can be used to index and at the same
time to spread all that information across people, libraries, and hospitals.
INTRODUCTION information among people who are interested in

learning. Knowledge management allows ev-
During the last 10 to 15 years, knowledge eryone to reuse the knowledge (best practice)
management (KM) has become more popular or to create new ideas (innovation).
day by day. There is a lot of interest in the According to Syed Sibte Raza Abidi (2001,
concept of capturing and sharing knowledge p. 1), “Knowledge Management (KM) in
with technology as the enabler. This requires healthcare can be regarded as the confluence
the existence of a knowledge-sharing culture. of formal methodologies and techniques to fa-
The KM system stores historical knowledge cilitate the creation, identification, acquisition,
and knowledge created during exchanges of development, preservation, dissemination and
Goals and Benefits of Knowledge Management in Healthcare
finally the utilisation of the various facets of a Figure 1. Knowledge pyramid (Marco, 2003)
healthcare enterprise’s knowledge assets.”
People in their everyday practice collect
massive amounts of data and information that
are knowledge poor, a fact that makes their
decision about patients’ cures more compli-
cated. Knowledge in healthcare is deemed a
high-value form of information that is neces-
sary for healthcare professionals to act. For
that matter, with the emergence of KM, the
raw empirical data can be changed into empiri-
cal knowledge and provide professionals with a
decision-support tool (Syed Sibte Raza Abidi,
2001). Knowledge is very difficult to define; it is
KM in healthcare presents further interest not just a simple document or something that
for all those who are involved in the delivery of someone told us. In order for a person to gain
health services. KM allows rapid access to a knowledge, there are three stages to progress
knowledge treasure. The KM model goes be- through as the pyramid (see Figure 1) indicates.
yond the need to manage data or information The actual content in each stage becomes
overload. It satisfies the requirements for imple- smaller, starting with data and finishing with
menting best practice and supplying high-qual- knowledge.
ity health services, which increase patient sat-
isfaction. The model aims at greater efficiency, 1. Data: Documents, unorganized and un-
coordination, and cost reduction. It is a portfolio processed (raw material)
of knowledge that increases ealth-care profes- 2. Information: Selected data è Interpre-
sionals’ effectiveness and productivity. A KM tation of the data (processed data)
system offers them the opportunity to learn 3. Knowledge: Selected information è In-
how other colleagues successfully carried out terpretation of the information
similar problems (De Lusignan, Pritchard, &
Chan, 2002). Example: In order to complete an aca-
demic assignment, some steps need to be taken:
WHAT IS KNOWLEDGE? 1. Research at libraries and on the Internet;

Collect some documents è Data
According to ITIL People (http:// 2. Interpretation of that data è Information
www.itilpeople.com/Glossary/Glossary_k.htm), 3. Interpretation and evaluation of the infor-
“Knowledge is part of the hierarchy made up of mation è Knowledge
data, information and knowledge. Data are raw
facts. Information is data with context and
perspective. Knowledge is information with TYPES OF KNOWLEDGE
guidance for action based upon insight and
experience.” Two types of knowledge exist in an organiza-
tion.
194
1. Explicit (“Know that”): Something that Figure 2. Nonaka and Takeuchi’s model
is written down (informative texts) and
can be easily understood if read, for ex- Tacit Dialogue
Dialogue Tacit
ample, technical reports and books

Socialization Externalization
2. Tacit (“Know how”): Something that is • Brainstorming
Socialization
• Discussion
• Write it down
Externalization
• Dialogue
Tacit -Brainstorming -Write it down Explicit
written in the mind and cannot be easily -Discussion -Dialogue
expressed, for example, experience gained

from a job
Internalization Combination
Tacit Explicit
Knowledge management refers to the Internalization
-Access to
• Access to
Codified
Combination
-Adding
• Adding
-Categorizing
codified • Categorizing
knowledge -Best
• BestPractices
knowledge of a company as an asset, the same knowledge Practices
as land, for example (Ahmad, 2001).

Learn doing
Learn by doing
Explicit Explicit
TRANSFORMATION OF
KNOWLEDGE
KNOWLEDGE MANAGEMENT
The transformation of knowledge is very im-
portant. The following explains Nonaka and Peter Drucker first introduced the terms knowl-
Takeuchi’s (1995) model of the four methods edge work and knowledge worker in the 1960s.
of knowledge conversion: Knowledge management as a term was coined
back in the 1990s due to the fact that Japanese
1. Socialization: Convert tacit knowledge companies were at the top compared to other
into tacit knowledge; share experiences companies globally. Knowledge management
(tacit knowledge) is about organizing and managing the knowl-
Example: Two people discussing edge of workers.
2. Externalization: Convert tacit knowl- To begin with, in order for a company to
edge into explicit knowledge; write it down introduce knowledge management, it should
Example: Writing a report first develop a culture of knowledge sharing
3. Combination: Convert explicit knowl- within the organization. However, in order to
edge into explicit knowledge; combine develop a culture of sharing in a company, the
explicit knowledge employees need to trust each other and the
Example: Reading two theories and from company in advance. After the preparation of
them creating a new one the ground, the IT part can take place. Workers
4. Internalization: Convert explicit knowl- then must take a training course on the KM
edge into tacit knowledge; gain experi- software. Additionally, good management of
ence the knowledge database is vital. This can be
Example: Learning from a book achieved by categorizing the knowledge data-
base by subjects and dates, and by means of a
Figure 2 demonstrates the conversion of search engine (Ahmad, 2003).
knowledge according to Nonaka and Takeuchi’s Imagine a library (physical) with a list of
model. books that can be traced via computer software
that indicates each bookshelf’s column code
195
and book details in order for a person to locate sharing and creating knowledge. It requires
and borrow a book. Additionally, there are effort to develop what we call “knowledge
people (librarians) that make sure that the right pull”—a grassroots desire among employees
book is in the right location. A knowledge- to tap into their company’s intellectual
management system without knowledge shar- resources. (Knowledge Portal, http://
ing is like a virtual library that needs librarians www.knowledge-portal.com/people.htm)
to take care of it.
Knowledge management is completely based KM aims at keeping knowledge up to date
on a knowledge-sharing culture inside a com- and correct, providing knowledge in the right
pany or between companies. People, by inter- location, applying knowledge of the most suit-
acting with each other, create communities of able type, and providing knowledge at the time
practice. These people have common interests at which it is needed.
and aims; so, by cooperation, they develop their In other words, KM allows for securing and
own kind of communication and rules. distributing knowledge in order to assure and
People must share and exchange knowl- optimise its availability (Montani & Bellazzi,
edge in order to find the best practice for a 2002).
subject orto innovate knowledge. Once the
knowledge is in the knowledge base, the knowl-
edge analysts can begin seeking, studying, and KNOWLEDGE MANAGEMENT IN
analyzing the information in order to pull the HEALTHCARE
quality knowledge. Afterward, the workers or
professionals can go through the information, According to the Royal College of Surgeons in
and pull and reuse the knowledge. This avoids Ireland (http://www.rcsi.ie), there are 10,000
the duplication of effort and dramatically re- different diseases and syndromes, 3,000 types
duces decision-making time. Knowledge man- of drugs, 1,100 different types of laboratory
agement supports the requirement that suc- tests, and finally 400,000 articles added per
cesses and failures have to be recorded. The annum to the biomedical literature.
knowledge can then be accessed by the use of Furthermore, internal medicine includes 2
technology (Ahmad, 2003). million facts (Wyatt, 2003). The growth rate in
Knowledge management refers to knowl- biomedical literature doubles every 19 years
edge as an asset that the company owns, like, (Wyatt, 1991).
for example, land and machinery. KM supports Doctors’ decisions determine three quar-
the fact that the most important resource in an ters of healthcare costs and depend critically on
organization is the people’s knowledge. As a medical knowledge (Tierney, Miller, Overhage,
popular song says, “people have the power…” & McDonald, 1993). Once knowledge has been
A McKinsey survey outlined in the article captured in some form such as a guideline, it
“Creating a Knowledge Culture” illustrates the can be managed (Wyatt, 2001).
importance to create a need for knowledge: Knowledge in medicine arrives from re-
ports, libraries, experience, guidelines, labora-
Less successful companies tend to take a tory experiments, protocols, practice, group
top-down approach: pushing knowledge to meetings, and so forth. By adding interaction
where it is needed. Successful companies, between these forces, then the outcome can be
by contrast, reward employees for seeking, superior.
196
Because of the above reasons, the use of Professionals can explore and learn faster
knowledge management is critical in the by going through the knowledge database and
healthcare sector. By having a sharing culture studying the cases of other patients and the
with a common aim—the best possible patient specific treatments that were applied to them.
care—and by using an information system, not Furthermore, professionals can apply to evi-
for data warehousing but instead for interpreta- dence-based clinical guidelines.
tion and annotation to create knowledge, then Besides this, each hospital may be posi-
the healthcare sector can be dramatically im- tioned in different places globally, which makes
proved. it difficult for a person to access another
Healthcare organizations need to improve hospital’s knowledge. With KM, research
the quality of patients’ treatment. This includes (healthcare) departments that are carrying out
a decision-support system (best practice) and experiments can provide other hospitals with
the reduction of errors in patients’ diagnoses knowledge by using a KM system.
and treatment. Imagine that a conference on a new infec-
Two kinds of healthcare information exist: tion is to take place, but not all doctors across
information about patients and information about the country are able to attend. By using a
cases. In order for professionals to be accu- knowledge-management system, the doctors
rate, both kinds of information are needed. By that were absent could log in and see a video or
combining these two kinds of information, pro- a tape on a specific subject and contribute later
fessionals are able to come up with the right to that problem by using the KM system.
solution (UCL, http://www.ucl.ac.uk/kmc/re- Moreover, other problems based on the
sources/top_tips.html). management of the hospital’s staff and strate-
By using a successful knowledge-manage- gies could be discussed between managers or
ment system tool, professionals are able to doctors in order to improve the quality of ser-
make conclusions by interpreting and under- vices offered. Afterward, the amount of infor-
standing patients’ data and by carrying out a mation could be written down (explicit) for
successful diagnosis in order for the right ac- future reuse by other hospitals.
tions to be applied. According to Wyatt (2001, p. 8):
Furthermore, by using a KM system, profes-
sionals are able to access the right information The future of knowledge management in
and advice at the right time. However, if there is health is bright. We already have adequate
a difference between a theory in the knowledge technology in the shape of the Internet and
base and the related practice, professionals can a good intellectual framework in evidence-
add comments for later improvement. based health, which are being used to
By implementing knowledge management, improve each other. We also have many
professionals can have access to best prac- health librarians who are knowledge
tices, and if they are able to understand the management professionals.
importance of a knowledge-sharing culture,
then there is also some space for development The importance of knowledge management
and innovation. Furthermore, a KM system can can also be justified by presenting Table 1,
improve the decision-making time and so also which illustrates the differences between the
reduce the costs of a hospital (as a famous old and the new healthcare elements.
slogan says, “time is money”).
197
Table 1. Old and new health-care paradigms • Improved cost management by reducing
(Olson, 2004) the time of patients’ hospital treatment
• Reduced time of decision making by using
Old Health-Care Paradigm New Health-Care Paradigm
• HIS System • Enterprise-Wide a knowledge database and consequently
• Operations Oriented Knowledge-Management
• Hierarchical Database System the reduction of the time of patients’ resi-
• Programmer Centric • Analysis Oriented dence in hospitals
• Silo Based • Relational Database
• Character Based • End-User Centric • Reduced error rate (defects) in decision
• Fragmented Decision • Process Based
Support • GUI (Graphical User making
Interface) and Web Based • Reduced inconvenience
• Holistic Decision Support
• Increased patient satisfaction
Example: Alerts or recommendations de-

CONCLUSION
scribed in literature (McDonald, 1976)
Without doubt, the KM model constitutes a
If treatment includes cardiac glycoside
hopeful innovation in the health sector with
and last premature ventricular
more possibilities and uses. It is about time for
systoles/minute > 2
a qualitative upgrade of health-services provi-
then “Consider cardiac glycoside as
sion to take place. It is obvious that the direct
cause of arrhythmia”
access o all healthcare professionals in KM will
minimise medical errors, decrease the medium
Result: When data matches events, a rec-
duration of hospitalisation, reduce the cost of
ommendation is printed out for the doctor.
hospitalisation, increase the patient’s satisfac-
tion, and have a positive contribution to the
Impact: The frequency with which doctors
relation of cost and effectiveness.
responded to target events doubled from 22%
People have to bear in mind that the health-
to 51%.
services sector has a specific particularity:
Patients are human beings with particular needs,
special characters, and different interactions in
GOALS AND BENEFITS OF therapeutic interventions. KM is a tool that
KNOWLEDGE MANAGEMENT helps healthcare professionals to implement the
IN HEALTHCARE best practices while considering the special
needs of each patient, and sometimes to inno-
• Support in decision making about patients, vate. IT in knowledge management is just the
interference, and the evaluation of re- enabler; the idea is for people to start sharing
search experiences and ideas with a common aim.
• Enhanced healthcare system by improv- Finally, healthcare remains a human science
ing communication between professionals with a strong scientific basis; its consultations
in the decision-making process have such high levels of complexity that they
• Improved patient healthcare probably can never be completely computer-
• Improved quality, access time, and port- ised and automated.
ability of healthcare
• Increased communication between pro-
fessionals and hospitals
198
REFERENCES Tierney, W. M., Miller, M. E., Overhage, J. M.,

& McDonald, C. J. (1993). Physician order
Abidi, S. S. R. (2001). Knowledge manage- writing on microcomputer workstations. JAMA,
ment in healthcare: Towards “knowledge- 269, 379-383.
driven” decision-support services. Interna-
Wyatt, J. (1991). Use and sources of medical
tional Journal of Medical Informatics, 63, 5-
knowledge. Lancet, 338, 1368-1373.
18.
Wyatt, J. (2001). Clinical knowledge and
Ahmad, K. (2001). The knowledge of
practice in the information age: A handbook
organisations and the organisation of knowl-
for health professionals. London: RSM Press.
edge. UK: University of Surrey.
Wyatt, J. (2003). When do we need support
De Lusignan, S., Pritchard, K., & Chan, T.
systems? London: National Institute for Clini-
(2002). A knowledge-management model for
cal Excellence.
clinical practice. J Postgrad Med, 48, 297-303.
Godbolt, S. (n.d.). Moving into knowledge
KEY TERMS
management in the NHS, issue 3. London:
London Library.
Best Practice: The distillation of accumu-
Marco, D. (2003). A meta-data repository is lated wisdom about the most effective way to
the key to knowledge management. carry out a business activity or process. How-
Brookfiled: Enterprise Warehousing Solutions. ever, what is best is highly subjective and
context dependent, so the term implies that no
McDonald, C. J. (1976). Protocol-based com-
further improvements are possible.
puter reminders: The quality of care and the
non-perfectability of man. New England Jour- Community of Practice: A group of people
nal of Medicine, 295, 1351-1355. who share and develop their knowledge in
pursuit of a common purpose or task, even
Montani, S., & Bellazzi, R. (2002). Supporting
though they do not necessarily work in the same
decisions in medical applications: The knowl-
department or organization.
edge management perspective. International
Journal of Medical Informatics, 68, 79-90. Explicit Knowledge: Knowledge that is
codified and articulated. It appears in the form
Myers, F. (2003). Knowledge management in
of documents and procedures and can be found
healthcare: Succeeding in spite of technol-
in databases.
ogy. FL: AFSM International.
IT: Information technology.
Nanaka, I., & Takeuchi, H. (1995). The knowl-
edge-creating company: How Japanese com- Knowledge Analyst: A person or business
panies create the dynamics of innovation. that interprets the needs of a knowledge seeker
New York: Oxford University Press. and finds the most suitable sources.
Olson, E. (2004). Knowledge management & Knowledge Base: A computer-held data-
business intelligence in healthcare. Coppell: base that records knowledge in an appropriate
The Shams Group. format for later extraction. It may take various
forms depending on whether it supports an
199
expert system or contains documents and tex- Tacit Knowledge: Knowledge that is not
tual information for human retrieval. codified but held in people’s heads. Intuitive,
experiential, judgmental, and context sensitive,
Knowledge Management: The explicit
it may be difficult to articulate.
and systematic management of vital knowledge
and its associated processes of creation, gath-
ering, organization, diffusion, use, and exploita-
tion in pursuit of organizational objectives.
200
201
Chapter XXVI
Knowledge Management
in Medicine
Nikolaos Giannakakis
Efstratios Poravas
ABSTRACT
In the last decades, the amount of information has risen because of the technology revolution.
The need for organizing information, in a way that the staff and the managers of a hospital
require, lead to the generation of a new value, the knowledge management. Its benefits are
sensible, not only for the staff, but also for the hospital as an entity. Many techniques are
applied to solve all the daily problems in the health sector.
INTRODUCTION management is an organism that is constituted

of small parts that aim to collect, assess, unify,
During the last three decades of the previous improve, and produce value from intellectual
century, there was a revolution in technology and information-based resources (Association
and its applications in medicine and the field of of State and Territorial Health Officials
information. The promotion of knowledge and [ASTHO], 2005).
its communicability certainly have profited all Progress in medicine is essential; there has
the scientific sectors, both in the increase of always been the mass production of knowl-
efficiency and productivity, and in the growth edge, and those who are related to this science
of innovations. should take it into account, develop it, and apply
Knowledge management is a notion that is it. Traditional sources of information are avail-
difficult to define. A lot of definitions have been able, but they usually fail to provide answers
formulated, one of which is that knowledge whenever and wherever they are needed. Thus,
Knowledge Management in Medicine
roughly two thirds of problems in clinical prac- clinical process in the providing of medical care
tice remain unsolved (Gale Group, 2001). Un- is rather obvious (Stefanelli, 2002). Thus, knowl-
fortunately, the information and knowledge that edge management in the field of medicine fo-
are available to doctors nowadays are poorly cuses on the knowledge of technologies used in
organized and old. clinical, administrative, and demographic activ-
In the healthcare field, doctors and patients ity. Today, the pressure of cost also influences
need help with the choice of better actions for the sector of health. The adoption of informa-
a given situation. The rate of growth and change tion technology is considered one of the basic
in worldwide biomedical knowledge leads to mechanisms for the reduction of cost (“Data
the fact that no one is able to know the current and Knowledge Management in Healthcare,”
practices in a sector without any kind of support 2005).
(Purves & Robinson, 2003). A partial solution The society of public health is continuously
to this problem can perhaps be brought by focused on digital communication for the fulfill-
medical knowledge management. Its aim is the ment of different kinds of tasks. Although
regrouping, incorporation, and connection of technology has improved for the possibility of
any medical knowledge that was produced in the collection, analysis, and dissemination of
the past in order for one to reach a reasonable data, there are still obstacles in the use of
decision in the present and useful study in the information, such as the existence of informa-
future (Quantum Enterprises, Inc., 2003). tion that is not well organized and systems that
Generally, knowledge management in the are not complete. The continuous improvement
medical field can ensure the effective growth of technology, the lack of resources, the failure
and dissemination of better practices, and a to confirm the requirements of data, and com-
continuous assessment aiming at their improve- plicated data have led experts to the use and
ment. Knowledge is created within time. All the exploitation of existing knowledge for the pro-
data that constitute its base become informa- motion of health (Data and Knowledge Man-
tion when they can be summarized and orga- agement for Public Health, 2005).
nized under reasonable models. Information Through knowledge management we can
becomes knowledge when it can be managed certainly reduce the gap between the lack of
for active decision making, and knowledge can data and the lack of systems that develop those
be turned into perspicacity when it is well data. Starting with the presumption that every
developed within regular periods of time problem has a solution, the effective manage-
(Lobodzinski & Criley, n.d.). ment of knowledge in the health sector can
constitute the base of knowledge, which is
essential for the presentation of its innovations
USE AND VALUE OF and distribution in a dynamically regenerative
KNOWLEDGE MANAGEMENT process (Bailey, 2003). Generally, there is a
framework in hospitals that can be used as a
The continuous effort toward efficiency and driver for the management of knowledge. This
economic effectiveness creates a balance is a methodology that helps with the designing
among the quality of provided services, and it of a strategy and its processes, and that en-
includes costs that lead to the more effective riches the transmission of knowledge and tools
management of medical knowledge that is de- that support the collection and analysis of knowl-
rived from biomedical research. The need for a edge, and the storage and search of informa-
202
Table 1. Causes of errors (Institute of Medicine [IOM], 2000)
• Technical errors 44%

• Incorrect diagnosis 17%
• Failure in damage prevention 12%
• Errors in pharmaceutical contact ion 10%
Table 2. Most frequent errors (IOM, 2000)

• Incorrect issuing of medicines 28%
• Errors in medical protocols 22%
• Administrative errors 4%
tion. All these occur in collaboration with the technological and scientific growth, which has
personnel, the economic resources, suitable been observed during the last few decades in all
systems that allow communication in all kinds the scientific sectors, the medical errors, and,
of situations, and the infrastructure so that the generally speaking, all kinds of errors, that
maximum effectiveness, efficiency, and cre- occur in hospitals remain frequent and rather
ativity are ensured (Managing Knowledge to expensive. In the year 2000, the Medical Insti-
Improve Reproductive Health Programs, tute reported that medical errors cost the medi-
2004). cal industry and the pharmaceutical industry
$37 billion annually, and 100,000 people lose
their lives each year because of these errors
BENEFITS OF (Detmer, 2001).
KNOWLEDGE MANAGEMENT Table 1 presents the basic reasons that led
to medical errors.
In a hospital, even more than in any other Table 2 presents the types of errors that
organization, knowledge management becomes occur most frequently.
a necessity since a vast number of research by Table 3 presents errors due to the medical
global organizations, both in Europe and the personnel.
USA, has proven that a lot of people have died It is obvious that there are plenty of medical
and a great part of the government budget was errors, they cost the state a lot of money, they
spent due to errors, which would probably have can be dangerous for the lives of patients, and
been avoided with the use of knowledge man- they basically occur due to human factors.
agement. Even though there has been major About 70% of these errors could have been
203
Table 3. Errors due to medical personnel (IOM, 2000)
• Negligence/carelessness 29%
• Inexperienced/uneducated personnel 14%
• Communication 12%
• Tiredness of personnel 8%
• Illegible recipe or incorrect issuing of medicines 6%
• Other errors 14%
avoided or even prevented. According to the how we can facilitate the transport of knowl-
Medical Company of the USA, in the year 2000, edge to it. It is crucial to improve the perception
250,000 deaths were caused due to medical of all the factors involved in the field of health
factors (JAMA, 2000). Even though the num- and to support the new methods and technolo-
ber seems rather small compared to the entire gies, which are not always accepted easily by
population of the USA (about the 0.07% of the the personnel of hospitals.
population), it is very important and should have The main objective of knowledge manage-
called to action all experts in the health sector. ment is to help in the briefing of our knowledge,
Knowledge management basically aims to re- and to aid its enrichment and exploitation. To-
duce expenses and costs, as well as to increase day there are systems that aim at improving and
productivity and efficiency, which will lead promoting knowledge. This can be achieved
indirectly to the reduction of human losses. through the review of practical progress in
Each employee should consider his or her knowl- order to find, develop, disseminate, and use the
edge so we can trust him or her when it comes knowledge for the profit of doctors and pa-
to the resolution of problems, the avoiding of tients. Furthermore, these systems achieve an
errors, and the assuring of positive results and approach among the experts in different kinds
practices. of sciences facilitating exchange between the
existing and the new techniques in the field of
health (Advances in Clinical Knowledge
OBJECTIVES OF Management Workshops, n.d.).
KNOWLEDGE MANAGEMENT: Table 4 presents examples of techniques in
PROBLEM SOLVING knowledge management in the medical field.
From Table 4 comes the conclusion that the
Knowledge management has a very positive main objectives of knowledge management are
impact on all sciences. Its aim is to serve each the isolation and later development of the struc-
science separately and minimize the errors tures of knowledge, the production of its gradual
derived from the lack of knowledge. Generally, structures for the description of a future con-
we should choose which information is essen- nection, the proposal to create functional sys-
tial in an organization, such as a hospital, and tems for the development of medical knowl-
204
Table 4. Examples of techniques in knowledge management
• Best practices in the management of special medical situations

• Analysis of indications for those situations
• Reports on existing and new medicines
• Development of direct and explicit information regarding complicated medical
subjects
• Use of special vocabulary and a common language for names of medicines,
illnesses, and so forth
• Determination of the available sources
• Analysis of data, taking into consideration the danger factors
• Ability for direct access to the data anytime, anyplace
• Information regarding actions that should be done in case of emergency
• Creation of protocols under which people in the health field should work
• Research and better practices for the settlement of medical subjects
• Research for new diseases that appear in society
• Finding of information related to the health conditions of a population in a given
geographical zone
• Thorough interconnection of information among multiple choices, for example,
handbooks and depictions
• Access to descriptions of diseases, practices, and research reports
• Knowledge regarding treatments that can be applied
• Access to informative pages of the hospital, and more important for other
hospitals, the acquisition of data and comparison of situations
• Unified communication and informative sending-receiving data systems
(ASTHO, 2005)
edge, better comprehension of problems and and professional management of information is

decision making, and the application of knowl- required by all professionals (Sozou, 1998).
edge management in medical information tech- Knowledge is undefined, is precious in its
nology in all clinical departments and health acquisition, and can be easily lost or stolen in
systems. order to be converted or erased afterward. In
the world of technologies nowadays, where
everything tends to function digitally, a lot of
SAFETY OF mechanisms as well as structures are used for
KNOWLEDGE MANAGEMENT the protection of knowledge. A model that can
ensure knowledge management should ensure
The information conveyed to doctors is turned its protection as well. In other words, it is
later into knowledge, either through the devel- essential in an informative system to ensure the
opment of computer-based systems or through authenticity of knowledge; the safety of data
the development of tools that will be used by from any kind of involuntary or voluntary modi-
experts in the field of information in collabora- fication, destruction, or revelation of its ele-
tion with clinical groups. The knowledge and ments; the integrity of data; and the general
information by themselves are not in a position application of political safety to information
to improve clinical practices. The organized (Mundy & Chadwick, n.d.).
205
Table 5. Advantages of knowledge management for hospital personnel
• To avoid errors that happened in the past

• To reduce the time needed for the detection of information, mainly for urgent
incidents
• To assure the best clinical decisions by inexperienced employees
• To find alternative procedures of care in emergency situations
• To identify a lack of information, which can lead to errors
• To encourage the flow of ideas, which leads to innovations
• To avoid unnecessary procedures and increase cohesion and collaboration among
employees
• To increase productivity and efficiency
Table 6. Advantages of knowledge management in a hospital
• Improvement of quality of sanitary care through the increased efficiency and

productivity of personnel
• Facilitation of communication between medical personnel
• Reduction of expenses mainly from the pointless use of medicines and
unnecessary insertion of patients in the hospital
• Creation of strategies for a vast number of data
• Contribution of hospital profits and better use of money, which would be
unavailable without knowledge management
ADVANTAGES IN THE USE OF DISADVANTAGES IN THE USE

KNOWLEDGE MANAGEMENT OF KNOWLEDGE MANAGEMENT
Knowledge management has a lot to offer to a Knowledge management has a lot of advan-
hospital, and even more to its personnel. A tages, but also presents a lot of disadvantages.
technical and organizational infrastructure can If there is not an explicit determination of
be achieved. Table 5 presents its advantages knowledge and we do not recognize its impor-
for the personnel of a hospital. tance, then we will not be able to achieve its
Table 6 presents the major advantages of appropriate management. In some cases, we
knowledge management in a hospital. focus on the past and the present and care less
206
Table 7. Disadvantages presented by knowledge management due to certain reasons
• Knowledge in medicine is incomplete, vague, and inaccurate, and requires many

years for its acquisition.
• There is a continuous alteration of knowledge as technology develops.
• There is excessive emphasis on the reserve of knowledge and not on its flux.
• Doctors usually forget the aim of knowledge management and do not realize the
complexity of the knowledge that is distributed.
about the future, giving emphasis on counting as well as established policies and processes.
knowledge and not on its results. This access will decrease the possibility of
Generally, despite the increasing power of costly medical errors and will promote healthcare
information technology, knowledge manage- for patients in the hospital (Detmer, 2001).
ment in the form of supporting decisions and It is very important to create a network
information concerning doctors has minimum aiming at collaboration among experts and at
impact on the results of healthcare. This occurs the improvement of methods for the collection
due to the fact that knowledge management and analysis of data both for the present needs
addresses the relationship between patients and the forthcoming ones. The future of medi-
and personnel; it does not provide harmony cal knowledge management is in the present.
between the complexity of individuals and their We should take advantage of the revolution of
regular actions that form special practices technologies and information that has begun
(Purves & Robinson, 2003). during the last decades (Gale Group, 2001). In
Usually the efforts toward the medical man- the future, in order to reach the use and appli-
agement of knowledge are underestimated. As cation of knowledge, we should analyze the
far as the systems are concerned, an uncon- existing aims, the types of knowledge, the users
trolled increase of medical knowledge can lead and their sources, the processes, and the tech-
to disaster (Quantum Enterprises, Inc., 2003). nologies (Bouthillier & Shearer, 2002).
Table 7 presents some of the reasons why
knowledge management is not efficient and
presents disadvantages. CONCLUSION
Knowledge management is a basic tool for all

THE FUTURE OF those working in the health field and for hospi-
KNOWLEDGE MANAGEMENT tals. It helps in sending the right information to
the right person at the right time so that the right
Through the practices and techniques, doctors decisions can be made depending on the exist-
can have easier access to new information ing problems. It is certain that with the help of
concerning medicine and their applications, di- knowledge management, effectiveness in the
agnostic tests, and the way diseases are treated, health field will be increased through unified
207
systems, processes, and methods; the cultiva- Lobodzinski, S. M., & Criley, M. (n.d.). Medi-
tion of exchanging knowledge; and the promo- cal knowledge management. Torrance: Cali-
tion of the effective use of available informa- fornia State University Long Beach & Harbor
tion. UCLA Medical Center.
Managing knowledge to improve reproduc-
tive health programs (MAQ Paper No. 5).
REFERENCES
(2004).
Advances in clinical knowledge manage- Mercola. (2000). Journal of the American
ment workshops. (n.d.). Retrieved from http:/ Medical Association (JAMA), 284(4), 483-
/www.ucl.ac.uk 485.
Association of State and Territorial Health Mundy, D., & Chadwick, D. W. (n.d.). Secure
Officials (ASTHO). (2005). Knowledge man- knowledge management for health care or-
agement for public health professionals. ganizations. Retrieved from http://
sec.isi.salford.ac.uk/download/SecureKM.pdf
Baily, C. (2003). Bulleting of the World Health
Organization, 81(11). Purves, I., & Robinson, P. (2003). Knowledge
management for health. Medical Education,
Bouthillier, F., & Shearer, K. (2002). Under-
37, 429-433.
standing knowledge management and informa-
tion management: The need for an empirical Purves, I., & Robinson, P. (2004). Knowledge
perspective. Information Research, 8(1). management for health programs. Medinfo,
Retrieved from http://Information.net/ir/8-1/ 678-682.
paper14
Quantum Enterprises, Inc. (2003). Knowledge
Data and knowledge management in healthcare. integration and management. Retrieved from
(2005). Proceedings of the 38th International http://www.quantument.com
Conference on System Sciences.
Sozou, P. (1998). Advances in clinical knowl-
Detmer, W. M. (2001). Medical knowledge edge management. London: University Col-
management solutions: Revolutionizing the lege.
delivery of medical information to the point
Steffanelli, M. (2002). Knowledge manage-
of need. Retrieved from http://
ment to support performance-based medicine.
www.unboundmedicine.com/
Methods of Information in Medicine, 41(1),
healthcarereview.htm
36-43.
Gale Group. (2001). Healthcare review. Re-
trieved April 2005, from http://
KEY TERMS
www.findarticles.com
Institute of Medicine (IOM). (2000). To err is Data: The representation of facts, con-
human: Building a safer health system. Re- cepts, or instructions in a formalized manner
trieved from http://www.iom.edu./object.file/ suitable for communication, interpretation, or
Masater/4/117/ToErr-8paper.pdf processing by humans or by automatic means.
208
They are also any representations such as knowledge base consists of the explicit knowl-
characters or analog quantities to which mean- edge of an organization, including troubleshoot-
ing is or might be assigned. ing, articles, white papers, user manuals, and
others. A knowledge base should have a care-
Empirical or A Posteriori Knowledge:
fully designed classification structure, content
Prepositional knowledge obtained by experi-
format, and search engine.
ence. It is contrasted with a priori knowledge,
or knowledge that is gained through the appre- Knowledge-Based System: A program
hension of innate ideas, intuition, pure reason, for extending and querying a knowledge base.
or other nonexperiential sources.
Knowledge Management: Caters to the
Information System: A system, whether critical issues of organizational adaptation, sur-
automated or manual, that comprises people, vival, and competence in the face of increas-
machines, and methods organized to collect, ingly discontinuous environmental change. Es-
process, transmit, and disseminate data that sentially, it embodies organizational processes
represent user information. that seek the synergistic combination of data
and the information-processing capacity of in-
Knowledge: A fluid mix of framed experi-
formation technologies, and the creative and
ence, values, contextual information, and ex-
innovative capacity of human beings.
pert insight that provides a framework for
evaluating and incorporating new experiences Knowledge-Management System: A dis-
and information. It originates and is applied in tributed hypermedia system for managing knowl-
the minds of knowers. In organizations, it often edge in organizations.
becomes embedded not only in documents or
Knowledge Relativity: The relation be-
repositories, but also in organizational routines,
tween a form of representation, two sorts of
processes, practices, and norms.
intent (communication and use goals), and three
Knowledge Base: A special kind of data- subjects (one who knows, one who is informed,
base for knowledge management. It is the base and one who observes and confirms).
for the collection of knowledge. Normally, the
209
210
Chapter XXVII
Knowledge Management
in Telemedicine
Jayanth G. Paraki
Telemedicine Research Laboratory, India
ABSTRACT
Knowledge management (KM) can be defined as the discovery and dissemination of new
knowledge. It has also been defined as the efficient utilisation of the existing intangible
knowledge-related resources available in every sector of the economy to enhance the
productivity of all factors of production. Telemedicine is a tool to enhance equitable
distribution of healthcare across the world. In this chapter the author discusses the various
aspects of knowledge management and telemedicine and proposes to globalize telemedicine.
INTRODUCTION charter points. Leaders of Internet business

houses will derive rich benefits from applying
Organizations all over the world are adapting to Deming’s TQM to their organization, while
rapid changes in many ways, and an approach investors will find it comfortable to interact with
that has made significant contributions to the such organizations.
resurrection of postwar Japanese industry is There is an urgent need for healthcare pro-
Deming’s philosophy of total quality manage- fessionals to be multiskilled. Doctors and nurses
ment (TQM). Deming’s TQM is indeed appli- have to introspect periodically and adopt a plan
cable to the management of Internet busi- for self-improvement that should include ac-
nesses and eminently suitable for telemedicine quiring new knowledge, developing a positive
and data-mining projects in life sciences. The mental attitude, and learning new skills to meet
apparent complexity of the system may appear the rapidly changing health needs of people
as a deterrent to many, but the inherent simplic- across the globe.
ity is clear on examination of the Deming
Knowledge Management in Telemedicine
The methods of diagnosis and treatment of • The right process,

human diseases have altered significantly in the • To the right population,
last decade. In the early 1990s, the first signals • At the right time,
of change began to surface with patients ex- • Through the right channel.
pressing dissatisfaction and discontent with one
or more aspects of their care. Unknown to their It is obvious that there are many elements of
primary-care physicians, they began to talk equal importance that have to be brought to-
about their experiences with the hope that a gether in order to succeed at biomedical com-
degree of attitudinal change would emerge in puting and data mining. What then has to be
some physicians, and permanent solutions to done to amalgamate these elements into a
chronic health problems would start emerging. single whole in order to begin the process of
Direct patient-doctor communication was not data mining? Simply stated, if we have a fool-
in evidence, and institutions committed to proof system and the necessary human intelli-
healthcare chose not to encourage free discus- gence, it is possible to build a data warehouse
sions with their clients. This pattern persisted and then mine it into saleable data marts for
over the best part of that decade and led to global consumption. I believe that the global
strained communications between physicians scientific community possesses both the sys-
and patients. This at one time resembled an tem and the intelligence to do this successfully.
estranged love affair, with neither party willing The purpose of this chapter is to highlight
to move forward to resolve the conflict. How- the capability of Deming’s total quality man-
ever, before long, the Internet revolution be- agement to provide the directions and impetus
came a phenomenon and made its presence felt to establish a framework for data mining in life
in healthcare activities. Curiously, the initial sciences, and to demonstrate through a simple
recognition of the value of this phenomenon application the different ways to apply Deming’s
was made by suffering patients who realized 14 charter points in actual practice (Creech,
the Internet was a handy tool to share their tales 1995). The future possibilities are varied and
of woe amongst themselves and to obtain a some thoughts will be shared toward creating
degree of temporary relief and comfort. It was an international consortium for the research
fortunate that some physicians had woken up to and development of telemedicine applications
this phenomenon simultaneously and had begun (Paraki, 2001).
to address the different scientific aspects in a
proactive manner.
During the period of 1998 to 2000, the THE CONTEXT
undercurrents of rapid information exchange
were being felt, with different nations discover- In the last decade, observations of results of
ing its effects through a variety of experiences. treatment with allopathy show it is incomplete
Limiting the discussion on the influence of the and inadequate. This applies to both the outpa-
Internet phenomenon on healthcare, we begin tient and inpatient care of those with acute and
to see certain common patterns emerging across chronic ailments. Patient dissatisfaction is evi-
a wide spectrum of diseases across the globe, dent in patients with chronic diseases such as
making data mining an exciting and profitable arthritis, bronchial asthma, and many functional
activity. The need is to automate: disorders like migraines, obstinate constipation,
211
and irritable bowel syndrome. Furthermore, ex- based medical charts to large segmented health
tensive interaction with professionals from other systems that fail to provide for adequate com-
disciplines like management, IT, telecommuni- munications among caregivers. Medical care
cations, fundamental sciences, psychology, and today can be a complex combination of thera-
philosophy has provided a solid basis for the peutic treatments and sophisticated drug regi-
development of holistic medicine as a preferred mens administered by many different healthcare
approach to the prevention and treatment of professionals at different sites. Under these
acute and chronic disease. This is keeping in conditions and given more than ample opportu-
tune with the current changes in developed nity for human error, it is no mystery we have
nations such as the USA and those in Europe a problem. Long-term solutions require consid-
where holistic medicine is gaining popularity. erable political will and the ability to attract
Management in this context not only implies champions who can help craft answers that will
the way hospitals and health universities func- not cripple the process with malpractice litiga-
tion, but also the need for a change in the tion. The Institute of Medicine has sounded the
healthcare systems and processes themselves. alarm. Are we prepared to answer it?
Data mining in life sciences will truly and
correctly provide the right evidence and direc-
tion to nations keen on making holistic medicine WHY DEMING?
a viable, cost-effective alternative to meet their
healthcare challenges and needs. The 1994 Survey on Change Management
In the year 2000, in a report titled To Err is published by the AMA and Deloitte & Touche
Human: Building a Safer Health System, the says, “It seems that many organizations have to
Institute of Medicine estimated that as many as change in order to change. Their present struc-
98,000 deaths per year occur in the United tures and cultures tend to disallow the success-
States because of medical errors that could ful implementation of change initiatives.” I find
have been prevented. The additional cost in the same in the healthcare industry, too. This
pain and suffering as well as in dollars—esti- change is of a global dimension, and Deming’s
mated between $17 billion and $29 billion—as a TQM (Creech, 1995) will be able to bring many
result of preventable medical errors is uncon- organizations and nations together on a com-
scionable in an age where the technology exists mon platform to strive for a higher global cause.
to virtually eliminate such occurrences. Worse, Telemedicine technology coupled with Deming’s
these studies account only for hospitals. They TQM will ensure that change initiatives will
do not begin to address the errors and their meet with success. Networked organizations
resulting consequences in other healthcare set- thrive better than others and are ideal for live
tings such as nursing homes, home care, day projects in life sciences (Woodcock, 2001).
surgery, outpatient clinics, and care delivered in
doctors’ offices. Application of Deming
According to the Institute of Medicine, the Charter Points
majority of medical errors do not result from
individual recklessness, but from “basic flaws Point 1: The constant improvement of health
in the way the health system is organized.” services is a must. Technology integration
These flaws reach into every aspect of the speeds up the process. Holistic medicine offers
healthcare enterprise, from illegible, paper- paths to total quality health for those with the
212
following diseases: cancer, AIDS (acquired What is data mining? Data mining, by its
immunodeficiency syndrome), and mental de- simplest definition, automates the detection of
pression. TQM is ideal for securing global relevant patterns in a database. However, data
cooperation to fight these diseases with vigor mining is not magic. For many years, statisti-
and vitality. cians have manually mined databases, looking
Point 2: The philosophy of holistic medicine for statistically significant patterns. In tradi-
is the need of the hour. There are different tional business use, data mining uses well-
schools of thought, and each nation has to established statistical and machine-learning
harmonize their philosophical thinking with sci- techniques to build models that predict cus-
entific temper to allow proven benefits to reach tomer behavior. When discussing models in life
the masses. sciences, the definition of a customer under-
Point 3: Dependence on laboratory tests goes a bit of a modification in the sense that a
alone is to be avoided, and close personal patient is not viewed as a customer, and a
interaction with patients is a must. Further- customer need not be a patient. The need for
more, the careless repetition of tests, too-fre- building models in life sciences exists in diverse
quent tests, and inappropriately timed tests are situations, and when viewed globally, is a healthy
some of the sources of avoidable drain of exercise physically, mentally, and financially.
financial resources to the individual and the
organization. Herein lies one of the values of a Data Mining and Data Warehousing:
robust, correctly designed informational data- The Connection
base. Later, we will see its relationship to an
operational database, which is a part of a data Data mining describes a collection of tech-
warehouse and its different uses. niques that aim to find useful but undiscovered
Point 4: I have deliberately left out ad- patterns in collected data. The goal of data
dressing this point at this stage. I believe that mining is to create models for decision making
this is a very personal issue at this stage of that predict future behavior based on analyses
global telemedicine technology development, of past activity. Data mining supports knowl-
and while there are several financial and rev- edge discovery, defined by William Frawley
enue models that have proved successful, none and Gregory Piatetsky-Shapiro (1991) as “the
have evolved fully to become a reference model nontrivial extraction of implicit, previously un-
for global following. Continued learning and known, and potentially useful information from
mature knowledge sharing is essential to enable data” This definition holds the key to the sub-
unshakeable clarity of perception to be gained stantial interest in data warehousing and data
in the economics of telemedicine technology,1 mining. The method used today in data mining,
and this may well prove to become a major when it is well thought out and well executed,
aspect of global research study in the future. consists of just a few very important concepts.
Point 5: Improve constantly and forever The first of these is the concept of finding a
the diagnostic and therapeutic processes. To do pattern in the data. In many cases, this just
this successfully, baseline data have to be re- means any collection of data that occurs a
corded, stored, and retrieved at will. Database surprising number of times. Usually, surprising
systems and data-mining methodology will in- is better defined, but in general it means any
fluence the outcome of disease research and sequence or pattern of data that occurs more
hence is of overriding importance. often than one would expect it to if it were a
213
random event. For example, the occurrence of the use of data-mining technology to gather
breast cancer in unmarried women in the age data from many different breast-cancer cen-
group 30 to 40 is significantly higher than in ters around the world.
married women of the same age group. What is Point 6: Institute training for both vendor
the relationship between marriage and the inci- and client organizations is of vital importance
dence of breast cancer? The answer lies in the vital importance if the intended product or
simple fact that in unmarried women, the in- service is to be marketed successfuly.
tended natural function of pregnancy and sub- Healthcare professionals engaged in
sequent child rearing being denied renders them telemedicine activity can ill-afford to neglect
more susceptible to the development of breast this vital exercise. The ranges of skills are
cancer. This piece of data makes it easy to varied and are determined by functional needs.
suspect early breast cancer in susceptible Training can be as simple as learning how to use
women in the age group 30 40, and alerts Microsoft Word or as complex as learning
doctors and/or patients to request for further Oracle Database Management. Nevertheless,
tests to confirm or refute the suspicion. Data training flows seamlessly with the rest of the
mining will throw further light on various sub- activities of the organization and does not usu-
groups within this larger main group, allowing ally pose a serious problem.
far greater precision and control in designing Point 7: Adopt and institute leadership
breast-cancer screening programs and the clini- (Adair, 1993). Successful Leaders are those
cal evaluation of various treatment protocols who convert ideas, contexts, and discoveries
including that drawn from alternative and into practical undertakings. Greater success
complementary medicines. Another very im- will be achieved through the optimal use of
portant concept associated with data mining is technology and team functioning. Highly fo-
that of validating the predictive models that cused efforts with minimal scattering of ener-
arise out of data-mining algorithms. For ex- gies are more likely to produce noticeable re-
ample, if lesbian women remain unmarried in sults than diverse activities aimed at several
their lifetime and if it is proved that the inci- goals. Scaleable-systems development is the
dence of breast cancer is statistically higher in need of the hour in life sciences, and it is
them, then an intervention program based on essential to recognize what the ideal fundamen-
alternative and complementary medicines is tal system is.
already available for implementation and may Point 8: Drive out fear. Fear as an incen-
prove valuable in the age group 20 to 30 to ward tive to improve work performance has been
off future threat from this dreaded disease. explored at length. Universally, it has a finite
There is a growing body of scientific literature limit to induce consistent elevation of perfor-
attesting to this fact, and multination clinical mance or to produce quality products and ser-
trials are justified to validate the data. If the vices. Deming realized this during his experi-
four parts of data-mining technology are pat- ence in Japan after World War II and stressed
terns, sampling, validation, and the choosing of the need for leaders to arrive at suitable meth-
the model, then the study of the incidence of ods to drive out fear from within their organiza-
breast cancer in lesbian women and its diagno- tions. One of the possible methods is to struc-
sis and treatment is warranted and justified, and ture the primary organization with care with
is suitable for multination trial since the disease attention to minute details as regards the func-
is ubiquitous in distribution. Equally suitable is tional abilities of the hired or chosen people, and
214
to put in place a system of reward and punish- workforce, and this reminds us of the statement
ment that is attitude based rather than incentive of Raymond Miles: “The greatest barrier to
based. Team working and collective conscious- success will be outmoded views of what an
ness also help to minimize and eliminate fear ‘organization’ must look like and how it must be
progressively. managed.”
Point 9: Break down barriers between staff Points 11, 12, and 13 are a cohesive
areas.Academic and research activities are interlinked movement of earlier thought pro-
often slow due to several barriers that exist cesses toward the strong-willed decisive action
within a group. While some are necessary and of the transformation and implementation of a
healthy, others are simply laborious and en- process that has proven benefits in global
ergy sapping and are better removed than healthcare. Action at the end will serve to
allowed to exist undetected. In the modern demonstrate the validity of the expressed sci-
information age, intellectual property rights entific thinking and stimulate others to follow
have assumed greater relevance than ever similar lines of thinking, paving the way for the
before and appear to govern conscious and Internet revolution to make a rapid foray into
unconscious thinking to a large extent. Never- the domain of life sciences in a convincing and
theless, it is mandatory for decision makers to incisive manner.
be alert to new opportunities in their primary Point 14: Take action to accomplish the
or related disciplines, and one way to do this transformation.
successfully is to have highly trained and
trustworthy team members to function unin- Data-Warehouse Architecture
hibitedly and spontaneously at all times. for a Data-Mining Project in
Point 10: Eliminate slogans, exhortations, Life Sciences
and targets for the workforce. The workforce
looks at patient care as a divine activity and is Operational and External Data
prepared to go the extra mile. The attitude of
the doctors and nurses is crucial in the affairs of See Figure 1. Example outputs from a data
a hospital. They are the pillars of strength of warehouse include the following:
any hospital, and active participation at all
levels of the workforce is the key to success. 1. Homeopathic history to maintain transac-
This is determined by the nature of the relation- tions
ship that exists between top management and 2. Analysis of homeopathic histories to iden-
the lower levels of professional and nonprofes- tify patterns
sional staff in a hospital. Charter Point 10 is 3. Identifying progressive disease patterns
probably the most vexatious issue in any orga- to initiate critical treatment measures
nization and it is true for hospitals also. A 4. Mobilizing work-flow processes for emer-
practical approach to minimize this problem is gency and critical situations like natural
to provide the ownership of smaller satellite disasters
hospitals and clinics to teams of doctors and 5. Recording the evolution of cancers from
nurses and encourage them to independently many patients within a nation and from
manage and produce results. The structure of many nations to create a data background
the parent organization is critical to enable such for the comparison and prediction of out-
a relationship to be developed with the comes of treatment plans, for the study
215
Figure 1.
Data
Information
Extract Delivery
Data Systems
Cleanup
Data Load 3 Data marts
4
1 Tools
and analysis of demographic patterns, and trolled and unrestrained biological activity that
so forth alters the basic human functions and eventually
6. Creating self-help treatment processes kills the host. Hitherto, animal models were
for widespread ailments like chronic nasal employed to understand the biology of cancer
allergies and extrapolate the results to humans. The
7. Creating products and services for ho- animal model has failed to make us fully aware
meopathic students in training of the origins of human cancer.
8. To initiate global action against wide- Several models of breast cancer are cur-
spread epidemics like gastroenteritis, mal- rently in use, each of which embodies some
nutrition, tuberculosis, and so forth accepted criterion. However, a universal model
does not exist simply because the criterion
Relevance of Data Mining to adopted varies enormously from center to cen-
Cancer Research ter, and also because the nature of the criterion
selected is difficult to determine all over the
Models of Human Breast Cancer world. Therefore, it is necessary to establish a
set of criteria that is fairly universal and funda-
The lack of absolute knowledge of the developmental in order to proceed with a meaningful
ment and progression of cancer has resulted in research program involving large numbers of
many empirical and hypothetical treatment people and many breast-cancer research cen-
models. The number and type of questions that ters.
can be answered by controlled trials are limited
by logistical and ethical considerations, and it is Types of Models Available
for this reason that models of human cancer are
sought. Cancer is neither a single disease nor a The following models are commonly employed
group of diseases. It is a phenomenon of uncon- in many research centers. They are the canine
216
model, rodent model, and human model. Models The nature of experimental data is different
are necessary to formulate a scheme for ob- from either epidemiological or clinical data.
taining, analyzing, and testing data in appropri- Epidemiological data is easy to collect and
ate environments in order to apply the results to should be drawn from many nations to provide
a larger population. a definite and conclusive base to expand on
seamlessly. The attributes of this database are
The Canine Model a combination of general and peculiar charac-
teristics spread across several age groups. The
Breast tumors occur commonly in dogs, and peculiar characteristics determine how an inju-
these tumors represent a close model of human rious agent produces its effect over time. Con-
breast cancer in morphology, clinical behavior, trary to the traditional allopathic model of can-
and incidence. However, by virtue of its close- cer causation, in which there is not much em-
ness to the human situation, this model includes phasis on a continuous cause operating in the
all the variables encountered clinically and ne- background, the holistic integrated model of
cessitates the same staging of tumors and cancer causation believes in the existence of
randomization of subjects as applied in human smaller cause-effect changes occurring all the
clinical trials. In addition, the time required for time, albeit imperceptible to normal human
the evaluation of treatment modalities is long, senses. This resembles a chain of events that
and together these factors make the cost of occurs in all the links of a computer network
using the model to obtain statistically significant when the main switch is turned on.
results too great. A. M. Neville (1981) has some insights into
the developments in research with relevance to
The Rodent Model human breast cancer. One overriding conclu-
sion is that we must try to escape from viewing
Aside from using different methods to study breast disease in a conventional sense as this
cancer in rodents, the overall results are not approach to date has failed. It seems more
significantly different from the canine model important to detect those factors that are in-
and hence are not suitable for incorporating into volved intimately in the control of the growth of
a data warehouse. breast cells, both normal and neoplastic. An
understanding of those factors, and how and
The Human Model where they act will allow us to gain a greater
appreciation of the progression of this disease
While it is reasonable to assume that the basic and how it may be modulated to the patients’
mechanisms of hormone action and metastasis advantage. Growth factors and their receptors
may be similar in human and rat breast cancers, draw attention once more to the properties of
it is less probable that they should show similar the cell surface. Increased appreciation of this
sensitivities to chemotherapeutic agents. It is aspect of cell biology and biochemistry may
for this reason that a model system that enables well have value in the diagnosis and monitoring
human material to be used for experimentation of the disease. Without such knowledge, it is
would be of value. Significant developments in unlikely that we will be able to devise meaning-
this direction are taking place in many centers ful and rational approaches to therapy in the
across the world, and a large amount of scat- coming decade.
tered data is available for incorporating into a Hopefully, with the advent of RAD tools,
central data warehouse. research time can be reduced and results made
217
more meaningful and of a uniform quality in all there is no other activity as important as deci-
the participating centers in a multination clinical sion making in our daily lives. The varied and
trial. There is reason to believe that light can be complex process of decision making is at the
shed on perplexing and controversial aspects of soul of all activities in life sciences.
breast-cancer treatment within 3 years of be- Data mining in life sciences with emphasis
ginning an accurately designed study. on cancer and AIDS will prove to be a profit-
able e-business opportunity for the current
millennium. One of the many interesting busi-
SOME ISSUES ADDRESSED AT ness aspects of this opportunity is the pro-
THE TELEMEDICINE RESEARCH found transformation it can bring about in
LABORATORY personal and business relationships, replacing
old and worn out ideas with fresh and revolu-
1. Infrastructure development for e-business tionary thinking, and paving the way for a new
and e-health practical approach to the economics of Web
2. Research into health information systems businesses. E-business economy will alter the
and the development of prototypes lifestyles of many people, and the only at-
3. Data warehousing and data mining in life tribute that may leave behind some losers in
sciences the bargain is the relative apathy and lethargy
4. E-publishing of some organizations and an unwillingness to
5. E-learning adapt speedily to rapidly changing business
6. E-communications processes. “The future allows unlimited and
7. Mobile technologies and healthcare infinite opportunities to those who can discern
in the present the mistakes of the past and
clearly commit to themselves that they will
CONCLUSION never again repeat the same or other mistakes
in the present or the future.” Considering the
The last few years have seen a growing recog- global epidemic nature of cancer and AIDS,
nition of information as a key business tool. an international consortium for the research
Those who successfully gather, analyze, un- and development of telemedicine applications
derstand, and act upon the information are is a distinct reality in the not-too-distant fu-
among the winners in this new information age. ture.
Translating this to the domain of life sciences,
biological organisms obtain information through
sensory and extrasensory means and methodi- REFERENCES
cally unravel the continuous flow of information
for purposeful behavior. With behavior ranging Belbin, M. (2001). Team roles at work. Butter
from simple survival to a life of uninhibited and Worth-Heinemann.
spontaneous activities, human beings are a
Berson, A., Smith, S., & Thearling, K. (2000).
different class of biological organisms. This is
Building data mining applications for CRM.
reflected in one’s choice of behavior when
India: Tata McGraw-Hill.
faced with many conflicting choices. Of all the
physical and mental activities we can perform, Creech, B. (1995). The five pillars of TQM.
Truman Tally Books, Penguin Group.
218
Paraki, J. G. (2001). Abstract: Telemedicine Epidemiology: Is the study of the distribu-

2001, 6th International Conference on the Medi- tion, effects, and causes of diseases in popula-
cal Aspects of Telemedicine. Technology and tions and the means by which they may be
Health Care, 9(4), 366. treated or prevented.
Woodcock, J. (2002). Step up to networking. Insight: Is the realization of the relation-
India: Prentice Hall. ship of all factors in a given situation to one
another.
KEY TERMS Knowledge Management: Emerging sci-
entific discipline of new knowledge discovery
Cancer: Is defined as uncontrolled prolif- and dissemination.
eration of cancer cells that impair the function
Network: Is defined as a set of nodes and
of normal organs by local tissue invasion and
connecting lines to describe intricate structures
metastic spread to distant anatomic sites.
(e.g., neuronal network, computer network,
Data Mining: By its simplest definition, data network).
means to automate the detection of relevant
Telemedicine: Telemedicine is defined as
patterns in a database.
the actual delivery of healthcare using audio,
Decision-Making: Is a process through visual, and data communication.
which a correct decision is reached. For a
decision to be correct, one must be able to say
“yes” or “no” confidently.
219
221
Chapter XXVIII
Use of Telemedicine Systems
and Devices for Patient
Monitoring
Dionisia Damigou
Fotini Kalogirou
Georgios Zarras
ABSTRACT
Today’s health standards demand a high quality and efficiency as a major characteristic of
every health service provided to the public, even in cases where patients have to be treated
from a distance. The combination of medicine and information technology (telecommunications)
led to the introduction of the term telemedicine. Telemedicine services are used in assisting
remote patients. Interaction and feedback through patient monitoring systems and devices
allow the health providers interfere when necessary, so medical maintenance can be
guaranteed. This chapter deals with the different kinds of such systems and devices. The
contribution of old and new telecommunication technologies is currently being discussed. The
individual needs of every remote patient are taken into account, thus, several devices and
systems are used for telemonitoring. This chapter indicates characteristics and features of the
various kinds of patient monitoring systems and devices.
Use of Telemedicine Systems and Devices for Patient Monitoring
INTRODUCTION ample, home health tele-assistance systems

could provide a large range of services that
Due to the technological revolution and the would permit the user or patient to remain in his
entry of informatics in our everyday lives, the or her normal environment:
expansion of telecommunications became a
reality. Challenging telecommunication appli- • Emergency alarm systems (tele-alarms),
cations such as cellular telephony, communica- • Post-hospital treatment monitoring,
tions through the intervention of satellites, the • Social assistance (24-hour tele-assistance)
Internet, and so forth, support the extensive and so forth (Linkous, 2003).
exchange of information. Several scientific ar-
eas have become related to telecommunication Remote medical services can also be used in
applications, including the field of health in other areas as medical care is provided to
which the term telemedicine was first intro- prisons, on board commercial aircraft and ships,
duced. in the military, and even at the South Pole.
Telemedicine involves the transfer of medi-
cal information for use in diagnosis, treatment,
and education over distances and brings medi- TELEMEDICINE SYSTEMS AND
cal services directly to the point of need. TELEHOMECARE
A monitoring system adapts to the needs of
each and every patient, care unit, and hospital The most common utilization of telemedicine
equipped with hardware and software designed for patients’ monitoring is telehomecare. With
to give maximum functionality, flexibility, and more technologies moving into home care, and
responsiveness. Concerning the monitoring more and sicker patients being treated outside
devices, severall types are available, including of the hospital environment, the home-care
pulse oximeters, spirometers, glucose monitors, approach to health care is here to stay.
and so forth (Demiris, 2004). Telehomecare is viewed as a method that uses
telecommunication and videoconferencing tech-
nologies to enable a health-care provider at the
TELEMEDICINE SYSTEMS FOR clinical site to communicate with patients at
PATIENT MONITORING their homes. It is one of the brightest examples
of the new frontier of health care. These
There was a big demand for devices for patient telehomecare solutions are low cost compared
monitoring, especially for people who are far to the classic way of monitoring (Linkous,
away from health providers. The population is 2003).
growing old, and the high health standards of
nowadays claim that patients should be cared
for and rehabilitated at their homes. Several TELEMEDICINE SYSTEMS AND
telemonitoring systems were developed. The INTERACTION
primary aim of such systems is to maintain the
autonomy, independence, and quality of life for The telemedicine systems permit monitoring
the frail elderly, disabled persons, remote pa- devices to interact with doctors’ displaying
tients, and their informal family careers by the systems. Such an interaction is called a virtual
application of telematic technology. For ex- visit. These devices collect information and
222
signs from the patient, and the resulting data heart failure revealed that there is a need
can then be transferred through telephone lines for further investigation (Demiris, 2004).
(regular, ISDN, DSL, or T1), the Internet, or
the wires of a LAN (local area network) or Systems for Pulmonary Function
WAN (wide area network) to the doctor or
other health scientists (Kyriacou, Voskarides, a. Systems for Asthma Patients:
Pattichis, Istepanian, Pattichis, & Schizas, 2002). Telemonitoring systems for asthma pa-
Recent technologies such as Bluetooth might tients provide assistance in daily routines;
also be used (Roke Manor Research Limited, the systems alert health-care providers
http://www.roke.co.uk). The diffused data, in when necessary.
many cases, has the potential of updating the b. Systems for Lung-Transplant Patients:
doctor’s displaying devices in real time. When- Systems for the monitoring of pulmonary
ever there is a need (urgent case), the medical function in lung-transplant recipients via
personnel might immediately interfere and get the Internet are feasible and accurate.
in contact with the patient.
There are several types of telemonitoring Systems for the Management of
system applications; their purposes differ in Insulin-Dependent Diabetes
proportion to the concerning situation. Examples
of telemonitoring systems are provided in the Distributed computer-based systems for the
following sections. management of insulin-dependent diabetes were
developed utilizing Internet technology and
Systems for Monitoring Heart monitoring devices to support the normal activi-
Function and Circulation ties of physicians and diabetic patients by pro-
viding a set of automated services enabling data
a. Systems for Monitoring Cardiac-Re- collection, transmission, and analysis, and deci-
habilitation Patients: The first applica- sion support (Demiris, 2004).
tions that were evaluated concerned sys-
tems that monitor cardiac-rehabilitation Systems to Assist Post-Transplant
patients unable to return to a hospital- Patients
based program. Devices for these pa-
tients could detect arrhythmia and were Other Web-based telemonitoring systems were
found to be more effective than ambula- developed to assist post-transplant patients; for
tory electrocardiography. The most re- example, for lung-transplant recipients, regular
cent systems are much more complicated spirometry monitoring can be used for the early
and sophisticated. detection of acute infection and rejection of the
b. Systems for Hypertensive Patients: allograft. A Web-based telemonitoring system,
Other disease-management applications for these cases, provides direct transmission of
support hypertensive patients, enabling home spirometry to the hospital (Demiris, 2004).
them to control their blood pressure, and
are found to be efficient at the evaluation Systems for Emergency Response
process.
c. Systems for Chronic Heart Failure: Telemonitoring systems have also been intro-
Studies for home monitoring of chronic duced to provide emergency response to disas-
223
ter situations. The goal is to offer quality health- • Germany: A telemonitoring project was
care services to persons who are victims of a designed in Germany and included exten-
disaster. Portable telemedicine instrumentation sive home monitoring of the vital param-
packages can provide a compact, integrated eters of infants at risk for sudden infant
suite of tools such as data-acquisition devices death syndrome (SIDS). The project moni-
for ear, nose, throat, and skin imaging, lung tored breathing movements, ECG, heart
sound auscultation, and so forth (Demiris, 2004). rate, and oxygen saturation. All sensors
Similar system projects were developed in were noninvasive and integrated with the
many countries for several purposes. babies’ pajamas. The parents were trained
in assessing the physiological status of the
• United Kingdom: In Oxford, a remote children and in emergency intervention
physiological-monitoring network was es- measures. Information was also transmit-
tablished to evaluate cardiorespiratory ted to the test and research telemedical
function during sleep in a number of in- laboratory for further analysis.
fants in their homes for research pur- • A Four-Country Project (Greece,
poses. The center designed and devel- Great Britain, France, Germany): This
oped portable monitors for the continuous project was related to patients with renal
measurement of vital signs, which allowed failure requiring home hemodialysis. The
downloading data from monitor memory goal of the project was to develop, apply,
to the hospital. This information was then and evaluate telematics monitoring and
further analyzed. consultation services for enabling the su-
• Israel: A program was developed to gain pervision of each hemodialysis session
access to cardiac patients. In the process and possible intervention (Demiris, 2004).
of a typical patient call, a health profes-
sional collects descriptive information
about the patient’s condition while simul- TELEMEDICINE DEVICES FOR
taneously receiving and recording the 12- PATIENT MONITORING
lead electrocardiograph (ECG) transmit-
ted by the client-managed portable ECG Portable Monitoring Devices
device. The ECG results are displayed
and analyzed, and compared with previ- There are several commercially available por-
ous ECGs using proprietary transtelephonic table monitoring devices that are approved by
ECG-management software. the FDA, including pulse oximeters, blood-
• The Netherlands: One of the areas that pressure monitors, weight scales, and glucose
seem to be exceptionally well developed monitors. In some cases, data are stored in the
in the Netherlands is the use of handheld device and retrieved at a later point or are
computers for electronic case manage- displayed on a monitor at the completion of the
ment. After data are collected, they can test session (Demiris, 2004).
be immediately transferred, using dial-up
networking capacity, from the point of Wearable Sensors
care to the central home-care database.
Records are regularly updated and are There are three kinds of wearable sensors:
available to different health professionals physical, chemical, and biological. These sen-
within the continuum of care. sors produce a signal in response to an event,
224
which is then transferred to a circuit and be- heart rate, ECG, respiration, temperature, and
comes digitised. A physical sensor measures vital functions, alerting the wearer or physician
physical parameters such as temperature or if there is a problem. The Smart Shirt also can
pressure, whereas a biological or chemical be used to monitor the vital signs of military
sensor involves a receptor that binds with an personnel, chronically ill patients, firemen, and
analyte. The resulting digital data can be stored frail elderly persons living alone (Demiris, 2004).
and/or displayed. Also, for military purposes, the U.S. mili-
The concept of wearable sensors is based tary is developing innovative applications for
on the incorporation of sensors into watches, advanced sensors and smart materials. De-
items of clothing, and eye glasses. Thus, one vices resembling wristwatches will be worn
could argue that wearable sensors can func- by all soldiers as part of the combat uniform.
tion as noninvasive, in vitro diagnostic tools as These devices will monitor the soldier’s vital
they are capable of analyzing, among others, signs continuously by monitoring parameters
human sweat, tears, stress, strain, and pH such as noninvasive blood pressure, pulse
increases. oximetry, and medical imaging (Garshnek &
There exists a knee-wearable sensor that is Burkle, 1999).
also known as the intelligent knee sleeve, and it
was first designed for football players. It moni- Robots
tors knee strain or injury. Originally, this device
was strapped to the knees, and its sleeve pro- It is impressive that robots have started claim-
vided feedback to users by emitting an audio ing doctors’ duties. At present, these robots are
tone. It can be a useful application for home- navigated by an operator using a joystick at a
care patients with mobility impairments or for control station and can perform rounds within
those at the rehabilitation phase. the hospital. A camera and microphone are
Another example of a wearable sensor is a mounted above the computer screen, allowing
small, portable detector in the form of a wrist- the operator to see and hear. A similar camera
watch that provides test results for cystic fibro- and microphone at the control station transmit
sis in minutes, rather than the 24 hours that is the operator’s face and voice. It is surprising
the typical response time for a laboratory. that a number of doctors perform rounds on
Another wristwatch device uses an electric their patients with these robots, particularly
field to push pilocarpine nitrate into the skin, from their homes. The full potential of robots is
thereby dilating the pores. Sweat is absorbed beginning to be explored. Future generations of
and stored in a duct in the watch. The sample is control stations are expected to be portable and
analyzed by a sensor, and the levels of sodium, operational from virtually anywhere in the world
chloride, and potassium ions are recorded. Other (Norris, 2004).
devices in the form of wristwatches include
glucose meters that measure glucose in the
interstitial fluid as a low electric current pulls ADVANTAGES OF
glucose through the skin, and a blood oxygen TELEMEDICINE
monitor.
In the last few years, the Smart Shirt was The categories of telemedicine are the follow-
introduced that incorporates technology into ing:
the design of clothing to monitor the wearer’s
225
• Teleconsultation In general, telemedicine offers the follow-

• Telediagnosis ing:
• Telecare
• Remote clinical sessions and tele-educa- • Reduction of health-care costs
tion • Access to health services in previously
• Remote data access unserved or underserved areas
• Teleradiology • Easy cooperation between health profes-
• Home care sionals
• Telemonitoring (Mantas & Hasman, 2002) • Improved quality of care
Telemedicine’s objective is to provide users

with an integrated health-information service CONCLUSION
through an expert system, which gives access
to existing information related to health, social The development of information-systems tech-
care, and other general issues that is now nology has led us to an increased number of
distributed in dispersed databases. After achiev- telemonitoring applications to help with patient
ing this objective, we expect the following health care. These applications enable health
advantages. professionals to carry out home health visits
(virtual visits). Such applications and systems
1. Advantages in Monitoring the Patient: complete the management of data collection
The use of telemedicine is followed by an and reinforce data analysis. Telemedicine ser-
improvement of the quality of the monitor- vices can be shared among patients and several
ing of patients and an increase in the regional hospitals and other specialized health
number of patients being monitored. More- centers. Thus, patients are allowed to ask for
over, tools are provided for medical deci- advice, and, as a response, health professionals
sion support and guided monitoring work may interfere when necessary. It appears that
is allowed. the use of telemonitoring systems can cut the
2. Advantages in the Management of cost of medical care for rural and urban areas.
Treatment and Training: Telemedicine Interesting research and implementations
improves the communication between have been developed to evaluate systems deliv-
doctor and patient, promotes patients’ self- ering assistance to different scientific health
management and training, and decreases fields, contributing to cardiology, neurology,
the response time in the treatment. surgery, orthopedics, pediatrics, and so forth. A
3. Advantages in Telecare: Telecare re- lot of them were found to be accurate and
duces the number of visits to hospitals, efficient enough, although the future promises
facilitates patients in finding information more potent and sophisticated systems for pa-
regarding their illnesses, allows the patient monitoring.
tients autonomy, and decreases short-term However, there are still barriers to the wider
as well as long-term complications (Man- adoption of telemedicine, affecting both health-
tas, Aguilera, del Pozo Guerrero, Arredondo care professionals and their patients. The gen-
Waldmeyer, & Martínez Fernández, eral public is not mature enough to get involved
2000). in such procedures due to the fact that it is not
well acquainted with the subject. In the future,
226
a lot of these difficulties will be overcome, and KEY TERMS

we expect that the public will accept the chal-
lenges of telemedicine and telemonitoring. Home-Care Technology: Any technol-
ogy used to implement home-care telemedicine
services. It can also incorporate any device or
REFERENCES instrument for patient monitoring, therapy, or
environmental control (Mantas & Hasman,
Demiris, G. (2004). Electronic home 2002).
healthcare: Concepts and challenges. Co-
Telecare: The use of telecommunication
lumbia: University of Missouri.
systems to provide remote assistance in therapy
Gantenbein, R. (1992). Telehealth technology: to patients (Mantas & Hasman, 2002).
Wyoming’s efforts to facilitate research and
Teleconsultation: Remote access to a
service in a rural state. American Stroke As-
specialist’s knowledge. This type of service is
sociation Conference, Ft. Lauderdale, FL.
seen as a particular case of cooperative work
Center for Rural Health Research and Educa-
or cooperative diagnosis (Mantas & Hasman,
tion.
2002).
Garshnek, V., & Burkle, F. (1999). Applica-
Telediagnosis: Diagnosis of a patient by a
tions of telemedicine and telecommunications
remote physician. This kind of service does not
to disaster medicine. Journal of American
operate directly between the patient and re-
Medical Informatics Association, 6, 26-37.
mote doctor (Mantas & Hasman, 2002).
Kyriacou, E., Voskarides, S., Pattichis, C.,
Telehomecare: Uses telecommunication
Istepanian, R., Pattichis, M., & Schizas, C.
and videoconferencing technologies to enable a
(2002). Wireless telemedicine systems: A brief
health-care provider at the clinical site to com-
overview. IEEE Antennas & Propagation
municate with patients at their homes (Demiris,
Magazine, 44(2), 143-153.
2004).
Linkous, J. (2003). Advances in telemedicine
Telemedicine: The use of electronic in-
technology. Helsinki, Finland: American
formation and communication technologies to
Telemedicine Association (ATA).
provide and support health care when distance
Mantas, J., Aguilera, E. G., del Pozo Guerrero, separates the participant. It emphasizes appli-
F., Arredondo Waldmeyer, M. T., & Martínez cations that link clinician to patient or one
Fernández, A. (2000). Health informatics. clinician to another (Gantenbein, 1992).
Textbook in health informatics. Athens: IOS
Telemonitoring: Remote monitoring of
Press.
patients’ physiological value. This kind of ser-
Mantas, J., & Hasman, A. (2002). Textbook in vice is used with chronic and/or high-risk pa-
health informatics. Athens: IOS Press. tients (Mantas & Hasman, 2002).
Norris, K. (2004). Medicine and technology:
Robot makes the rounds. Detroit, MI: Detroit
Free Press.
227
228
Chapter XXIX
Current Telehealth Applications
in Telemedicine
Georgios Economopoulos
ABSTRACT
Rapidly emerging information and communication technologies (ICT) have spurred the recent
escalation of various telehealth applications. There is an enormous interest in finding new
ways to apply telehealth as much as telemedicine as a special part of telehealth. This chapter
has along with providing a better understanding of what telehealth is, investigated the ways
in which such an avant-garde, advancing, and newly emerging technology could be used in
order to be available in an upper-healthcare level.
INTRODUCTION which such an avant-garde, advancing, and

newly emerging technology could be used in
Rapidly emerging information and communica- order to raise the level of healthcare.
tion technologies (ICTs) have spurred the re- First of all, it is necessary to clarify the issue
cent escalation of various telehealth applica- regarding the confusion between the terms
tions (Lehoux, Battista, & Lance, 2000). It is telehealth and telemedicine, and even telecare
true that there is an enormous interest in finding and e-health.
new ways to apply ICT as much as telemedicine
as a special part of telehealth. This chapter, • Telemedicine involves the use of modern
along with providing a better understanding of information technology, especially two-
what telehealth is, investigates the ways in way interactive audio and video communi-
Current Telehealth Applications in Telemedicine
cations, computers, and telemetry, to de- integration of communications systems into the
liver healthcare to remote patients and to practice of protecting and promoting health,
facilitate information exchanges between while telemedicine is the incorporation of these
primary healthcare physicians at some terms into curative medicine” (WHO, 1997).
distance from each other (Bashur & Lovett,
1997). It existed long before the Internet.
It has been said that telemedicine was THE HISTORICAL EVOLUTION
present when the term was first estab- OF TELEHEALTH
lished nearly 30 years ago (Willemian et
al., as cited in Maheu, Whitten, & Allen, Before an in-depth analysis of current telehealth
2001). applications is conducted, it is necessary to
• Telecare refers to services that provide examine the historical evolution of telehealth.
care for people away from institutions, Nowadays, we consider telehealth as being a
typically in their own homes, for example, matter of high technology, especially associ-
in the monitoring of elderly people as they ated with the revolution of ICTs such as televi-
lead their normal lives. In other words, it sion, fax, or the Internet. Although we cannot
refers to services that provide healthcare ignore the dependence of telehealth on concur-
no matter where the doctor, the patient, or rent developments, it is true that humans have
his or her medical records are (Cornford since the early 1900s been communicating in-
& Klecun-Dabroeska, 2001). formation about health long before such new
• E-health is a relatively new term that first technologies arose (Darkins & Cary, 2000;
appeared in 1999, and it refers to Internet- Singh, 2002).
based healthcare delivery (McClendon, The first stage of telehealth was in the
2000). It includes all forms of healthcare 1970s when scientists made the first effort to
over the Internet (Singh, 2002). transmit health information in order to provide
• Telehealth could be considered an um- healthcare to people who were travelling by
brella term because the it is seen by au- ships (Maheu et al., 2001). There were also
thors as being more encompassing of the more efforts, especially in the United States, to
above terms whereas telemedicine is reuse telehealth in a productive way, but the
stricted toward interactive patient-doctor efforts were aborted due to high costs and
teleconsultations, and e-health refers solely some other reasons, such as insufficient image
to Internet-based healthcare delivery. It quality or insufficient acceptance from doctors
covers a number of different technolo- (Darkins & Cary, 2000; Singh, 2002).
gies, services, and professions including The second stage was in the beginning of
medicine, health promotion, health admin- the 1990s. Almost 20 years after the first
istration, social services, and information attempt, the conditions were good enough for
systems (Cornford & Klecun-Dabroeska, the telehealth sector to expand. The most im-
2001; Singh, 2002). portant reason for this regeneration of telehealth
was the lower cost of technology. Many coun-
In order to avoid a misunderstanding, the tries took advantage of that low cost, and a
World Health Organisation (WHO) distin- great deal of telehealth applications was adopted.
guished the terms telehealth and telemedicine The leading countries were the United States
in 1997: “Telehealth is understood to mean the and Norway, but Australia was not far away
229
from developing telehealth applications for its pation of their parents instead of in hospitals.
rural areas (Maheu et al., 2001). As we can see, there is a huge space for
The third stage of the telehealth evolution telehealth’s applications regarding telehomecare
began in the mid-1990s and culminates today. with multiple benefits: The low cost for institu-
The evolution of technology and communications, the ability for the patient to remain at
tions via the Internet has been really rapid home, and the existence of family support are
recently, and this seems to have a direct im- enough to give a boost for telehomecare.
pulse on telehealth (Poulis, 2002). One of the applications of telehealth is to
store and forward data and images such as X-
rays, CT (computed tomography) scans, and
CURRENT TELEHEALTH skin images (Matrini, 2003). By storing and
APPLICATIONS forwarding data and images, doctors have the
ability to improve diagnoses and to communi-
There is a great range of telehealth applica- cate with specialists in order to get a second
tions. First of all, we should mention that every- opinion. This procedure can be done in real time
thing that can provide health information in and could prove to be quite vital in case of an
rural areas could be seen as a telehealth appli- emergency.
cation. This means that even the simple and Telehealth poses a critical role respecting
very common household telephone is such an information access, especially in education and
application. We can use the telephone in order training (Matrini, 2003). It is important because
to keep in communication with the oldest pa- through it, everybody can have access to the
tients to remind them of simple things like taking source of the knowledge no matter if this is a
their pills or having their examinations (Singh, professional doctor, a student, or even a patient
2002). Of course, telehealth is applicative in or a whole community.
many ways regarding voice and video commu-
nication. Videoconference is one of them. In
addition, we can use video and voice communi- BENEFITS OF TELEHEALTH
cation for clinical procedures, particularly for
emergency procedures (Matrini, 2003). The revolution of telehealth is being followed
In general, doctors have the ability to com- by a number of benefits. Some of them are
municate with their patients, to support them mentioned as follows:
psychologically, and to educate them on how to
take care of themselves (Bergman, 1993). Apart • Decreased costs for patients because they
from that, videoconferencing can be used for do not have to travel in order to get to a
meetings between doctors in a way that reas- hospital
sures the knowledge flow will be constant • Lower costs for hospitals
(Meyer, 1996). • Availability of a wide range of services
Another issue of increasing interest is difficult to access
telehomecare. There has been an effort in • Potential constant education
preventing unnecessary hospitalizations, and • Prevention of social exclusion, for ex-
this can be achieved through telehealth by ample, the ability to serve people who live
monitoring patients with chronic conditions or in isolated areas
treating sick children at home with the partici-
230
• Many other parallel benefits such as bet- 150 that are now in orbit (Subba Rao, 2001).
ter feedback for epidemiological research Another important issue that is going to affect
the future of telemedicine is the relationship
between costs and outcomes (Grigsby & Sand-
PROBLEMS WITH TELEHEALTH ers, 1998). For the time being, telemedicine
seems to deserve the cost, but we should be
Although there are many benefits, the evolution careful in order to develop it according to the
of telehealth faces some problems. patients’ needs and not to the demand of the
industry. Apart from that, telemedicine may
• The situation regarding the reimburse- result in changes in the relationships between
ment of telehealth could be prescribed as healthcare professionals (Subba Rao). On the
dim. one hand, doctors and other scientists will have
• There is no adequate planning for the from now on many ways to communicate and
diffusion of telehealth. discuss issues about a patient. Additionally,
• There are some legal aspects that need to they have at their disposal a great amount of
be examined. clinical information that can be delivered ev-
erywhere in just a few seconds. On the other
hand, the role of the teleconsultant seems to
THE FUTURE OF TELEMEDICINE draw the attention, so he or she has to depend
on high standards. Even more importantly,
The evolution of telemedicine is without prece- changes are going to happen in the functioning
dent. It has been spread all over the medical of the whole healthcare system. It must be
practise so that we can talk about telecardiology, mentioned that many changes will involve pri-
teleneurology, teleradiology, teledermatology, mary and secondary health are. Using
telesurgery, and many other specialities as be- telemedicine, patients can avoid being in a
ing different fields in medicine (Grigsby & hospital and, thus, administrative change that
Sanders, 1998). Just a few years before, only a supports the growth of primary healthcare will
few could imagine that doctors could conduct a be conducted. Another issue that will probably
surgery while at a distance away. The close configure the future of telemedicine is that of
relationship between telemedicine and evolu- security. There has been a great discussion
tion in technology, involving both information about security, and it began right after the
and communication, leads us to the thought that discussion about the security of the Internet.
the evolution can be even broader. The impor- The truth is that a great amount of personal data
tant thing is to determine the changes that are is being transferred via telemedicine through
going to occur. There are many issues that need the Internet, and nobody is able to reassure us
examination so that we will be able to use that it will be safe. The abuse of these personal
telemedicine in the most productive and effec- data could lead to situations like the denial of an
tive way. insurance company to provide its services. This
First of all, satellite technology is going to issue is really crucial and demands very careful
make telemedicine available to even the most consideration.
isolated areas on the planet. Around 1,700 The future of telemedicine depends on the
commercial satellites are scheduled for launch answer about reimbursement. There are still
in the next decade worldwide compared to the some unsolved problems about reimbursement
231
and the lack of payment (Greenpope, 2001). Cambridge Quarterly of Healthcare Ethics,
The technology of telemedicine and its use 10(2), 161-169.
generally can be considered avant-garde, so it
Darkins, A., & Cary, M. (2000). Telemedicine
is understandable that there is still not the right
and telehealth. Springer Publishing Company.
legislation that will provide everybody with the
ability to use telemedicine. Greenpope, D. (2001). Telemedicine: What it
is and how we can use it to achieve better
management of health care.
CONCLUSION
Grigsby, J., & Sanders, J. (1998). Telemedicine:
Where it is and where it’s going. Annals of
The use of telehealth and telemedicine seems
Internal Medicine, 129.
to have spread like wildfire. It is true that due
to the latest technology, many patients who live Lehoux, P., Battista, R., & Lance, J.-M. (2000).
in rural and isolated areas can have much better Telehealth: Passing fad or lasting benefits?
healthcare. However, that fast growing must Canadian Journal of Public Health, 91, 277-
be followed by a number of changes involving 280.
administration and reimbursement, which will
Maheu, M., Whitten, P., & Allen, A. (2001). E-
undoubtedly lead to a better level of services.
health, telehealth and telemedicine. Jossey
Bass Publishers.
REFERENCES Matrini, A. (2003). Bridging the gap between
community participation and policy imple-
American Nurses Association. (2001). Devel- mentation in health related communication.
oping telehealth protocols: A blueprint for suc- Unpublished doctoral dissertation.
cess. Proceedings of the 38th Hawaii Inter-
McClendon, K. (2000). E-commerce and HIM:
national Conference on System Sciences.
Ready or not, here it comes. Journal of Ameri-
Bashur, R. L., & Lovett, J. (1997). Assessment can Health and Information Management,
of telemedicine: Results of initial experience. 71, 22-23.
Space and Environmental Medicine, 48, 65-
Meyer, K. (1996). Can telemedicine deliver
70.
what it promises? American Academy of Fam-
Bergman, R. (1993). Letting telemedicine do ily Physicians.
the walking: Rural projects use video communi-
Poulis, S. (2002). Telemedicine in cardiol-
cations to enhance access to care. Hospitals &
ogy. Unpublished master’s thesis.
Health Networks.
Singh, H. (2002). Can telehealth prevent so-
Committee on Evaluating Clinical Applications
cial exclusion and create a fairer health
of Telemedicine. (1996). Telemedicine: A guide
service? Brunel University, Department of IS
to assessing telecommunications in health care.
and Computing.
Proceedings of the 38 th Hawaii Interna-
tional Conference on System Sciences. Subba Rao, S. (2001). Integrated health care
and telemedicine. Retrieved from http://
Cornford, T., & Klecun-Dabrowska, E. (2001).
emerald.com
Ethical perspectives in evaluation of telehealth.
232
World Health Organisation (WHO). (1997). Teleconferencing: Interactive electronic

Health care systems in transition. communication between two or more people at
Copenhagen, Denmark: WHO Regional Office two or more sites that makes use of voice,
for Europe. video, and/or data-transmission systems such
as audio, audio graphics, and computer and
video systems.
KEY TERMS
Telehealth: The removal of time and dis-
Electronic Health Record (EHR): An tance barriers for the delivery of healthcare
electronic health record provides an individual services or related healthcare activities (Ameri-
in Canada with a secure and private lifetime can Nurses Association, 2001).
record of his or her key health history and care
Telemedicine: The use of electronic in-
within the health system. The record is avail-
formation and communications technologies to
able electronically to authorized healthcare pro-
provide and support healthcare when distance
viders and the individual anywhere, anytime in
separates the participants (Committee on Evalu-
support of high-quality care.
ating Clinical Applications of Telemedicine,
Information and Communications Tech- 1996).
nology (ICT): The application of modern elec-
Teletriage: A means of providing health
tronic and computing capabilities (technology)
information and advice on preferred courses of
to the creation and storage of meaningful and
treatment, usually over the telephone using
useful facts or data (information), and to their
computerized protocols or algorithms devel-
transmission to users by various electronic
oped by clinical experts.
means (communication). The ultimate goal is
for ICT to transform data into information, and
information into knowledge.
233
234
Chapter XXX
Mobile Telemonitoring Insights
Pantelis Angelidis
Vidavo Ltd., Greece
ABSTRACT
Technology advances create new possibilities for healthcare monitoring, management, and
support, focusing on prevention rather than disease management. The provision of personalized
healthcare applications is also greatly supported. Developments in the wireless and mobile
markets are capitalized by the medical device industry. Services are becoming personalized
and location independent to fulfill the increasing patient needs for self-empowerment and
quality in the healthcare delivery away from the traditional nursing areas. This overview
discusses the new opportunities for the healthcare domain in the mobile times we live.
INTRODUCTION delivery’s potential without geographical limi-

tations.
The healthcare industry is experiencing a sub- The concept of prevention prevails now against
stantial shift to care delivery away from the disease management and treatment plans. As
traditional nursing areas due to the conver- patient-centric processes emerge, the citizens
gence of several technology areas. Increas- and patients undertake an active role in moni-
ingly capable health-monitoring systems are toring their health status. Meanwhile, e-wellness
moving the point of care closer to the patient, evolves to address the rising expectations of
while the patient, better informed and aware the e-health consumers, who are better in-
now, undertakes an active role to self-care and/ formed, more demanding, and empowered. The
or -prevention. Emerging ICTs in conjunction empowered, worried-well consumers require
with the medical device industry development quality health services on the spot. The drivers
(intelligent devices, biosensors, novel software, are now connectivity, speed, and personaliza-
etc.) demonstrate personalized healthcare tion (McKnight, 2000).
MOBILE HEALTHCARE vide functionality that reduces costs, improves

PROVISION the quality of care, and improves the ease with
which caregivers can perform their everyday
Waves of technology incorporation and scien- tasks (Wolf, 2001).
tific discoveries have driven the sector from The most significant challenge posed by
reliance on direct communication and physician mobile technology is the seamless integration of
experience to a higher reliance on technology multiple hardware and software platforms with
and community information. This new Web- reliable, uninterrupted wireless services in a
enabled environment has taken healthcare from secure manner, which will become mission
local areas, where telemedicine left it, literally critical to successful healthcare organizations,
into the patient’s home and, more recently with payers, and providers (Wolf, 2001).
m-Internet, to wherever the patient might be The current state-of-the-art technology in
and whenever he or she needs it (Simão, 2001). medical sensors allows for the easy and unob-
M-Internet enables information exchange trusive electronic measurement of several health
and promotes the availability of services and conditions. The sensors are often stand-alone
communication modes to serve working teams devices and sometimes comprised of two or
with increasing mobility requirements. more elements connected by a cable or wire-
Services are becoming personalized and less technology. Medical sensors have the
location independent to serve increasing patient capability to measure vital signs such as blood
needs for self-empowerment and quality in pressure, pulse rate, respiration frequency, and
healthcare delivery away from the traditional so forth. Based on these medical parameters,
nursing areas. the medical professionals can monitor the
Furthering the new approaches in the provi- patient’s health condition and act in case of an
sion of healthcare services in the frame of e- anomaly.
health, wireless developments create new op- The application areas of the medical-device
portunities for healthcare professionals, indi- wireless telemonitoring capabilities include the
viduals and organizations, patients, and health following:
authorities. The scope of mobile health ad-
dresses clinical, administrative, and consumer 1. Assistance in case of accidents and emer-
health-information applications and, as it could gencies
contribute to the improvement of health out- 2. Increased capacity and lower costs for
comes, m-health may be utilized to measure hospitals
health status and population welfare. 3. Assistance and monitoring in a home-care
Many healthcare organizations are invest- setting
ing in IT projects that take advantage of new 4. Monitoring of chronically ill patients
technologies in the mobile healthcare applica- 5. Patient involvement in setting a diagnosis
tion space. Functionality that augments the 6. Medicine dosage adjustment
capture of evidence-based patient plans of care 7. Physical-state monitoring in sports
is essential and must map and bridge the infor- 8. Monitoring of sporadically occurring symp-
mation flow for both inpatient and outpatient toms
work-flow clinical-practice guidelines. As the 9. Emergency alarms (Fosse & Haug, 2003)
medical community continues to embrace these 10. Improved health management
new technologies, system integrators must pro-
235
As a result, citizens can enjoy quality cially in the area of prevention, and (c) con-
healthcare provision and an elevated quality of cerned about the quality of care offered by
life. As underlined by the European Council physicians and institutions, with a willingness to
objectives set in Lisbon, “effective integration select the highest level of services. In short, an
of healthcare and related support services by e-health consumer manages health, in all ex-
electronic means, including the widespread use tents possible, as the most important asset of his
of telecare, could improve the quality of life of or her family. The main objective is to maintain
citizens by enabling safer independent living the highest level of quality of life (Lerer, 2000)
and increased social inclusion.” The rapid proliferation of wireless personal
computers, phones, appliances, and other de-
vices will require organizations to look beyond
EVOLUTION FORCES AND single-platform solutions. System-integration
CHALLENGES activities have a new level of complexity and
cost to support rapidly changing technology
Empowered patients demand advanced wire- (Wolf, 2001). Mobile-health advances gener-
less health solutions. Similar to most authors, ate new capabilities in patient self-care and
Lerer (2000) suggests that the e-health con- health-practice administration and reimburse-
sumer is being empowered due to an increased ment. Cost-effective solutions minimize effort
ability to obtain health information and to seek in monetary and human-input terms, while the
health-related offerings via the Internet. A creation of new communication modes facili-
Deloitte Research (2000) study suggests the e- tates both the healthcare professionals and the
health consumer is a mix of an empowered and patients.
an engaged consumer. Recognizing that e- When it comes to investing in new technol-
health consumers’ empowerment can increase ogy solutions, affordability is a major milestone
efficiency and reduce health costs, Lerer ar- to consider. Budget allocation to mobile health
gues that consumers’ education and empower- applications can be easily influenced both by
ment should be a key concern for all health the technology cost and the user awareness of
players. E-health consumers, he suggests, are current and future cost benefits. The complex-
not just the ill, but the potentially ill, the worried, ity and fragmentation of the overall healthcare
and those adjacent to illness, patients, their sector (i.e., centralized vs. decentralized health
relatives, and friends. At the first level, e-health systems, variations in the public and private
services are information-driven activities, which funding mix, etc.) often leads to the implemen-
are mostly “event triggered.” The Deloitte tation of fragmented and disposable technologi-
Research study suggests that the demographic cal solutions. Interoperability thus is essential
profile of the e-health consumer population for large-scale applications with international
reveals a significant population group with eco- scope. Conformance to global (when available)
nomic clout, information sophistication, and tech- and/or U.S. and European standards enables
nological familiarity, and that is generally faster and ubiquitous communications, while
wealthy. also ensuring the compatibility and connectivity
An e-health consumer is an individual who is of systems and points of care.
(a) fully involved in the management of health According to CEN/TC 251 (2001), the
for himself or herself and his or her family, (b) present lack of standardized ICT communica-
proactively educated about health issues, espe- tion, which prevents appropriate access to health
236
records, may result in important clinical risks will be fed back to the individual (Istepanian et
for the patients. This is an important safety al., 2004)
issue that has not been recognized sufficiently. A fourth-generation m-health solution builds
Implemented standards are often crucial for upon the mobile information portal of a 3G
any communication, and they are especially solution by adding the multiple devices render-
important for open, very complex healthcare ing the capability of the 2G (second-generation)
systems with many different organizations and solutions. Now, an end user has the ability to
units, with information systems from different access any application with any device (“Going
suppliers, providing different parts of the total Mobile,” 2001). 4G solutions embrace the dis-
ICT support. tributed and loosely coupled HIS applications
Furthermore, the wider implementation of throughout a health unit. A 4G solution can
mobile solutions requires a robust security plan allow for the acquisition of data from various
to reassure the confidentiality of sensitive medi- sources and for the mobile end user to view,
cal data. analyze, manipulate, graph, and merge data
according to his or her needs right on the mobile
device.
M-HEALTH POTENTIAL In the home of the future, some devices will
contribute physiological information about the
The next few years will witness a rapid deploy- patient (e.g., heart rate, blood pressure), while
ment of both wireless technologies and mobile other devices in and around the home will
Internet-based m-health systems with perva- contribute information about the patient’s envi-
sive computing technologies. The increasing ronment (e.g., humidity, temperature, carbon-
data traffic and demands from different medi- monoxide level). In some cases, groups of
cal applications and roaming applications will devices will have enough collective awareness
be compatible with the data rates of 3G (third- to function autonomously based on sensor data.
generation) systems in specific mobility condi- The challenge for healthcare providers and
tions. The implementation and penetration of health authorities lies in the comprehension of
4G (fourth-generation) systems are expected the end users’ needs for the effective integra-
to help close the gap in medical care. Specifi- tion of new technological capabilities with ex-
cally, in a society penetrated by 4G systems, isting settings in order to leverage their capaci-
home medical care and remote diagnosis will ties and quality of service.
become common, checkups by specialists and
the prescription of drugs will be enabled at
home and in underpopulated areas based on CONCLUSION
high-resolution image-transmission technolo-
gies and remote surgery, and virtual hospitals Systematically sensitizing users and providing
with no resident doctors will be realized. Pre- them with specific information on new mobile
ventive medical care will also be emphasized: and wearable computing technologies will help
For individual health management, data will to discover possible fields of new applications.
constantly be transmitted to the hospital through The initiation of a dialogue between users in
a built-in sensor in the individual’s watch or healthcare and developers of mobile IT solu-
another item worn daily, and diagnosis results tions eventually may lead to the identification of
237
new application fields (i.e., medical specialties) Simão, C. M. V. Q. (2001). A study on Internet
and related practices in mobile healthcare pro- impact in business designs for the health
vision. sector.
A first step to this end is the identification
Wolf, J. (2001). MBA mobile health applica-
and definition of mobile-activities profiles, and
tions, Version 1.1.
stakeholder profiles and their level of involve-
ment, as well as mobile application scenarios.
Technologies should be designed for people KEY TERMS
rather than making people adapt to technolo-
gies in order to capitalize on the capabilities that 3G and 4G: Third- and fourth-generation
wireless technologies create in the healthcare wireless Internet devices. The major distinction
domain. of 4G over 3G communications is increased
data transmission rates. 4G is expected to
deliver more advanced versions of the same
REFERENCES improvements promised by 3G, such as en-
hanced multimedia, smooth streaming video,
CEN/TC 251. (2001). Report on the Mandate universal access, and portability across all types
M/255. of devices. 4G enhancements are expected to
include worldwide roaming capability and are
Deloitte Research. (2000). The emergence of
likely to incorporate global positioning services
the e-health consumer.
(GPSs). As was projected for the ultimate 3G
eHealth in 2010: Meeting the Lisbon objec- system, 4G might actually connect the entire
tives with ambient intelligence technologies? globe and be operable from any location on or
(2003). Seville. above the surface of the earth.
Fosse, B., & Haug, B. E. (2003). A feasibility Ambient Intelligence: The concept of
study and recommendation of technology ambient intelligence provides a vision of the
and solutions for wireless monitoring of information society in which the emphasis is on
biomedical data. Unpublished master’s the- user friendliness, efficient and distributed ser-
sis, Agder University College, Grimstad. vices support, user empowerment, and support
for human interactions. People are surrounded
Going mobile: From eHealth to mHealth. A
by intelligent, intuitive interfaces that are em-
Daou Systems White Paper. (2001).
bedded in all kinds of objects in an environment
Istepanian, R., et al. (2004). Non-telephone that is capable of recognizing and responding to
healthcare: The role of 4G and emerging mobile the presence of different individuals in a seam-
systems for future m-health systems. In Medi- less, unobtrusive, and often invisible way.
cal and Care Compunentics. IOS Press.
E-Health Consumer: Self-reliance and
Lerer, L. (2000). The healthcare 2020 plat- empowerment are the core characteristics of
form: The e-health consumer. INSEAD. the e-health consumer, who actively pursues
patient-centric quality services in a frame of
McKnight, L. (2000). Internet business mod-
information-supported activities.
els. Medford, MA: Tufts University.
238
Empowered Patient: A patient whose self- recorded to personal sensors and alarm notices
management is based on informed decisions in case of an emergency. Mobile health or m-
and who takes into account his or her quality of health is a step beyond electronic healthcare as
life, including both physical well-being and psy- it enhances ubiquitous health provision regard-
chological state, as well as other dimensions. less of the patient’s or physician’s geographic
location.
E-Wellness: The utilization of Internet
capabilities (information, Web-based health Telemonitoring: The science and tech-
services, etc.) in order to maintain a condition nology of automatic measurement via medical
of good physical and mental health. sensors and the transmission of data by radio or
other means from remote sources to receiving
Medical Sensor: A device, such as a pho-
stations for recording and analysis. Data trans-
toelectric cell, that receives and responds to a
fer can be achieved via wireless communica-
signal or stimulus.
tions means and/or via other media, such as a
M-Health: Mobile health refers to ambula- telephone, a computer network, or an optical
tory-care provision enabled by third-generation link.
devices that allow for the collection, manage-
Vital Signs: The pulse rate, blood pressure,
ment, and processing of the patient’s vital data.
body temperature, and rate of respiration of a
Mobile health services range from the record-
person. The vital signs are usually measured to
ing of the patient’s medical signs and the syn-
obtain a quick evaluation of the person’s gen-
chronous or asynchronous communication with
eral physical condition.
health professionals via mobile communication
means, to the automatic diagnosis of the data
239
240
Chapter XXXI
Telepathology
and Digital Pathology
Vincenzo Della Mea
University of Udine, Italy
ABSTRACT
The present chapter deals with state-of-the-art topics related to the application of information
and communication technologies to the field of pathology, in particular for what regards
telepathology and the so-called digital pathology. A classification of telepathology techniques
is provided together with their typical applications. Starting from a definition of virtual or
digital slide, digital pathology techniques and issues are then discussed.
INTRODUCTION telepathology have been the two forces driving

pathology toward the use of digital images.
In pathology, the sample subject of analysis by In particular, telepathology has been tradi-
the doctor is most often a biological tissue tionally constituted by the set of techniques for
specimen cut in very thin sections, disposed remotely transmitting images acquired from a
over a glass slide, and coloured with suitable glass slide through a microscope. The term
stainings in order to make morphological struc- telepathology was first referred to by R.
tures and biochemical components visually ap- Weinstein in 1986; he defined it as the “practice
parent. Such a specimen is observed by means of pathology over a long distance,” thus encom-
of an optical microscope with resolutions up to passing almost every possible application of
0.2 micron. digital images, including distant diagnosis, ex-
In the last 40 years, first morphometry (i.e., pert consultation, distant education, and remote
image analysis applied to the recognition and image processing and analysis.
quantification of biological shapes) and then
Telepathology and Digital Pathology
The two usual telemedicine categories of Flandrin, Got, & Hemet, 1998), or Web-based
real-time and store-and-forward systems apply applications like iPath (Brauchli et al., 2000).
to telepathology, too, with the same cost and Store-and-forward telepathology is the least
practical consequences but with peculiar as- expensive approach as it can be implemented
pects given by the features of the material to be with very basic technology, and thus it is the
represented by means of digital images. most diffused one despite some concerns on
In the very few last years and thanks to the the bias inducted by the image selector
innovations in the information technology field, (Mairinger, Netzer, Schoner, & Gschwendtner,
a novel acquisition technique has been created 1998).
that makes it possible to fully digitize a speci- In real-time telepathology, a robotized mi-
men, resulting in billions of pixels. The resulting croscope that can be remotely controlled is
image is called a virtual slide, virtual micro- needed. In this case, the remote user may move
scope, or digital slide, and can be used for most the stage, change objectives, and so forth,
of the usual telepathology applications plus receiving either still images (Demichelis,
others. This innovation led to the creation of a Barbareschi, et al., 2001) or live video (Dunn et
somewhat new discipline, that is, digital pathol- al., 1997; Nordrum et al., 1991). As live video
ogy, where the emphasis is no more on the has lower quality than still images, hybrid ap-
transmission of images, like in telepathology, proaches have been reported (Della Mea, 2000).
but more generally on digital-image use. Due to the need for quality of service, transmis-
In the next section, a brief overview of sion often occurs through ISDN lines; how-
telepathology and digital pathology techniques ever, in the past, attempts have been made to
and applications will be provided. use Internet protocols (Wolf, Petersen, Dietel,
& Petersen, 1998), which are at present more
easily usable.
TELEPATHOLOGY TECHNIQUES: From the diagnostic point of view,
APPLICATIONS AND ISSUES telepathology might be useful in two main ar-
eas: second-opinion consultation and intraop-
The approaches to telepathology are classically erative diagnosis. In the former case, an expert
divided into two categories: store-and-forward is requested for an opinion regarding a difficult
and real-time telepathology, respectively based case with loose time constraints. While ordi-
on the asynchronous delivery of images and narily the expert will receive a glass slide by
real-time transmission. Another classification, post, with telepathology, he or she may receive
often used in place of the former but with a a set of selected images (store and forward;
slightly different meaning, is such between Raab et al., 1997) or access to a real-time
static and dynamic telepathology, respectively system (Dunn et al., 1997). In fact, store and
based on still images and live video. forward is mostly used for expert consultation
In store-and-forward telepathology (Della as it overcomes limitations due to different time
Mea, 1999), the sender pathologist is supposed zones when requesting and consulting patholo-
to select some representative image from the gists are remotely separated (Della Mea &
specimen to be delivered or simply made avail- Beltrami, 1998). In the last years, two official
able to a remote recipient. Delivery may occur services have been founded for providing con-
by means of standard e-mail (Della Mea et al., sultation through store-and-forward systems:
1996), proprietary systems (Klossa, Cordier, the Armed Forces Institute of Pathology
241
telepathology service and the Union The basic equipment for doing this is similar
Internationale Centre le Cancer (UICC) con- to those needed for real-time telepathology,
sultation service. that is, a robotized microscope driven by soft-
Intraoperative diagnosis is made during sur- ware able to scan the whole glass-slide surface
gery as quickly as possible to guide the surgeon and acquire all fields. More sophisticated de-
in his or her work. A rural hospital might not vices have been developed in the form of either
have a pathologist available during operations, slide scanners or array microscopes (Weinstein
so telepathology can be used to deliver the et al., 2004).
same service from a distance provided that a The practice of digitizing large parts of a
real-time system is used. For this application, glass slide has been known since its beginnings
diagnostic performance is comparable to that as virtual microscopy (Ferreira et al., 1997),
of the microscope (Dunn et al., 1997; Kayser, and the digitized slide has thus been called a
Beyer, Blum, & Kayser, 2000; Nordrum et al., virtual slide or digital slide; sometimes the view-
1991); however, concerns are brought up when ing software is referred to as a virtual micro-
leaving the macrosampling task to the surgeon scope.
or to a technician. Typical applications of the digital slide in-
Other applications of telepathology include clude education (Ferreira et al., 1997), quality
remote image processing and evaluation, and control (Demichelis, 2002; Taylor, Gagnon,
education. In the former case, selected images Lange, Lee, Draut, & Kujawski, 1999), and
can be sent to a remote image-processing ser- image processing and analysis (on tissue
vice for automatic morphometry or cytometry microarray slides; Dell’Anna, Demichelis,
(Ferrer Roca, Ramos, & Diaz Cardama, 1995; Barbareschi, & Sboner, 2005); informal reports
Kunze, Boecking, Haroske, Kayser, Meyer, & have been made on their use for telediagnosis
Oberholzer, 1998). In the latter, digital imagery and long-term storage, too.
may substitute microscope sessions to teach The main problems with digital slides are
students histology and cytology (Harris, Leaven, related to their size. A full slide (e.g., a Pap test)
Heidger, Kreiter, Duncan, & Dick, 2001). is made of up to 30GB of data when
uncompressed, which when compressed may
become just 1GB. This is an issue of storage;
DIGITAL PATHOLOGY for remote viewing this is not a great problem as
TECHNIQUES: APPLICATIONS often a slide is only partially viewed to render a
AND ISSUES diagnosis (Tsuchihashi et al., 1999).
The size influences acquisition time, too,
In a very few years, advances in information which when using tools based on robotized
technology, including the increase in hard-disk microscopes is measured in hours, although
capacity and faster network connections, led to new developments like the array microscope
the possibility of digitizing the whole glass slide, (Weinstein et al., 2004) strongly reduce it. A
thus avoiding the image-selection bias reported secondary yet important problem is given by the
for traditional store-and-forward telepathology. fact that, although rarely, there is not a single
The digitized slide is then viewed by means of focus plane. So, at least in some visual fields, it
suitable software that emulates the features of is necessary to acquire more than one image to
a real microscope and is often able to connect capture all information.
to remote cases through the Internet.
242
DISCUSSION to properly handle digital images in large scale

tissue microarray experiments. Computer
Telepathology and digital pathology provide for Methods and Programs in Biomedicine, 79(3),
three telemedicine techniques still in use due to 197-208.
their different features and issues.
Della Mea, V. (1999). Store-and-forward
In fact, while real-time telepathology and
telepathology. In B. Hernandez & R. Wootton
digital slides may provide the same amount of
(Eds.), European telemedicine 1998/99. Lon-
information to the distant user, the former is
don: EHTO, RSM Press, & Kensington Publi-
needed in urgency while the latter needs time
cations.
for acquisition. On the other hand, digital slides
automatically provide for long-term storage, Della Mea, V., & Beltrami, C. A. (1998).
while classical real-time pathology cannot in an Telepathology applications of the Internet mul-
easy way. timedia electronic mail. Medical Informatics,
In addition, store-and-forward telepathology 23, 237-244.
based on selected images still represents the
Della Mea, V., Forti, S., Puglisi, F., Bellutta, P.,
most used technique due to its low cost and
Finato, N., Dalla Palma, P., et al. (1996).
ease of use, which makes it suitable, for ex-
Telepathology using Internet multimedia elec-
ample, for supporting developing countries; good
tronic mail: Remote consultation on gastrointes-
examples are the UICC telepathology centre
tinal pathology. Journal of Telemedicine and
and the consultation communities born around
Telecare, 2, 28-34.
the free iPath software (Brauchli et al., 2000).
Demichelis, F., Barbareschi, M., Boi, S.,
Clemente, C., Dalla Palma, P., Eccher, C., et
REFERENCES al. (2001). Robotic telepathology for intraop-
erative remote diagnosis using a still-imaging-
Barbareschi, M., Demichelis, F., Forti, S., & based system. American Journal of Clinical
Dalla Palma, P. (2000). Digital pathology: Sci- Pathology, 116(5), 744-752.
ence fiction? International Journal of Surgi-
Demichelis, F., Barbareschi, M., Dalla Palma,
cal Pathology, 8, 261-263.
P., & Forti, S. (2002). A new method to com-
Brauchli, K., Christen, H., Meyer, P., Haroske, pletely digitise cytological and histological slides.
G., Meyer, W., Kunze, K. D., et al. (2000). Virchows Archive, 441(2), 159-164.
Telepathology: Design of a modular system.
Demichelis, F., Della Mea, V., Forti, S., Dalla
Analytical Cellular Pathology, 21(3-4), 193-
Palma, P., & Beltrami, C. A. (2002). Digital
199.
storage of glass slides for quality assurance in
Costello, S. S., Johnston, D. J., Dervan, P. A., histopathology and cytopathology. Journal of
& O’Shea, D. G. (2003). Development and Telemedicine and Telecare, 8(3), 138-142.
evaluation of the virtual pathology slide: A new
Dunn, B. E., Almagro, U. A., Choi, H., Sheth,
tool in telepathology. Journal of Medical
N. K., Arnold, J. S., Recla, D. L., et al. (1997).
Internet Research, 5(2), e11.
Dynamic-robotic telepathology: Department of
Dell’anna, R., Demichelis, F., Barbareschi, M., Veterans Affairs feasibility study. Human Pa-
& Sboner, A. (2005). An automated procedure thology, 28, 8-12.
243
Ferreira, R., Moon, B., Humphries, J., Sussman, Nordrum, I., Engum, B., Rinde, E., Finseth, A.,
A., Saltz, J., Miller, R., et al. (1997). The virtual Ericsson, H., Kearney, M., et al. (1991). Re-
microscope. Proceedings of AMIA Annual mote frozen section service: A telepathology
Fall Symposium (pp. 449-453). project in northern Norway. Human Pathol-
ogy, 22(6), 514-518.
Ferrer Roca, O., Ramos, A., & Diaz Cardama,
A. (1995). Immunohistochemical correlation of Raab, S., Robinson, R. A., Snider, T. E.,
steroid receptors and disease-free interval in McDaniel, H. L., Sigman, J. D., Leigh, C. J., et
206 consecutive cases of breast cancer: Vali- al. (1997). Telepathologic review: Utility, diag-
dation of telequantification based on global nostic accuracy, and interobserver variability
scene segmentation. Analytical Cellular Pa- on a difficult case consultation service. Mod-
thology, 9, 151-163. ern Pathology, 10, 630-635.
Foran, D. J., Meer, P. P., Papathomas, T., & Taylor, R. N., Gagnon, M., Lange, J., Lee, T.,
Marsic, I. (1997). Compression guidelines for Draut, R., & Kujawski, E. (1999). CytoView: A
diagnostic telepathology. IEEE Transactions prototype computer image-based Papanicolaou
on Information Technology in Biomedicine, smear proficiency test. Acta Cytologica, 43(6),
1, 55-60. 1045-1051.
Harris, T., Leaven, T., Heidger, P., Kreiter, C., Tsuchihashi, Y., Mazaki, T., Nakasato, K.,
Duncan, J., & Dick, F. (2001). Comparison of Morishima, M., Nagata, H., Tofukuji, I., et al.
a virtual microscope laboratory to a regular (1999). The basic diagnostic approaches used
microscope laboratory for teaching histology. in robotic still-image telepathology. Journal of
The Anatomical record, 265(1), 10-14. Telemedicine and Telecare, 5(Suppl. 1), S115-
S117.
Kayser, K., Beyer, M., Blum, S., & Kayser, G.
(2000). Recent developments and present sta- Weinstein, R. S. (1986). Prospects for
tus of telepathology. Analytical Cellular Pa- telepathology. Human Pathology, 17(5), 433-
thology, 21, 101-106. 434.
Klossa, J., Cordier, J. C., Flandrin, G., Got, C., Weinstein, R. S., Descour, M. R., Liang, C.,
& Hemet, J. (1998). A European de facto Barker, G., Scott, K. M., Richter, L., et al.
standard for image folders applied to (2004). An array microscope for ultrarapid
telepathology and teaching. International Jour- virtual slide processing and telepathology: De-
nal of Medical Informatics, 48, 207-216. sign, fabrication, and validation study. Human
Pathology, 35(11), 1303-1314.
Kunze, K. D., Boecking, A., Haroske, G.,
Kayser, K., Meyer, W., & Oberholzer, M. Wolf, G., Petersen, D., Dietel, D., & Petersen,
(1998). Remote quantitation in the framework I. (1998). Telemicroscopy via the Internet.
of telepathology. Advances in Clinical Pa- Nature, 391, 613-614.
thology, 2(2), 141-143.
Mairinger, T., Netzer, T. T., Schoner, W., & KEY TERMS
Gschwendtner, A. (1998). Pathologists’ atti-
tudes to implementing telepathology. Journal Digital Pathology: The set of image acqui-
of Telemedicine and Telecare, 4, 41-46. sition, storage, transmission, and processing
techniques based on the digital slide.
244
Real-Time Telepathology: Synchronous Telepathology: Practice of pathology over

form of telepathology in which the distant op- a long distance. It includes telediagnosis.
erator is able to drive a remote robotized micro-
Virtual Microscope: The system consti-
scope by using software simulating the behaviour
tuted by one or more virtual slides and the
of a microscope.
viewer software needed for their examination.
Store-and-Forward Telepathology: Asyn- The term is also used to identify the viewer
chronous form of telepathology based on the alone.
exchange of still images selected from a glass
Virtual Slide (Digital Slide): A digital
slide, often used for second-opinion consulta-
copy of a complete glass slide or of a substan-
tion.
tially large part of it.
245
246
Chapter XXXII
Collaborative Environments for
the Health Monitoring of
Chronically Ill Children
G. Ganiatsas
K. Starida
Dimitrios I. Fotiadis
Biomedical Research Institute—FORTH, Greece
ABSTRACT
A revolution is taking place in the healthcare field with information technology (IT) playing
an increasingly important role in its delivery. Healthcare providers are exploring IT
opportunities in reducing the overall costs of healthcare delivery while improving the quality
of its provision to citizens. Healthcare services have accumulated great benefits from the
application of information technologies, telecommunications and management tools. Internet,
wireless, and handheld technologies have the capability to affect healthcare by improving
quality, efficiency, and cost-effectiveness of work. Healthcare information systems include a
wide range of applications ranging from diagnostic tools to health management applications
and from inpatient to outpatient monitoring services. Home-care systems address patients and
their families and provide the means to manage their health status related to a specific health
problem. Home-care systems include a wide variety of offered services such as: (a) directory
services (hospital location, doctor specialties), (b) computer patient records (CPR) along with
interfaces for interoperability, (c) certified medical information provision, (d) interfacing to
specialized medical monitoring devices, and (e) synchronous and asynchronous collaboration
services. All these services are offered, most of the time, through secure and seamless
networks.
Collaborative Environments for the Health Monitoring of Chronically Ill Children
INTRODUCTION advanced development in healthcare, enabling

the deployment of healthcare systems for chil-
A revolution is taking place in the healthcare dren who are chronically ill and need continu-
field with information technology playing an ous monitoring of their health conditions (Southall
increasingly important role in its delivery. Fur- et al., 2000). Asthma, diabetes, and chronic
ther exponential growth is expected as the heart problems are among those that raised the
healthcare industry implements electronic medi- demand for continuous monitoring systems, and
cal records, upgrades hospital information sys- several services and products became avail-
tems, sets up intranets for sharing information able powered by different technologies.
among related participants, and uses public Asthma is defined as a chronic inflamma-
networks to distribute health-related informa- tory disease of the airways; this inflammation is
tion and provide remote diagnostics via responsible for the appearance of symptoms
telemedicine (Directorate-General Information associated with reversible airway obstruction,
Society & Immarsat Ltd, 1999). Today’s both spontaneous as well as following treat-
healthcare providers, faced with an unprec- ment, and is a determining factor of concomi-
edented era of competition, are exploring IT tant airway hyperresponsiveness. The clinical
opportunities in reducing the overall costs of situation of acute asthma may be classified as
healthcare delivery while improving the quality mild, moderate, or severe. Moderate asthma is
of its provision to citizens. accompanied by tachypnoea, the use of acces-
Healthcare services have accumulated great sory muscles for respiration, and the inhibition
benefits from the application of information of physical activity. In severe asthma, wheez-
technologies, telecommunications, and man- ing may no longer be audible, cyanosis and
agement tools. The Internet and wireless and shortness of breath are present with the con-
handheld technologies have the capability to traction of the thoracic muscles, and the patient
affect health care by improving the quality, is compelled to stay in bed; the heart rate
efficiency, and cost effectiveness of work (Eder exceeds 120 beats per minute, and PEF and
& Darter, 1998). By integrating data from FEV1 are lower than 50% (Lung Association,
disparate sources—various medical depart- n.d.).
ments, billing systems, insurers, and other medi- Diabetes mellitus is a condition in which the
cal information resources—into a single point amount of glucose (sugar) in the blood is too
of reference and making them available at high because the body cannot use it properly.
anytime and anyplace via the Internet and Glucose comes from the digestion of starchy
wireless technology on a handheld computer, foods such as bread, rice, potatoes, chapatis,
health-care professionals are able to provide yams, and plantain, from the digestion of sugar
the most effective patient care at the point of and other sweet foods; and from the liver,
care. which makes glucose. There are two main
Healthcare information systems include a types of diabetes. Type 1 diabetes, also known
wide range of applications ranging from diag- as insulin-dependent diabetes, develops if the
nostic tools to health-management applications, body is unable to produce any insulin. This type
and from inpatient to outpatient monitoring ser- of diabetes usually appears before the age of
vices. During the last few years, advances in 40. Type 2 diabetes, also known as non-insulin-
telecommunications, Internet technologies, and dependent diabetes, develops when the body
specialized lightweight medical devices have can still make some insulin, but not enough, or
247
when the insulin that is produced does not work directory services (Howes, Smith, & Good,
properly (known as insulin resistance). This 1998).
type of diabetes usually appears in people over
40 years old, though in South Asian and Afri- Computer Patient Records and
can-Caribbean people, it often appears in those Interoperability Mechanisms
over 25 years old (American Diabetes associa-
tion, n.d.). A computer-based patient-record system adds
information-management tools to provide the
maintenance of records, clinical reminders and
BACKGROUND alerts, linkages with knowledge sources for
health-care decision support, analyzes of ag-
Home-care systems address patients and their gregate data, and so forth. An important issue
families and provide the means to manage their when designing and implementing computer
health status related to a specific health prob- patient records lies in the interoperability with
lem (Tatman, Woodroffe, Kelly, & Harris, existing legacy systems.
2001). Home-care systems include a wide va- The most common method used for
riety of offered services such as (a) directory interoperability is the Health Level 7 (HL7)
services (hospital location, doctor specialties), protocol. HL7 is a messaging protocol for elec-
(b) computer patient records (CPRs) along tronic data exchange in health-care environ-
with interfaces for interoperability, (c) certified ments. It defines transactions for transmitting
medical information provision, (d) interfacing data about patient registration, insurance, bill-
to specialized medical monitoring devices, and ing, orders and results for laboratory tests,
(e) synchronous and asynchronous collabora- physiology, image studies, observations, diet
tion services. All these services are offered, orders, pharmacy orders, and supply orders.
most of the time, through secure and seamless The HL7 standard is message based and uses
networks. The following section provides an an event-trigger model that causes the sending
insight on the most commonly used home-care system to transmit a specified message to the
services and applications. receiving unit, with a subsequent response by
the latter (Heath Level 7, n.d.). HL7 handles
Directory Services transactions for exchanging information about
patient administration (admission, discharge,
A directory provides a consistent way to name, transfer, outpatient registration), patient ac-
describe, locate, access, manage, and secure counting (billing), appointment scheduling, prob-
information. A directory is used to store infor- lem lists, clinical-trial enrollments, patient per-
mation that must be shared between applica- missions, voice dictations, advanced directives,
tions in a network environment. Information is and physiologic signals.
organized as a hierarchy of objects with their Another method for interchanging medical
associated attributes. A directory service makes information relies on the Digital Imaging and
this information available as named objects that Communications in Medicine (DICOM) stan-
can be found using familiar white-page-style dard (American College of Radiology/National
lookups or yellow-page-style searches. There Electronic Manufacturers Association, 1983;
is currently a huge amount of interest, com- Digital Imaging and Communications in
ment, and speculation within the industry about Medicine, n.d.). This is a specification that
248
describes the formatting and exchanging of tively. Patients use specialized medical devices
medical images and associated information. It to acquire measurements related to their spe-
relies on standard communication protocols cific diseases, and they communicate the re-
and addresses the communication of images sults to medical professionals using either de-
from digital modalities such as CT, MR, ultra- vice software or communication-link applica-
sound, nuclear medicine, digital cardiology, and tions by inserting the acquired values to specific
angiography and RF equipment, as well as fields (Philips Medical Systems, n.d.). Existing
radiation-therapy devices. It also allows the home-care services often provide communica-
exchange of patient demographics, exam station with healthcare experts through various
tus, and scheduling information. means, although this is not mandatory.
Certified Medical Content Synchronous and Asynchronous

Collaboration Services
A widely used home-care service is found in
the provision of certified medical content through Online collaboration among health specialists
the delivery of personalized medical informa- can be based on advanced telemedicine and
tion or educational material for the patients and teleconference tools of high cost with which
their family. These services range from simple healthcare centers and organizations are
Web pages to more advanced, personalized equipped. Collaboration falls into two main
Web applications. They usually provide health- categories: synchronous collaboration, in which
care tips and instructions on how to manage users communicate with each other at the same
various situations related to the patient’s status. time, and asynchronous collaboration, in which
The increasing availability of large bodies of the communication of the users is achieved at
medical literature online (journal articles, ency- different times (Microsoft, n.d.). Based on
clopedias, patient records) fuels the develop- these categories, collaboration tools and ser-
ment of applications based on some kind of vices are further classified into the following
knowledge processing. Examples of such appli- functional categories.
cations include concept-based indexing and
retrieval, question answering, and, more gener- Synchronous Collaboration
ally, text understanding, as well as medical
corpus or database mining. A sound represen- • Electronic meeting systems.
tation of the medical domain, or ontology, is • Real-time conferencing systems (local and
needed by all of these applications (Medical remote) as well as collaborative presenta-
Ontology Research, n.d.). tion systems (Data Protocols for Multi-
media Conferencing, 1996).
Interfacing to Specialized Medical • Desktop video and real-time data
Monitoring Devices conferencing focusing on real time rather
than BBS or Notes. All products in this
These services include the monitoring of chronic category store documents, and/or allow
diseases through the use of appropriate equip- others to see and work on documents
ment, for instance, portable devices for peak simultaneously or on each other’s screen
expiratory flow or blood-glucose monitoring in or a whiteboard.
order to deal with asthma or diabetes, respec-
249
• Group document handling, including group • User Authorization: Whenever access

editing, shared-screen editing work, group to the services, the data repositories, or
document and image management, and the shared resources of the system is
document databases. required, an access-control mechanism is
invoked, involving access-control lists and
Asynchronous Collaboration security filters. A hierarchy of authoriza-
tion levels is built in order to model the
• Electronic mail and messaging, including different privileges of the system users.
messaging infrastructures and e-mail sys- • Privacy of Transmission: The secure
tems. exchange of health records, personal in-
• Group calendaring and scheduling, includ- formation, and management plans over
ing products for meeting and resource the Web is achieved by employing a PKI
coordination. scheme by means of secure network pro-
• Non-real-time data conferencing. Asyn- tocols like S-HTTP and SSL (Data Pro-
chronous conferencing is similar to a bul- tection Working Party, n.d.).
letin board, in which the user carries on a • Security Policies: Administrative pro-
conversation over time by leaving and cedures and physical safeguards secure
answering messages. These messages can data integrity, confidentiality, and avail-
be public, as in a BBS, or private, as in a ability.
Notes discussion database.
• Work-flow-process diagramming and
analysis tools, work-flow enactment en- MAIN THRUST
gines, electronic forms that route prod-
ucts. The existing home-care systems are either
Web-based or stand-alone applications that act
Security and Privacy as health-management tools for child patients
suffering from chronic diseases. The following
Security mechanisms include all technological section presents the main characteristics, tech-
and operational measures to safeguard the nical features, and services of the major home-
confidentiality and privacy of personal data, as care systems available.
well as to protect the services and the operators
from liability exposure and possible legal sanc- LifeChart
tions.
The usual security and privacy practices of LifeChart has created a system to collect and
most of the existing home-care systems involve store personal health information from individu-
a number of steps. als with chronic diseases. Individuals monitor
their health using personal electronic monitors
• Protection of Resources: To protect and transmit their results to LifeChart through
both data and services from a number of the public telephone network. The information
external threats, firewalls are incorpo- resides in a robust relational database that
rated into the system (Stallings, 1999). allows sophisticated query and analysis. Infor-
• User Authentication: The identities of mation collected from the remote monitor is
legitimate users are verified by means of formatted into reports available for transmis-
digital certificates.
250
sion by automated fax to the healthcare forcing successes and highlighting areas
professional’s fax number listed on the patient’s that need further work
enrollment in the LifeChart information ser- • Communicate with nurses and other
vice. The collected information can be made healthcare professionals via a secure mes-
available to health-care professionals to assist saging system for support, advice, and
clinicians in treating their patients’ diseases; it education
is also available for cross-sectional analysis to • Allow their physician to review their per-
assist in disease management. sonal charts and diaries
• Share their experiences and receive peer
Transtelephonic group support through bulletin boards
Home-Care Services
Transtelephonic monitoring (TTM) is a tech- CHILDCARE

nology that allows patients to report symptoms
and/or transmit data from a medical device The CHILDCARE system supports the out-of-
(usually a pacemaker or implanted cardioverter hospital, continuous health-care management
defibrillator) over the phone for medical evalu- of chronically ill children. CHILDCARE is a
ation by their physician. TTM may be used to complete home-care system designed to pro-
allow patients to have their pacemaker or ICD mote collaboration between health profession-
checks done over the phone, or to transmit data als. Enhancing this collaboration with intelligent
from an electrocardiograph (EKG) machine features can improve the prevention, early de-
over the phone for evaluation. Signals from the tection, prognosis, and treatment of children’s
devices are converted into sound waves, sent diseases while at the same time facilitating
over the phone, received by the monitoring access to best medical practice regardless of
station at the other end of the phone, and the child’s location (CHILDCARE Project,
reconverted back into the original signals n.d.).
(Mednet Healthcare Technologies, n.d.). Special focus is placed on the utilization of
the latest technology accomplishments toward
MyHealthyLife Network the incorporation of teleconference sessions,
collaborative spaces, and intelligent add-ons
Protocol Driven HealthCare Inc. (PDHI) has over a common secure platform. This combina-
developed 10 channels in major disease areas tion allows for (a) remote communication be-
that can be configured, privately labeled, and tween health experts and the child’s family, (b)
integrated with a customer Web strategy. Ac- efficient collaboration between the responsible
cess to any of these sites requires registration doctor and healthcare professionals, (c) the
and is password protected. Nonetheless, the interfacing of legacy hospital information sys-
service is free of charge. PDHI health chan- tems to obtain health records, (d) the setup,
nels allow users to do the following: execution, and monitoring of health-manage-
ment plans, (e) ubiquitous access from multiple
• Input data about their condition, symp- devices, (f) role-based management of the
toms, and lifestyle actors with respect to their actual involvement
• Receive immediate feedback in the form in the healthcare process, and (g) confidential-
of graphs, scores, and reminders, rein- ity and security at all levels.
251
CHILDCARE services are built reflecting ment, and have clearly defined parameters of
the most urgent needs of real users: electronic well-being. CHILDCARE also incorporates
medical records, communication between the the management of the normal development of
patient and the doctor, ability for remote exami- neonates, despite the fact that neonatal devel-
nation and online consultation, automated track- opment is far from a chronic condition. This
ing of the health condition based on monitored target group provides added value to the system
indicators, collaboration between the respon- since neonates are most common and have a
sible doctor and specialized health profession- great impact on the public.
als, delivery of personalized information about
the illnesses, provision of yellow-pages infor-
mation on healthcare resources, and unification CONCLUSION
and flexibility in accessing the services. Pri-
vacy and Web-based-service security mecha- It is easily understood why chronically ill chil-
nisms are integrated in the CHILDCARE plat- dren and adults feel more comfortable receiv-
form, thus safeguarding both the CHILDCARE ing care at home rather than having to visit a
data and its services. All these end up creating healthcare facility. Usually, they are frightened
a system that advances quality in pediatric and under stress, not only due to their medical
home care. condition, which might be a chronic disease, but
CHILDCARE is based on a modular and also because they are in a place completely
flexible design that focuses on the unified inte- foreign to them. Yet, providing these people
gration of technologies, the utilization of well- with a telemedicine tool is not an adequate
established experience, the adopting of power- solution since ongoing health care involves more
ful standards in the exchange of healthcare than just the patient and the doctor. The medi-
records, and the representation of medical in- cal professional needs to work in close coop-
formation, while at the same time advancing eration with a variety of health specialists and
current information-processing technologies. the patient’s family for delivering the best care.
CHILDCARE encourages the initiative and For this reason, a great variety of home-care
effort of all the involved actors. This is achieved systems and services has been developed. The
through ubiquitous and secure services at all most common services provided through these
levels that can be accessed through a vast systems include (a) electronic medical records,
variety of fixed and wireless network inter- (b) streamlined communication between pa-
faces. tient and doctor, (c) cooperation among health
The CHILDCARE framework addresses professionals over a virtual collaborative space,
the management of chronic conditions—cur- (d) the exchange of medical data, (e) the inter-
rently asthma and diabetes—but it can easily be facing of medical equipment that measure medi-
customized to other chronic diseases. The men- cal indicators, and (f) alerts and reminders that
tioned diseases have been most favored by relate to the obtained measurement. The re-
pediatricians (89% and 93%, respectively) as lated legal and ethical implications or con-
far as monitoring through the use of telemedicine straints that concern home-care systems and
services is concerned. The criteria for select- medical procedures are taken into account so
ing diabetes and asthma among other diseases as to ensure that all the appropriate technologi-
are that they are serious, common, amenable to cal and operational measures are delivered.
improved outcomes through better manage-
252
REFERENCES Medical Ontology Research. (n.d.). Retrieved

from February 7, 2006, http://lhnbc.nlm.nih.gov
American Diabetes Association. (n.d.). Re-
Mednet Healthcare Technologies. (n.d.). Re-
trieved January 15, 2005, from http://
trieved February 8, 2005, from http://
www.diabetes.org.uk/
www.mednethealth.net
CHILDCARE project. (n.d.). Retrieved Janu-
Microsoft. (n.d.). Retrieved February 20, 2005,
ary 10, 2005, from http://www.childcare-eu.com
from http://www.microsoft.com/ms.htm
Data Protection Working Party. (n.d.). Pri-
Philips Medical Systems. (n.d.). Retrieved
vacy on the Internet: An integrated EU ap-
March 5, 2005, from http://www3.medical.
proach to on-line data protection (5063/OO/
philips.com/
EN/FINAL, Article 29). Retrieved February 8,
2005, from http://europa.eu.int/comm/ Southall, D., Burr, S., Smith, R. D., Bull, D. N.,
internal_market/en/dataprot/wpdocs/ Randford, A., Williams, A., et al. (2000). The
wp37en.pdf child-friendly healthcare initiative (CFHI):
Healthcare provision in accordance with the
Data protocols for multimedia conferencing
UN Convention on the Rights of the Child.
(Recommendation T.120). (1996). Retrieved
Pediatrics, 106(5), 1054-1064.
February 5, 2005, from http://www.itu.int/rec/
recommendation.asp Stallings, W. (1999). Cryptography and net-
work security: Principles and practice (2nd
Digital imaging and communications in medi-
ed.). Upper Saddle River, NJ: Prentice Hall.
cine (DICOM). (n.d.). Retrieved from http://
medical.nema.org Tatman, M. A., Woodroffe, C., Kelly, P. J., &
Harris, R. J. (2001). Paediatric home care in
Directorate-General Information Society &
tower hamlets: A working partnership with
Immarsat Ltd. (1999). Telemedicine in the 21st
parents. Quality in Health Care, 1, 98-103.
century: Opportunities for citizens, society and
industry. An international space university
workspace. In Proceedings of Telemedicine KEY TERMS
in the 21st century. Strasbourg, France.
Chronic Diseases: Diseases that have
Eder, L. B., & Darter, M. E. (1998). Physicians
one or more of the following characteristics:
in cyberspace. Communications of the ACM,
They are permanent, leave residual disability,
41(3), 52-54.
are caused by nonreversible pathological alter-
Health Level 7. (n.d.). Retrieved January 26, ation, require special training by the patient for
2005, from http://www.hl7.org rehabilitation, and/or may be expected to re-
quire a long period of supervision, observation,
Howes, T., Smith, M., & Good, G. (1998).
or care.
Understanding and deploying LDAP direc-
tory services. London: MacMillan Technical Computer Patient Record (CPR): A re-
Publications. pository of electronically maintained informa-
tion about an individual’s lifetime health status
Lung Association. (n.d.). Retrieved February
and health care, stored in such a way so as to
16, 2006, from http://www.lung.ca/asthma
253
be able to serve the multiple legitimate users of Home-Care Services: Complete applica-
the record. tions that provide the means to allow medical
professionals to monitor and manage their pa-
CPR Interoperability: The ability of two
tients’ health status.
or more computer patient-record systems or
their components to exchange medical infor- Medical Ontology: The terminology used
mation and to use medical information that has to refer to the shared understanding of some
been exchanged. domain of interest, which may be used as a
unifying framework to solve the problem of
Home-Care Products: Medical devices
dispersed medical information.
that enable patients to monitor their health
status (e.g., by providing the ability to measure Real-Time Conferencing Systems: Com-
certain vital indicators), but that can also be posed of software and services that provide a
integrated with other applications that are de- real-time venue for the exchange, creation, and
veloped for this purpose. viewing of information by users in real time.
254
255
Chapter XXXIII
Electronic Submission of
New Drugs in Europe
A. Susanne Esslinger
Friedrich-Alexander-University, Germany
Daniela Marschall
Friedrich-Alexander-University, Germany
ABSTRACT
All over the world, drugs and drug applications have to be submitted to and approved by an
admission office before they may be sold on the market. All procedures are extensive, time-
consuming, and costly. To simplify the process, it could be organised electronically. In an
economic perspective, there are many benefits by using the electronic form for the
pharmaceutical industry: managing knowledge, cost advantages, and time savings. All,
pharmaceutical industry and institutions have undertaken lots of efforts to enforce the
electronic solutions. They focus on international standards in order to harmonise structures
and processes. It would be necessary to reduce paper and copies, especially if the electronic
solution takes place. This method will simplify the way to deal with data and documents and
reduce process time and costs.
INTRODUCTION (Bundesverband der Pharmazeutischen

Industrie e.V., 2004a). In the year 2004, almost
The pharmaceutical industry is characterised 9,000 drugs were listed in the Rote Liste, a
by high expenditures on research and develop- compendium of all medication patients may get
ment. In 2003 in Germany, over 3.56 billion (Bundesverband der Pharmazeutischen
euros were spent for the development of new Industrie e.V., 2004b). The developing process
drugs and new drug applications for a new product lasts on average between 8
Electronic Submission of New Drugs in Europe
and 11 years and costs about 800 million euros the drug. For some medicines, registration is
(Bundesverband der Pharmazeutischen sufficient for the firms: No clinical evaluations
Industrie e.V., 2004a). The duration of a patent or pharmacological-toxicological tests have to
for a new drug or new drug application lasts 20 be conducted by the producer. Within Europe,
years. Usually, in the product-development pro- several regulations, directives, legal decisions,
cess, the application for a letters patent takes and guidelines have to be considered
place early. Therefore, the patent time for the (Bundesverband der Arzneimittel-Hersteller,
producer, the time in which they can promote n.d.). They are especially concerned with tech-
the drug on the market and amortise the costs nical aspects and harmonisation, form, and
of research and development, is reduced to content. The detailed Common Technical Docu-
about 10 years (Bundesverband der ment (CTD) has existed since 2003. It is a
Pharmazeutischen Industrie e.V., 2004a). Thus, guideline for the technical documentation of
time to market is a critical issue for firms in any drug in Europe, the United States, or Japan
order to economically succeed in a long-term and was established in the International Con-
perspective. The duration for research and ference on Harmonisation (ICH; BfArM, 2003).
development of a drug usually is fixed. But the Formal aspects of the dossier are shown in the
time for the submission of a new medicine, European standardised Notice to Applicants
which takes about 2 years, may be reduced. If (NTA; Wagner, 2000). The content specifica-
firms do so, they may gain profit. tions are described in EU (European Union)
Electronic solutions are more and more com- guidelines or country-specific rules. In Ger-
mon in the healthcare sector (e.g., electronic many, for instance, firms must fulfill diverse
prescription), and it may be useful to submit criteria set by the Arzneimittelgesetz (AMG),
pharmaceutical products electronically. This the AMG-Einreichungsverordnung, the
article will give a closer look on the opportuni- Verwaltungsverfahrensgesetz, and the
ties of electronic submission concerning pro- Arzneimittelzulassungsgesetz. Products re-
cesses, time, and costs. leased in just one country might be accredited
just for this region. But drugs that are placed on
the European market have to fulfill European
SUBMISSION OF DRUGS standards. Depending on the requested licens-
ing, the submission dossier for a new drug or
All over the world, drugs and drug applications drug application will be made up of about 1,000
have to be submitted to and approved by an folders and more than 500,000 pages
admission office before they may be sold on the (Mitteleuropäische Gesellschaft für Regula-
market (Jordan, 2002). Usually, for the submis- tory Affairs, n.d.).
sion, the producer has to get in contact with the As one can imagine, all procedures are
local admission office (for example, in Ger- extensive, time consuming, and costly. To sim-
many, it is the Bundesinstitut für Arzneimittel plify the process, it could be organised elec-
und Medizinprodukte, BfArM, 2004) and with tronically. Doing so, in 1985, the United States
the European Medicines Evaluation Agency founded the Computer Assistance in New Drug
(EMEA; Europäische Arzneimittel-Agentur, Applications (CANDA) project. In 1994, the
2003). The firm has to send a submission dos- SMART (Submission Management and Track-
sier to the office. It contains information about ing System) project displaced the labor-inten-
the harmlessness, effectiveness, and quality of sive CANDA project. This system is based on
256
the guideline “Providing Regulatory Submis- harmonisation (ICH). Thus, as an example, the
sions in Electronic Format: General Consider- German BfArM would accept the format of the
ations.” It is a data-oriented approach with a DAMOS but prefers the electronic CTD
general and a specific section. Thus, the evalu- (eCTD) format. This format has assumed the
ation is based on data and supported by state- basic structure and specification of DAMOS
ments of experts and textual documentation. and will displace it completely in the close
The legal framework of SMART is the future. It can be observed that in Germany,
Electronic Records and Electronic Signatures nobody hands in DAMOS-based dossiers. In
Regulation (21 CFR Part 11; U.S. Department fact, the CTD format has been obligatory in
of Health and Human Services, Food and Drug Europe since July 2003.
Administration [FDA], CDER, & CBER, 1999).
Today in the United States, over 80% of all
submissions, including new-drug applications, The Electronic Common Technical
are conducted electronically, and the process Document
time has been reduced by more than half a year
in each of the cases (Food and Drug Adminis- According to eCTD (2005):
tration, 2004).
The eCTD is defined as an interface for
industry to agency transfer of regulatory
The DAMOS Initiative information while at the same time taking
into consideration the facilitation of the
In Europe in the year 1989, the DAMOS (Drug creation, review, lifecycle management and
Application Methodology with Optical Storage) archival of the electronic submission. The
project was the first to submit drugs electroni- eCTD specification lists the criteria that will
cally (Franken, 2003a). In 1993, the admission make an electronic submission technically
office started the pilot. Two years later, in 1995, valid. The focus of the specification is to
a software developer offered the first elec- provide the ability to transfer the registration
tronic dossiers on the market. At the same time, application electronically from industry to a
the firm started to invent a review tool for the regulatory authority.
admission office. In 1998, DAMOS was offi-
cially accepted as a possible way to hand in the Besides the transfer:
electronic dossier for the submission of a new
drug. Documents from experts are evaluated in the eCTD and related standards will help
the first step, and data in the second step (top- pharmaceutical companies to control global
down process). Therefore, the focus in elec- submissions, product life cycle, time to
tronic documents is on text besides data, which market and supply chain cost-effectiveness
is different from the SMART solution. DAMOS (eCTD, 2005).
is structured in two parts. One part only illus-
trates the structure of the complete dossier, and The guidelines are based, as mentioned above,
the other part is content (Franken, 2003a). on the CTD. In order to send data via Web
In Europe, submission processes are still browsers, the format of the data is XML (Ex-
country specific. However, lots of efforts are tensible Markup Language; Franken, 2003b).
undertaken concerning worldwide
257
“The consequence of this approach is that in ECONOMICAL ASPECTS

order to receive, validate and review eCTD
compliant applications, regulatory authorities Benefits
need to have a system in place.” In this context,
the most critical initial user requirement of such From an economic perspective, there are many
a system is actually the European Union Re- benefits by using the electronic form for the
view System (EURS, 2002). Besides the eCTD, pharmaceutical industry: managing knowledge,
firms need an information-management system cost advantages, and time savings.
(IMS). According to PIM (2005): First of all, it is of high importance to reveal
implicit knowledge and make it transparent.
PIM is a system to be introduced by the This is an essential success factor for pharma-
EMEA in the first instance in November ceutical firms as their success is mainly based
2005. It has been conceived as a means of: on research and development. Economies of
a) increasing the efficiency of the scope may be gained easily (Erhard, 2003), and
management and exchange of product organisational learning in a long-term perspec-
information (summary of product tive is possible. Also, the quality of the docu-
characteristics, package leaflet and ments will be improved (Vis-à-Vis, 2002) be-
labelling) by all parties involved in the cause of well-informed employees. It might
evaluation process through the structuring also be possible to standardise processes due to
of the information and its exchange by routines.
electronic means; and b) improving the Second, cost advantages may be realised:
quality and consistency of the published Costs of paper, copies, mailing, and storage
product information. may be saved. Firms using electronic submis-
sion do not need plenty of bookshelves to store
Within Europe, experts still have problems the documents. Above all, they do not need the
with complete electronic submission because laborious transportation of documents from their
of security concerns. In the EU Directive 1999/ offices to the archive or from expert to expert.
93/EG, the common frames for electronic sig- In consequence, logistical overhead will be
natures are set already. The directive may reduced.
differ in the various European countries (Hock Third, time savings are possible. The elec-
& Jostes, 2003). If a pharmaceutical producer tronic submission of new drugs allows the
decides to hand in the complete documents in admission office to access the complete docu-
an electronic form, he or she still has to hand in mentation with one click. Thus, searching for
a textual version. The latter is still the only one one specific aspect one might need for the
legally accepted (Europäische Arzneimittel- submission will proceed more quickly.
Agentur, 2004), and it is still necessary to sign All these benefits allow the firms to speed
some of the needed documents in most Euro- up the submission process significantly. Time to
pean countries. Besides this, the pharmaceuti- market for a new drug or a new-drug applica-
cal firms may send their electronic versions. In tion is improved (Vis-à-Vis, 2002). The phar-
order to do so, they have to consider, as it was maceutical producer will be able to profit from
said earlier, a wide range of regulations. this pioneering. Being the first in the market, the
company skims the payers’ high willingness to
258
pay for a new innovative drug. The medicine is vestment takes at least 12 months (Zimmer,
known as a brand. Thus, the firm gains market 2003).
shares. On the basis of pioneer experiences of Cost calculation concerned with
the drug, they may further improve the quality reorganisation and process optimisation is a
of other drugs and create bonding of the patient difficult task. Also, costs related to amortisa-
and doctors on the firm’s brands. tion, the maintenance of software, network
Actually, the measurement of the benefits in administration, and support for users are diffi-
firms and institutions is still difficult. There are cult to measure. Nevertheless, it is very impor-
some explorative data. It was measured that tant to compare the realisable benefits with the
the time of the submission process was reduced calculated costs.
by over 75% (Höniger, 2003) through elec-
tronic submission. Also, costs for the process
were reduced by over 60% (Witzel, Yamaguchi, CONCLUSION
& Lorenz, 1998). Have a product on the market
one day earlier creates an added daily average The electronic submission of drugs today is
turnaround of more than 1 million euro reality. All pharmaceutical industry and soft-
(Baldowski, 2000; Kainz & Harmsen, 2003; ware firms as well as institutions have under-
Schmitt, 2003). taken lots of efforts to enforce electronic solu-
tions. They focus on international standards in
Costs order to harmonise structures and processes.
The problem is still the electronic signature. It
The electronic submission of drugs does not will be necessary to reduce paper and copies,
only create cost advantages. Some investments especially if the electronic solution takes place.
have to be made. For realising electronic sub- Still, there are some pharmaceutical firms that
mission, the pharmaceutical firms have to imple- have no need to change to the electronic sys-
ment management information systems. For tem. But if they think it over, this method will
this reason, they have to invest in hardware and simplify the way they deal with data and docu-
software. These costs (e.g., for the search and ments, and reduce process time and costs.
selection of adequate software) are of high
importance especially for the pharmaceutical REFERENCES
industry because information techniques do not
necessarily belong to their core competencies. Baldowski, N. (2000). Elektronische
The configuration and maintenance of the sys- arzneimittelzulassung verkürzt die behandlung
tem may even require employing specialists. der anträge. Computerwoche, 20, 75-77.
Training is necessary in order to achieve the
Bundesinstitut für Arzneimittel und
organisational learning of the employees men-
Medizinprodukte (BfArM) (Ed.). (2003).
tioned above. Additionally, a professional sup-
Hinweise zum einreichen von
port system for the users has to be assured. The
zulassungsanträgen im CTD: Format beim
firms also have to install firewalls in order to
bundesinstitut für arzneimittel und
protect their data.
medizinprodukte. Retrieved November 10,
Depending on the implemented system, ex-
2004, from http://www.bfarm.de/de/
perts estimate that the investment at least takes
Arzneimittel/zul/CTD.pdf
250,000 euro, and the amortisation of the in-
259
Bundesinstitut für Arzneimittel und tion using the electronic common technical
Medizinprodukte (BfArM) (Ed.). (2004). document (“eCTD”) following the centralised
Erläuterungen zum vollzug der verordnung procedure. Retrieved from November 10, 2004,
über die einreichung von unterlagen in from http://esubmission.eudra.org/EMEA%
verfahren für die zulassung und verlängerung 20eCTD%20Guidance%200.4-Draft.doc
der zulassung von arzneimitteln (AMG-
European Union Review System (EURS).
Einreichungsverordnung: AMG-EV), Version
(2002). Requirements document for EURS
4.09. Retrieved November 10, 2004, from http:/
version 1: Telematics Implementation Group on
/www.bfarm.de/de/Arzneimittel/amg_ev/Erl-
Electronic Submissions (TIGes).
V4-9.pdf
Food and Drug Administration. (Ed.). (2004).
Bundesverband der Arzneimittel-Hersteller
Electronic common technical document
e.V. (Ed.). (n.d.). ICH: International confer-
(eCTD). Retrieved November 26, 2004, from
ence on harmonisation of technical require-
http://www.fda.gov/cder/regulatory/ersr/
ments for registration of pharmaceuticals
ectd.htm
for human use. Retrieved November 10, 2004,
from http://www.bah-bonn.de/arzneimittel/ Franken, A. (2003a). eSubmission manage-
europa/ich.htm ment im regulatorischen umfeld von
arzneimittelzulassungen. Pharmind: Die
Bundesverband der Pharmazeutischen Industrie
Pharmazeutische Industrie, 65(5a), 463-472.
e.V. (Ed.). (2004a). Pharmadaten 2004. 34.
Franken, A. (2003b). Zukunft des
Bundesverband der Pharmazeutischen Industrie
regulatorischen e-managements bei der
e.V. (Ed.). (2004b). Pharmadaten 2004
arzneimittelzulassung. Pharmind: Die
kompakt. Retrieved November 10, 2004, from
http://www.bpi.de/internet/download/
pharmadaten_2004_kompakt.pdf Hock, S., & Jostes, T. (2003). Gesetzliche
regelungen zu elektronischen unterschroften
Electronic common technical document
bei der einreichung von zulassungsunterlagen
(eCTD). (2005). Retrieved July 22, 2005, from
für arzneimittel. Pharmind: Die
http://www.ectd.com/about.htm
Erhard, D. (2003). Erfolgskriterien für die
Höniger, L. (2003). Einführung und nutzung
einrichtung eines elektronischen dokumenten-
von elektronischen verfahren bei der
managament-systems zur unterstützung der
arzneimittelzulassung: Dargestellt am beispiel
erstellung von zulassungsdossiers. Pharmind:
von regierungsaktivitäten in den länern mittel-
Die Pharmazeutische Industrie, 65(5a), 498-
und osteuropas. Pharmind: Die
502.
Europäische Arzneimittel-Agentur. (Ed.).
Jordan, H. (2002). Regulatory affairs. In O.
(2003). Explanatory note on fees payable to the
Schöffski, F. U. Fricke, W. Guminski, & W.
EMEA.
Hartmann (Eds.), Pharmabetriebslehre (pp.
Europäische Arzneimittel-Agentur. (Ed.). 177-193).
(2004). Practical guidance for electronic
Kainz, A., & Harmsen, S. (2003). Einsatz eines
submission of regulatory information in sup-
dossier-management-systems in der
port of a marketing authorisation applica-
260
arzneimittelindustrie: Auswahl, einführung und Zimmer, G. (2003). Dokumenten-management-

praktische erfahrungen. Pharmind: Die systeme als basis für die einreichung von
Pharmazeutische Industrie, 65(5a), 511-519. arzneimittel-zulassungsdossiers. Pharmind:
Die Pharmazeutische Industrie, 65(5a), 537-
Mitteleuropäische Gesellschaft für Regulatory
545.
Affairs e.V. (Ed.). (n.d.). Die aufgaben einer
zulassungsabteilung & arzneimittel und regu-
latory affairs. Retrieved November 10, 2004, KEY TERMS
from http://www.megra.org/html/affairs.htm
Benefits: The positive implications, both
PIM (Ed.). (2005). EU-telematics. Retrieved
direct and indirect, resulting from some action.
July 22, 2005, from http://pim.emea.eu.int/
They include both financial and nonfinancial
Schmitt, A. (2003). Umgang mit anforderungen information.
der elektronischen einreichung von arzneimitteln
Costs: The total spent for goods or services
beim pharmazeutischen unternehmner:
including money, time, and labor.
Dargestellt am beispiel von änderungsanzeigen.
Pharmind: Die Pharmazeutische Industrie, Drug: A substance that is used as a medi-
65(5a), 525-530. cine or narcotic.
U.S. Department of Health and Human Ser- Economical Aspects: Factors relating to
vices, Food and Drug Administration (FDA), an economy. Practically, it is using the mini-
CDER, & CBER (Eds.). (1999). Guidance for mum time or resources necessary for effec-
industry. Providing regulatory submission tiveness.
in electronic format: General considerations.
Electronic Signature: A paperless way to
Retrieved November 10, 2004, from http://
sign a document using an electronic symbol or
www.fda.gov/cder/guidance/2867fnl.pdf
process attached or associated with the docu-
Vis-à-Vis (Ed.). (2002). COI-PharmaSuite: ment.
Zulassung leicht gemacht! Retrieved August
Electronic Submission: Submission made
18, 2004, from http://www.visavis.de/
via a modem or computer disk. Very few
modules.php?name=News&file=article%sid=254
editors accept electronic submissions of unso-
Wagner, S. A. (2000). Europäisches licited manuscripts.
zulassungssystem für arzneimittel und
Guideline: A detailed plan or explanation
parallelhandel unter besonderer berück-
to guide one in setting standards or determining
sichtigung des deutschen arzneimittelrechts.
a course of action.
Witzel, W., Yamaguchi, A., & Lorenz, R.
Time to Market: The period between
(1998). Damos experience report.
product development and launch on the market.
261
262
Chapter XXXIV
Semantic Web Services
for Healthcare
Christina Catley
Carleton University, Canada
Monique Frize
University of Ottawa, Canada
Dorina Petriu
ABSTRACT
This chapter explores the ongoing efforts to integrate Web services and the Semantic Web for
the purposes of sharing knowledge, enabling access to services, and application integration
in distributed clinical environments. Combining the Semantic Web and Web services in relation
to the healthcare domain, results in Semantic Web services for healthcare, which will enable
intelligent interpretation of healthcare data by services such as clinical decision support
systems. Critical issues in ontology standardization and security are discussed. The multi-
disciplinary problem of service composition is presented with emphasis on the role healthcare
experts play in identifying value-added medical services.
NEXT-GENERATION INTERNET: changeable Semantic Web services used in

THE SEMANTIC WEB AND distributed but related medical domains. Ac-
WEB SERVICES cording to Hoffman (2003, p. 54), “Entire eco-
systems of electronic services will be built
The Semantic Web and Web services are two around specific industries, providing specific
complementary and evolving technologies that processes to solve specific problems for spe-
will change the face of healthcare delivery. cific types of customers, through specific trans-
Healthcare IT experts predict that in the future, action chains.”
healthcare services will be offered as inter-
Semantic Web Services for Healthcare
Web services constitute an important emerg- • Description: A Web-service interface

ing technology for which potential applications provides a collection of operations acces-
are unlimited. Stafford (2003, p. 27) explains sible through standardized XML messag-
that “if a provider can imagine a way of deliv- ing. This interface is described using the
ering something of value to a customer to Web services description language
provide some usefulness … they have a viable (WSDL), which specifies the operations
Web service.” Web services are based on a provided by a Web service (Graham, 2003).
service-oriented system architecture, in which • Discovery: The service requestor dis-
providers assess which applications they can covers the Web service via discovery
offer as services to different groups of potential agencies, such as universal description
users. There are three main roles in a service- discovery and integration (UDDI), which
oriented architecture: a provider, a consumer, allows service descriptions to be pub-
and a directory. In the case of a Web-services lished and discovered.
system, the provider publishes descriptions of
its Web services in a directory, which is acces- Adding semantics to the Web is a necessary
sible by the consumer. Once the consumer component toward realizing Web services’ goal
selects a service, the consumer and the service of application-to-application integration. To this
are dynamically bound, for example, at run end, the next-generation Internet will be the
time. The publish-find-bind model of interaction Semantic Web. The vision of the Semantic
enables the loose coupling of providers and Web is to associate meaning to all Web re-
consumers and, thus, increases the agility, flex- sources such that they can be discovered and
ibility, and adaptability of distributed systems consumed autonomously by applications
(Tosic, 2004). (Berners-Lee, Hendler, & Lasilla, 2001), mak-
Web services facilitate integration and ing the Semantic Web a meaningful indexed
interoperability because the underlying imple- repository of documents and services (Lee,
mentation and deployment platform are not Patel, Chun, & Geller, 2004). Currently, the
relevant to the application invoking the service. interpretation of Web-based information re-
There are three key components of Web-ser- quires human knowledge and intuition; both
vice systems with three major corresponding humans and machines could interpret the Se-
XML- (extended markup language) based stan- mantic Web.
dardization initiatives proposed by the World Schweiger, Brumhard, Hoelzer, and Dudeck
Wide Web Consortium (W3C) to support the (2005, p. 274) claim that because the Web and
interactions among Web services. healthcare systems are both “little organized
systems of distributed data,” innovations in
• Delivery: Comprises all technologies re- Web technology are particularly relevant to the
quired to transport a service request from healthcare industry. As such, the development
the client to the server, including XML for of the Semantic Web will impact healthcare in
message encoding, and SOAP (previously numerous ways, such as in retrieving informa-
known as the simple object access proto- tion from multiple disparate databases so that
col, now considered a misnomer) for han- patient mobility will not affect the continuity of
dling the transmission of XML-formatted individual care and the transfer of patient infor-
data. mation (Sun, 2004), and in enabling machines to
263
capture and provide clinicians with the informa- allied healthcare professionals, and facilitating
tion stored in clinical guidelines and scientific data processing by medical software.
publications. Although there is a need for clinical deci-
sion-support systems (CDSSs) to aid physi-
cians in making optimum diagnoses and reduc-
SEMANTIC WEB SERVICES ing medical errors (Committee on Quality of
FOR HEALTHCARE Health Care in America, Institute of Medicine,
2000; Wilson, Runciman, & Gibberd, 1995),
The healthcare domain is defined by a plethora from a technical perspective, the advancement
of distributed data and knowledge from which of CDSSs has been severely hampered by two
complex and timely high-value decisions must key factors. First, clinical data are intended for
be made in a low-tolerance environment (Turner, use by humans; even in the case of electronic
Rigby, et al., 2004). Wreder and Deng (1999, p. patient records (EPRs), the data have no at-
250) elaborate: “How to migrate from stove- tached Semantic meaning and are primarily
pipe systems to the next generation of open intended for human viewing. Second, the clini-
healthcare information systems that are cal data needed to make decisions are con-
interoperable, extensible and maintainable is stantly increasing, heterogeneous, fragmented,
increasingly a pressing problem for the and distributed geographically, making it diffi-
healthcare industry.” Semantic Web services cult to both train and deploy CDSSs. This
are being advocated as a logical means of problem is compounded by the implicitly private
achieving open healthcare systems. Current nature of healthcare data and the resulting
applications span a range of healthcare do- restrictions from an ethical viewpoint.
mains (Catley, Frize, Petriu, Walker, & Yang, Kwon (2003) reports that as decision-sup-
2004; Chatterjee, 2003; Lee et al., 2004; Turner, port-system environments are rapidly changing
Rigby, et al.) and encompass services with from centralized and closed to distributed and
diverse goals, from knowledge management, to open, scalability and interoperability features
application integration, to clinical decision sup- are becoming more crucial to CDSS develop-
port. ment. Combining the Semantic Web and Web
With the ability to offer sophisticated appli- services offers a solution, providing physicians
cations to consumers via Semantic Web ser- with instant access to knowledge, not just data,
vices, a new dimension emerges: applying the in real-time decision-making environments.
service-oriented concept to data and software. A sophisticated healthcare example involves
Turner, Budgen, and Brereton (2003) have using Semantic Web services to reduce the
coined the terms software as a service (SaaS) waiting times for noncritical surgery (Motta,
and data as a service (DaaS). The vision is to Domingue, Cabral, & Gaspari, 2003) by provid-
eliminate many of the problems that occur with ing five interacting Semantic Web services: (a)
developing, updating, and evolving software a diagnostic service to diagnose conditions based
systems (Turner, Rigby, et al., 2004) and, indi- on a set of symptoms, (b) a yellow-page service
rectly, the data accessed by these systems. The indicating which hospitals in Europe provide
implications for healthcare mean offering the specific medical services, (c) a cost-query ser-
potential to easily share distributed heteroge- vice to provide the cost of medical services on
neous medical data between researchers and a per-hospital basis, (d) an ambulance service
264
to determine the cost of transporting patients gies are needed to define healthcare standards
between hospitals, and (e) an exchange-rate and the mapping between them. In simplest
service for converting between European cur- terms, ontologies define the common words and
rencies. Semantic Web services are also being concepts used to describe and represent an
applied to artificial-intelligence-based CDSSs. area of knowledge (Daconta, Obrst, & Smith,
Other work describes a Web-services infra- 2003). Standard ontologies, schemas, and vo-
structure for linking obstetrical, perinatal, and cabularies are a prerequisite for the Semantic
NICU (neonatal intensive care unit) data with Web (Schweiger et al., 2005). In describing
CDSSs for the purposes of predicting preterm their experience in creating a healthcare infor-
birth, exploring indicators of cesarean birth mation broker, Turner, Rigby, et al. (2004) state
(Catley, Frize, Petriu, et al., 2004), and predict- that future work will require an ontology ser-
ing critical outcomes in the NICU, such as vice comprising multiple ontologies as this is
mortality, length of stay, and duration of venti- considered the only viable option for providing
lation (Frize, Ennett, Stevenson, & Trigg, 2001; a global view of an information space. While
Tong, Frize, & Walker, 2002). The CDSSs efforts such as Health Level 7 (HL7) and
offered as Web services in the infrastructure Digital Imaging and Communications in Medi-
include (a) trained artificial neural networks for cine are providing a common ground for de-
outcome prediction (Ennett & Frize, 2003), (b) scribing healthcare information, they are not
case-based reasoners for matching an indi- currently at a level to resolve all relevant
vidual patient’s condition to the most similar interoperability issues (Lee et al., 2004).
past cases (Frize & Walker, 2000), (c) alert While Web services offer benefits and op-
generation for notifying physicians of potential portunities, they also present many challenges
complications via mobile devices (Catley, Frize, in healthcare environments, such as the need to
Walker, & St-Germain, 2003), and (d) an ethi- comply with health-insurance regulations, and
cal decision-support tool for parents of very increased requirements for reliability, security,
sick infants to help them make difficult deci- and monitoring (Chatterjee, 2003). Currently,
sions, such as withholding or terminating criti- one of the biggest obstacles in deploying Se-
cal care (Frize, Yang, Walker, & O’Connor, in mantic Web services is security (Daconta et
press). al., 2003); this is of particular concern in
healthcare when maintaining the privacy of
sensitive patient data is paramount. Authenti-
CRITICAL ISSUES cation, authorization, confidentiality, data integ-
rity, and nonrepudiation are all security con-
While some large healthcare systems use a cerns that relate to Web services and must be
common data model to integrate information addressed before such services can be exposed
from multiple facilities, the majority of existing externally. The Web-services industry is work-
clinical information is stored in heterogeneous ing to overcome security concerns with initia-
databases using different names and different tives such as XML signatures for validating
data models (Sun, 2004). Although the Seman- message integrity and nonrepudiation, XML
tic Web provides a means to attach meaning to encryption for data confidentiality, the security
data, there is no guarantee that everyone agrees assertion markup language (SAML), an OA-
with this meaning; in order to deal with numer- SIS (Organization for the Advancement of
ous and heterogeneous data formats, ontolo- Structured Information Standards) standard that
265
provides assertions of trust between parties, A notable composition initiative is the work
and WS-Security, which combines XML signa- by Lee et al. (2004) on Semantic medical
tures and XML encryption with standard SOAP services. Their team is determining how het-
messaging. Given that Web-service security is erogeneous medical Web services interoperate
still evolving, the majority of applications de- in a medical service flow for the cardiovascular
scribed in the literature are based on the de- domain using three kinds of knowledge: syntac-
ployment of Web services on intranets, which tic, Semantic, and contextual. First the syntac-
nevertheless provides great potential for the tic constraints must be met; an example would
integration of data and services across large be obtaining a valid patient identifier before
healthcare organizations. allowing the invocation of a service. The Se-
mantic knowledge constrains the order in which
services are invoked and requires input from
FUTURE TRENDS domain experts, such as those with knowledge
of healthcare policies, health insurance, and
Automatic Web service composition has re- drug regulations. Contextual knowledge identi-
cently taken center stage as an emerging re- fies the situations in which a service should be
search area (Medjahed, Bouguettaya, & used; for example, if a service for choosing a
Elmagarmid, 2003). As more Web services blood lab returns multiple options, the system
become available, it is possible to offer con- could select the lab closest to the patient.
sumers more complex services by combining Contextual constraints are sophisticated and
simple ones. Candidate Web services can be involve detailed knowledge of the patient’s
categorized as being either core or composite preferences.
services. A core Web service offers basic Efforts are also under way to deploy mobile
functionality that will potentially be required by applications based on loosely coupled Web
multiple higher level applications. Composite services. Chatterjee (2003) claims that the
Web services represent these higher level ap- combination of lightweight and almost ubiqui-
plications, which combine two or more core tous Web services together with around-the-
services to offer a complete system application clock access to mobile devices represents a
as seen from the user’s perspective; the com- powerful platform for the development and
plete application is referred to as a composi- delivery of pervasive and cost-effective
tion scenario. healthcare applications and systems.
While Web-service delivery, description, and
discovery are largely technical problems under
the domain of IT experts, service composition CONCLUSION
represents a complex multidisciplinary prob-
lem. Healthcare experts are needed to identify Based on the acceptance of Web services
value-added medical services and to determine predicted by IT experts (Daconta et al., 2003;
appropriate service composition scenarios, in- Kreger, 2003; Lea & Vinoski, 2003), many
cluding contextual knowledge that indicates healthcare providers are starting to leverage
when and why certain services should be com- Web services as a solution to data and applica-
bined. Due to the more complex nature of interdis- tion integration (Chatterjee, 2003). The Seman-
ciplinary work, service composition in healthcare tic Web provides a mechanism for adding mean-
is currently an evolving research area. ing to data, essential in healthcare when the
266
same clinical information can have many dif- Committee on Quality of Health Care in America,
ferent representations. Combining the Seman- Institute of Medicine. (2000). To err is human:
tic Web and Web services in relation to the Building a safer health system. Washington,
healthcare domain results in semantic Web DC: National Academy Press.
services for healthcare, which will ultimately
Daconta, M. C., Obrst, L. J., & Smith, K. T.
enable the automated interpretation of clinical
(2003). The Semantic Web: A guide to the
data. Semantic Web services for healthcare
future of XML, Web services, and knowl-
open new possibilities in knowledge manage-
edge management. Indianapolis, IN: Wiley
ment, clinical decision support, and application
Publishing, Inc.
integration. Semantic Web services have the
potential to support an advanced healthcare Ennett, C. M., & Frize, M. (2003). Weight-
environment, offering new applications and elimination neural networks applied to coronary
services to health networks. Primary issues of surgery mortality prediction. IEEE Transac-
concern are data security and confidentiality. tions of Information Technologies in Bio-
medicine, 7(2), 86-92.
Frize, M., Ennett, C. M., Stevenson, M., &
REFERENCES
Trigg , H. C. E. (2001). Clinical decision-
support systems for intensive care units using
Berners-Lee, T., Hendler, J., & Lasilla, O.
artificial neural networks. Medical Engineer-
(2001). The Semantic Web. Scientific Ameri-
ing & Physics, 23, 217-225.
can, 284(5), 34-43.
Frize, M., & Walker, C. R. (2000). Clinical
Catley, C., Frize, M., Petriu, D. C., Walker, C.
decision-support systems for intensive care
R.., & Yang, L. (2004). Towards a Web ser-
units using case-based reasoning. Medical
vices infrastructure for perinatal, obstetri-
Engineering and Physics, 22(9), 671-677.
cal, and neonatal clinical decision support.
Proceedings of the 26th Annual International Frize, M., Yang, L., Walker, R. C., & O’Connor,
Conference of the IEEE Engineering in Medi- A. (in press). Conceptual framework of knowl-
cine and Biology Society, San Francisco. edge management for ethical decision-making
support in neonatal intensive care. Transac-
Catley, C., Frize, M., Walker, C. R., & St-
tions of Information Technology in Biomedi-
Germain, L. (2003). Integrating clinical alerts
cine.
into an XML-based healthcare framework for
the neonatal intensive care unit. Proceedings Graham, S. (2003). Building Web services
of the 25th Annual International Conference with Java: Making sense of XML, SOAP,
of the IEEE Engineering in Medicine and WSDL, and UDDI. Indianapolis, IN: SAMS
Biology Society (pp. 1276-1279). Publishing.
Chatterjee, S. (2003). Developing enterprise Hoffman, D. (2003). Marketing + MIS = e-
Web services and applications: Opportunities service. Communications of the ACM, 46(6),
and best practices for the healthcare industry. 29-34.
Fifth International Workshop on Enterprise
Kreger, H. (2003). Fulfilling the Web services
Networking and Computing in Healthcare
promise. Communications of the ACM, 46(6),
Industry (p. 159).
29-34.
267
Kwon, O. B. (2003). Meta Web service: Build- nal of Biomedical Informatics, 37(3), 162-
ing Web-based open decision support system 178.
based on Web services. Expert Systems with
Tong, Y., Frize, M., & Walker, R. (2002).
Applications, 24(4), 375-389.
Extending successful predictions of ventilation
Lea, D., & Vinoski, S. (2003). Middleware for needs using artificial neural networks from
Web services. IEEE Internet Computing, 7(1), adult to neonatal ICU patients. IEEE Transac-
28-29. tions on Information Technology in Bio-
medicine, 6(2), 188-191.
Lee, Y., Patel, C., Chun, S. A., & Geller, J.
(2004). Compositional knowledge management Tosic, V. (2004). Service offerings for XML
for medical services on Semantic Web. Pro- Web services and their management appli-
ceedings of the 13 th International World cations. Unpublished doctoral dissertation,
Wide Web Conference on Alternate Track Carleton University, Tosic, Ottawa, Canada.
Papers & Posters (pp. 498-499).
Turner, M., Budgen, D., & Brereton, P. (2003).
McIlraith, S. A., & Martin, D. L. (2003). Bring- Turning software into a service. IEEE Com-
ing semantics to Web services. IEEE Intelli- puter, 36(10), 38-44.
gent Systems, 18(1), 90-93.
Turner, M., Rigby, M., Zhu, F., Kotsiopoulus, I.,
Medjahed, B., Benatallah, B., Bouguettaya, A., Russell, M., Budgen, D., et al. (2004). Using
Ngu, A. H. H., & Elmagarmid, A. K. (2003). Web service technologies to create an informa-
Business-to-business interactions: Issues and tion broker: An experience report. Proceed-
enabling technologies. The International Jour- ings of the 26th International Conference on
nal on Very Large Data Bases, 12(1), 59-85. Software Engineering (pp. 552-561).
Medjahed, B., Bouguettaya, A., & Elmagarmid, Wilson, R. M., Runciman, W. B., & Gibberd, R.
A. K. (2003). Composing Web services on the W. (1995). The quality in Australian health care
Semantic Web. The International Journal on study. Medical Journal of Australia, 163,
Very Large Data Bases, 12(4), 333-351. 458-471.
Motta, E., Domingue, J., Cabral, L., & Gaspari, Wreder, K., & Deng, Y. (1999). Architecture-
M. (2003). IRS-II: A framework and infra- centered enterprise system development and
structure for Semantic Web services. Interna- integration based on distributed object technol-
tional Semantic Web Conference 2003 (pp. ogy standard. Proceedings of the 23 rd Annual
306-318). International Computer Software and Ap-
plications Conference (pp. 250-258).
Schweiger, R., Brumhard, M., Hoelzer, S., &
Dudeck, J. (2005). Implementing health care
systems using XML standards. International KEY TERMS
Journal of Medical Informatics, 74(2-4),
267-277. Clinical Decision-Support System: A
computer program designed to aid physicians
Stafford, T. (2003). E-services. Communica-
and others (parents, patients, clinicians) in the
tions of the ACM, 46(6), 26-28
decision-making process. Applications include
Sun, Y. (2004). Methods for automated con- generating alerts, diagnostic assistance, and
cept mapping between medical databases. Jour- therapy planning.
268
Integration: Implies both semantic and capabilities and content in an unambiguous,

technical interoperability, as well as a logical computer-interpretable language (McIlraith &
integration flow between interoperable mod- Martin, 2003).
ules.
Web Service: Software applications that
Interoperability: The ability of two or can be discovered, described, and accessed
more systems or components to exchange in- based on XML and standard Web protocols
formation and to use the information that has over intranets, extranets, and the Internet
been exchanged. (Daconta et al., 2003).
Ontology: An explicit, formal specification World Wide Web Consortium (W3C): A
representing the knowledge entities in a do- forum for information, commerce, communica-
main, such as objects and concepts, and the tion, and collective understanding that develops
relationships between them. open Web standards and guidelines with the
goal of achieving Web interoperability.
Semantic Web: The next-generation
Internet in which semantic meaning will be XML: A standard from the World Wide
associated with all Web resources, both data Web Consortium that provides the tagging of
and services, such that they can be discovered information content within documents. XML
and consumed autonomously by applications. offers a means for representing content in a
format that is both human and machine read-
Semantic Web Service: Using Semantic
able.
Web technology to describe a Web service’s
269
271
Chapter XXXV
Imaging Technologies and their
Applications in Biomedicine
and Bioengineering
Efstratios Poravas
ABSTRACT
New developments are making the technology faster, more powerful, less invasive, and less
expensive. While the technology evolves, new devices are developed, in purpose to be used in
the hospitals. Many new imaging methods are used in biomedical applications today and can
predict the growth of a tumor or detect a disease. The advantages are numerous, but the
problems, during the acquisition and use by the staff, are also remarkable.
INTRODUCTION advantage of the power of computing and tech-

nology so as to manage and analyse data.
We have come from the family doctor’s signa- Imaging technologies save day to day more and
ture black bag in the first half of the 20th century more people.
to the powerful scanning equipment of the X-rays, endoscopes, CT (computed tomog-
modern medical center, from tens of thousands raphy) scans, MRI (magnetic resonance imag-
dying in influenza epidemics to hundreds of ing), digital mammography—these imaging tech-
thousands of seniors receiving their annual flu nologies make it possible for medical scientists
shots, and from an average life expectancy of to peer into the body without cutting through the
about 50 years to our present expectancy of 75 skin. With video monitors and robotic equip-
years. The biomedical community is taking ment, surgery becomes less invasive and less
Imaging Technologies and their Applications in Biomedicine and Bioengineering
traumatic to the body (Sawchuck, 2000). techniques is known as nuclear medicine imag-
Noninvasive means of looking into the human ing.
body are now being used to diagnose a wide Although the mathematical sciences were
variety of diseases, including cancer, used in a general way for image processing,
Alzheimer’s disease, stroke, heart failure, and they were of little importance in biomedical
vascular disease (President’s Committee of work until the development in the 1970s of
Advisors on Science and Technology, 2000). computed tomography for the imaging of X-
The first imaging technologies, the X-ray (dis- rays (leading to the CAT scan) and isotope-
covered by W. K. Roentgen) and EEG (elec- emission tomography (leading to positron-emis-
troencephalogram), were primitive by today’s sion tomography [PET] scans and single-pho-
standards, but both have been considerably ton-emission computed tomography [SPECT]
improved and provided the conceptual base of scans). In the 1980s, MRI eclipsed the other
the other amazing imaging technologies that modalities in many ways as the most informa-
have recently emerged. tive medical imaging methodology (Webb,
The most common, CAT (computer-assisted 1988).
tomography) scans, combine X-rays with com- Table 1 summarises some of the imaging
puter technology to create cross-sectional im- methods used in biomedical applications.
ages of the patient’s body, which are then Technologies such as those in Table 1 are all
assembled into a three-dimensional picture that being investigated in small-animal models. The
displays organs, bones, and tissues in great goal is to marry fundamental advances in mo-
detail. MRI scanners use magnets and radio lecular and cell biology with those in biomedical
waves instead of X-rays to generate images imaging to advance the field of molecular imag-
that provide an even better view of soft tissues, ing (TA-Datenbank-Nachrichten, 2001). The
such as the brain or spinal cord (President’s two basic starting points in evaluating the over-
Committee of Advisors on Science and Tech- all utility of a medical technology are efficacy
nology, 2000). and safety. If a technology is not efficacious, it
Much of today’s imaging technology re- should not be used. In addition, efficacy and
lies on microprocessors and software. In addi- safety data are needed to evaluate the cost
tion, the great advances in noninvasive sensing,
tomography, and imaging technologies now al-
low repeated studies with minimal stress and Table 1. Imaging methods used in biomedical
damage (National Research Council, & Insti- applications
tute for Laboratory Animal Research, 2002).
X-ray projection imaging (discovered in 1895)
Medical imaging is often thought of as a way X-ray CT (1972)
of viewing anatomical structures of the body. MRI (1980)
Magnetic resonance spectroscopy (MRS)
Indeed, X-ray computed tomography and mag- SPECT
netic resonance imaging yield exquisitely de- PET (1974)
Gamma camera (1958)
tailed images of such structures. It is often
Nuclear magnetic resonance (NMR, 1946)
useful, however, to acquire images of physi- Ultrasonics
ologic function rather than of anatomy. Such Electrical source imaging (ESI)
Electrical impedance tomography (EIT)
images can be acquired by imaging the decay of Magnetic source imaging (MSI)
radioisotopes bound to molecules with known Medical optical imaging
Micro computerised axial tomography (MicroCAT)
biological properties. This class of imaging Optical and thermal diagnostic imaging (OCT, DOT)
272
effectiveness of a technology (Banta, Clyde, & purpose of the biomedical imaging techniques is
Williams, 1981). the early detection, clinical diagnosis, and stag-
Biomedical imaging devices have been ing of a disease, and therapeutic applications
used to obtain anatomical images and to provide (Biomedical Imaging Symposium: Visualiz-
localised biochemical and physiological analy- ing the Future of Biology and Medicine,
sis of tissues and organs. The ability of these 1999). Imaging technologies have many appli-
devices to provide anatomical images and physi- cations in biomedicine. Oncology, cardiology,
ological information has provided unparalleled and ophthalmology are only some of its sections
opportunities for biomedical and clinical re- that use these technologies, which everyday
search, and has the potential for important are developed more and more.
improvements in the diagnosis and treatment of
a wide range of diseases (National Institute of
Biomedical Imaging and Bioengineering PROBLEMS AND
[NIBIB], 2002). DISADVANTAGES OF
Technological devices visualise and en- IMAGING TECHNOLOGIES
large somatic space, rendering images of our
most infinitesimal cells, molecules, and genetic Despite all the promises, the use of imaging
structures, which allows for a more precise technologies in biomedicine and bioengineering
manipulation of our muscles, tissues, and bones evoke many problems. All biomedical imaging
(Sawchuck, 2000). Imaging tests now provide devices suffer from various limitations that can
much clearer and more detailed pictures of restrict their general applicability. Some major
organs and tissues. New imaging technology limitations are sensitivity, spatial resolution, tem-
allows us to do more than simply view anatomi- poral resolution, and the ease of the interpreta-
cal structures such as bones, organs, and tion of data. One way to circumvent these
tumours. Functional imaging—the visualisation limitations is to develop technological and meth-
of physiological, cellular, or molecular pro- odological approaches that improve and extend
cesses in living tissues—enables us to observe the sensitivity and the information content of
activity such as blood flow, oxygen consump- individual imaging techniques. Another way is
tion, or glucose metabolism in real time. to combine two or more complementary bio-
Imaging technology already has had life- medical imaging techniques (like MRI and PET,
saving effects on our ability to detect cancer MRI and MEG, and optical MRI).
early and more accurately diagnose the disease Table 2 summarises some problems of the
(especially the PET device). Generally, the imaging technologies.
Table 2. Problems and disadvantages of imaging technologies
The high cost of equipment and their maintenance, which aggravates the national
economy for medicine
Wasteful expenditures because of bad usage by users and technical staff (20 to 40%)
The technology changes rapidly and devices may become out of date
Users need education to learn how to break the new technologies in
Physicians and the nursing staff must continuously be acquainted through articles
related to the new technologies and equipment
New technologies cause disruption and disappointment for staff
They are venturous for patients because the levels of radiation they are exposed to
may be too high
273
There is no crystal ball to predict the future toward the development of multifusion optical
of medical imaging technologies. New applica- sensing imaging systems and techniques for
tions continue to be explored for both diagnosis efficient disease detection (Giakos, 2003).
and treatment (Canadian Institute for Health Today, as for all products and services in
Information, http://www.cihi.ca). Biomedical all sectors, there exists the DICOM (Digital
imaging has seen truly exciting advances in Imaging and Communications in Medicine) Stan-
recent years. New imaging methods can now dards Committee. Its purpose is to create and
reflect internal anatomy and dynamic body maintain international standards that help the
functions heretofore only derived from text- allocation of medical pictures (like radial to-
book pictures, and applications to a wide range mographies, magnetic tomographies, etc.), and
of diagnostic and therapeutic procedures can the communication of biomedical diagnostic
be envisioned. Not only can technological ad- and therapeutic information in disciplines that
vances create new and better ways to extract use digital images and associated data (DICOM,
information about our bodies, but they also 2004). DICOM is used or will be used by every
offer the promise of making some existing medical profession that utilises images within
imaging tools more convenient and economical. the healthcare industry.
Advances based on medical research prom-
ise new and more effective treatments for a
wide variety of diseases. New noninvasive CONCLUSION
imaging techniques for the earlier detection and
diagnosis of disease are essential to take full The rapid progress in imaging technologies
advantage of new treatments and to promote during the last decades has stimulated many
improvements in healthcare. The development developments and applications in medicine, bi-
of advanced genetic and molecular imaging ology, industry, aerospace, remote sensing,
techniques is necessary to continue the rapid meteorology, oceanography, and the environ-
pace of discovery in molecular biology. Several ment.
breakthrough imaging technologies, including New developments are continually mak-
MRI and CT, have been developed primarily ing the technology faster, more powerful, less
abroad (American Institute for Medical and invasive, and less expensive. Imaging technol-
Biological Engineering, http://www.aimbe.org). ogy was primarily used in medical diagnosis
Key paradigms of emerging imaging tech- initially, but it is being increasingly used in pure
nologies from different technological areas will neuroscience, psychological research, and many
be presented, and the engineering principles other fields. The quantitative nature of data will
and research findings leading to the design of be relevant for the effective diagnosis as well
efficient bioimaging technologies will be intro- as therapeutic management of patients, which-
duced and analysed. Specifically, imaging tech- ever disease they have (“Nuclear Medicine
nologies from space or aerospace research Sextet,” 1999).
have been identified and successfully applied
toward the development of novel high-resolu-
tion, multisensor medical imaging systems, with REFERENCES
potential applications in digital radiography and
CT. Similarly, experimental research findings Banta, H. D, Clyde, J. B., & Williams, J. S.
for defence applications have been applied (1981). Toward rational technology in medi-
274
cine: Considerations for health policy. New Science and Technology Policy. Retrieved from
York: Springer Verlag. http://www.ostp.gov
Biomedical Imaging Symposium: Visualiz- Sawchuck, K. (2000). Digibodies online ex-
ing the Future of Biology and Medicine. hibition and Synapse online forum. Retrieved
(1999, June 25-26). Natcher Conference Cen- from http://www.digibodies.org/synapse/
ter, National Institutes of Health, Bethesda, intro.html
MD. Sponsored by the NIH Bioengineering
TA-Datenbank-Nachrichten. (2001). Nr. 1/
Consortium (BECON), the American Institute
10. Jahrgang: März, S. 13-22. Retrieved
for Medical and Biological Engineering
from http://www.itas.fzk.de
(AIMBE), and the Radiological Society of North
America (RSNA). Retrieved from http:// Webb, S. (1988). The physics of medical imag-
www.becon.nih.gov/report_19990625.pdf ing. Magnetic Resonance Imaging and Bio-
physics and Medical Imaging. Retrieved from
Digital Imaging and Communications in Medi-
http://newton.ex.ac.uk/teaching/modules/
cine (DICOM). (2004). Scope of DICOM:
PHYM433.html
Strategic document, Version 4.0. VA: Na-
tional Electrical Manufacturers Association.
KEY TERMS
Giakos, G. (2003). Emerging imaging tech-
nologies: Technology identification trans-
Assessment of Imaging Technology:
fer and utilization for bioengineering appli-
Research on and development of methods for
cations. McMaster University. Retrieved Oc-
the evaluation and comparison of new and
tober 28, 2003, from http://www.ece.mcmaster.
existing imaging technologies to establish their
ca/news/seminars/giakos_sem.htm
effectiveness, robustness, and range of appli-
National Institute of Biomedical Imaging and cability.
Bioengineering (NIBIB). (2002). Improvements
Bioengineering: The application of engi-
in imaging methods and technologies. Re-
neering principles to the fields of biology and
trieved from http://www.nibib1.nih.gov
medicine, as in the development of aids or
National Research Council (U.S.), & Institute replacements for defective or missing body
for Laboratory Animal Research (U.S.) (2002, organs. It is also called biomedical engineering.
April 17-19). International perspectives: The
Biomedicine: A branch of medical sci-
future of non-human primate resources. Pro-
ence concerned especially with the capacity of
ceedings of the Workshop. Retrieved from
human beings to survive and function in abnor-
http://www.nap.edu/catalog/10774.html
mally stressful environments and with the pro-
Nuclear medicine sextet. (1999, August 21). tective modification of such environments.
The Lancet, 665. Broadly, it is medicine based on the application
of the principles of the natural sciences, espe-
President’s Committee of Advisors on Science
cially biology and biochemistry.
and Technology (2000, Spring). Chapter 3.
Biomedical technologies. In Wellspring of Development of Imaging Devices: Re-
prosperity: Science and technology in the search and development of generic biomedical
U.S. economy. Washington, DC: Office of imaging technologies before specific applica-
tions are demonstrated.
275
Diagnostic Imaging: A study section re- vention, monitoring, treatment, alleviation, or

views applications dealing with the develop- investigation of a disease, injury, or handicap.
ment and evaluation of new technology for
Medical Imaging: Term describing the
imaging, including instrumentation and soft-
various technologies that produce pictures or
ware for producing, evaluating, storing, and
images of the body and its structures. Imaging
transmitting images for anatomical, physiologi-
technologies include X-ray, CT scanning, PET
cal, metabolic, diagnostic, and therapeutic in-
scanning, and ultrasound. This term also in-
formation.
cludes technology such as digital cameras, which
Image Exploitation: Development, design, produce digital images.
and implementation of algorithms for image
Medical Imaging Technologies: A study
processing and information analysis, including
section reviews all modalities of medical imag-
advanced methods for the acquisition, storage,
ing, including gamma ray; MRI; functional MRI;
and display of images; research and develop-
PET; SPECT; X-ray; CT; visible, infrared, and
ment on image-guided procedures; and tech-
ultraviolet photons; and optical, photo-acoustic,
niques for using multidimensional images to
microwave-acoustic, and exotic imaging meth-
understand physiology and normal and abnor-
ods.
mal function.
Minimally Invasive Technologies: Ba-
Medical Device: Any instrument, appara-
sic research involving the use of robotics tech-
tus, appliance, material, or other article, whether
nologies for actuation, sensing, control, pro-
used alone or in combination, including the
gramming, and the human-machine interface,
software necessary for its proper application,
and the design of mechanisms to determine
intended for the purpose of the diagnosis, pre-
research end points such as diagnosis and the
automated or remote treatment of disease.
276
277
Chapter XXXVI
Medical Image Compression
Using Integer Wavelet
Transforms
B. Ramakrishnan
M.I.T. Manipal, India
N. Sriraam
ABSTRACT
In this chapter, we have focused on compression of medical images using integer wavelet
transforms. Lifting transforms such as S, TS, S+P(B), S+P(C), 5/3, 2+@, 2, 9/7-M and 9/7-F
transforms are used to evaluate the performances of lossless and lossy compression. Four
medical images, namely, MRI, CT, ultrasound, and angiograms are used as test data sets. It
is found from the experiments that, among the different transforms, the 9/7-M wavelet
transform is identified as the optimal method for lossless and lossy compression of medical
images.
INTRODUCTION pose a problem during transmission. Hence,

there is a need to achieve high compression and
Compression of medical images is an area of at the same time preserve the image quality. In
discussion among the medical community due telemedicine applications, the compression of
to the fact that compressing an image could medical images plays a paramount role in re-
lead to vital diagnostic information being lost. ducing the image file size, thereby reducing the
On the other hand, images obtained from imag- bandwidth for transmission over a network.
ing modalities such as computed tomography There has been a tremendous increase in
(CT), magnetic resonance imaging (MRI), ul- the use of wavelets as image-compression tools
trasound (US), and other modalities require due to their ability to achieve high compression
large amounts of space for storage and also while sustaining image quality (Antonini,
Medical Image Compression Using Integer Wavelet Transforms
Barlaud, Mathieu, & Daubechies, 1992; The image is first transformed from a spatial
Averbuch, Lazar, & Israeli, 1996). One of the domain to a wavelet domain using two-dimen-
advantages of wavelet-based compression is sional lifting wavelet transform (2D-LWT).
that it supports progressive lossy to lossless The resulting wavelet coefficients are entropy
reconstruction (Adams & Kossentini, 2000; coded to obtain a compressed image. In this
Said & Pearlman, 1996a, 1996b; Sheng, Bilgin, chapter, SPIHT is used for entropy coding as it
Sementilli, & Marcellin, 1998). The Joint Pho- is fast and easy to implement and provides
tographic Experts Group (JPEG2000) com- superior compression among the wavelet-based
pression standard is based on wavelets. Wave- coders. To reconstruct the image, the process
let-based entropy coders have been developed is reversed. The compressed image is first
to enhance compression that exploits the spatial entropy decoded and then two-dimensional in-
similarities among the wavelet coefficients. verse lifting wavelet transform (2D-ILWT) is
These coders, called zero-tree coders, include applied to obtain the original image.
embedded zero-tree wavelet (EZW; Shapiro, The chapter is arranged as follows. In the
1993), set partitioning in hierarchical trees next section, a brief description of the lifting
(SPIHT; Said & Pearlman, 1996a, 1996b), an wavelet transform will be presented, followed
advancement of EZW, and embedded block by the description of the SPIHT coder. A
coding with optimized truncation (EBCOT; number of integer wavelets for various medical
Taubman, 2000). With the development of inte- images are then analyzed based on their lossless
ger wavelets, lossless compression could be and progressive lossy compression performance.
realized (Calderbank, Daubechies, Sweldens,
& Yeo, 1997, 1998). Integer wavelets generate
integer coefficients, whereas the conventional LIFTING WAVELET TRANSFORM
wavelets generate floating-point coefficients.
Lossless compression could not be achieved The lifting approach of constructing wavelets
with the conventional wavelets due to having to was proposed by Sweldens (1998). It allows
round off these floating-point values. Integer fast, efficient, and in-place calculation of the
wavelets are possible with the construction of wavelet transform. Besides this, it is feasible to
wavelets based on the lifting scheme (Sweldens, construct integer wavelets based on this
1998). The lifting scheme provides fast, effi- scheme.
cient, and in-place calculation of the wavelet Consider an image I of size N x N with N
transform. being an integer power n of 2. and each row and
The main objective of this work is to evalu- column consisting of 2n pixels. The two-dimen-
ate different integer wavelets on the basis of sional wavelet transform is performed by first
their lossy and lossless compression perfor- applying one-dimensional wavelet transform
mance for various medical images. The core along the columns and then applying one-di-
idea behind compression is that in an image, mensional wavelet transform along the rows or
there exists some correlation among the neigh- vice-versa. Therefore, we can treat the image
borhood pixels. The task is to decorrelate the as N x N one-dimensional vectors with each
image data so as to eliminate the redundancy vector consisting of 2n pixels. Hereafter, we
and reduce the entropy of the image. The refer to the one-dimensional vector as a signal
general architecture for a lifting-based wavelet xn. Applying wavelet transform to xn divides it
compression scheme is depicted in Figure 1. into coarse sn–1 values and detail dn–1 values,
278
Figure 1. Compression scheme
Original SPIHT
2D-LWT Encoder
Medical
Image
Compressed
Image
Original
Medical SPIHT
2D-ILWT Decoder
Image
Figure 2. Different levels of dyadic (pyramid) decomposition
LL2 HL2
LL1 LH1 LH2 HH2 LH1

LL HL HL HL
LH1 HH1 LH1 HH1
LH HH LH HH LH HH
both consisting of 2n–1 pixels. This step is per- • Split: In this operation, the signal xn is split
formed for each column and the result is used according to even x2j and odd x2j+1 ele-
as input for row transformation. After one ments (0 ≤ j ≤ 2n–1). The even elements
sequence, the image will be divided into four represent the coarse sn values while the
bands, LL, LH, HL, and HH, with each column odd elements represent the detail dn val-
and row in a band consisting of 2n–1 pixels. The ues. This step is also referred to as lazy
wavelet transformation is then performed on wavelet transform.
the LL sub-band, resulting in the division of the • Predict: In the prediction or dual lifting
band into four sub-bands. This procedure can step, the odd elements are predicted from
be repeated until the LL sub-bands consist of the even elements. Then the detail dn–1 is
only one pixel. This decomposition is referred the difference between its prediction and
to as dyadic or pyramid decomposition (Figure the odd element. The odd element is re-
2) and provides a multiresolution representation placed by its detail. The procedure is
of the image. described by the equation
The lifting approach of wavelet transform
consists of the following operations: dn–1 = dn – P(sn) (1)
where P is the prediction operator.
279
Figure 3. Forward lifting wavelet transform
sn–1
xn Split Predict Update
dn–1
Figure 4. Inverse lifting wavelet transform

sn-1
Update Predict Merge xn
dn-1
• Update: In the update or primal lifting dn = dn–1 + P(sn). (4)

step, the even elements are replaced by an
average. The coarse sn–1 is calculated as • Merge: With both even and odd elements
follows: recovered, the original signal xn is ob-
tained by combining the even and odd
sn–1 = sn + U(dn–1) (2) elements.
where U is the update operator. The preceeding process is depicted in Fig-

ure 4.
The preceeding process is illustrated in Figure 3. Figures 3 and 4 represent a one-pair predict
The inverse lifting wavelet transform is per- and update process. This process is repeated
formed by the following operations: for L pairs until the even elements become the
coarse values (low-pass coefficients) and the
• Undo Update: Given sn–1 and dn–1, the odd elements become the detail values (high-
even elements are recovered as described pass coefficients). The predictor and update
in the equation operators can be thought of as filters with their
values constructed depending on the choice of
sn = sn–1 – U(dn–1). (3) the wavelet. In general:
• Undo Predict: Once the even elements di , j = di −1, j − ∑ pi ,k si −1, j −k (5)

k
are recovered, the odd elements are found
as follows: and
280
si , j = si −1, j − ∑ ui ,k di , j −k , (6)
and
k
 
where the following applies: si , j = si −1, j −  ∑ ui ,k di , j −k + 1/ 2  . (8)
 k 
• k: filter coefficient
• i: level of decomposition or reconstruc- The wavelet transforms are either repre-
tion, 1 ≤ i ≤ n sented by their vanishing moments (W, ), where
• j: element in the vector, 1 ≤ j ≤ 2n W is the number of vanishing moments of the
analyzing high-pass filter and is the number
The integer version of the lifting process of vanishing moments of the synthesizing low-
can be built by truncating the filter output to the pass filter, or by their filter lengths f/r, where f
nearest integer in both the predict and update represents the low-pass filter length and r
operations. Equations 5 and 6 then becomes represents the high-pass filter length. For ex-
ample, (2, 2) transform is also referred to as 5/
3 transform. Table 1 lists the wavelet trans-
 
di , j = di −1, j −  ∑ pi ,k si −1, j −k + 1/ 2 (7) forms used for evaluation along with their for-
 k 
ward lifting equations (Calderbank et al., 1998).
Table 1. Transforms and their forward lifting equations
Transform Forward Lifting Equation

d j = x2 j +1 − x2 j
(1,1) or S
s j = x2 j + d j / 2
d1, j = x2 j +1 − x2 j
(3,1)or TS s j = x2 j +  d1, j / 2 
d j = d1, j + 1/ 4(s j −1 − s j +1 ) + 1/ 2
d1, j = x2 j +1 − x2 j
(2,1) or S+P(B) s j = x 2 j +  d 1, j / 2 
d j = d1, j + 1/ 8(2s j −1 + s j − 3s j +1 + 2d j +1 ) + 1/ 2
d1, j = x2 j +1 − x2 j
(2,1) or S+P(C) s j = x2 j +  d1, j / 2 
d j = d1, j + 1/16(− s j − 2 + 5s j −1 − 4s j − 8s j +1 + 6d j +1 ) + 1/ 2
d j = x2 j +1 − 1/ 2( x2 j + x2 j + 2 ) +1/ 2
(2,2) or 5/3
s j = x2 j + 1/ 4(d j −1 + d j ) + 1/ 2
d1, j = x2 j +1 − 1/ 2( x2 j +2 + x2 j ) 
(4,2) or 2+2,2 s j = x2 j + 1/ 4(d1, j −1 + d1, j ) + 1/ 2 
d j = d1, j − 1/16( − s j −1 + s j + s j +1 − s j + 2 ) + 1/ 2 
d j = x2 j +1 − 9/16(x2 j + x2 j +2 ) − 1/16( x2 j − 2 + x2 j + 4 ) + 1/ 2
(4,2) or 9/7-M
s j = x2 j + 1/ 4(d j −1 + d j ) + 1/ 2
d1, j = x2 j +1 +  203 /128(− x2 j +2 − x2 j ) + 1/ 2 
s1, j = x2 j +  217 / 4096( −d1, j −1 − d1, j ) + 1/ 2 
(4,4) or 9/7-F
d j = d 1, j + 113 /128( s1, j + s1, j +1 ) + 1 / 2 
s j = s1, j + 1817 / 4096( d j −1 + d j ) + 1/ 2 
281
SET PARTITIONING IN also probably be significant in magnitude. The

HIERARCHICAL TREES SPIHT algorithm takes advantage of the spatial
similarity present in the wavelet space to opti-
The final step in compression is the entropy mally find the location of the wavelet coeffi-
coding of the decorrelated image data. The cients that are significant by means of a binary
entropy coder should take advantage of this search algorithm.
decorrelation. SPIHT is a progressive trans- The SPIHT algorithm sends the top coeffi-
mission coder and produces embedded bit cients in the pyramid structure using a progres-
streams. It works on the principle that there sive transmission scheme. This scheme is a
exists a spatial relationship among the wavelet method that allows the obtaining of a high-
coefficients at different levels and frequency quality version of the original image from the
sub-bands in the pyramid structure. A wavelet minimal amount of transmitted data. The pyra-
coefficient at location (i,j) in the pyramid rep- mid wavelet coefficients are ordered by magni-
resentation has four direct descendants (off- tude and then the most significant bits are
spring) at locations: transmitted first, followed by the next bit plane
and so on until the lowest bit plane is reached.
O(i,j) = This reduces the mean square error (MSE) for
{(2i,2j), (2i,2j+1), (2i+1,2j), (2i+1,2j+1)}, every bit plane sent.
(9) To take advantage of the spatial relationship
among the coefficients at different levels and
and each of them recursively maintains a spa- frequency bands, the SPIHT coder algorithm
tial similarity to its corresponding four off- partitions all the coefficients Ci,j according to a
spring. This pyramid structure is commonly number of sets Tk and performs the significance
known as the spatial orientation tree. For ex- test:
ample, Figure 5 shows the similarity among
sub-bands within levels in the wavelet space. If
max Ci , j ≥ 2 n (10)
a given coefficient at location (i,j) is significant i , j∈Tk
in magnitude, then some of its descendants will

on each set Tk, with n being the bit plane. If the
result of the significance test is yes, then, using
the same rule, Tk is partitioned into subsets, and
Figure 5. Offspring dependencies in the
the same significance test is performed on all
pyramid structure
the subsets. This partitioning is continued until
all the significance tests are reduced to size 1.
The significance test performed on a set T can
be summarized by:
1, max Ci , j ≥ 2 n

S n (T ) =  i , j∈Tk
. (11)
0, otherwise
Wavelets coefficients that are not signifi-

cant at the nth bit-plane level may be significant
282
Figure 6. Test images of (a) MRI brain, (b) CT abdomen, (c) ultrasound, and (d) angiogram
(a) (b) (c) (d)
at the (n-1)th bit plane or lower. This informa- Bit Rate =

no. of bits in the original image × resolution of the image
no. of bits in the compressed image
tion is arranged, according to its significance, in
three separate lists: the list of insignificant sets (12)
(LIS), the list of insignificant pixels (LIP), and
the list of significant pixels (LSP). In the de- We can infer from Table 2 that no single
coder, the SPIHT algorithm replicates the same transform performs best for all medical images.
number of lists. In the case of the MRI image, the S+P(C)
transform yields the best compression due to
the fact that MRI images have high contrast but
EXPERIMENTAL RESULTS give poor results for other images. For CT and
angiogram images with low contrast, the 9/7-M
The analysis was performed on MRI brain, CT transform performs the best. For the ultrasound
abdomen, ultrasound, and angiogram images image, the 2+2,2 transform works better than
(Figure 6). All images are of size 512x512 with other transforms. The 9/7-M transform per-
8-bit resolution. forms consistently well for all images, and the
9/7-F, S, and TS transforms perform the worst.
Lossless Compression
Lossy Compression
The test images were evaluated for lossless
compression using the transforms listed in Table The test images were evaluated for progres-
1. The performance results are defined in terms sive lossy compression by varying the bit rates.
of their bit rates (Equation 12) and tabulated in The results are tabulated in Table 3. Unlike in
Table 2. the conventional transform where the magni-
Table 2. Lossless compression results
Bit rate (bpp)

Image
S TS S+P(B) S+P(C) 5/3 2+2,2 9/7-M 9/7-F
br 3.931 3.671 3.539 3.443 3.569 3.474 3.476 3.546
abd 1.799 1.658 1.589 1.591 1.583 1.588 1.568 1.733
us 3.631 3.178 3.111 3.129 3.064 3.019 3.040 3.315
angio 4.473 4.364 4.292 4.315 4.241 4.210 4.189 4.296
283
Table 3. Lossy compression results

MRI
Bit Rate PSNR (dB)
(bpp) S TS S+P(B) S+P(C) 5/3 2+2,2 9/7-M 9/7-F
0.1 29.45 30.74 31.15 30.94 31.81 31.74 31.21 31.81
0.5 34.19 35.09 34.80 34.26 36.26 36.26 35.96 36.43
1.0 36.61 37.58 37.17 36.76 38.98 39.06 38.73 38.87
2.0 39.49 40.41 40.19 39.95 42.70 42.65 42.26 41.55
CT
Bit Rate PSNR (dB)
(bpp) S TS S+P(B) S+P(C) 5/3 2+2,2 9/7-M 9/7-F
0.1 32.79 35.68 35.56 34.52 36.34 36.09 36.97 36.89
0.5 38.23 41.94 40.97 39.54 45.05 44.53 43.81 43.68
0.7 41.02 44.32 43.81 42.82 47.38 46.88 46.28 45.25
1.0 44.53 47.29 46.72 45.97 50.04 49.59 49.14 46.11
US
Bit Rate PSNR (dB)
(bpp) S TS S+P(B) S+P(C) 5/3 2+2,2 9/7-M 9/7-F
0.1 29.76 30.58 30.52 29.74 31.22 30.93 30.46 31.25
0.5 34.25 35.06 36.59 36.58 37.86 37.70 37.13 37.84
1.0 37.74 38.27 38.58 38.25 41.07 41.38 41.41 40.11
2.0 40.19 41.51 41.88 41.32 46.87 46.47 46.01 43.17
ANGIO
Bit Rate PSNR (dB)
(bpp) S TS S+P(B) S+P(C) 5/3 2+2,2 9/7-M 9/7-F
0.1 32.92 30.25 32.62 32.21 33.16 32.11 32.65 33.26
0.5 34.24 32.33 34.75 34.08 35.08 34.61 34.71 35.10
1.0 36.61 35.92 37.14 36.79 40.78 38.40 38.98 38.47
2.0 38.68 38.06 39.41 38.91 44.11 43.51 43.75 40.04
tudes of coefficients are unitary, the lifting where max(I) is the maximum pixel value of the
approach produces coefficients that are not original image I. The MSE is given by:
unitary. SPIHT works on unitary transforma-
tion, wherein the larger magnitudes at the higher
MSE =
∑[I − I ] ' 2
, (14)
sub-bands are transmitted first, thereby reduc- N2
ing MSE. The lifting wavelet transform is thus
adapted for SPIHT by scaling the wavelet where I' is the reconstructed image and N 2 is
coefficients in each sub-band of each level the total number of pixels in the image.
accordingly. The scaling is done intrinsically in From Table 3, the results indicate that at
the SPIHT coder, as reported by Said and lower compression ratios, the 5/3 transform
Pearlman (1996a). performs better than other transforms, whereas
The lossy compression performance is evalu- at higher compression ratios, the 9/7-F trans-
ated in terms of the peak signal-to-noise ratio form gives the best result. The 9/7-F, 9/7-M,
(PSNR) described by Equation 13. 2+2,2, and 5/3 transforms have the best perfor-
mance, whereas the S+P(B), S+P(C), TS, and
[max(I )]2 S transforms have the worst performance.
PSNR(dB) = 20 log10 , (13)
MSE
284
CONCLUSION and lossy compression. IEEE Transactions on

Image Processing, 5(9), 1303-1310.
A number of integer wavelets were evaluated
Said, A., & Pearlman, W. A. (1996b). A new
for various medical images based on their
fast and efficient image codec based on set
lossless and progressive lossy performance.
partitioning in hierarchical trees. IEEE Trans-
No one transform has superior lossless or lossy
actions on Circuits and Systems for Video
compression performance. In general, the 9/7-
Technology, 6(3), 243-249.
M transform gives consistent results for both
lossless and lossy compression and could there- Shapiro, J. M. (1993). Embedded image coding
fore be accepted as the optimum wavelet for using zerotrees of wavelet coefficients. IEEE
medical image compression. Transactions on Signal Processing, 41(12),
3445-3462.
Sheng, F., Bilgin, A., Sementilli, P. J., &
REFERENCES
Marcellin, M. W. (1998). Lossy and lossless
image compression using reversible integer
Adams, M. D., & Kossentini, F. (2000). Re-
wavelet transforms. Proceedings of IEEE
versible integer-to-integer wavelet transforms
International Conference on Image Pro-
for image compression: Performance evalua-
cessing (Vol. 3, pp. 876-880).
tion and analysis. IEEE Transactions on Im-
age Processing, 9(6), 1010-1024. Sweldens, W. (1998). The lifting scheme: A
construction of second generation wavelets.
Antonini, M., Barlaud, M., Mathieu, P., &
SIAM Journal on Mathematical Analysis,
Daubechies, I. (1992). Image coding using
29(2), 511-546.
wavelet transform. IEEE Transactions on
Image Processing, 1(2), 205-220. Taubman, D. (2000). High performance scal-
able image compression with EBCOT. IEEE
Averbuch, A., Lazar, D., & Israeli, M. (1996).
Transactions on Image Processing, 9(7),
Image compression using wavelet transform
1151-1170.
and multiresolution decomposition. IEEE Trans-
actions on Image Processing, 5(1), 4-15.
KEY TERMS
Calderbank, A. R., Daubechies, I., Sweldens,
W., & Yeo, B.-L. (1997). Lossless image com-
Bandwidth: The amount of data that can be
pression using integer to integer wavelet trans-
transferred in a given time period.
forms. Proceedings of IEEE International
Conference on Image Processing (Vol. 1, pp. Decoder: Algorithm that does the reverse
596-599). of an encoder, undoing the encoding so that the
original information can be retrieved. The same
Calderbank, A. R., Daubechies, I., Sweldens,
method used to encode is usually just reversed
W., & Yeo, B.-L. (1998). Wavelet transforms
in order to decode.
that map integers to integers. Applied and
Computational Harmonic Analysis, 5(3), 332- Encoder: Algorithm used to encode a bit
369. stream into a form that is acceptable for trans-
mission or storage.
Said, A., & Pearlman, W. A. (1996a). An
image multiresolution representation for lossless
285
Entropy: Minimum channel capacity re- should be communicated over a channel such
quired to reliably transmit the source as en- that the source (input signal) can be recon-
coded binary digits. structed at the receiver with given distortion.
Multiresolution Analysis: The study of Source Coding: Process of encoding in-
signals at various resolutions and scales. A formation using fewer bits.
signal, when wavelet transformed, provides
Telemedicine: The use of modern tele-
this representation.
communication and information technologies
Progressive Transmission: Type of transfor the provision of clinical care to individuals
mission in which the image is displayed from a located at a distance and for the transmission of
low-quality scale to a high-quality scale. information to provide that care.
Rate-Distortion Theory: Branch of in- Wavelet Transform: Transformation to
formation theory addressing the problem of basis functions that are localized in scale and in
determining the minimal amount of entropy that time as well.
286
287
Chapter XXXVII
Three Dimensional
Medical Images
Efstratios Poravas
Dimitra Petroudi
ABSTRACT
The revolution of technology has lead to a change; from the analogic to the digital function
of medical devices. Some of them were produced in the last years to improve the quality of
images. Although the procedure of acquiring and using the devices has been very complicated,
the analysis of the images is so dependable that a big amount of the annual budget is spent
for their acquisition.
INTRODUCTION try, physics, microbiology, physiology, pharma-

cology, and so forth enforced medical research,
The rapid development of science and the which resulted in the continuous discoveries in
continuous manufacture of pioneer medical medical technology. In 1895, W. K. Roentgen
technological products, together with the mod- discovered X-rays, which was a turning point
ern requirements for high-quality medical ser- for medical imaging and diagnostics in general.
vices, led to the growth and introduction of In the 1950s, there was the development of
modern technologies in the health sector. This computerized systems, while in the 1960s, there
development has been really impressive and were applications such as the transport of
rapid. biological signals from equipped space missions
At the beginning of the 20th century, the and teletransfers of information.
progress of applied sciences, such as chemis-
Three Dimensional Medical Images
Figure 1. Three dimensional imaging for only 15 years. Its development is revolution-
ary in the health field and constitutes a major
contribution for the promotion of health.
The vast amount of optical information and
the need for elaborating them led scientists and
technicians to the discovery of storage means
for the images and their elaboration by using
computers combined with the development and
improvement of biomedical equipment.
Digital elaboration, as the title itself de-
clares, deals with the digital registration of
images and their elaboration by computers. The
objective of elaboration can be the quality im-
To be more specific, the dynamic entry and provement of images, the straining of recording
the enforced intervention of sciences such as or transmission noises, the compression of the
informatics gave an enormous impulse to the amount of information, the storage of images,
medical field and created new data for treat- and their digital transmission and depiction.
ment and diagnosis. Thus, we now have the In the picture below, we schematically rep-
interaction and participation of many different resent all the necessary equipment that is needed
scientific sectors, which aim at the best medical to fulfill a digital registration of images to be
care. One of these sectors is the imaging and elaborated by the appropriate personnel in or-
treatment of medical pictures, whose develop- der to give all the necessary information and
ment and use are a crucial point in modern conclusions.
therapeutics. Thus, everyone can understand that for the
Medical imaging is related to issues such as elaboration of three-dimensional medical im-
the descriptive principles of medical images ages, a device for the admission of images is
and their elaboration, together with whatever required, and this can be an axial tomographer,
they include. Therefore, it is easy for someone a magnetic tomographer, an ultrasound device,
to understand the importance and the role of or even more developed depiction systems. A
imaging in diagnosis, treatment, and recovery in computer is also required, with all the neces-
general. sary equipment in order to elaborate the im-
This chapter will discuss the basic concepts, ages, analyse them, and store and transmit
such as the analysis of medical images, their them, and finally all the appropriate exit devices
elaboration, various uses and applications, as that will project the images will be needed,
well as an analysis of the functional require- which can be terminal stations, special films,
ments of such applications in order for them to and printers. Finally, it is very important that the
be fulfilled. whole system will support network communi-
cations so that all information can be available
to the scientific community fast and safely with
DIGITAL ELABORATION both quality and reliability. Before we introduce
OF IMAGES some of the depiction devices of three-dimen-
sional medical images that are widely used by
The digital elaboration of images is actually a the scientific community nowadays, it is impor-
new sector of informatics and has been applied tant to study some necessary notions for the
288
comprehension of the way this elaboration of Figure 2. Image elaborator

three-dimensional images occurs.
An image is a two-dimensional signal. Thus,
for the analysis and elaboration of this signal, all
the techniques and mathematic relations of
digital signal elaboration can be used. Specifi-
cally, in the health field, we introduce the notion
of modality, which is a biomedical signal that
represents one view or one function of the
organization. Therefore, the depiction devices
receive modalities and with the appropriate
processes turn them into images. The dimen-
sion of a signal depends on the number of evolution of medical depiction devices and
independent variables it has. Consequently, the machines, something that proved to be benefi-
two-dimensional signals have as independent cial for health services. Today, there is a large
variables the two dimensions of the surface. number of medical apparatuses that fills any
Therefore, the elements that should be in- kind of diagnostic demands. More specifically,
cluded in a digital system for the elaboration of in the three-dimensional depiction area, there
images are the following: are now many options that suit every demand.
The categories of medical images are as fol-
• Image Elaborator: This is the hardcore lows.
of the system that receives the image,
temporarily stores and elaborates it on a • Digital abstraction angiography
low level, and finally demonstrates it on a • Ultrasound system
primary level. • Computed tomography (CT)
• Digitizer: It arithmetically represents the • Magnetic resonance imaging (MRI)
image in order for it to be received by a • Gamma camera
computer. • Computed tomographies of nuclear medi-
• Digital Computer: It performs an exten- cine (single-photon-emission computed
sive elaboration. tomography [SPECT] and positron-emis-
• Storage Devices sion tomography [PET])
• Screens and Recording Devices
Digital Abstraction Angiography
We will now proceed to analyse the above
elements in order to realize their participation in It is a medical diagnostic examination that
the elaboration of images, and at the same time depicts the condition of the vessels. It is ex-
we will present all the processes that occur at tremely useful in the sector of cardiology and
each level. helps to with the prevention of many diseases of
the cardiac system. Its characteristic is that it
Image Elaborator produces a vast number of data, especially as
we move from one-dimensional systems to
The evolution of technology, as mentioned above, two-dimensional ones and from a still image to
was the vaulting bar for the development and a moving one (video).
289
Figure 3. Digital angiography Figure 4. MRI depiction
more, the patient does not have to be exposed

to harmful radiation. It is based on the applica-
tion system of a magnetic field in order to obtain
magnetic tomographies.
Ultrasound System By applying a magnetic field, the cells orient
to the rotation frequency and thus we have the
It depicts the resonance of high-frequency production of resonance through the pulse of
sound waves, which depend on the auditory radiofrequency. Afterward, we have a rest so
properties of tissues that are produced by dif- that the cells can reorient to their initial posi-
ferent organs and are examined as brightness in tions. The produced signal depends on the
the image. One application of the ultrasound density and type of cells and, thus, magnetic
system is the ultrasound of the heart and ves- tomographies show a high contrast on the soft
sels (Doppler), which is developed in real time tissues.
and depicts the flux speed of the blood.
Gamma Camera
Computed Tomography
It is a diagnostic method that belongs to the field
It uses X-rays for the creation of images, which of nuclear medicine. It is based on the calcula-
means that the patient is exposed to radiation tion of the position and concentration of a
during the examination. The part of the body radioactive isotope that is provided to the pa-
that needs to be examined receives X-rays tient before the examination. It gathers clinical
through different transmission angles. With the information about the normal function while its
use of mathematic transformations and calcu- discernibility is very low.
lations on the counted prices from these differ-
ent angles, we receive images that are cross Computed Tomographies of
and plane sections. Each ray penetrating the Nuclear Medicine (PET and SPECT)
body is recording densities of tissues.
This section is about diagnostic examinations
Magnetic Tomography that are related to the physiology of the organi-
zation and are the most modern in technological
It is a nonpenetration diagnostic method that terms in the field of diagnostics. SPECT in-
produces a series of images that represent volves the computed tomography of photon
biological differences among tissues. Further- transmission, something that shows that the
290
Figure 5. Depiction in SPECT Compression is based on redundant infor-

mation that is included in the images. The more
redundancy there is, the more is the achieved
compression. The whole process is called source
coding, and it is based on a system that includes
the following:
• Transformer: It transforms the initial

image into a more appropriate one for
compression.
• Quanter: It quants the transformed im-
age either with graduation or with vector.
signal-production process is based on photon • Coder: A sequence of bits corresponds
transmission. PET involves tomography with with each quant level. The code can be of
positron transmission. a specific or variable length.
The diagnostic devices mentioned are the
means of acquiring medical images in analo- Compression with no losses does not use a
gous or digital forms. The device used may quanter system because a quant always shows
have the ability to store digitally; otherwise, the a loss of information.
digitalization occurs with the use of a digitizer The used patterns of image compression
before storage. Next we have the process of are as follows:
compressing images.
• G3, G4, GBIG—binary images
Digital Image Compression • JPEG—firm images
• H.261 MPEG1, MPEG2, MPEG4,
The term compression is mentioned in a number MPEG7—movable images
of technicalities and algorithms aiming to re-
duce the required memory for the representa- After the compression process, we have the
tion and storage of digital images. The storage process of storage to the PACS system and
of a digitalized image leads to a squandering of recuperation from RIS.
the memory of the computing system. Thus, in
the medical field, where the amount of image Storage of Medical Images and
information is enormous, a major problem rises Their Recuperation
because of the amount of memory required for
image storage. For this reason, faster compres- Medical images, according to the diagnostic
sion and decompression algorithms for images device used, have large sizes due to the need
have been searched for. Besides sparing for high-quality analysis and clearness, thus
memory, compression provides a reduction in demanding a large storage space. The recu-
the time and the width needed for transmission. peration time differs according to the kind of
The techniques for the compression of digi- examination and the demanded number of im-
tal images can be divided into two categories. ages in each examination.
In medicine, where even the smallest detail When the image reaches the final stage of
in depiction can be of vital importance, we elaboration, the goals are then the improvement
choose lossless techniques.
291
of the quality of the given information with the these tools provided by the graphic software for
use of the appropriate graphic-elaboration soft- quality improvement.
ware (e.g., clearness, brightness, contrast),
and also the exploitation of specific information Image Division
with the use of special operations and transfor-
mations like division and recombining. By dividing a medical image, we are able to
divide the image in different sections that show
Quality Improvement of Images some cognitive interest according to the speci-
ality under examination. Thus, according to the
The quality improvement of images is achieved reason why the diagnostic examination occurs,
with the use of software that helps to reach the the scientist can focus on a specific part of the
desired levels of analysis aiming to exploit and image to process suitably and find the needed
evaluate the information given from an image. information without the redundancy of infor-
During the process of recording images, certain mation to trouble him or her.
deformities appear. These are: Furthermore, with the appropriate software,
we can store to the computed system only those
• Dimness, parts of the image that interest us, thus sparing
• Noise of recording, and storage space and making it more easy to
• Geometrical deformities. recuperate the images from the used medical
databases.
All these deformities should be corrected in
order to eliminate all kinds of alterations. The Medical Image Recombination
elimination of dimness is done with the process
of restoration. The process of restoration is With medical image recombination, we can
extremely important in the elaboration of mov- have a combination of different views from
ing images because movement creates dim- various depiction examinations (e.g.,
ness. In most cases, a filtering of the image is ultrasound+CT, MRI+PET, MRI+SPECT), giv-
also required in order to remove sound. This ing us a three-dimensional representation of
can be achieved with various linear or nonlinear composition.
filters. Usually, nonlinear filters are preferred Indeed, medical image recombination is ex-
because they preserve the contrast of outlines, tremely important in medical depiction. Each
which are an important factor for human vision. diagnostic examination allows the scientist to
Moreover, the general contrast of the image
can be improved with special nonlinear tech-
niques.
Figure 6. Division of image received by an
One other important procedure for image
MRI
analysis is the recognition of outlines. Usually,
the areas of an image are coloured with false
colours.
Finally, we achieve improvement of quality
by adjusting the proper clearness, brightness,
and contrast with the use of appropriate tech-
nical equipment that allows us to exploit all
292
Figure 7. Image recombination MRI + PET Figure 8. Image recombination SPECT + CT

+ CT
extract specific information through it. With • Safe use

this recombination method, the scientist can • Is easy to learn
have the elements and information provided by
different diagnostic examinations in only one
image, can elaborate them and analyse them in CONCLUSION AND
any way he or she desires, and, finally, can PERSPECTIVES
reproduce them in a three-dimensional environ-
ment. The whole undertaking occurs through Three-dimensional image elaboration is a very
complex graphic software, which should have strong tool in medical diagnostics. Many of the
certain characteristics in order to be an actual diseases of tissue and microscopic interest are
diagnostic tool in the hands of the scientist vs. able to be diagnosed early and therefore can be
an obstacle in his or her effort to do the job. cured. Especially benefited are the fields of
The characteristics of the elaboration soft- research in medicine, pharmacology, and ge-
ware include the following: netics, where the elaboration and depiction of
three-dimensional models is extremely crucial
• Friendly user interface for their development. The appearance of more
• Large number of choices that are easy to and more evolved depiction devices and the
overcome growth of more powerful and more complete
• Support of the established models of com- software for digital graphic elaboration will
pression and image management predispose even bigger steps of evolvement in
• Support of future applications the near future.
293
294
Chapter XXXVIII
Imaging the Human Brain with
Magnetoencephalography
Dimitrios Pantazis
University of Southern California, USA
Richard M. Leahy
University of Southern California, USA
ABSTRACT
Magnetoencephalography is a relatively new medical imaging modality for the monitoring

and imaging of human brain function. Extracranial magnetic fields produced by the working
human brain are measured by extremely sensitive superconducting sensors, called SQUIDs,
enclosed in a liquid helium-filled dewar. Mathematical modeling allows the formation of
images or maps of cortical neuronal currents that reveal neural electrical activity, identify
cortical communication networks, and facilitate the treatment of neuronal disorders, such as
epilepsy.
INTRODUCTION The temporal resolution of MEG is in the

millisecond (ms) range, the timescale at which
Magnetoencephalography (MEG) is a neurons communicate. Therefore, we can fol-
noninvasive technique for measuring neuronal low the rapid cortical activity reflecting ongoing
activity in the human brain. Electrical currents signaling between different brain areas. This is
flowing through neurons generate weak mag- a great advantage compared to other medical
netic fields recorded using magnetic sensors imaging modalities such as functional magnetic
surrounding the head. The MEG method is part resonance imaging (fMRI) and positron emis-
of a broad area of research referred to as sion tomography (PET), where temporal reso-
biomagnetism, which involves studies of mag- lution is on the order of seconds. Furthermore,
netic fields emanating from several organs of unlike other methodologies that measure brain
the human body, notably the brain and heart. metabolism or the relatively slow hemodynamic
Imaging the Human Brain with Magnetoencephalography
response, MEG directly measures electrical Figure 1. Cerebral frontal cortex drawn by
brain activity. Electroencephalography (EEG) Ramón y Cajal using a Golgi staining
is a complimentary method to MEG, measuring technique. Pyramidal (A, B, C, D, E) and
electrical scalp potentials rather than magnetic nonpyramidal (F, K) cells are clearly
fields. It offers similar temporal resolution to depicted. Currents flowing in the dendritic
MEG, but the spatial resolution is less accurate trunks of pyramidal cells are believed to be
because electrical potentials measured on the the primary generators of magnetic signals
scalp are heavily influenced by strongly inho- outside the head.
mogeneous conductivity of the head, whereas
magnetic fields are mainly produced by cur-
rents that flow in the relatively homogeneous
intracranial space.
NEURAL BASIS OF
ELECTROMAGNETIC SIGNALS
A neuron consists of the cell body (or soma),

which contains the nucleus; branching den-
drites, which receive signals from other neu-
rons; and a projection called an axon, which
conducts the nerve signal. When a pulse arrives
at an axon of a presynaptic cell, neurotransmit-
ter molecules are released from the synaptic
vesicles into the synaptic cleft. These mol-
ecules bind to receptors located on target cells,
opening ion channels (mostly Na+, K +, and Cl-)
through the membrane. The resulting flow of currents as to produce measurable external
charge causes an electrical current along the fields. Clusters of thousands of synchronously
interior of the postsynaptic cell, changing the activated pyramidal cortical neurons are be-
postsynaptic potential (PSP). When an excita- lieved to be the main generators of MEG signals
tory PSP reaches the firing threshold at the (Figure 1). In particular, the currents associ-
axon hillock, it initiates an action potential that ated with large dendritic trunks, which are
travels along the axon with undiminished ampli- locally oriented in parallel and perpendicular to
tude. the cortical surface, are believed to be the
The conservation of electric charge dictates primary source of the neuromagnetic fields
that intracellular currents, commonly called outside the head. In contrast, the temporal
primary currents, give rise to extracellular cur- summation of currents for action potentials,
rents flowing through the volume conductor. which have duration of only 1 ms, is not as
Both primary and volume currents contribute to effective as for dendritic currents flowing in
magnetic fields outside the head; however, only neighboring fibers, so action potentials are be-
locally structured arrangements of cells can lieved to contribute little to MEG measure-
achieve sufficient coherent superposition of ments.
295
INSTRUMENTATION measurements proportional to the spatial gradi-

ent of the magnetic field, thus offering robust-
Empirical observations indicate that we ob- ness to interference from distant magnetic field
serve sources on the order of 10 nA-m, and sources.
consequently, the neuromagnetic signals are Modern MEG systems consist of a few
typically 50 to 500 fT, that is, 10 9 or 10 8 times hundred SQUID sensors placed in a liquid-
smaller that the geomagnetic field of the earth helium-filled dewar, with the flux-transformer
(Hämäläinen, Hari, Ilmoniemi, Knuutila, & pickup coils surrounding a helmet structure
Lounasmaa, 1993). The only detector that of- (Figure 2). Worldwide, three companies build
fers sufficient sensitivity to measure such fields the majority of whole-head MEG systems: 4-D
is the superconducting quantum interference Neuroimaging (formerly Biomagnetic Tech-
device (SQUID) introduced in the late 1960s by nologies Bti), Elekta Neuromag Oy, and VSM
James Zimmerman (Zimmerman, Thiene, & MedTech Ltd. (manufacturers of the CTF Sys-
Harding, 1970). The first measurement of brain tems). In recent years, all three vendors have
magnetic fields using a SQUID magnetometer introduced dense arrays comprising over 200
was carried out by David Cohen (1972) at the SQUID channels.
Massachusetts Institute of Technology, and it Brain magnetic signals are very weak com-
consisted of the spontaneous alpha activity of a pared to ambient noise. Outside disturbances
healthy participant and abnormal brain activity include fluctuations of the earth’s geomagnetic
in an epileptic patient. field, power-line fields, electronic devices, el-
The SQUID is a superconducting ring evators, and radio-frequency waves. Nearby
interrupted by thin insulating layers to form artifacts are caused by instrumentation noise
one or two Josephson junctions (Barone & and body interference, such as heart, skeletal
Paterno, 1982). One important property as- muscle, and spontaneous or incoherent back-
sociated with Josephson junctions is that ground brain activity. Shielded rooms made of
magnetic flux is quantized in units of successive layers of mu-metal, copper, and
Φ 0 = 2.0678 x10 −15 tesla ⋅ m 2 . If a constant bi- aluminum effectively attenuate high-frequency
asing current is maintained in the SQUID de-
vice, the measured voltage oscillates as the
magnetic flux increases; one period of voltage Figure 2. Whole-head CTF Omega MEG
variation corresponds to an increase of one flux system with 275 axial gradiometers (left),
quantum. Counting the oscillations allows one and MEG sensors using low-temperature
to evaluate the flux change that has occurred, electronics cooled by liquid helium (right)
and therefore detect magnetic fields on the
order of a few fT. The sensitivity of the SQUID
can be increased to 1 fT by attaching a coil of
superconducting wire or flux transformer. The
latter is placed as close to the human head as
possible, and depending on its shape, it can be
configured as a first-order planar or axial gra-
diometer, a second-order axial gradiometer, or
a simple magnetometer (Hämäläinen et al.,
1993). The gradiometer configurations produce
296
disturbances. Furthermore, gradiometer flux fields inside the head is estimated based on the
transformers cancel distant noise sources that conductivities of the scalp, skull, gray and white
produce magnetic fields with small spatial gra- matter, cerebrospinal fluid, and other tissue
dients. types. Head models that consist of a set of
nested concentric spheres with isotropic and
homogeneous conductivities have closed-form
MODELING solutions. Even though spherical head models
work surprisingly well, more accurate solutions
To estimate the neural sources of magnetic use realistic head models based on anatomical
fields, one must first solve the associated for- information from high-resolution magnetic reso-
ward problem, that is, the forward model that nance (MR) or x-ray computed tomography
maps sources of known location, strength, and (CT) volumetric images. To estimate the pa-
orientation to the MEG sensors. The most rameters of these models, numerical solutions
common source model is the current dipole using boundary-element methods (BEMs), fi-
(Baillet, Mosher, & Leahy, 2001), used to ap- nite-element methods (FEMs), or finite-differ-
proximate the flow of an electrical current in a ence methods (FDMs) are necessary (Darvas,
small area of the brain. The typical strength of Pantazis, Kucukaltun-Yildirim, & Leahy, 2004).
a current dipole, generated by the synchronous To make inferences about the brain activity
firing of thousands of neurons, is 10 nA-m. that gives rise to a set of MEG data, we must
Alternatively, to avoid the identifiability prob- solve the inverse problem, that is, find a neu-
lem that arises when too many small regions ronal current-source configuration that explains
and their dipoles are required to represent a the MEG measurements. Inverse methods for
single large region of coherent activation, we MEG can be roughly categorized into two
can use multipolar models, consisting of di- classes: imaging methods and dipole-fitting or -
poles, quadrupoles, octupoles, and so on scanning methods. The imaging approaches
(Mosher, Leahy, Shattuck, & Baillet, 1999). are based on the assumption that the primary
Since the useful frequency spectrum for sources are intracellular currents in the den-
electrophysiological signals is largely below dritic trunks of cortical pyramidal neurons that
100 Hz, the physics of MEG can be described are aligned normally to the cortical surface.
with the quasistatic approximation of Maxwell’s Consequently, a tessellated representation of
equations. The propagation of electromagnetic the cerebral cortex is extracted from a
Figure 3. MEG model depicting: (a) Sensor arrangement of a 275-channel CTF MEG system,
(b) topography of sensor measurements, and (c) minimum-norm inverse solution on a
tessellated cortical surface
297
coregistered MR image, and the inverse prob- STATISTICAL ANALYSIS

lem is solved for a current dipole located at
each vertex of the surface. In this case, since Given the large number of localization method-
the position and orientation of the dipoles are ologies, it is important to perform validation and
fixed, image reconstruction is a linear problem statistical analysis under different experimen-
and can be solved using standard techniques. tal settings, such as the number, location, and
The dipole-fitting or -scanning methods assume time series of neuronal sources. Furthermore,
that the sources consist of only a few activated several methods require the fine-tuning of pa-
regions, each of which can be represented by rameters, such as the subspace correlation
an equivalent current dipole of unknown loca- threshold for the MUSIC algorithm. The re-
tion and orientation. The standard approach to ceiver operating characteristic (ROC) curve is
localization is to perform a least-squares fit of a standard tool to evaluate the trade-off be-
the dipole model to the data (Lu & Kaufman, tween sensitivity and specificity, and to com-
2003). More recently, scanning methods have pare different inverse methods. By varying a
been developed that are also based on the threshold applied to localization maps, we can
dipole model, but involve scanning a source estimate two performance measures: the sensi-
volume or surface and detecting sources at tivity or true positive fraction (TPF), and 1-
those positions at which the scan metric pro- specificity or false positive fraction (FPF). The
duces a local peak (Baillet et al., 2001). Ex- ROC curve is a plot of the TPF vs. FPF as a
amples of these methods include the MUSIC detection threshold is varied. When comparing
(multiple signal classification) algorithm two detection methods, the one whose ROC
(Mosher, Leahy, & Lewis, 1992) and the LCMV curve gives higher sensitivity at matched speci-
(linearly constrained minimum variance) ficity, and vice versa, for all points on the curve
beamformer (VanVeen, van Drongelen, is the better detector. A simple metric to com-
Yuchtman, & Suzuki, 1997). pare methods is the area under the ROC curve
Due to intrinsic spatial ambiguities of the (AUC), where the method with the largest
electromagnetic principles that underlie MEG, AUC is superior. The use of free-response
the spatial resolution is lower than that of PET ROC, an ROC variant that can handle the
and fMRI. These ambiguities force a choice presentation and detection of multiple targets
between low-resolution linear cortical imaging per image, is demonstrated in Yildirim, Pantazis,
methods, or potentially higher resolution meth- and Leahy (in press) for the evaluation of
ods based on parametric models, or Bayesian minimum-norm and scanning-inverse methods.
or other nonlinear imaging methods incorporat- In addition to evaluating the relative perfor-
ing physiological priors that reflect the ex- mance of different methods, it is important to
pected characteristics of neural activation. A establish some degree of confidence in the
consensus is developing in the research com- results of real data analysis. Dipole-scanning
munity that no single method suits all MEG methods often produce unstable solutions, and
applications; each method has strengths and the reproducibility of the reconstructed dipoles
weaknesses, reflecting the ill pose of the in- is not guaranteed. A number of different ap-
verse problem. The characteristics of expected proaches have been investigated for assessing
neural activation, as well as model-fitting tech- dipole-localization accuracy, including Cramer
niques, can facilitate the proper choice of in- Rao lower bounds, perturbation analysis, and
verse methodology. Monte Carlo simulation. To avoid strict distri-
298
butional assumptions, a resampling alternative APPLICATIONS

based on bootstrap theory was proposed in
Darvas, Rautiainen, et al. (2005). The principle Applications in MEG include both basic and
underlying the bootstrap theory is that although clinical research. One of the most important
the distribution of the data is unknown, it can be clinical applications is the detection, classifica-
approximated by the empirical distribution of a tion, and localization of abnormal neuronal ac-
set of independent trials. By sampling with tivity in epilepsy patients. MEG has been suc-
replacement over independent trials collected cessfully used to localize three different spon-
during an event-related MEG study, the posi- taneous interrictal signal components: epileptic
tion, variance, and time series of current dipoles spikes, slow-wave activity, and fast-wave ac-
can be estimated reliably. tivity (Lu & Kaufman, 2003). The neurosurgi-
In contrast to dipole-scanning methods, im- cal planning of medically intractable epilepsy
aging methods are hugely underdetermined, often includes the identification of epileptoge-
resulting in low-resolution localization maps; nic lesions with MEG (Ossadtchi, Baillet,
interpretation is further confounded by the pres- Mosher, Thyerlei, Sutherling, & Leahy, 2004;
ence of additive noise exhibiting a highly non- Stefan et al., 2003). Furthermore, recent litera-
uniform spatial correlation. In this case, we ture investigates the possibility of seizure pre-
need a mechanism to decide which features in diction based on a drop in the complexity of
the data are indicative of true activation vs. neural activity immediately before seizures
those that are noise artifacts. To determine a (Maiwald, Winterhalder, Aschenbrenner-
suitable threshold for detecting statistically sig- Scheibe, Voss, Schulze-Bonhage, & Timmer,
nificant activation, the familywise error rate 2004).
(FWER), that is, the chance of any false posi- In addition to the diagnosis of epilepsy, MEG
tives under the null hypothesis of no activation is currently used for functional brain mapping.
(Type 1 error), is typically controlled. Paramet- Evoked response fields have been used to
ric random-field methods and nonparametric identify somatosensory-, motor-, and vision-
permutation methods are used to estimate related activity (Lu & Kaufman, 2003). Several
familywise-corrected thresholds in Pantazis, MEG studies (Pantazis, Merrifield, Darvas,
Nichols, Baillet, and Leahy (2005). Alterna- Sutherling, & Leahy, 2005) have localized lan-
tively, the control of the false discovery rate guage-specific cortical activity using either
(FDR), that is, the proportion of false positives equivalent current dipoles or distributed corti-
among those tests for which the null hypothesis cal imaging, with promising results for clinical
is rejected, can produce more sensitive thresh- application in neurosurgery. Time-frequency
olds. analysis of MEG oscillatory-evoked responses
Recent literature in MEG statistical analysis (Pantazis, Weber, Dale, Nichols, Simpson, &
has been mostly limited to pairwise compari- Leahy, 2005) has detected networks of cortical
sons at each cortical surface element for event- interactions and determined the functional
related averages. However, extensions of this specificity of several frequency bands. A wide
methodology to the investigation of multiple range of signal-processing techniques including
effects using analysis of variance (ANOVA) image modeling and reconstruction, blind source
and analysis of covariance (ANCOVA) in indi- separation, phase synchrony estimation, nonlin-
viduals and groups is possible, as, for example, ear analysis, and chaos theory are under inves-
in Brookes et al. (2004).
299
tigation to reveal complex cognitive processes Barone, A., & Paterno, G. (1982). Physics and
such as attention and working memory. applications of the Josephson effect. Wiley.
Recent literature investigates how evoked
Brookes, M. J., Gibson, A. M., Hall, S. D.,
response fields relate to neuronal disorders,
Furlong, P. L, Barnes, G. R., Hillebrand, A., et
such as Alzheimer’s disease, autism, dyslexia,
al. (2004). A general linear model for MEG
brain tumors, and Parkinson’s disease. Fur-
beamformer imaging. Neuroimage, 23, 936-
thermore, MEG has been used in conjunction
946.
with transaxial magnetic stimulation to amelio-
rate abnormal brain activity (Anninos, Tsagas, Cohen, D. (1972). Magnetoencephalography:
Sandyk, & Derpapas, 1991). Detection of the brain’s electrical activity with
a superconducting magnetometer. Science,
175, 664-666.
CONCLUSION
Darvas, F., Pantazis, D., Kucukaltun-Yildirim,
E., & Leahy, R. M. (2004). Mapping human
Magnetoencephalography is a relatively new
brain function with MEG and EEG: Methods
medical imaging modality for the monitoring
and validation. Neuroimage, 23(Suppl. 1),
and imaging of human brain function. While
S289-S299.
spatial resolution is significantly lower than that
of PET and fMRI, the ability to monitor neu- Darvas, F., Rautiainen, M., Pantazis, D., Baillet,
ronal activation at the millisecond time scale S., Benali, H., Mosher, J. C., et al. (2005).
makes this modality, together with EEG, a Investigations of dipole localization accuracy in
unique window on the human brain. Recent MEG using the bootstrap. Neuroimage, 25(2),
developments in instrumentation have lead to 355-368.
the manufacture of whole-head MEG arrays
Hämäläinen, M., Hari, R., Ilmoniemi, R. J.,
with an excess of 300 magnetometers. Coupled
Knuutila, J., & Lounasmaa, O. V. (1993).
with new data-analysis tools for mapping brain
Magnetoencephalography: Theory, instrumen-
function from MEG data, these systems will
tation, and applications to noninvasive studies
lead to important new insights into the workings
of the working human brain. Reviews of Mod-
of the human brain with applications in both
ern Physics, 65(2), 413.
clinical and cognitive neuroscience.
Lu, Z.-L., & Kaufman, L. (Eds.). (2003). Mag-
netic source imaging of the human brain.
REFERENCES Mahwah, NJ: Lawrence Erlbaum Associates,
Inc.
Anninos, P. A., Tsagas, N., Sandyk, R., &
Maiwald, T., Winterhalder, M., Aschenbrenner-
Derpapas, K. (1991). Magnetic stimulation in
Scheibe, R., Voss, H. U., Schulze-Bonhage,
the treatment of partial seizures. International
A., Timmer, J. (2004). Comparison of three
Journal of Neuroscience, 60, 141-171.
nonlinear seizure prediction methods by means
Baillet, S., Mosher, J. C., & Leahy, R. M. of the seizure prediction characteristic. Physica
(2001). Electromagnetic brain mapping. IEEE D, 194, 357-368.
Signal Processing Magazine, 18(6), 14-30.
300
Mosher, J., Leahy, R., & Lewis, P. (1992). electrical activity via linearly constrained mini-
Multiple dipole modeling and localization from mum variance spatial filtering. IEEE Transac-
spatiotemporal meg data. IEEE Transactions tions on Biomedical Engineering, 44(9), 867-
on Biomedical Engineering, 39, 541-557. 880.
Mosher, J. C., Leahy, R. M, Shattuck, D. W., & Yildirim, E. K., Pantazis, D., & Leahy, R. M.
Baillet, S. (1999). MEG source imaging using (in press). Task-based comparison of inverse
multipolar expansions. Proceedings of the 16 th methods in MEG. IEEE Transactions on Bio-
Conference on Information Processing in medical Engineering.
Medical Imaging, IPMI’99 (pp. 15-28).
Zimmerman, J. E., Thiene, P., & Harding, J. T.
Ossadtchi, A., Baillet, S., Mosher, J. C., Thyerlei, (1970). Design and operation of stable rf-based
D., Sutherling, W., & Leahy, R. M. (2004). superconducting poinst-contact quantum de-
Automated interrictal spike detection and source vices and a note on the properties of perfectly
localization in magnetoencephalography using clean metal contacts. Journal of Applied Phys-
independent components analysis and spatio- ics, 41, 1572-1580.
temporal clustering. Clinical Neurophysiol-
ogy, 115(3), 508-522.
KEY TERMS
Pantazis, D., Merrifield, W., Darvas, F.,
Sutherling, W., & Leahy, R. M. (2005). Hemi- ANOVA and ANCOVA: Analysis of vari-
spheric language dominance using MEG corti- ance or covariance is a collection of statistical
cal imaging and non-parametric statistical analy- models and their associated procedures that
sis. WSEAS Transactions on Biology and compare means by splitting the overall ob-
Biomedicine, 2(3), 318-325. served variance into different parts.
Pantazis, D., Nichols, T. E., Baillet, S., & Current Dipole: Popular source model in
Leahy, R. M. (2005). A comparison of random MEG, representing a point’s current source. It
field theory and permutation methods for the is a convenient representation for the coherent
statistical analysis of MEG data. Neuroimage, activation of a large number of pyramidal cells,
25(2), 383-394. possibly extending over a few square centime-
ters of gray matter.
Pantazis, D., Weber, D. L., Dale, C. C., Nichols,
T. E., Simpson, G. V., & Leahy, R. M. (2005). EEG: Electroencephalography measures
Imaging of oscillatory behavior in event-related neuronal activity by recording electrical poten-
MEG studies. In C. A. Bouman & E. L. Miller tials with electrodes attached on the human
(Eds.), Proceedings of SPIE, Computational scalp. The resulting waveforms are used to
Imaging III, 5674 (pp. 55-63). localize brain activity and assess brain damage,
epilepsy, or even in some cases brain death.
Stefan, H., Hummel, C., Scheler, G., Genow,
A., Druschky, K., Tilz, C., et al. (2003). Mag- fMRI: Functional magnetic resonance im-
netic brain source imaging of focal epileptic aging uses powerful magnets to create a field
activity: A synopsis of 455 cases. Brain, that resonates the nuclei of atoms in the body.
126(11), 2396-2405. The oscillating atoms emit radio signals that are
converted by a computer into 3-D images of the
VanVeen, B., van Drongelen, W., Yuchtman,
human body and cerebral blood flow.
M., & Suzuki, A. (1997). Localization of brain
301
LCMV Beamformer: Linearly constrained space to identify the origin of signals. It is often
minimum-variance beamformer applies spatial used in MEG to estimate the location, orienta-
filtering to sensor array data to discriminate tion, and strength of current dipoles.
between signals from a location of interest and
PET: Positron emission tomography is a
those originating elsewhere. In the application
noninvasive imaging modality that measures
to MEG, the goal is to find a spatial filter that
the distribution of radioactive-labeled molecules
minimizes the output power of the beamformer
inside a biological system. By using molecular
subject to a unity gain constraint at the desired
probes that have different rates of uptake de-
location on the brain.
pending on the type of tissue involved, PET can
MUSIC: Multiple signal classification is a localize lesions, and detect regional blood flow
localization algorithm that uses the subspace and gene expression among others.
correlation between the data and model sub-
302
303
Chapter XXXIX
Region of Interest Coding in
Medical Images
Sharath T. Chandrashekar
Sarayu Softech Pvt Ltd., India
Gomata L. Varanasi
Samskruti, India
ABSTRACT
To provide efficient compression of medical images, identifying and extracting the region of
interest from the entire image and coding the specific region to accuracy is important. This
chapter introduces the basics of region of interest coding, an overview of the coding methods
available and their main features for the benefit of learners and researchers. The special
focus is on JPEG-2000-based algorithms.
INTRODUCTION What is the Region of Interest?
One of the main aims in medical image process- ROI is the region of image that is of clinical or
ing is to extract important features from radio- diagnostic interest to the doctor, radiologist, or
logical image data, called the region of interest image analyst. Its shape may be regular, as
(ROI), for accurate diagnostic analysis, inter- shown in Figure 1, or arbitrary and irregular, as
pretation, and better patient treatment. Coding in Figure 2.
the region of interest is significant for easy,
rapid transmission, and also for efficient stor- Multiple ROIs
age. This is useful in the application areas of
teleradiology, picture archiving and communi- There could be more than one region of interest
cation systems (PACSs), and hospital informa- within a given image, leading to multiple ROIs
tion systems (HISs; http://www.dclunie.com). as shown in Figure 3.
Region of Interest Coding in Medical Images
Figure 1. UltraSound image ROI is fetus Figure 2. CT image—ROI is heart—arbitrary

zone—regular shape shape
Identifying and Extracting Lossy Compression Schemes
Identifying and extracting the region of interest Totally lossy schemes result in image alter-
is required before compressing and coding the ation, which might entail a loss of diagnostic or
image that includes the region of interest. Iden- scientific utility. Sometimes lossy compression
tifying a region is done by manual and/or auto- may deliver exquisite quality and yet can intro-
matic segmentation methods. The segmenta- duce medically unacceptable artifacts into the
tion procedure used is based on the input image image. These are less popular in medical situ-
data, the nature of the information sought by the ations.
end user from the segmented image, and the
application (Grimes, 2004). Lossless Compression Schemes
ROI Compression and Coding The following are categorized as lossless com-
pression schemes in different contexts based
With the growing interest in the areas of on the end user or observer:
telemedicine and health informatics, compress-
ing and coding ROI is a necessity. The follow- • Visually Lossless: Nonclinical human
ing are some of the compression schemes observer
employed in this area of image coding.
Image compression usually can be lossy or
lossless. Lossless compression methods are Figure 3. Lesions are the ROIs in brain MRI
preferred for high-value content, such as medi- slice
cal imagery or image scans made for archival
purposes. Lossy methods are especially suit-
able for natural images in applications where
minor (sometimes imperceptible) loss of fidelity
is acceptable to achieve a substantial reduction
in bit rate.
304
Figure 3a. ROI, the cube representing the wavelet coefficients of Figure 5. Surface of the cube
represents the most significant bit, MSB of the coefficients. Each of the coding method scales
this height for ROI.
• Diagnostically Lossless: Clinical ob- degradation of data in unimportant regions.

servers, significant degrees of
interobserver and intra-observer varia- Uses of Compression
tions may exist
• Quantifiably Lossless: Mostly nonhuman In radiology, compression can be done in many
observers, computer-assisted detection ways and has the following uses:
Lossless schemes cannot offer high com- • Compression before primary diagnosis (for
pression ratios, yet they are preferred in certain rapid transmission)
situations where the data is needed without • Compression after primary diagnosis (for
much loss (Clunie, n.d.). long-term archiving in a PACS)
• Compression for database browsing
Regionally Lossless (where progression in quality and resolu-
Compression Schemes tion would be useful)
These are a special case of lossless coding Region of Interest Coding

schemes based on a certain region of interest
(Christopoulos, Askelöf, & Larsson, 2000). A ROI coding is a process performed mostly at
chosen ROI is encoded with higher quality than the encoder end. The encoder decides which
the background (BG). The ROI can either be ROI is to be coded with better quality than the
static or dynamic. Static ROIs are defined at background. If the encoder does not know the
coding time. The user defines dynamic ROIs ROIs in advance, there is still a possibility for
interactively while they are progressively trans- the decoder to receive only the data that is
mitted and decoded. requested using some end-user specification.
After segmenting an image into regions This case is also explained in one of the coding
(either automatically or manually), it is possible algorithms mentioned in later sections.
for a compression algorithm to deliver different
levels of reconstruction quality in different spa- Coding Methods: Encoding End
tial regions of the image. One could accurately
preserve the features needed for medical diag- The JPEG2000 standard (Christopoulos,
nosis or for scientific measurement while achiev- Skodras, & Ebrahimi, 2000; Clunie, n.d.;
ing high compression overall by allowing the Marcellin, Gormish, Bilgin, & Boliek, 2000) is a
305
wavelet-based still-image compression stan- the least significant bit plane (LSB) so that
dard. This compression standard is adopted in the ROI-associated bits are placed in the
DICOM format (Digital Imaging and Commu- higher bit planes. This can also be viewed
nications in Medicine; http://www.dclunie.com; as ROI mask coefficients shifted up (Fig-
http://www.rsna.org), commonly used for medi- ure 4). This relative scaling value between
cal images. Two of the ROI coding methods ROI and BG is recorded.
defined in JPEG2000 are the following: • Multiple ROIs: The generic scaling
method allows for multiple ROIs; each
• Generic scaling-based method ROI is upshifted by a different value. The
• Maximum shift (Maxshift) method upshift value is recorded in the code-
stream header for each ROI.
Other than these, the following are also • Multiple Overlapped ROIs: If the dif-
being used and have different advantages based ferent ROIs overlap, the overlapped re-
on the application: gion is coded as belonging to the highest
quality ROI. Because this method does
• Most significant bit-plane shift (MSBShift) not result in an easily identifiable ROI
method mask, it must be made available to the
• Bit-plane-by-bit-plane shift (BbBShift) decoder.
method • Fine Control: As any scaling value is
supported, the generic scaling-based
Coding algorithms used in any of these method allows fine control on the relative
methods are derived based on the standard importance between ROI and BG.
image-compression schemes (Bhaskaran & • Shape Coding Needed: The ROI shape
Konstantinides, 1997). Along with the color information has to be coded, and this sig-
transform, discrete wavelet transform, quanti- nificantly increases the complexity and
zation, image coding, rate control, and entropy reduces the coding efficiency.
coding, these methods include additional pro- • Restricted Shapes: Regular ROI shapes
cessing related to ROI compression (Figure 5 (rectangle and ellipse) are supported in
and Figure 6). The information provided in this this method. This restriction may limit the
article is only at an overview level, and it is real application of ROI coding and the
highly suggested that the readers consult the compression efficiencies that can be
references for all the details. achieved.
Generic Scaling Method Coding Algorithm

The following is the overview of this method The general scaling-based method is imple-
and its features (Christopoulos, Askelöf, et al., mented as follows, and it is very familiar amongst
2000): image-compression communities.
Encoding ROI:
• Part 2 of the JPEG2000 standard
(Christopoulos, Skodras, et al., 2000) sup- 1. The color transform is performed on the
ports the general scaling method. entire image, and the discrete wavelet
• ROI in Higher Bit Planes: The bits of transform is calculated as in JPEG2000
BG coefficients are downshifted toward (Bhaskaran & Konstantinides, 1997).
306
Figure 4. Generic scaling method. ROI-1-star: coefficients are raised to higher scale—2 and
ROI-2—the elliptical brain slice. Different ROIs can be raised to different scales.
ROI-1
ROI-2
scale-2
scale-1
2. If the ROI is identified, then an ROI mask to the shape information of the ROI included in
is derived extracting the region, indicating the code stream, the BG coefficients are lo-
the set of coefficients that are required for cated and scaled up to their original places
lossless ROI reconstruction. before the inverse wavelet transform is applied
3. The wavelet coefficients are quantized. to reconstruct the image.
These coefficients are stored in a sign
magnitude representation. Magnitude bits Maxshift Method
comprise the most significant part of the The following is the overview of this method
implementation precision used. and its features:
4. The coefficients that are out of the ROI
are scaled up or down by a specific scaling • Maxshift scaling is supported in Part 1 of
value. If there are more than one ROI, the JPEG2000 standard.
these can be multiple coded with different • Special Case of General Scaling: This
scaling values. can be viewed as a particular case of the
5. The resulting coefficients are progressively generic scaling-based method when the
entropy encoded (with the most significant scaling value is so large that there is no
bit planes first). As overhead information, overlap between BG and ROI bit planes,
the scaling value assigned to the ROI and as seen in Figure 7. After scaling, all bits
the coordinates of the ROI are added to the of the ROI coefficients will be in higher bit
bit stream. The decoder performs also the planes than all the bits associated with the
ROI mask generation, but scales up the BG.
background coefficients in order to recre- • Larger Scaling and Less Compres-
ate the original coefficients. sion: Compared with the generic scaling-
based method, the Maxshift method uses
Decoding ROI: The decoder reverses the larger scaling values and reduces the com-
steps of encoding to reconstruct the image as in pression efficiency by introducing more
Figure 6 (Bhaskaran & Konstantinides, 1997). bit planes.
At the decoder, the bit planes are first recon- • Multiple ROIs: A single scale value is
structed from the most significant bit plane to used for all the ROIs. When multiple
the least significant bit plane. Then, according ROIs with different priorities may be in-
307
Figure 5. Encoding algorithm
Shape Mask
& Scale Factor
Preprocessing Forward Inter Forward Intra Quantization

Component Component
Transform Transform
Input Image
Code Stream
Rate Control
EBCOT code
Figure 6. Decoding algorithm
ROI-data
Decode with De-quantization Inverse Inverse Inter

options Intercomponent Component
Code Stream Transform Transform
Reconstructed Image
volved, the Maxshift method cannot sup- perform ROI mask generation, either (this
port the concept like the generic scale- might still be needed at the encoder).
based method.
• Arbitrary-Shape Coding Support: Ar- Coding Algorithm
bitrary shapes are supported. The shape Encoding ROI: The coding algorithm is almost
of the ROI is encoded and sent. It is the same as the generic scaling method. It
available implicitly for the decoder. differs after the first two steps as follows:
• No Fine Control: Unlike the generic
scaling-based method, the Maxshift • The wavelet coefficients are quantized.
method cannot control the relative impor- The encoder scans the quantized coeffi-
tance between ROIs and the BG. No BG cients and chooses a scaling value S such
information can be received until all ROI that the minimum coefficient belonging to
coefficients are decoded fully. the ROI is larger than the maximum coef-
• No ROI Mask Generation in Decoder: ficient of the background (non-ROI area).
The ROI mask is easy to find because of • It shifts the ROI mask coefficients up
the scaling value mentioned along with the such that their LSBs are higher than the
location and shape of the ROI. This means most significant non-zero bit of all back-
also that the decoder does not have to ground coefficients.
308
Figure 7. MaxShift method. ROI-1 and ROI-2 both are scaled up to the MSB level. Green
cylinder and star show the location. But no ROI coefficients are there below MSB level.
Coefficient value
MSB
scale
scale
LSB
ROI Coefficients
Decoding ROI: At decoding, the BG and • Fine Control: This is the same as in the
ROI coefficients can be identified easily ac- generic scaling-based method. The
cording to the bit-plane positions. Every coeffi- MSBShift method can flexibly adjust the
cient that is smaller than the scale belongs to the relative importance between ROI and BG
background and is therefore scaled up. The by using different scaling values.
decoder needs only to upscale the received
background coefficients. Coding Algorithm
Steps 1 and 2 of the generic scaling method are
MSBShift Method used:
• Arbitrarily Shaped ROI Coding: The • This method (Liu & Fan, 2003) removes
MSBShift method not only supports arbi- all the overlapped bit planes between the
trarily shaped ROI coding without coding ROI and BG coefficients, and relatively
the shape, but also enables the flexible modifies the quantization step size of the
adjustment of compression quality in the coefficients. This reduces the final ROI
ROI and background. quality.
• Multiple ROIs: Additionally, the new • We can isolate a certain number of bit
method can efficiently code multiple ROIs planes of the ROI bits in the most signifi-
with different priorities in an image.
Figure 8. MSBShift method. Some of the coefficients of ROI-1 and ROI-2 can be raised to
different scales. But the MSB of the background is shifted down.
Coefficient value
MSB
scale
scale
LSB
ROI Coefficients
309
Figure 9. ROIs and coded images using generic scaling method (a) original, (b) and (c) ROIs
(a) (b) (c)
cant bit planes to adjust the importance • No Multiple ROI Coding: Moreover,
between the ROI and BG. the BbBShift method does not support
• We only need to shift part of the most multiple ROI coding.
significant bit planes of the ROI coeffi- • Compression Efficiency: This is similar
cients instead of shifting all of the bit to that of the Maxshift method.
planes of the ROI coefficients as the
standard methods do. Coding Methods: Decoding End
• Complexity and Coding Efficiency:
Since it is not necessary for the MSBShift All of the above algorithms cater to the coding
method to code ROI shapes, the complex- schemes at the encoder end. On the decoder
ity is less than the generic scaling-based end, the operation that is performed is just the
method, and the coding efficiency is higher reverse of the encoding operation:
when the same scaling value is used. If the
point of lossless coding is reached, the bit • We can exploit the feature of JPEG2000
rate produced by the MSBShift method is that enables one to compress once and
not larger than the Maxshift method be- decompress in many ways according to
cause the MSBShift method encodes less the user’s requirement at the decoder
or at most the same number of bit planes. end. The authors have done some re-
search work on these lines, and the results
BbBShift Method obtained are in the process of publication.
• The encoding operation is done using the
• This method (Wang & Bovik, 2002) shifts baseline algorithm of JPEG2000.
the bit planes on a bit-plane-by-bit-plane • Using the features of random access to
basis instead of shifting them all at once the encoded data, the required region of
like in the Maxshift method. interest can be progressively decoded.
• Finer Control: The BbBShift method • Manual data input is required for identify-
supports arbitrarily shaped ROI coding ing the location of the likely region of
without coding shape information, and it interest.
can offer finer control of the ROI and BG • This method has applications in the areas
quality than the Maxshift method. of telemedicine, mobile communications,
and PDAs (personal digital assistants).
310
Coding Metrics and Compression Christopoulos, C., Askelöf, J., & Larsson, M.
Efficiency (2000). Efficient methods for encoding regions
of interest in the upcoming JPEG 2000 still
• PSNR at a Given Bit Rate: Other than image coding standard. IEEE Signal Process-
subjective judgments, the parameter that ing Letters, 7(9), 247-249.
measures the coding efficiency is the
Christopoulos, C., Skodras, A., & Ebrahimi, T.
PSNR in dB (peak signal-to-noise ratio of
(2000). The JPEG 2000 still image coding sys-
the reconstructed image) at a given num-
tem: An overview. IEEE Transactions on
ber of bits per pixel.
Consumer Electronics, 46(4), 1103-1127.
• Complexity and Coding Efficiency: The
MSBShift method has less complexity than Clunie, D. A. (n.d.). Lossless compression of
the generic scaling method as the ROI grayscale medical images: Effectiveness of
shape need not be coded for transmission. traditional and state of the art approaches.SPIE,
Coding efficiency is higher when the same Proceedings, Medical Imaging, 3980, 74-
scaling is used. The complexity of the 84. Retrieved from http://www.dclunie.com
MSBShift method is more than Maxshift,
Grimes, S. L. (2004). Clinical engineers: Stew-
but the bit rates achieved are comparable.
ards of healthcare technologies. IEEE EMBS
Magazine, 23(3), 56-58.
CONCLUSION Kalawsky, R. S. (2000). The validity of pres-
ence as a reliable human performance metric in
ROI coding is a very significant and promising immersive environments. Proceedings of the
research area that can benefit many medical Third International Workshop on Presence,
computing applications. The generic scaling Delft, Netherlands.
and Maxshift methods are popular amongst
Liu, L., & Fan, G. (2003). A new method for
image-processing communities and are well
jpeg2000 region of interest image coding: Most
established. This article has limited scope and is
significant bitplanes shift. IEEE Signal Pro-
only an introduction for curious developers.
cessing, 10(2), 35-39.
The details can be explored and expanded
further based on the references given. Marcellin, M. W., Gormish, M., Bilgin, A., &
Boliek, M. (2000). An overview of JPEG 2000.
Proceedings of IEEE Data Compression Con-
REFERENCES ference, Snowbird, UT.
Wang, Z., & Bovik, A. C. (2002). Bitplane-by-
Bhaskaran, V., & Konstantinides, K. (1997).
bitplane shift (BbBShift): A suggestion for
Image and video compression standards:
JPEG2000 region of interest coding. IEEE Sig-
Algorithms and applications (2nd ed.). Bos-
nal Processing Letters, 9(5), 160-162,
ton: Kluwer Academic Publishers.
Burak, S., Tomasi, C., Girod, B., & Beaulieu, C.
INTERNET REFERENCES:
(2001). Medical image compression based on
region of interest with application to colon CT
http://www.dclunie.com
images. IEEE EMBS Proceedings (pp. 2453-
2456). http://www.nema.org
311
http://www.rsna.org Hospital or Radiology Information Sys-

tem (HIS or RIS): While PACS applications
http://www.wikipedia.org
manage images, HIS and RIS applications man-
age patients, studies, and results.
KEY TERMS
Image Coding: A definition of a set of
rules to map one set of image data onto another
Color Transform: The RGB components
set to make the image or signal more suitable
in natural images are more correlated than the
for an intended application. To suit the applica-
components in a luminance-chrominance space.
tion, the image-coding scheme may optimize an
Color transform from one space to the other
image for transmission, improving transmission
(typically from RGB space to YCrCb or lumi-
quality and fidelity, modifying the image for
nance-chrominance space) is to decorrelate
providing error detection and/or correction, pro-
the components, reduce the redundancy, and
viding data security, and so forth. Different
thereby achieve energy compaction. This is a
coding schema may have different sets of
standard approach for compression.
advantages and disadvantages.
Data Compression (Encoder and De-
Image Compression: The application of
coder): It is the process of encoding informa-
data compression on digital images. In effect,
tion using fewer bits (information-bearing units)
the objective is to reduce the redundancy of the
than a detailed representation through the use
image data in order to be able to store or
of specific encoding schemes to optimize the
transmit data in an efficient form.
data. The follow-up compressed-data commu-
nication can work when both the sender (en- Picture Archiving and Communication
coder) and receiver (decoder) of information System (PACS): The Picture Archiving and
share and understand the coding scheme. Communication System is an image-based in-
formation system for the acquisition, storage,
Digital Imaging and Communications
communication, archiving, display, and manipu-
Medicine (DICOM): Digital Imaging and
lation of medical digital images and other rel-
Communications in Medicine is an imaging
evant data.
standard that allows the exchange of data
between different hosts and equipment across Quantization: It is the process of approxi-
the network in a heterogeneous environment. It mating continuous data or signals by a set of
includes a file format definition and a network discrete symbols and values rather than con-
communication protocol that uses TCP/IP tinuous representation.
(transmission-control protocol/Internet proto-
Telemedicine: In Greek, tele means far.
col) to communicate between systems.
Telemedicine is the delivery of medicine or
Entropy Coding: It is a coding scheme clinical care to remote individuals with the use
that assigns codes to symbols so as to match of modern telecommunication and information
code lengths with the probabilities of the sym- technologies.
bols. Three of the most common entropy en-
Teleradiology: A method of distributing
coding techniques are Huffman coding, range
digital diagnostic radiological information such
encoding, and arithmetic encoding.
as images using radiological systems based on
312
x-rays, ultrasound, magnetic resonance, and let transform is set of filter coefficients ob-
other related information through local area or tained over the set of basis wavelets functions.
wide area networks between remotely located These are localized in scale and in time, as well
facilities. as in frequency. The frequency is derived from
the scale. These functions are scaled and con-
Wavelet Transform: Wavelets in general
volved with the function being analyzed all over
are functions that can be used to efficiently
the time axis to get to transform space inviting
represent other functions. The discrete wave-
compression.
313
314
Chapter XL
Imaging the Human Brain with
Functional CT Imaging
Sotirios Bisdas
Johann Wolfgang Goethe University Hospital, Germany
Tong San Koh

Nanyang Technological University, Singapore
ABSTRACT
Recent advances in multi-detector computed tomography (CT) have revitalized its role in the
clinical routine. In the field of cerebral perfusion, CT provides a rapid, low-cost functional
imaging, which by the utilization of a suitable tracer kinetic analysis can provide valuable
information in many clinical applications, like acute stroke, cerebrovascular reserve capacity,
vasospasm after subarachnoidal hemorrhage, cerebral trauma, tumor imaging, and brain
death diagnosis. The limitations of the existing commercially available post processing
software are discussed and a new distributed-parameter tracer kinetic model for generating
more accurate perfusion parametric maps is introduced.
INTRODUCTION and those under intensive care. CT perfusion is

a brilliant example of a protocol that has under-
Recent advances in multidetector computed gone significant improvements bearing consid-
tomography (CT) have resulted in subsecond erable impact on patient care in the acute
volumetric patient scanning, revitalizing the role setting.
of CT in the clinical routine. Such quantum In the field of cerebral perfusion, CT over-
technological laps combined with the availabil- comes other imaging modalities (e.g., magnetic
ity and turnaround efficiency of CT have had a resonance imaging [MRI], xenon computed
notable impact on the imaging paradigms of the tomography, positron-emission tomography
critically ill such as stroke and trauma patients [PET], and single-photon-emission computed
Imaging the Human Brain with Functional CT Imaging
tomography [SPECT]) through the rapid, low- used for parametric fitting of the dynamic imag-
cost generation of meaningful images and avail- ing data are usually linear compartmental mod-
able user-friendly postprocessing software that els that can be further classified as conven-
interprets the data in a clinically relevant man- tional compartmental (CC) or DP models.
ner. The theory behind perfusion CT is a tracer Three main models have been used for
kinetic analysis of the rapidly intravenously model-independent approaches including the
injected contrast material. maximum-slope model, the equilibring-indica-
The purpose of this chapter is to revisit the tor model, and the central-volume principle.
assessment of perfusion CT imaging in many The maximum-slope model requires very short
clinical applications, to suggest new clinical injection times, which are not always tolerable
fields (e.g., brain-death diagnosis) for possible by the patients, and tends to underestimate the
future applications of perfusion CT, and to cerebral blood flow (BF). Moreover, there is no
emphasize the pitfalls of the method, suggesting general consensus regarding the reference ar-
a new distributed-parameter (DP) kinetic model terial input function, which is necessary for the
for generating more accurate perfusion para- calculation of the perfusion parameters. On the
metric maps. other hand, the equilibring-indicator model and
the central-volume model require a mathemati-
cal operation called deconvolution (Axel, 1981;
IMAGING TECHNIQUES, Ostergaard et al., 1996). Deconvolution can be
INDICATORS, AND TRACER realized in two ways. The first method to solve
KINETICS MODELS the deconvolution problem is a parametric
method in which supplemental hypotheses re-
The measurement of cerebral perfusion relies garding the anatomical structure or the
on a triad including an imaging technique, an behaviour of the indicator are taken into ac-
indicator, and a model. In the case of CT count. The second method, called nonparamet-
imaging, the sequential acquisition of cerebral ric, can be performed with single-value
sections (one or more sections per second) is deconvolution and involves less source hypoth-
performed during the rapid (6 mL/sec in our eses. Deconvolution-based CT perfusion of-
clinical protocol) intravenous administration of fers a fast and robust analysis of neurovascular
an iodinated contrast agent (50 mL of a 400 mg/ disorders and has been an important strategy to
dL non-ionic iodinated contrast material in our assess stroke and to characterize intracranial
protocol). tumors.
The iodinated contrast agent is restrained in On the other side, comprehensive studies
the cerebral vascular bed, at least at first pass have indicated that linear compartmental mod-
and in healthy cerebral tissue. Tracer kinetic- els could allow a more complete analysis of the
analysis methods for estimating microcircula- kinetic parameters. Also, as illustrated in vari-
tory parameters can generally be grouped un- ous works, DP models possess more realism
der two classes (Larson, Markham, & Raichle, than CC models (Koh, Cheong, Hou, & Soh,
1987; Lee, 2002): (a) model-independent ap- 2003). The reason for this is that the tracer
proaches that include numerical deconvolution- concentration gradients within compartments
based methods and (b) model-dependent or are taken into account in DP models, while in
parametric-fitting approaches. In the model- CC models, tracer concentration gradients are
dependent approach, the tracer kinetics models assumed to be zero at all times and thus the
315
tracer is assumed to distribute instantaneously unbiased and is supposed to overcome the

on arrival in each compartment. problem arising from the individual determina-
tion of the arterial input function (AIF), for
which no consensus exists regarding the most
CLINICAL APPLICATIONS appropriate location.
Identifying the volume of the dead brain
Acute Stroke tissue has important prognostic value, but iden-
tifying the volume of the potentially salvageable
Unenhanced CT, which provides morphologic brain tissue could improve the course and ag-
information about brain parenchyma, has long gressiveness of acute–stroke treatment even
been the cornerstone for evaluating acute stroke. more. The proposed DP model peers into the
A normal CT scan of the brain excludes almost cellular biology of the infarcted brain and, in the
always intracranial hemorrhage, which, in evi- case of a substantial tracer leakage from the
dence of an acute infarction, would herald the vessels, quantifies the extravascular volume
need for rapid medical therapy and potentially (v2) and the permeability surface product (PS).
intravascular intervention. Nevertheless, the The loss of integrity of the blood-brain bar-
alteration of the brain morphology due to cyto- rier (BBB) resulting from ischemia or
toxic edema is not evident until some significant reperfusion is believed to be the precursor to
time after the stroke onset, leading many au- hemorrhagic transformation and poor outcome.
thors to advocate the use of diffusion MRI to Since two major therapy strategies for acute
evaluate acute infarctions. Perfusion CT has stroke are strongly related to the hazardous
been stratified to assess acute cerebral is- effect of hemorrhagic transformation, namely,
chemia, and threshold infarcted and non-inf- anticoagulation and thrombolysis, PS values
arcted tissue, and to predict tissue and clinical can serve as a tool for selecting treatment
condition outcome (Bisdas et al., 2004). In paradigms. Recent retrospective observations
contrast to other imaging modalities, like PET using MRI showed an early disruption of the
and MRI, it is well tolerated by patients with BBB in acute stroke, whereas CT can only
acute stroke symptoms and delivers quantita- confirm it at a later stage, missing the initial
tive results. Nevertheless, as aforementioned, signs of microbleeding or BBB disruption (Latour
the maximum-slope model may deliver unreli- et al., 2004).
able results, and together with deconvolution- Furthermore, in tracer kinetic models in
based analysis, it is extremely robust and, thus, which numerical deconvolution is used, the
is not realistic. MTT is calculated as the area under the time-
With our proposed DP tracer kinetic model, density curve, and blood volume (BV) is the
we calculate the perfusion (F); vascular mean convolution of BF and MTT. The potential PS
transit time (MTT; t1), accounting for the mean disorders are not accounted for in this case, and
transit time of the contrast agent in the intravas- MTT corresponds to t1 in our model while BV
cular compartment; and intravascular volume is equal to v1. Undoubtedly, at the presentation
(v1). Secondly, we introduce in perfusion CT of BBB disruption, BV parametric maps based
the use of lag time (tlag), which estimates the on numerical deconvolution or conventional com-
delay of the peak of the tissue residue function, partment models cannot be used to indicate low
aiming at the rapid identification of areas with perfusion. Thus, our DP model can provide
varying degrees of hypoperfusion. Tlag is user valuable and precise physiologic data (Figure 1).
316
Figure 1. DP-model analysis of functional CT data obtained from a patient with a sudden onset
(<2h) of right hemiparesis
a. b.
c. d.
a. The perfusion t1 parametric map shows a time

delay of the contrast agent’s arrival in the
e. anterior portion of the middle cerebral-ar-
tery territory on the right side.
b. The time delay is better recognized in the tlag
map, which is independent of the subjectively
determined AIF.
c. The v1 parametric map demonstrates reduced
intravascular volume, reflecting the ischemic
lesion margins.
d. The delineation of the ischemic lesion is also
recognized in the F parametric map.
e. Subtle permeability surface product abnor-
malities are demonstrated in the PS map,
indicating a disturbance of the brain-blood
barrier’s integrity.
317
Cerebrovascular Reserve Capacity the patients undergoing bypass surgery and can
also be used to evaluate the efficacy of
Patients with chronic steno-occlusive disease revascularization procedures postoperatively.
of the extracranial carotid represent a complex
subgroup of patients with few proven therapeu- Vasospasm
tic options. Neurological symptoms, produced
by thromboembolic stroke and by altered cere- Vasospasm is a frequent complication in the
bral hemodynamics, are very common in these early course of subarachnoidal hemorrhage
patients. In fact, collateral supply, mostly via (SAH). A progression to infarction occurs in
the circle of Willis, by extracranial-to-intracra- approximately half of the symptomatic cases;
nial collaterals and leptomeningeal anastomoses thus, measurements of cerebral blood flow can
maintains normal perfusion in patients with be useful in identifying patients at increased
carotid stenosis. Nevertheless, a percentage of risk of cerebral infarction by guiding therapeu-
patients have an insufficient collateral supply, tic decisions and monitoring responses to therapy.
which leads to hemodynamic compromise PET, SPECT, xenon CT, and transcranial Dop-
(Derdeyn, Grubb, & Powers, 1999). The im- pler sonography have already been employed
paired cerebral hemodynamics can be identi- for this purpose. Perfusion CT offers a rapid
fied by demonstrating an impaired vasodilatatory and reliable measurement of blood flow as well
response after acetazolamide challenge-perfu- as of BV and MTT, increasing thus the under-
sion studies. The acetazolamide test is a reli- standing of the impairment of the brain auto-
able predictor of critically reduced cerebral regulation. ROI analysis in our patients with
perfusion and may unmask cerebrovascular SAH demonstrated prolonged times of t1 and F
reserve deficits even in patients with asymp- values under 12 mL/min/100g of tissue (Nabavi
tomatic carotid stenosis who have hemody- et al., 2001). Nevertheless, the presence of
namic compromise. massive subarachnoidal blood in the brain cis-
The status of cerebrovascular autoregula- terns did not allow perfusion measurements of
tory control has been examined with xenon, large brain territories as the volume averaging
PET, single-photon-emission, and perfusion CT; of pathologic tissue and subarachnoidal blood
perfusion-weighted MRI; and transcranial Dop- led to false increased F values. In this case, t1
pler ultrasonography. Recent studies on perfu- values are more reliable and useful for therapy
sion CT have demonstrated significant correla- monitoring.
tion between the measured perfusion values
and the estimated cerebrovascular reserve with Traumatic Cerebral Contusions
PET measurements (Kudo et al., 2003; Bisdas
et al., 2006). With our proposed DP model, the Patients with traumatic brain injury frequently
evaluation of the cerebrovascular response as exhibit cerebral contusions, which may swell
(a) a less-than-expected augmentation relative and cause increased intracranial pressure with
to the contralateral side, (b) an absent augmen- secondary ischemia. It is still controversial
tation, or (c) a paradoxical reduction in flow whether traumatic contusions represent irre-
(steal phenomenon) could be easily performed. versibly infarcted focal lesions and how their
Thus, the visualization of cerebral perfusion presence can predict outcome in patients with
symmetry and the obtained quantitative perfu- severe head trauma. Conventional CT at ad-
sion results can substantially assist in selecting mission usually underestimates the extent and
318
severity of the cerebral parenchymal lesions, of tumor mitotic monitoring, visualization of the
while initial experience with perfusion mea- most malignant tumour portions, and therapy
surements, in terms of blood flow, showed the response. Apart from the patient radiation ex-
irreversibly damaged zone and the potential posure during the perfusion CT imaging, a
viability of tissue in the pericontusional zone possible limitation of the method is the limited
(von Oettingen, Bergholt, Gyldensted, & Astrup, anatomic coverage. The latter results in an
2002). In our clinical setting, we were able to inadequate estimation of the tumor volume (in
acquire absolute perfusion values of acute trau- case of large tumors) and a possibly insufficient
matic contusions and to predict later infracted arterial input function since a large vessel is not
areas. Analogously to the peri-infarction pen- in the examined slab of brain tissue. The intro-
umbral zones in acute ischemic stroke, the duction of 64-detector multislice CT units may
pericontusional tissue damage was outlined and solve the problem of anatomic coverage and
the infarction risk was assessed. provide reliable tumor perfusion measurements.
Tumors Brain-Death Diagnosis
Expansive masses in the brain are readily rec- The diagnosis of brain death must be certain to
ognized with CT imaging in the case of perifo- allow organ transplantation and the discontinu-
cal oedema and contrast-enhancement pat- ation of artificial ventilation. Brain death is
terns that are also used to differentiate differ- present when all functions of the brain stem
ent types of brain tumours. The use of perfusion have irreversibly ceased. Clinical and electro-
CT in this field offers a numerically solid basis physiological criteria may be misinterpreted
for differential diagnosis and assessment of the due to drug intoxication, hypothermia, or tech-
tumor characteristics beyond that of visual nical artefacts. Thus, if clinical assessment is
assessment (Roberts, Roberts, Lee, & Dillon, suboptimal, reliable early confirmatory tests
2002). In terms of BF and BV values, cerebral may be required for demonstrating the absence
perfusion assesses the increased tumor-inher- of intracranial blood flow. Cerebral angiogra-
ent angiogenic activity and neovascularization. phy, MRI, CT imaging after the inhalation of
The hyperpermeability related to the immature stable xenon, electroencephalography, brain-
vessels can also be assessed by the PS para- perfusion SPECT measurements, and scintig-
metric maps. PS measurements need a perfu- raphy are possible methods for providing brain-
sion-CT technique modification in the case of death diagnosis in comatose patients (Bonetti,
the central-volume models in order to account Ciritella, Valle, & Perrone, 1995, Kurtek, Lai,
for the extravasation of contrast material due to & Kay-Yin, 2000). Perfusion CT proved to be
the disrupted BBB. The use of DP models a reliable, safe, and cost-effective method for
offers the additional advantage of measuring defining brain death in five patients in our
the extravascular blood volume as well as the institute. Perfusion CT was easily carried out
first-pass extraction ratio of the tracer into the and interpreted in the comatose patients with
extravascular extracellular space from the in- brain damage without discontinuing therapy.
travascular space (E). A significant correlation Brain death was diagnosed by recognizing the
between the v2 and E parameters with tumour absence of brain perfusion, as shown by no
growth is still not observed as it is demonstrated intracranial arrival of the contrast material or
with the PS values. The latter provide a means by extremely prolonged MTT and low BF in-
319
compatible with life. The possibility of perform- Transactions on Biomedical Engineering,

ing subsequent perfusion measurements, ap- 50, 159-167.
plying each time a new contrast-agent bolus,
Kudo, K., Terae, S., Katoh, C., Oka, M., Shiga,
offers the advantage of greater anatomic cov-
T., Tamaki, N., et al. (2003). Quantitative cere-
erage, although it should be considered that the
bral flow measurement with dynamic perfusion
used contrast material may damage the har-
CT using the vascular-pixel elimination method:
vested organs.
Comparison with H215O positron emission to-
mography. American Journal of
Neuroradiology, 24, 419-426.
REFERENCES
Kurtek, R., Lai, K., & Kay-Yin, B. S. (2000).
Axel, L. (1981). A method of calculating brain Tc-99m hexamethylpropylene amine oxime
blood flow with a CT dynamic scanner. Ad- scintigraphy in the diagnosis of brain death and
vanced Neurology, 30, 67-71. its implications for the harvesting of organs
used for transplantation. Clinical Nuclear
Bisdas, S., Donnerstag, F., Ahl, B., Bohrer, I.,
Medicine, 25, 7.
Weissenborn, K., & Becker, H. (2004). Com-
parison of perfusion computed tomography with Larson, K. B., Markham, J., & Raichle, M. E.
diffusion-weighted magnetic resonance imag- (1987). Tracer kinetic models for measuring
ing in hyperacute ischemic stroke. Journal of cerebral blood flow using externally detected
Computer Assisted Tomography, 28, 747- radiotracers. Journal of Cerebral Blood Flow
755. Metabolism, 7, 443-463.
Bisdas, S., Nemitz, O., Berding, G., Latour, L. L., Kang, D. W., Ezzeddine, M. A.,
Weissenborn, K., Ahl, B., Becker, H., et al. et al. (2004). Early blood-brain barrier disrup-
(2006). Correlative assessment of cerebral blood tion in human focal brain ischemia. Ann Neurol,
flow obtained with perfusion CT and positron 56, 468-477.
emission tomography in symptomatic stenotic
Lee, T. Y. (2002). Functional CT: Physiological
carotid disease. European Radiology (in
models. Trends Biotechnol, 20(8), S3-S10.
press).
Nabavi, D. G., Le Blanc, L. M., Baxter, B., et
Bonetti, M. G., Ciritella, P., Valle, G., & Perrone,
al. (2001). Monitoring cerebral perfusion after
E. (1995). 99mTc HM-PAO brain perfusion
subarachnoidal hemorrhage using CT.
SPECT in brain death. Neuroradiology, 37,
Neuroradiology, 43, 7-16.
365-369.
Ostergaard, L., Sorensen, A. G., Kwong, K.
Derdeyn, C. P., Grubb, R. L., Jr., & Powers,
K., et al. (1996). High resolution measurement
W. J. (1999). Cerebral hemodynamic impair-
of cerebral blood flow using intravascular tracer
ment: Methods of measurement and associa-
bolus passages: Part I. Mathematical approach
tion with stroke risk. Neurology, 53, 251-259.
and statistical analysis. Magn Reson Med, 36,
Koh, T. S., Cheong, L. H., Hou, Z., & Soh, Y. 715-725.
C. (2003). A physiologic model of capillary-
Roberts, H. C., Roberts, T. P., Lee, T. Y., &
tissue exchange for dynamic contrast enhanced
Dillon, W. P. (2002). Dynamic contrast-en-
imaging of tumour microcirculation. IEEE
320
hanced CT of human brain tumors: Quantitative used to reverse the effects of convolution on
assessment of blood volume, blood flow, and recorded data. The concept of deconvolution is
microvascular permeability-report of two cases. widely used in the techniques of signal process-
American Journal of Neuroradiology, 23, ing and image processing. Since these tech-
828-832. niques are in turn widely used in many scientific
and engineering disciplines, deconvolution finds
Von Oettingen, G., Bergholt, B., Gyldensted,
many applications. In general, the object of
C., & Astrup, J. (2002). Blood flow and is-
deconvolution is to find the solution of a convo-
chemia within traumatic cerebral contusions.
lution equation of the form f * g = h. Usually,
Neurosurgery, 50, 781-790.
h is some recorded signal, and f is some signal
that we wish to recover but has been convolved
KEY TERMS with some other signal g before we recorded it.
The function g might represent the transfer
Acetazolamide Test: Test to determine function of an instrument or a driving force that
the inherent vasodilatatory activity of the brain was applied to a physical system. If we know g,
vasculature. Acetazolamide, a carbonic anhy- or at least know the form of g, then we can
drase inhibitor, has been shown to increase perform deterministic deconvolution.
cerebral perfusion in healthy subjects.
Equilibring-Indicator Model: Perfusion
Arterial Input Function (AIF): The input model that applies mainly to diffusible indica-
arterial concentration of the contrast agent in a tors (e.g., those used in nuclear medicine and in
reference artery needed for the deconvolution xenon CT) and considers a balance of the
process in the central-volume-based perfusion indicator concentrations between blood and
model. cerebral tissue.
Central-Volume Principle: General per- Extravascular Volume (v2): Fractional
fusion model that assumes an arterial input and contrast-agent (tracer) distribution volume in
a venous output of the injected tracer. The extravasular space.
deconvolution of the measured parenchymal
Intravascular Volume (v1): Fractional
and arterial time-curve concentrations gives
contrast-agent (tracer) distribution volume in
the mean transit time of the tracer in the
intravascular space.
vascular bed and subsequently the other perfu-
sion parameters. Iodinated Contrast Agent: Under normal
conditions and in healthy subjects, it is a nondif-
Cerebral Contusions: Bruises on the brain,
fusible, iodinated non-ionic tracer whose con-
usually caused by a direct, strong blow to the
centration in the tissue is linearly proportional to
head.
the tissue enhancement, the X-ray attenuation,
Convolution and Deconvolution: In math- and the intensity of the CT image (expressed as
ematics and in particular functional analysis, Hounsfield units).
convolution is a mathematical operator that
Lag Time (tlag): The delay of the peak of
takes two functions f and g and produces a third
the tissue residue function between different
function that in a sense represents the amount
brain regions.
of overlap between f and a reversed and trans-
lated version of g. Deconvolution is a process
321
Maximum-Slope Model: Perfusion model Perfusion (F): A constant flow of con-

that considers a complete extraction of the trast-enhanced blood that is assumed to supply
tracer at first pass. Perfusion is proportional to the intravascular space (unit of measure is mL
the total amount of tracer accumulated in a per 100g tissue•min).
given area as well as to the rate of accumula-
Permeability Surface Area Product (PS):
tion (slope of the accumulation curve).
Transfer (diffusion) of the contrast agent
Neovascularization: A newly formed tu- (tracer) between the intravascular space and
mor may induce the formation of new capillar- the extravascular extracellular space (unit of
ies and start to invade the surrounding tissue. measure is mL per 100g tissue•min).
The newly developed tumor vessels often dis-
Tissue Residue Function: Function that
play an abnormal architecture, characterized
describes the elimination of a theoretically unique
by collapsing or poorly differentiated fragile
and instantaneous unit bolus of contrast agent
and leaky vessels, which are frequently unable
once it has entered the vascular system.
to meet the rapid growth of tumor cells, result-
ing in local hypoxia and necrosis. The Vascular Mean Transit Time (t1): The
neovascularization of a primary tumor increases mean time taken by the contrast (tracer) to
the possibility that cancer cells will enter the traverse the intravascular space (units of mea-
blood stream and spread to other organs, and is sure are minutes and seconds).
also necessary for the growth of metastases in
Vasospasm: A dangerous side effect of
distant organs.
subarachnoid hemorrhage that irritates the blood
vessels on the surface of the brain, causing
them to constrict erratically, cutting off blood
flow.
322
324
Chapter XLI
Nonlinear Signal Processing
Techniques Applied to
EEG Measurements
Christos L. Papadelis
Aristotle University of Thessaloniki, Greece
Chrysoula Kourtidou-Papadeli
Greek Aerospace Medical Association and Space Research, Greece
Panagiotis D. Bamidis
Nicos Maglaveras
ABSTRACT
The electrical activity of the brain is sensitive to its oxygen supply, and electroencephalography
(EEG) has been proposed as a suitable measurement to detect brain activity alterations
induced by hypoxia. Since, linear processing techniques that have been used so far in hypoxia
studies are based on false linearity assumptions about the generation of the EEG signal, there
is a definite need for nonlinear approaches to be applied on EEG data derived from hypoxic
conditions. The aim of the present study is to compare nonlinear techniques’ effectiveness to
identify significant variations in EEG due to hypoxia. EEG data from two channels were
derived from ten healthy subjects participated in the present study. Oxygen and nitrogen
mixture was used to simulate hypoxic conditions that correspond to an altitude of 25.000 feet.
Non-linear measurements such as correlation dimension, approximate entropy, Lyapunov
exponent and detrended fluctuation analysis (DFA) parameters were estimated for EEG
signals. The results of the present study confirm the effectiveness of nonlinear techniques to
identify significant variations in EEG, which reflect alterations in cerebral function induced
by cerebral hypoxic conditions.
Nonlinear Signal Processing Techniques Applied to EEG Measurements
INTRODUCTION by nonlinear dynamic systems governed by

chaotic attractors (Henry, Lovell, & Camacho,
It has been well over a century since it was 2001). The contribution of models derived from
discovered that the mammalian brain generates deterministic chaotic systems, such as the
a small but measurable electrical signal (Gloor, Lorentz, Rössler, and Chua attractors, was of
1969). The technique of measuring this electri- fundamental importance for the development
cal activity, called electroencephalography of nonlinear signal-processing techniques.
(EEG), is a sensitive but nonspecific measure Nonlinear dynamics help us to understand
of brain function, and it is widely applicable in that irregular and seemingly unpredictable time
the clinical diagnosis of brain disorders and in evolutions do not necessarily have to be a result
brain physiological processes research. Never- of pure randomness, but on the contrary, can be
theless, in most of the applications used so far, the result of completely deterministic and fairly
the EEG recordings result in long traces with simple (low-dimensional) dynamical systems.
marked interobserver variability (Williams, Moreover, the unpredictability of these sys-
Luders, Brickner, Goormastic, & Klass, 1985). tems can be explained as their dynamics are
The need for more specific, compact, and reli- strongly dependent on the starting condition
able medical information derived from EEG has (Lorenz, 1963). The theoretical description of
been partially satisfied by quantitative EEG this dynamic behaviour is called deterministic
analysis (qEEG). This technique has been con- chaos (Signorini & Cerutti, 1999).
fined to feature analysis, conventional power The essential problem in deterministic chaos
spectrum analysis, parametric description of is to determine whether or not a given time
EEG, or frequency analysis (Geocadin et al., series is a deterministic signal from a low-
2000). The theory of linear stochastic pro- dimensional dynamical system. Grassberger
cesses had led to the development of a collec- and Procaccia (1983) provided a simple algo-
tion of tools and techniques for the analysis of rithm for the estimation of the correlation di-
EEG. Basically, these were tools for a precise mension from time series. Since then, many
description of the deterministic and stochastic scientists have presented a large number of
aspects of given time series, and, unfortunately, studies reporting low dimension in EEG mea-
were based on very simple assumptions about surements (Babloyantz & Destexhe, 1986;
the system (brain) that produced the EEG sig- Babloyantz, Salazar, & Nicolis, 1985). It has
nal (linearity assumption). Although linear tech- also been supported that a chaotic and rather
niques contribute a lot in EEG applicability in high-dimensional EEG characterizes the healthy
clinical practice, the brain’s electrical activity state of the brain, whereas a reduction of
presents aperiodic waveforms that suggest its dimension and a tendency toward nonchaotic,
origin in chaotic dynamics (Galka, 2000). periodic dynamics is characteristic of present
One of the most important mathematical or imminent pathologies (Galka, 2000). How-
discoveries of the past few decades is that ever, the reports of finite dimension in EEG
random behaviour can arise in deterministic recordings were and are still received with
nonlinear systems with just a few degrees of scepticism by several authors working in the
freedom. The broad spectra and aperiodic os- field of nonlinear time-series analysis. First, it is
cillations that are observed in recordings of hard to believe that a highly complex system as
brain activity have suggested to many research- the brain should exhibit as little complexity as,
ers the possibility that this activity is generated for example, the Lorentz system (Kantz &
325
Schreiber, 1995). The second reason for scep- system. The active sites on the scalp were
ticism comes from the fact that the dimension referenced to linked mastoids. Oxygen satura-
estimators that produced the finite estimates tion (SaO2) was recorded with a polarographic
were simplified (Grassberger, 1986; Procaccia, pulse oximeter, which was placed in one of the
1988). Moreover, most of the nonlinear para- fingers of the left hand. The target SaO2 for the
metric measurements (such as correlation and mixture was 70.
information dimensions, Lyapanov exponents, At sea level, participants performed a motor
etc.) are noise sensitive, and the stationarity task as a warm-up. The motor task was also
requirement is not fulfilled. performed in three conditions of 100% oxygen,
The aim of the present study is to compare hypoxia (gas mixture), and recovery at 25 feet.
nonlinear techniques’ effectiveness to identify Each session lasted 3 minutes. The participants
significant variations in EEG, which reflect were seated in front of a computer screen, a
alterations in cerebral function induced by hy- response keypad, and a tracking-control joy-
poxia. Cerebral hypoxia is caused by the failure stick.
of the human body’s systems to deliver ad- During the experimental period, EEG and
equate oxygen to the brain (Marianai & Wright, EOG were monitored via Ag-AgCl electrodes.
1998), and it is a significant clinical condition, The EOG electrodes were positioned above
especially in intensive-care units. and below the left eye and at the outer canthus
of both eyes. EOG signals were filtered at 1 to
13 Hz and amplified by Grass P511 amplifiers.
MATERIAL AND METHODS During recording periods, qualified personnel
monitored the signals on screen to ensure that
Participants, Equipment, and any controllable artifacts from participant move-
Experimental Protocol ments or eye blinks could be eliminated.
Ten healthy, drug-free people (six men, four Artifacts Rejection

women) participated in the study. Their ages
ranged from 22 to 27 years old (mean age was EEG recordings were first band-pass filtered
24.3). An oxygen and nitrogen mixture was (low pass was 0.5 Hz, high pass was 35 Hz),
used to simulate an altitude of 25 feet (10.9% and then the Infomax independent component
oxygen and 89.1% nitrogen). Neither the par- analysis (ICA) technique was applied in order
ticipant nor the observer was aware of the gas to remove artifacts from the data, including eye
mixture. The mixture was administered to the movements, eye blinks, cardiac signals, and
participants by an aviation mask (type MBU- muscle noise (Jung et al., 2000). ICA decompo-
2OP Gentex) completely sealed for any leaks; sition was performed on EEG measurements
a control valve was connected to the mixture using EEGLAB software coded in Matlab
bottles for the right selection of oxygen and (available from http://sccn.ucsd.edu/eeglab;
nitrogen content. Delorme & Makeig, 2004). Components con-
The Biologic-Brain Atlas III was used for taminated by artifacts were rejected, and the
the EEG recordings, connected to the ECI remaining components were mixed and pro-
Electro-Cap Electrode system. Two Ag-AgCl jected back onto the scalp channels. A more
electrodes were positioned on the scalp sites detailed description of the method followed can
C3 and C4 according to the 10-20 international be found in Jung et al.
326
State-Space Reconstruction autocorrelation function. These include the

earliest time τ at which the autocorrelation
The precondition of a successful reconstruc- drops to a fraction of its initial value (Albano,
tion of the state space of the underlying EEG Muench, Schwartz, Mees, & Rapp, 1988) or
generation process is typically required for the has a point of inflection (Fraser & Swinney,
analysis of possible deterministic properties of 1986).
a time series (Galka, 2000). For discretely The methodology of finding an optimal em-
sampled data, all approaches to state-space bedding dimension has been provided by Sauer,
reconstruction can be regarded as variants of Yorke, and Casdagli (1994). An embedding
the central technique of time-delay embedding. dimension of m ≥ 2Do + 1, where Do is the box-
It is possible to perform an analysis of the counting dimension of the attractor, is sufficient
underlying process, starting from the evaluation to ensure that the reconstruction is a one-to-
of an observed time series x(n), (n = 1, 2, ..., N) one embedding. In an alternative approach, if
of length N (Parker & Chua, 1987). According the attractor has a correlation dimension D2,
to the time-delay embedding theory, we obtain then an embedding dimension of m ≥ D2 is
multivariate vectors in an m-dimensional space, sufficient to measure the correlation dimension
each defined as: from the embedding. More recently, Cao (1997)
proposed a method for the optimal embedding
y (n ) = {x(n ), x(n + τ ),..., x(x + (m − 1)τ )}, dimension that is based on the property of
n = 1, N − (m − 1)τ , (1) chaotic attractors in that their orbits should not
intersect or overlap with each other. Such an
where τ is an appropriate time lag and m is the intersection or overlap may result when the
embedding dimension (Abarbanel, Brown, attractor is embedded in a dimension lower than
Sidorowich, & Tsimiring, 1993). the sufficient one stated by the delay-embed-
Theoretically, any time delay τ will yield an ding theorem.
embedding (Signorini & Cerutti, 1999). How-
ever, if we have only finite precision, both too Correlation Dimension
small and too large values of τ cause failures of
the reconstruction. If the time delay is small, the The estimation of the fractal dimension for
values x(t) and x(t + τ) will be almost equal, the experimental data sets can be performed using
system did not have time to change its state the correlation-dimension parameter D2. D2
significantly, and there is little gain of informa- provides an interior bound for the fractal di-
tion between them. Due to the little gain of mension D of the system attractor. The
information between consecutive components, Grassberger-Procaccia algorithm has been pro-
the impact of noise will be large (redundancy). posed for the estimation of the correlation
If τ is large and the dynamical system is cha- dimension (Grassberger & Procaccia, 1983).
otic, the reconstruction attractor becomes a This algorithm is based on the assumption that
largely featureless mess since the reconstruc- the probability that two points of the set are in
tion vectors contain components with hardly the same cell of size r is approximately equal to
any dynamical correlation (overfolding effect). the probability that two points of the set are
There have been various proposals for separated by a distance ρ less than or equal to
choosing an optimal delay time for topological r (Henry et al., 2001). Thus, the correlation
properties based on the behaviour of the dimension is approximately given by:
327
of divergence of two neighboring trajectories.

N
H H
∑ Θ(r − ρ ( xi, x j )) A sufficient condition to recognize a chaotic
i =1, j >1
C (r ) ≈ , (2) system is the presence of at least one positive
1
N ( N − 1) exponent (Signorini & Cerutti, 1999). A nega-
2
tive exponent implies that the orbits approach a
common fixed point. A zero exponent means
where the distance τ is typically measuring the
the orbits maintain their relative positions; they
Euclidian metric
are on a stable attractor.
The algorithm proposed by Wolfe, Swift,
H H
ρ (xi , x j ) = ∑ (xi (k ) − x j (k ) )2 ,
m
Swinney, and Vastano (1985) is used to get the
(3)
k =1 largest LE (LLE) estimation from EEG data. In
order to calculate Lyapunov exponents from a
and the Heaviside function is defined as time series, an appropriate embedding of the
experimental time series has to be constructed
by using the time-delay embedding methodol-
1 if s ≥ 0
Θ(s) =  . (4) ogy. The EEG time series x(t) for an m-dimen-
0 if s < 0 sional phase space with delay coordinate τ that
is a point on the attractor is given by
The precise calculation of the correlation
dimension is performed in the limit Ν → ∞. {x(t ), x(t + τ ),..., x(t + (m − 1)τ }. (6)
However, this limit cannot be realized in prac-
tical applications. Grassberger and Procaccia The nearest neighbor to the initial point
(1983) propose the approximate evaluation of should also be located:
C(r) over a range of values of r, and then
deduce D2 from the slope of the straight line of {x(to ), x(t o + τ ),..., x(to + (m − 1)τ )}. (7)
best fit in the linear scaling region of a plot of
logC(r) vs. logr. The distance between these two points is
There have been several estimates of the denoted as L(to). At a later time t1, the initial
minimum number of data points Nmin required length will evolve to length L'(t1). The mean
for estimates of D to be reliable using the exponential divergence of two initially close
Grassberger-Procaccia algorithm. Ruelle orbits is characterized by:
(1990) has proposed one of them:
M
1 L' (t k )
N min = 10 ( D / 2)
. (5) λ= ∑
t M − to k =1
log 2
L' (t k −1 )
. (8)
Lyapunov Exponents
The set of numerical parameters m, τ, T,
Lyapunov exponents (LE) describe system Smax, and Smin has to be chosen, where m is the
dynamics, giving a quantitative measure of the embedding dimension, T is the evaluation time,
attractor stretching and folding mechanism and Smax and Smin are the maximum and mini-
(Eckmann & Ruelle, 1985). It is a quantitative mum separations of the replacement point,
measure of the sensitive dependence on the respectively. According to Das, Das, and Roy
initial conditions, and it defines the average rate (2002), an embedding dimension between 5 to
328
20 and a delay of 1 should be chosen when Detrend Fluctuation Analysis

calculating LE for EEG data.
The detrend fluctuation analysis (DFA) technique
Approximate Entropy (Peng, Havlin, Stanley, & Goldberger, 1995) is a
modification of the root-mean-square analysis of
Approximate entropy (ApEn) is a statistic that a random walk for nonstationary data (Walleczek,
can be used as a measure to quantify the complex- 2000). DFA was proposed for the investigation of
ity (or irregularity) of a signal. It was first pro- correlation properties in nonstationary time series
posed by Pincus (1991a) and was then used and was applied to the studies of heartbeat (Peng,
mainly in the analysis of heart-rate variability Havlin, et al.), DNA nucleotides (Peng, Buldyrev,
(Huikuri, Makikallio, & Perkiomaki, 2003; Pincus, Havlin, Simons, Stanley, & Goldberger, 1994),
1991b, 1994), endocrine-hormone release pulsatility and EEG (Watters, 1998).
(Pincus, 1996), and EEG (Burioka et al., 2005; According to Hwa and Ferree (2002), the
Natarajan, Acharya, Alias, Tiboleng, & EEG time series y(t) is divided into B subsets of
Puthusserypady, 2004). independent segments of the same size k. Within
ApEn presents salient features that make it each subset labelled b, perform a least-square
attractive for use in biomedical signal process- fit of EEG time series y(t) by a straight line
ing. A robust estimator of ApEn can be ob- yb (t ) , which is the semilocal trend for the bth
tained by using shorter data (in the range of 100 subset. Combine yb (t ) for all B subsets and
to 5,000 points); it is highly resistant to short denote the B straight segments by:
transient interference, and the influence of
noise can be suppressed by properly choosing B
the relevant parameter in the algorithm. y (k , t ) = ∑ yb (t )θ (t − (b − 1)k )θ (bk − t )
b =1
Two parameters m and r must be chosen (10)
prior to the computation of ApEn, where m
specifies the pattern length and r is the effective for 1 ≤ t ≤ kB . Define
filter. Assuming that the original data are
1 kB
x(n ) = x(1), x(2 ),..., x(N ) , where N is the total F 2 (k ) = ∑ [y (t ) − y (k , t )] ,
2
(11)
kB t =1
number of data points, the correlation integral Cm
(r) (with the embedding dimension m and time lag where F(k) is the root-mean-squared fluctuation
1) should be estimated. The approximate en- from the semilocal trends in B subsets each
tropy measure is finally obtained as follows: having k time points, and is also a measure of the
fluctuation in each subset averaged over B sub-
sets. The study of the dependence of F(k) on the
1 L−m 1 L − m+1
ApEn(m, r , L) = ∑ log Cim+1 (r ) − L − m + 1 ∑ log Cim (r ) .
L − m i =1
subset size k is the essence of DFA (Peng,
i =1
(9) Buldyrev, et al., 1994; Peng, Havlin, et al., 1995).
If there is a power-law dependence:
The ApEn quantifies the likelihood that sets
of patterns that are close remain close on the F(k) ∞ k a, (12)
next incremental comparison. For the present
study, m is set to 2 and r is set to 25% of the then the scaling exponent a is an indicator of the
standard deviation of each time series. power-law correlations of the fluctuations in
EEG (Hwa & Ferree, 2002).
329
Figure 1. The autocorrelation function vs. the time lags for the experimental data of the 100%-
oxygen session (Participant 3, channel C3). The horizontal green line is the value of 1/e.
RESULTS bedding dimension ranging from 4 to 7 was

revealed, and a fixed embedding dimension of
In order for the precondition of a successful d=5 was used, which also satisfies uelle’s
reconstruction of the state space of the under- (1990) criterion of Equation 5.
lying EEG generation process to be satisfied, A statistically significant decrease of the
the optimal time delay τ and the optimal embed- mean value of the correlation dimension for all
ding dimension were estimated. The analysis of participants was observed during the hypoxia
the optimal time delay and embedding dimen- session compared to the 100%-oxygen session
sion was performed for the EEG measurements for both channels (Figure 3): channel C3
of each participant and each channel for the (p=0.0418, df=9, t=1.943) and channel C4
three experimental sessions. For the optimal (p=0.029, df=9, t=2.175). No statistically sig-
time-delay selection, the criterion of the earliest nificant differences were observed between
time τ at which the autocorrelation drops to 1/ the recovery session and the 100%-oxygen
e of its initial value was used (Figure 1). Thus, session, neither for channel C3 (p=0.1162, df=9,
the autocorrelation function analysis revealed t=1.7378) nor for channel C4 (p=0.1822, df=9,
that the optimal time delay ranged from five to t=-1.445). Analysis of variance (ANOVA) did
nine samples. For the purpose of comparison not reveal significant differences between the
between attractors estimated for a large num- three experimental sessions for both channels:
ber of different segments, it would not be channel C3 (F=1.225, df=2, p=0.309) and chan-
advisable to readjust the time delay for each nel C4 (F=2.111, df=2, p=0.14).
epoch. Thus, the fixed value of τ =8 samples Although, Das et al. (2002) have proposed a
was used in the present study as the optimal time delay of 1 for the LE calculation of EEG
delay time for all EEG segments. For the opti- data, in the present study, a time delay of 8 has
mal embedding-dimension estimation, Cao’s been used for the LE estimation. A decrease,
(1997) technique was used in the present study but not statistically significant, of the largest
(Figure 2). From Cao’s methodology, an em- Lyapunov exponent for all participants was
330
Figure 2. Cao’s (1997) method was used for the optimal embedding-dimension estimation with
a delay time of 8, a maximal dimension of 10, three nearest neighbors, and 1,000 reference
points. The correlation dimension vs. embedding dimension for the experimental data of the
100%-oxygen session (Participant 3, channel C3) are presented here.
Figure 3. Mean value of D 2 correlation-dimension estimation for all participants. Although a

statistically significant decrease of D2 was observed during the hypoxia session for channel
C3, an increase of the corresponding correlation dimension was observed for channel C4.
observed during the hypoxia session compared reveal significant differences between the three
to the 100%-oxygen session for both chan- experimental sessions for both channels: chan-
nels—channel C3 (p=0.0668, df=9, t=-1.123) nel C3 (F=2.225, df=2, p=0.122) and channel
and channel C4 (p=0.0796, df=9, t=1.671)—as C4 (F=2.554, df=2, p=0.0972).
well as between the recovery session and the The approximate entropy was estimated for
100%-oxygen session for channel C3 (p=0.4162, segments with 5-second duration (500 samples).
df=9, t=2.7348) and for the channel C4 The parameter m was set to 2 and r was set to
(p=0.4822, df=9, t=1.332). ANOVA did not 25% of the standard deviation of each time
331
Figure 4. Time trends of approximate entropy for electrodes C3 and C4 for all participants.
The 100%-oxygen session corresponds to segments 1 to 36, the hypoxia session at a simulated
altitude of 25 feet to segments 37 to 72, and the recovery session to segments 73 to 108. The
red line represents a regression polynomial trend line of order 6.
Figure 5. Mean values of approximate entropy for the experimental sessions of 100% oxygen,
hypoxia, and recovery for all participants and for both EEG channels C3 and C4. A
statistically significant decrease of ApEn was observed during hypoxia.
series. Time trends for the ApEn of all partici- df=9, t=1.9438; Figure 5). A linear regression
pants during all sessions—100% oxygen, hy- analysis of the ApEn time trend was performed
poxia, and recovery—were evaluated, and the for each experimental session and for all par-
mean value and standard deviation of the ApEn ticipants, and the slope of each trend line was
for each EEG segment were calculated at 25. estimated. ANOVA between the slopes of the
Time trends for electrodes C3 and C4 with the three experimental conditions revealed statisti-
mean values of all participants are presented in cally significant differences for C3 (F=5.849,
Figure 4. Paired t-tests between the 100%- df=2, p=0.007753) and for C4 (F=4.3299, df=2,
oxygen session and the hypoxia session re- p=0.0233).
vealed a statistically significant decrease of For the DFA, the values of F(k) were calcu-
ApEn during the hypoxia session for C3 lated for all participants and for all experimental
(p=0.0321, df=9, t=2.10837) and C4 (p=0.0418, sessions. The value of k ranged from 3 to 7. The
332
Figure 6. F(k) vs. k of EEG channel C3 for the mean values of the three experimental sessions
Figure 7. Scatter plot of a2 vs. a 1 for all participants and for the three experimental sessions.
The left panel corresponds to EEG channel C3 and the right one to the EEG channel C4.
corresponding values of F(k) for the channel statistically significant decrease of the a1 slope
C3 are shown in the log-log plot in Figure 6. during the hypoxia session compared to the
Evidently, the striking feature is that there are 100%-oxygen session for channels C3 (p<0.001,
two scaling regions with a discernible bend df=9, t=6.17) and C4 (p=0.0017, df=9, t=3.9363).
when the two slopes in the two regions are No statistically significant differences were
distinctly different. This feature was found in observed between the recovery session and the
both channels for all participants. In order to 100%-oxygen session for both channels.
quantify the scaling behaviour, we perform a ANOVA of the slope a 1 between the three
linear fit in the region for 3.8 < ln k < 4.4 and experimental sessions revealed statistically sig-
denote the slope by a1, and similarly for 4.5 < ln nificant differences for both channels: for C3
k < 5.2 with the slope denoted by a2. The slopes (df=2, F=25.467, p<0.001) and for C4 (df=2,
a1 and a2 for all participants and for the three F=10.674, p<0.001). No statistically significant
experimental sessions are estimated and the differences were observed for the slope a 2.
results of this analysis are presented in Figure
7 for both channels C3 and C4. We observed a
333
CONCLUSION Dimension measurements and Lyapanov

exponents require a sufficient number of data
The results confirm the effectiveness of nonlin- to process and considerable computation time.
ear techniques to identify significant variations Moreover, these measurements are also sensi-
in EEG, which reflect alterations in cerebral tive to the influence of noise. From the statisti-
function induced by cerebral hypoxic condi- cal analysis of the present study, approximate
tions. Parameters measuring the dimension of entropy and DFA seem to overcome these
the state space, like the correlation dimension difficulties. ANOVA between the slopes of the
and the Lyapanov exponents, confirm the pres- three experimental conditions for ApEn and
ence of a low-dimensional system, even if DFA revealed the effectiveness of the scaling
corrupted by a large amount of noise. Statistical exponents of DFA to identify EEG variations
analysis revealed a decrease of the correlation induced by hypoxia.
dimension during the hypoxia session compared We conclude that the parameters arisen
to the 100%-oxygen session, which returned to from nonlinear signal-processing techniques and
its initial levels in the recovery session. It is well obtained from the EEG measurements may be
known that the dimension of EEG time series is effectively employed for the advanced moni-
closely related to the cognitive activity of the toring of cerebral function. The statistical analy-
brain (Bruce, 1990). The correlation dimension sis also suggests that complexity measure-
increases with the degree of cognitive activity. ments such as approximate entropy and detrend
Nonlinear analysis of EEG signals during dif- fluctuation analysis present the highest effec-
ferent mental tasks revealed a decrease of the tiveness to identify EEG variations induced by
correlation dimension during reflexological hypoxia.
stimulation, indicating that the brain is not in-
volved in cognitive tasks or thinking rigorously,
meaning that the brain is in a stoic, stolid, REFERENCES
passive state of relaxation (Natarajan et al.,
2004). Moreover, a reduction of dimension and Abarbanel, H. D. I., Brown, R., Sidorowich, J.
a tendency toward nonchaotic behaviour has J., & Tsimiring, L. S. (1993). The analysis of
been presented as a characteristic of present or observed chaotic data in physical systems.
imminent pathologies (Galka, 2000). This no- Reveiws Modern Physics, 65, 1331-1392.
tion is also supported by the present study, in
Albano, A. M., Muench, J., Schwartz, C., Mees,
which hypoxia can be characterized as a patho-
A. I., & Rapp, P. E. (1988). Singular-value
logical condition that leads to EEG correlation-
decomposition and the Grassberger-Procaccia
dimension decrease. Although, the dimension
algorithm. Physical Review A., 38, 3017.
decrease was statistically significant during
hypoxia, ANOVA for D2 and Lyapanov expo- Babloyantz, A., & Destexhe, A. (1986). Low-
nents did not reveal significant differences dimensional chaos in an instance of epilepsy.
between the three experimental conditions. Proceedings of the National Academy of the
Thus, these two nonlinear measurements did Scinces, 83, 3513-3517.
not reach significance to distinguish the differ-
Babloyantz, A., Salazar, J. M., & Nicolis, C.
ent experimental conditions. This could be at-
(1985). Evidence of chaotic dynamics of brain
tributed to the small number of data processed
activity during the sleep cycle. Physical Let-
in the present study.
ters A, 111, 152-156.
334
Bruce, J. W. (1990). Fractal physiology and Grassberger, P. (1986). Do climatic attractors

chaos in medicine. London: World Scientific exist? Nature, 323, 609-612.
Press.
Grassberger, P., & Procaccia, I. (1983). Char-
Burioka, N., Miyata, M., Cornelissen, G., acterization of strange attractors. Physical
Halberg, F., Takeshima, T., Kaplan, D. T., et al. Review Letters, 50, 346-349.
(2005). Approximate entropy in the electroen-
Henry, B., Lovell, N., & Camacho, F. (2001).
cephalogram during wake and sleep. Clinical
Nonlinear dynamics time series analysis. In M.
EEG and Neuroscience, 36(1), 21-24.
Akay (Ed.), Nonlinear biomedical signal pro-
Cao, L. (1997). Practical method for determin- cessing: Vol. 2. Dynamic analysis and mod-
ing the minimum embedding dimension of a elling (pp. 71-78). Upper Saddle River, NJ:
scalar time series. Physica D, 120, 43-50. John Wiley and Sons.
Das, A., Das, P., & Roy, A. B. (2002). Appli- Huikuri, H. V., Makikallio, T. H., & Perkiomaki,
cability of Lyapunov exponents in EEG data J. (2003). Measurement of heart rate variability
analysis. Complexity International, 9, 1-20. by methods based on nonlinear dynamics. Jour-
nal of Electrocardiology, 36(1), 95-99.
Delorme, A., & Makeig, S. (2004). EEGLAB:
An open source toolbox for analysis of single- Hwa, R. C., & Ferree, T. C. (2002). Fluctuation
trial EEG dynamics including independent com- analysis of human electroencephalography.
ponent analysis. Journal of Neuroscience Nonlinear Phenomena in Complex Systems,
Methods, 134, 9-21. 5(3), 302-307.
Eckmann, J. P., & Ruelle, D. (1985). Ergodic Jung, T. P., Makeig, S., Humphries, C., Lee, T.
theory of chaos and strange attractors. Re- W., Mckeown, M. J., Iraqui, V., et al. (2000).
views of Modern Physics, 57, 617-656. Removing electroencephalographic artifacts by
blind source separation. Psychophysiology,
Fraser, A. M., & Swinney, H. (1986). Indepen-
37, 163-178.
dent coordinates for strange attractors from
mutual information. Physical Review A., 33, Kantz, H., & Schreiber, T. (1995). Dimension
1134-1139. estimates and physiological data. Chaos, 5,
143-154.
Galka, A. (2000). Topics in nonlinear time
series analysis: With implications for EEG Lorenz, E. N. (1963). Deterministic non-peri-
analysis. London: World Scientific Press. odic flow. Journal of Atmospheric Science,
20, 130-141.
Geocadin, R., Ghodadra, R., Kimura, K., Lei,
H., Sherman, D., Hanley, D., et al. (2000). A Marianai, J. J., & Wright, L. A. (1998). Acute
novel quantitative EEG injury measure of global respiratory failure. In G. L. Baum et al. (Eds.),
cerebral ischemia. Journal of Clinical Neu- Textbook of pulmonary diseases (6th ed.,
rophysiology, 111, 1779-1787. Vol. 1, pp. 919-940). Philadelphia: Lippincott-
Raven.
Gloor, P. (1969). Hans Berger on the electro-
encephalogram of man. Natarajan, K., Acharya, U. R., Alias, F.,
Electroencephalography and Clinical Neu- Tiboleng, T., & Puthusserypady, S. K. (2004).
rophysiology, 28, 1-36 Nonlinear analysis of EEG signals at different
335
mental states. Biomedical Engineering Sauer, T., Yorke, J., & Casdagli, M. (1994).
Online, 3(1), 7. Embedology. Journal of Statistical Physics,
65, 579-616.
Parker, T. S., & Chua, L. O. (1987). Chaos: A
tutorial for engineers. Proceedings of IEEE, 75(8), Signorini, M. G., & Cerutti, S. (1999). Nonlin-
982-1008. ear properties of cardiovascular time series. In
M. Di Rienzo, G. Mancia, G. Parati, A. Pedotti,
Peng, C. K., Buldyrev, S. V., Havlin, S., Simons,
& A. Zanchetti (Eds.), Methodology and clini-
M., Stanley, H. E., & Goldberger, A. L. (1994).
cal applications of blood pressure and heart
Mosaic organization of DNA nucleotides.
rate analysis (pp. 282-287). Amsterdam: IOS
Physical Review E, 49, 1685-1689.
Press.
Peng, C. K., Havlin, S., Stanley, H. E., &
Walleczek, J. (Ed.). (2000). Fractal mecha-
Goldberger, A. L. (1995). Quantification of
nisms in neuronal control: Human heartbeat
scaling exponents and crossover phenomena in
and gait dynamics in health and disease. In
nonstationary heartbeat time series. Chaos, 5,
Self-organized biological dynamics and
82-87.
nonlinear control (pp. 66-96). Cambridge,
Pincus, S. M. (1991a). Approximate entropy as UK: Cambridge University Press.
a measure of system complexity. Proceedings
Watters, P. A. (1998). Fractal structure in the
of the National Academy of Sciences of the
electroencephalogram. Complexity Interna-
United States of America, 88, 2297-2301.
tional, 5. Retrieved from http://
Pincus, S. M. (1991b). Heart rate control in www.complexity.org.au/ci/vol05/watters/
normal and aborted-SIDS infants. American watters.html
Journal of Physiology, 33, R638-R646.
Williams, G. W., Luders, H. O., Brickner, A.,
Pincus, S. M. (1994). Quantification of evolu- Goormastic, M., & Klass, D. W. (1985). Inter-
tion from order randomness in practical times observer variability in EEG interpretation. Neu-
series analysis. Methods in Enzymology, 240, rology, 35, 1714-1719.
68-89.
Wolfe, A., Swift, J. B., Swinney, H. L., &
Pincus, S. M. (1996). Older males secrete Vastano, J. A. (1985). Determining Lyapunov
luteinizing hormone and testosterone more ir- exponents from a time series. Physica D, 16,
regularly and joint more asynchronously, than 285-317.
younger males. Proceedings of the National
Academy of Sciences of the United States of
KEY TERMS
America, 93, 14100-14105.
Procaccia, I. (1988). Complex or just compli- Electroencephalography (EEG): The
cated? Nature, 333, 498-499. recording of electric currents created in the
brain by using electrodes placed on the scalp or
Ruelle, D. (1990). Deterministic chaos: The
surface of the brain.
science and the fiction. Proceedings of the
National Academy of the Sciences, 427, 241- Cerebral Hypoxia: The pathological con-
248. dition that is caused by the failure of the human
body’s systems to deliver adequate oxygen to
the brain.
336
Dynamical System: Any system that Fractal Dimension: It is any dimension

evolves over time. Dynamical systems whose measurement that allows noninteger values.
behaviour changes continuously over time are
Nonlinear: A property of a system whose
mathematically described by a coupled set of
output is not proportional to its input. It behaves
first-order autonomous, ordinary differential
in an erratic and unpredictable fashion and is
equations.
unstable.
Entropy: The degree of randomness or
Phase Space: An abstract mathematical
disorder in a system.
space spanned by the dynamical variables of a
system.
337
339
Chapter XLII
Medical and Biomedical Devices
for Clinical Use
Evangelos K. Doumouchtsis
ABSTRACT
Medical technology has been rapidly growing over the last decades. It is characterized by a
constant flow of innovations and a high level of research and development. Many medical and
biomedical devices have changed dramatically the way that medicine diagnoses and treats
human disease, such as getting three-dimensional images of the internal human body. This
chapter describes medical and biomedical devices, the regulatory framework about them, as
well as the most active areas of research of medical technology. It also discusses the future
trends of the medical industry and biosciences that constantly provide new possibilities of
improving health care and patient quality of life.
INTRODUCTION impossible to detect before. Improved healthcare

technology has presented many revolutionary
Medical technology is a science discipline that medical devices that have reduced mortality
has been rapidly growing over the last decades. and morbidity.
It is characterized by a constant flow of innova- Medical devices range from simple ones like
tions and a high level of research and develop- first-aid bandages to more sophisticated ones
ment. Many technological achievements have like positron-emission tomography (PET) scan-
changed dramatically the way that medicine ners. Their main purpose is to improve the
diagnoses and treats human disease. For ex- health status of patients and to support the
ample, the invention of computed tomography prevention, diagnosis, and treatment of disease.
(CT), a noninvasive diagnostic technique, al- There are thousands of medical and bio-
lowed clinicians to get three-dimensional im- medical devices, and this number is rapidly
ages of the inside of the human body, and thus increasing. Therefore, a regulatory framework
they can detect early many diseases that were is essential to ensure the safety and efficiency
Medical and Biomedical Devices for Clinical Use
of the medical devices. In Europe, three direc- • Aids for disabled persons, for example,
tives have been applied by the European Com- wheelchairs, crutches, standing supports,
mission in order to provide the guidelines for the electrical beds, hearing aids, and stoma
development of new medical devices. These appliances
are the 93/42/EE directive on medical devices • Active and nonactive implantable devices,
(MDD), the 90/385/EEC directive on active for example, stents, cardiac pacemakers,
implantable medical devices (AIMDD), and hip implants, neurostimulators, and insulin
the 98/79/EC in vitro diagnostic medical-device pumps
directive (IVDMDD). • Anaesthetic or respiratory equipment, for
example, oxygen masks, anaesthesia
breathing circuits, and gas delivery units
SOME EXAMPLES OF MEDICAL • Orthopaedic devices, for example, knee
DEVICES prostheses, orthopaedic shoes, and spinal
corsets
The 93/42/EEC directive defines a medical • Dental devices, for example, dentistry tools
device as: and drills, alloys and resins, dental floss,
and toothbrushes
[a]ny instrument, apparatus, appliance, • Electromedical and imaging equipment,
material or other article, whether used alone for example, x-ray machines, scanners,
or in combination, including the software electrocardiographs, monitors, lasers, and
necessary for its proper application intended microscopes
by the manufacturer to be used on human • In vitro diagnostics, for example, devices
beings for the purpose of: for clinical chemistry, microbiology, im-
munology, and genetic tests
• diagnosis, prevention, monitoring, • Ophthalmic devices, for example, contact
treatment or alleviation of disease, lenses, optometers, optical lenses, eye
• diagnosis, monitoring, treatment, alle- glasses, and ophthalmoscopes
viation or compensation for an injury • Surgical instruments, for example, scal-
or handicap, pels, surgical drills, forceps, tubes, drains,
• investigation, replacement or modifi- sutures, and masks
cation of the anatomy or of a physi- • Biotechnological products, for example,
ological process, tissue-engineered bones, cartilage, and skin
• control of conception, • medical disposables, for example, ban-
dages, dressing, and syringes
and which does not achieve its principal
intended action in or on the human body by An active implantable medical device
pharmacological, immunological or (AIMD) according to the 90/385/EEC directive
metabolic means, but which may be assisted is a medical device as defined above that is at
in its function by such means. the same time both active and implantable.
A medical device is active if it “relies for its
According to the Global Medical Devices functioning on a source of electrical energy or
Nomenclature (GMDN), the product range in- any source of power other than that directly
cludes the following categories: generated by the human body or gravity.” This
340
includes, for instance, devices activated by AREAS OF MEDICAL DEVICES:

means of pressure, unless this effect is achieved NEW TECHNOLOGIES
by energy resulting from the body of the pa-
tient. The definition implies that the function of The most active areas of medical-device re-
the device involves using a source of power to search are given in the following sections.
perform useful work. The mere transmission of
heat, light, pressure, or vibration does not mean Diagnostic Imaging
that a device is active.
An active medical device is defined as im- Diagnostic imaging systems such as x-ray and
plantable if it is “totally or partly introduced, ultrasound have been in use for decades; other
surgically or medically, into the human body or systems, including computed tomography, mag-
by medical intervention into a natural orifice, netic resonance imaging (MRI), and nuclear or
and which is intended to remain after the pro- positron-emission tomography, are newer
cedure.” technologies. Recently, a revolutionary four-
Examples of AIMDs are as follows: dimensional CT scanner was developed, and it
brings new possibilities to diagnosis and treat-
• Implantable pulse generator for pacing ment. Three-dimensional technology only pro-
including an electrode vides a static image at one instance in time.
• Implantable pulse generator without an However, when dealing with parts of the body
electrode that move, such as the abdomen, pelvis, or
• An electrode chest, there is the need for a technology that
• Implantable drug-administration device accounts for time as well. This scanner adds
with or without a catheter the fourth dimension of time, allowing clinicians
• Catheter for implantable drug-administra- to track moving organs almost in real time and,
tion device for example, to treat tumors by radiation therapy
more precisely while limiting the exposure of
An in vitro diagnostic medical device is any healthy tissue to radiation (Fontanazza, 2004).
medical device that is a reagent, reagent prod-
uct, calibrator, control material, kit, instrument, Cardiology
apparatus, piece of equipment, or system,
whether used alone or in combination, intended Cardiovascular disease is a major cause of
by the manufacturer to be used in vitro for the death across Europe and the USA. Intervention
examination of specimens, including blood and cardiology, such as coronary angiography, coro-
tissue donations derived from the human body, nary stents, arrhythmia, and stroke manage-
solely or principally for the purpose of providing ment, provides an example of the contribution
information related to the following: of medical technology to patient care. There
has been a continuous trend of innovation in
• Concerning a physiological or pathologi- these areas. Advanced techniques, such as
cal state cardiac rhythm management (CRM), help to
• Concerning a congenital abnormality avoid complications and improve the quality of
• To determine the safety and compatibility life for patients.
with potential recipients CRM consists of the following three
• To monitor therapeutic measures (98/79/EC) key areas:
341
• Pacing systems eluting stents may be beneficial to certain high-

• Implantable cardiac defibrillators (ICDs) risk patients (MDDI, 2004).
• Automatic external defibrillators (AEDs) One trend driving growth in the device in-
dustry is the ability to give patients more control
Pacing systems consist of pacemakers and over their own healthcare. The FDA (Food and
programmers that are needed to monitor and Drug Administration) approved over-the-
correct slow heart rhythms. The following pacing- counter sales of a home defibrillator (Swain,
system components work together to relieve 2005). Another device for home use is a re-
symptoms of bradycardia (slow heart rate). cently developed biosensor that monitors a
couple’s vital signs as they sleep, notifying one
• Pacemaker: Implantable device that elec- person when it detects abnormal signs in the
trically stimulates the heart to pump other (Engel & Cook, 2005).
blood. It contains an electronic circuit and
a battery. Endocrinology
• Pacing Lead: An insulated wire that
carries a tiny electrical pulse to the heart Endocrinology is another important area in which
to initiate heartbeat. medical devices play a vital role in patient care.
• Programmer: Monitors the pacemaker. It Devices for diabetes treatment, for example,
is usually kept in the hospital or a clinic. include insulin-delivery products such as sy-
ringes, pens, automatic injectors, insulin patches,
Implantable Cardiac Defibrillators (ICDs) and external or implantable pumps. Precise blood
are a new technology introduced in 2004 (MDDI, glucose monitoring can substantially reduce the
2004). They are similar to pacing systems in risk of developing complications and slow the
that they continuously monitor the heart’s progression of the disease. Many diabetic pa-
rhythm. Specifically, ICDs treat tachyarrhythmia tients also rely on sophisticated dialysis equip-
(fast heart beat). If the heart beats too ment (Eucomed Medical Technology, 2004).
quickly, the ICD issues a lifesaving jolt of elec- A characteristic example is monitoring glu-
tricity to restore the heart’s normal rhythm and cose levels in blood to regulate an infusion of
prevent sudden cardiac death. Like pacing insulin. The FDA approved a hybrid device that
systems, ICDs have similar components, in- included a wireless link between an insulin
cluding a defibrillator, lead, and a programmer. pump and a blood glucose monitor (Wilson,
An AED is used to jump-start a heart that has 2004). This, in combination with an insulin
failed due to a heart attack or cardiac dosing controller, could be described as an
arrest. Designed so that even nonexperts can artificial pancreas.
operate them without much difficulty, AEDs are
rapidly gaining popularity. They can even be found Electromedical Devices
in public places such as airports, train stations, and
airplanes (Altera, http://www.altera.com). The electromedical segment comprises the fol-
In 2004, the industry provided many other lowing three key areas:
new devices that offer improved clinical utility
in cardiology. Drug-eluting coronary stents are • Patient monitoring
one of the most significant technologies of the • Ventilation and life support
year. A number of studies showed that drug- • Anesthesia
342
Patient-monitoring equipment capture and bilities for improving healthcare and patient
analyze a patient’s vital information for clinical quality of life (Albrecht, 2004).
decision makers. Advances in patient monitor-
ing include new form factors that facilitate
patient transport. THE FUTURE OF
Ventilation and life support is another im- MEDICAL DEVICES
portant area in the medical sector. These equip-
ment primarily consist of ventilators and drug- The field of research on medical technology
delivery systems. These are very tightly inte- constantly expands. There are already hun-
grated with the central monitoring system. dreds of revolutionary medical devices under
Anaesthesia is a life-critical application that development. Some examples of the medical
needs the utmost human attention, especially devices of the future, expected to open new
while delivering it to the patient. Technology horizons to medicine, are listed as follows:
plays a key role in anaesthesia-delivery equip-
ment by delivering the precise dosage to the • Neurostimulation devices
patient (Altera, http://www.altera.com). • Nanoscale biosensors
• Tissue-engineered components
Life-Science and • Less invasive hip replacements
Hospital Equipment • Advanced biomaterials
• Ultrahigh-resolution imaging systems
Life-science and hospital equipment such as • Drug-device hybrids
spectrometers, centrifuges, protein analyzers, • Biosensors
powered beds, surgical instruments, radiation • Nanodevices
equipment, and endoscopes are some examples • Virtual reality
of medical devices that are used in healthcare. • Biorobotics
• Brain-machine interfaces
Wireless Technologies • The computerization of medicine
(Bronzino, 2004)
Many new medical devices use more and more
wireless technologies, such as Bluetooth. Sev-
eral device manufacturers have already imple- CONCLUSION
mented Bluetooth technologies and received
FDA approval. Some examples are a wireless Devices of the future will incorporate informa-
defibrillator and monitor that allows the trans- tion technology, nanotechnology, and bio-
mittal of vital-sign reports to a PDA (personal sciences. In the near future, a convergence
digital assistant), a pulse oximeter that offers between engineering and biology is expected.
patient mobility within 10 meters of the moni- Medical technology should develop new skills
toring device, and a wireless device that is used in molecular and cell biology in order to corre-
for monitoring patient weight and blood pres- spond to the demands of medical science. As
sure in the home while data is transmitted to a industry trends move toward smaller, lower
home gateway for remote retrieval and review. power, sensor-driven devices, new sciences
All these technologies provide amazing possi- such as protein-based therapies present new
targets and new opportunities (Conroy, 2005).
343
REFERENCES Fontanazza, M. (2004). Imaging technology

moves to the fourth dimension. Medical De-
Albrecht, M. (2004). Bluetooth and medical vice & Diagnostic Industry, 26. Retrieved
devices. Breakthrough Technologies, LLC. from http://www.devicelink.com/mddi/archive/
Retrieved from http://www.breaktech.com/ 04/12/010.html
content.php?article.60
IBM. (n.d.). Medical technology equipment
Bronzino, J. (2004). The medical device in- modality definitions/glossary. Retrieved from
dustry in Southern New England’s I-91 cor- http://www-1.ibm.com/financing/pdf/us/igf4-
ridor: Potential for growth and the role of a042.pdf
Beacon. Retrieved from http://www.cbia.com/
MDDI. (2004). Focus on technology: Providing
business/Presentations/2004/Sept/bronzio%20
improved utility at reduced costs. Medical
presentation.ppt
Device & Diagnostic Industry. Retrieved from
Conroy, S. (2005). Devices of the future must http://www.devicelink.com/mddi/archive/04/12/
incorporate nanotech and biotech. Medical 016.html
Device & Diagnostic Industry, 32. Retrieved
Swain, E. (2005). Making news in 2004. Medi-
from http://www.devicelink.com/mddi/archive/
cal Device & Diagnostic Industry, 63. Re-
05/01/028.html
trieved from http://www.devicelink.com/mddi/
Directive 98/79/EC of the European Parlia- archive/05/02/004.html
ment and of the Council on in vitro diagnostic
Wilson, D. (2004). MedEdge 2004: Conver-
medical devices: L 331/5. (1998). Official Jour-
gence of life sciences and medical devices.
nal of the European Communities, 331, 5.
Breakthrough Technologies, LLC. Retrieved
Dorland’s pocket medical dictionary. (1998). from http://www.breaktech.com/content.php?
W. B. Saunders Company. article.63
Engel, D., & Cook, C. (2005). Biosensor plat-
form offers real-time monitoring of vital signs. KEY TERMS
European Medical Device Manufacturer.
Retrieved from http://www.devicelink.com/ Computerized Tomography: A CT scan
emdm/archive/05/01/002.html is an x-ray procedure that is enhanced by a
computer. This results in a three-dimensional
Eucomed Medical Technology. (2004). Medi-
view (referred to as a slice) of a particular part
cal technology brief. Retrieved from http://
of the body. Typical applications include view-
www.eucomed.be/docs/Brief%202004%20
ing the chest, abdomen, and spinal cord.
Final.pdf
Coronary Stent: A coronary stent is an
European Commission DG Enterprise Direc-
artificial support device that is inserted in the
torate G. (1994). Unit 4: Pressure equipment,
coronary artery to keep the vessel open.
medical devices, metrology. Guidelines re-
lating to the application of the council di- Defibrillator: An electronic apparatus used
rective 90/385/EEC on active implantable to counteract atrial or ventricular fibrillation by
medical devices: The council directive 93/ the application of a brief electroshock to the
42/EEC on medical devices. heart, either directly or through electrodes placed
on the chest wall.
344
MRI Scanner: Magnetic resonance imag- that have been injected into the body to provide
ing is an imaging technique used primarily in information on function rather than structure,
medical settings to produce high-quality images and to help differentiate normal tissue from
of the inside of the human body. MRI produces cancer.
images that are the visual equivalent of a slice
Pulse Oximeter: An oximeter that mea-
of the anatomy, and it is also capable of produc-
sures the oxygen saturation of arterial blood by
ing those images in an infinite number of projec-
passing a beam of red and infrared light through
tions through the body. To produce its images,
a pulsating capillary bed, the ratio of red to
MRI uses radio frequencies, a computer, and a
infrared transmission varying with the oxygen
large magnet that surrounds the patient.
saturation of the blood; because it responds
PET Scanner: PET stands for positron- only to pulsatile objects, it does not detect
emission tomography and is a method of body nonpulsating objects like skin and venous blood.
scanning that detects radioactive compounds
345
346
Chapter XLIII
Artificial Intelligence in
Medicine and Biomedicine
Athanasios Zekios
Dimitra Petroudi
ABSTRACT
Man has always strived to augment his abilities by inventing tools. Artificial intelligence in
medicine (AIM), has taken up the challenge of creating and distributing advanced tools,
utilising technical developments aimed at augmenting man’s reasoning. Increasing quality
healthcare needs and advances in medical and pharmaceutical sciences, yet restrictions on
physicians’ time for learning while practicing, indicate these tools will prove invaluable in
effecting changes (i.e. Simpler organising, storing, and retrieving of important medical facts/
new findings) especially when treating difficult cases; continual availability of same for
learning purposes; assisting with appropriate diagnostic, prognostic and therapeutic decisions/
decision making techniques, using databases, flowcharts and decision theory. Proof of these
tools’ indispensability through actual trials, is pending.
INTRODUCTION the threshold of new technical developments

that will augment people’s reasoning; the com-
Men and women strive to augment their abili- puter and the programming methods being de-
ties by building tools. From the invention of the vised for this are the new tools to effect this
club to lengthen their reach and strengthen their change.
blows to the refinement of the electron micro- Medicine and biomedicine are fields in which
scope to sharpen their vision, tools have ex- the help of such tools is critically needed. Our
tended humans’ ability to sense and to manipu- increasing expectations of the highest quality
late the world about them. Today we stand on healthcare and the rapid growth of ever more
Artificial Intelligence in Medicine and Biomedicine
detailed medical knowledge leave the physician from the introduction to an early collection of
without adequate time to devote to each case AI papers.
and struggling to keep up with the newest
developments in the field. There is also another Is it Possible for Computing
huge problem. Continued training and recertifi- Machines to Think?
cation procedures encourage the physician to
keep more of the relevant information con- If one defines thinking as an activity peculiarly
stantly in mind, but fundamental limitations of and exclusively human then the answer is no.
human memory and recall coupled with the Any such behaviour in machines, therefore,
growth of knowledge assure that most of what would have to be called thinking-like behaviour.
is known cannot be known by most individuals. If someone postulates that there is something in
Currently, there is the opportunity for new the essence of thinking which is inscrutable,
computer tools to help organize, store, and mysterious, and mystical then the answer is
retrieve appropriate medical knowledge needed again no. But it the opposite answer if one
by the practitioner in dealing with each difficult admits that the question is to be answered by
case, and to suggest appropriate diagnostic, experiment and observation, comparing the
prognostic, and therapeutic decisions and deci- behaviour of the computer with that behaviour
sion-making techniques. of human beings to which the term “thinking” is
A field that is now taking up the challenge of generally applied.
creating and distributing the tools mentioned
above is artificial intelligence in medicine (AIM). ARTIFICIAL INTELLIGENCE IN
MEDICINE
ARTIFICIAL INTELLIGENCE AIM is AI specialized to medical applications.

Researchers in AIM need not engage in the
What is artificial intelligence in medicine? One controversy introduced above. Although we
introductory textbook defines artificial intelli- employ humanlike reasoning methods in the
gence this way: “Artificial Intelligence is the programs we write, we may justify that choice
study of ideas which enable computers to do the either as a commitment to a human-computer
things that make people seem intelligent. The equivalence sought by some or as a good engi-
central goals of Artificial Intelligence are to neering technique for capturing the best-under-
make computers more useful and to understand stood source of existing expertise on medicine:
the principles which make intelligence pos- the practice of human experts. Most research-
sible.” This is a rather straightforward defini- ers adopt the latter view.
tion, but it embodies certain assumptions about Another currently much smaller use of com-
the idea of intelligence and the relationship puters in medicine is their application to the
between human reasoning and computation, substance rather than the form of healthcare. If
which are, in some circles, quite controversial. the computer is a useful manager of billing
Historically, researchers in AI have had to records, it should also maintain medical records,
defend this linkage against humanist attacks on laboratory data, data from clinical trials, and so
the reduction of the human intellect to compu- forth. And if the computer is useful to store
tational steps. The debate has sometimes been data, it should also help to analyze, organize,
heated, as exemplified by the following quote and retrieve it. Three main approaches to this
347
second type of medical computing have so far tic and therapeutic program for glaucoma and
been used. related diseases of the eye that describes EX-
PERT, a somewhat simpler and more widely
Flowcharts applied system that is being used in the analysis
of thyroid disorders and in rheumatology.
A flowchart is conceptually the simplest deci- CASNET identified the fundamental issue of
sion-making tool. It encodes, in principle, the causality as essential in the diagnostic and
sequences of actions a good clinician would therapeutic process.
perform for any one of some population of The MYCIN system was developed at
patients. Stanford University originally for the diagnosis
and treatment of bacterial infections of the
Databases blood and later extended to handle other infec-
tious diseases as well. The fundamental insight
Large databases of clinical histories of patients of the MYCIN investigators was that the com-
sharing a common presentation or disease are plex behaviour of a program, which might re-
now being collected in several fields. The growth quire a flowchart of hundreds of pages to
of data capture and storage facilities and their implement as a clinical algorithm, could be
co-occurring decline in cost make attractive the reproduced by a few hundred concise rules and
accumulation of enormous numbers of cases, a simple recursive algorithm (described in a
both for research and clinical uses. one-page flowchart) to apply each rule just
when it promised to yield information needed by
Decision Theory another rule.
INTERNIST-I uses a problem-formulation
Decision theory is a mathematical theory of heuristic to select from among all its known
decision making under uncertainty. It assumes diseases the set that should be considered as
that one can quantify the a priori and conditional competing explanations of the currently known
likelihoods of existing states and their manifes- abnormal findings in a case. A distinction is
tations, and can similarly determine an evalua- made between the tasks of formulating such a
tion (utility) of all contemplated outcomes. differential problem and of solving it. Formulat-
ing the problem is what might be called an ill-
structured task, similar to the problem of mak-
SYSTEMS REFER TO AIM ing up an interesting mathematical theorem or
designing a house; solving the differentiation
The Present Illness Program (PIP) is able to problem once formulated is well structured,
infer that if a patient passed a military physical inviting the application of numerous conven-
or a life-insurance company’s health examina- tional methods. The simple heuristic of INTER-
tion, then neither blood, sugar, nor protein was NIST-1 is seen to do well in many complex
present at that time in the urine. This is a widely cases, but falters in cases requiring an analysis
known heuristic among physicians, being one of from several different viewpoints, for example,
the many ways that past data can be inferred in an interaction between the causal mechanism
the absence of definitive reports. of the disease and the organ systems involved
The CASNET system, developed at Rutgers in it. Based on such deficiencies, a new, ex-
University, in its major incarnation is a diagnos- tended method of medical knowledge repre-
348
sentation and problem formulation is presented 1. Human cognition is still superior to ma-
that is intended to form the basis for CADU- chine intelligence.
CEUS, the second-generation follow-up to IN- 2. Decisions about whether to treat a given
TERNIST-1. patient are often value laden and must be
made relative to treatment goals.
3. Applying computational operations on ag-
PROBLEMS gregate data to individual patients runs the
risk of including individuals in groups they
The significant questions facing the field of resemble but to which they do not actually
artificial intelligence in medicine are “Who is an belong.
appropriate user of a healthcare-related com-
puter application?” and “How and when should In regard to the last problem, clinicians run
computers be used in clinical practice?” the risk of including individuals in the wrong
Perhaps one of the more difficult questions groups all the time. It is a long-standing logic
to answer is “Who should use a healthcare- challenge, trying to infer correctly that an indi-
related computer application?” One of the early vidual belongs in a particular set, group, or
papers on ethical issues in informatics reported class. Computers have not been able to solve
that the potential users of informatics systems this problem, yet.
included physicians, nurses, physicians’ assis-
tants, paramedical personnel, students of the
health sciences, patients, and insurance and FUTURE OF ARTIFICIAL
government evaluators. As discussed among INTELLIGENCE IN MEDICINE
the major groups of people involved in medical-
informatics ethics, other groups that should be The field of artificial intelligence in medicine
included are nurse practitioners, pharmacists, has been slow to make its mark on medicine;
managers, administrators, scientists, research- however, this may soon change. With the em-
ers, applied computer professionals, other an- phasis in medicine shifting to more evidence-
cillary healthcare personnel, and patients and based practice, the increasing reliance on com-
their employers. puters, the increasing volume of information for
Trying to determine who should be allowed clinicians to assimilate, and the many pressures
to use a healthcare-related computer applica- to practice medicine more efficiently, those in
tion will be an ethical challenge, one that is the field of AIM may find themselves thrust into
already under way. Use by physicians in prac- the forefront of medicine as they will be provid-
tice, and medical and nursing students seems ing computer-based solutions for this ever-
plausible. However, before using a diagnostic changing field.
decision-support system, the user must be able
to recognize when there is an error and when it
is providing accurate information. POTENTIAL OF ARTIFICIAL
The key problems encountered when using INTELLIGENCE IN MEDICINE
clinical computer programs to determine policy
or aid in practice are the following.: The potential of AI in medicine has been ex-
pressed by a number of researchers. The po-
349
tential of AI techniques in medicine are as puter program that considers clinical responses
follows: (a) They provide a laboratory for the to digitalis. Amer. J Med, 64, 452-460.
examination, organization, representation, and
Morris, A. H. (2000). Developing and imple-
cataloguing of medical knowledge, (b) produce
menting computerized protocols for standard-
new tools to support medical decision making,
ization of clinical decisions. Ann Intern Med.,
training, and research, (c) integrate activities in
132, 373-383.
medical, computer, cognitive, and other sci-
ences, and d) offer a content-rich discipline for Pauker, S. G., Gorry, G. A., Kassirer, J. P., &
future scientific medical specialties. Schwartz, W. B. (1976). Toward the simulation
of clinical cognition: Taking a present illness by
computer. Amer. J Med, 60, 981-995.
CONCLUSION
Schwartz, W. B. (1970). Medicine and the
computer: The promise and problems of change.
We must realize that although current AIM
New Engl. J. Med., 283, 1257-1264.
programs already give quite impressive demon-
strations of the success of the techniques used Shortliffe, E. H., et al. (1979). Knowledge
and of the dedication of the investigators, none of engineering for medical decision making: A
the programs reported on here or developed by review of computer-based clinical decision aids.
other similar efforts is in current clinical use. Proceedings of the IEEE, 67(9), 1207-1224.
Perhaps, as it has been argued, programs will
Slack, W. V., & Van Cura, L. J. (1968). Patient
only be clinically accepted once their indispens-
reaction to computer-based medical interview-
ability is established, and only when successful
ing. Comput. Biomed Res, 1, 527-531.
demonstrations exist that physicians or other
medical personnel working with such programs Szolovits, P., & Pauker, S. G. (1978). Categori-
are more successful than those working without cal and probabilistic reasoning in medical diag-
them. Alternatively, social and administrative nosis. Artificial Intelligence, 11, 115-144.
mechanisms may be more responsible for the
Szolovits, P., & Pauker, S. G. (1979). Comput-
ultimate utilization or abandonment of these tools.
ers and clinical decision making: Whether, how,
and for whom? Proceedings of the IEEE,
67(9), 1224-1226.
REFERENCES
Tautu, P., & Wagner. G. (1978). The process
Bleich, H. L. (1972). Computer-based consul- of medical diagnosis: Routes of mathematical
tation: Electrolyte and acid-base disorders. investigations. Meth, Inform. Med., 7(1).
Amer. J. Med., 53, 285.
Coiera, E. W. (1996). Artificial intelligence in URL REFERENCES
medicine: The challenges ahead. J Am Med
Inform Assoc, 3, 363-366. http://compubiosys.medsch.ucla.edu/AI/
Doyle, J. (1978). A truth maintenance system. http://www.abcnews.go.com/ABC2000/
Artificial Intelligence, 12, 231-272. abc2000tech/geek38.html
Gorry, G. A., Silverman, H., & Pauker, S. G. http://www.coiera.com/ailist/list.html
(1978). Capturing clinical expertise: A com-
350
http://www.coiera.com/aimd.html Biomedicine: The branch of medical sci-

ence that applies biological and physiological
http://www.cs.washington.edu/research/jair/
principles to clinical practice.
home.html
Clinical Decision-Support System: Ac-
http://www.csd.abdn.ac.uk/research/
tive knowledge systems that use two or more
ai_in_medicine.html
items of patient data to generate case-specific
http://www.elsevier.com/locate/artmed advice.
http://www.hi-europe.info/files/2002/9980.htm Database: An organized body of related
information.
http://www.journeyofhearts.org/jofh/jofh_old/
minf_528/ai.htm Decision Theory: A branch of statistics
concerning strategies for decision making in
http://www.medg.lcs.mit.edu/ftp/psz/AIM82/
nondeterministic systems. Decision theory seeks
ch1.html#c1_definitions
to find strategies that maximise the expected
http://www.openclinical.org/aiinmedicine.html value of a utility function measuring the desir-
ability of possible outcomes.
KEY TERMS Expert System: Computer programs that

use artificial-intelligence strategies such as
Artificial Intelligence: A collection of symbolic representation, inference, and heuris-
mathematical and computing methods for pre- tic search to perform sophisticated tasks once
dicting complex real-world processes, such as thought possible only for human experts.
the behaviour of complicated games, like chess;
Flowchart: A diagram of the sequence of
the behaviour of experts, who might predict
operations in a computer program.
future real-estate values, render medical diag-
noses, or translate foreign languages; the
behaviour of complex biological systems; or
ordinary human behaviour, such as perceiving
objects visually or auditorily.
351
352
Chapter XLIV
Comparative Genomics and
Structure Prediction in
Dental Research
Andriani Daskalaki
Jorge Numata
Free University, Germany
ABSTRACT
Since the completion of the Human Genome Project (HGP) in 2003, the 3.2 billion basepairs
which make up the human genome have been sequenced. These sequences contain the plan for
the mechanisms controlling the behavior of each cell. The small variations in the DNA
sequence that lead to different characteristics, such as facial features, or color, are known
as polymorphisms, which also can cause oral diseases. Periodontitis is a chronic infective
disease of the gums caused by bacteria present in dental plaque. Severaazl techniques have
been developed to regenerate4 periodontal tissues including guided tissue regeneration
(GTR), and the use of enamel matrix derivative (EMD). EMD is an extract of enamel matrix and
contains amelogenins. This is evidence to show that amelogenins are involved not only in
enamel formation, but also in the formation of the periodontal attachment during tooth
formation. Comparative sequence analysis is an approach for detecting functional regions in
genomic and protein sequences. Motifs, conserved domains, secondary structure
characteristics, and functional sites of proteins related to oral health may be compared,
revealing the degree of sequence conservation during vertebrate evolution. Secondary and
tertiary structures are important in understanding the function of a protein. In a comparative
sequence analysis, the most well-known bioinformatics tools that are used are: basic local-
alignment search tool (BLAST), multiple-sequence alignment software (ClustalW), and PROSITE,
a database of proten families and domains. The PROSITE database consists of biologically
significant sites, patterns, and profiles that help to reliably identify to which known protein
Comparative Genomics and Structure Prediction in Dental Research
family a sequence belongs. Phylogeny Inference Package (PHYLIP) can be used for building
phylogenetic trees and a Python-enhanced molecular graphics program (PyMOL) for 3D
visualization of proteins.
INTRODUCTION The extracellular matrix of dentin primarily

consists of Type I collagen, noncollagenous
Dental researchers collaborating with matrix proteins, and proteoglycans.
bioinformaticians have achieved advances in Amelogenin and osteocalcin are
oral-health research by actualizing the impact noncollagenous matrix proteins secreted by the
of genetics in oral health. With the help of ameloblasts and odontoblasts, respectively.
bioinformatics, a spectrum of questions in den- These proteins primarily function in enamel
tistry can be addressed. mineralization.
Comparative genomics approaches are used Krishnaraaju et al. (2003) have used
to identify the functional domains of a protein bioinformatics tools for multiple-sequence
and suggest similarities for assigning 3-D struc- analysis of these proteins in different species.
tures by homology modeling. It is then possible Phylogenetic analysis using sequence data
to use classical molecular dynamics simulations is used to study sequence relatedness.
to account for the dynamic behavior in protein
function. Multiple-sequence analysis of pro-
teins in different species reveals the degree of STRUCTURE PREDICTION
sequence conservation at the nucleotide and
protein levels. Motifs, conserved domains, sec- Secondary and tertiary structures are impor-
ondary structure characteristics, and functional tant in understanding the function of a protein.
sites of proteins related to oral health may be Frequently, however, such information is not
compared, revealing the degree of sequence available because neither crystallographic nor
conservation during vertebrate evolution. nuclear magnetic resonance (NMR) structure
Three-dimensional structure predictions determination has been carried out. In this
developed by the modeling of conserved do- case, structure-prediction methods may help.
mains of proteins support a key role for specific Homology-based methods are not perfect, and
residues in processes like mineralization. depend on the following:
1. One or more known crystal or NMR 3-D

COMPARATIVE GENOMICS structures
2. Strong sequence similarity of the unknown
Comparative sequence analysis is an approach structures (>25%)
for detecting functional regions in genomic and
protein sequences. It facilitates the identifica- The secondary structure elements may be
tion of conserved domains, motifs, and distantly predicted with good accuracy. However, side-
related sequences of different organisms, and chain rotameters and loop insertions may be far
provides evolutional insights into the underlying from reality.
biology of organisms (Rubin et al., 2000).
353
Figure 1. Three-dimensional structure of osteocalcin. The graphic was generated by a PyMOL

viewer. Calcium (CA) binding with gamma-carboxyglutamic acid (Cgu) is shown. Residues
forming the alpha-helices (secondary structure) are highlighted with ribbons.
If the sequence or structural similarity is ers motifs in the sequence by a search of

established between the protein of interest and the protein database.
the sequence with a known 3-D structure, it is 4. PROSITE: It identifies posttranslationally
possible to predict the 3-D structure in the modified sites such as phosphorylation,
conserved domain database. The 3-D structure glycosylation, and N-myristoylation on
(Figure 1) provides evidence that the protein homologous sequences.
osteocalcin is involved in mineralization in den- 5. PHYLIP: Phylogeny Inference Package.
tal and bone structures by similar mechanisms. A phylogenetic tree can be generated to
find the closely related organisms from
multiple-sequence alignment.
BIOINFORMATIC TOOLS 6. PyMOL: PyMOL is a molecular graph-
ics program with an embedded Python
1. BLAST: Basic local-alignment search interpreter designed for the real-time vi-
tool. This tool is used to find protein se- sualization and rapid generation of high-
quences that are similar to query in the quality molecular graphics images and ani-
protein database. mations.
2. CLUSTALW: Multiple-sequence align-
ment software. In the alignment, similarity
among amino acids can be determined CONCLUSION
based on specific alignment parameters
(for example, BLOSUM matrices). Predicting the three-dimensional structures of
3. MEME: Multiple-expectation maximiza- protein from sequence data by comparative
tion for motif elicitation. This tool discov- modeling provides information on which ex-
354
periments can be planned. If the sequence of study of oral oncology. Adv Dent Res, 17, 104-
structural similarity is established between the 108.
target (protein of interest) and the template
Sigrist, C. J., Cerutti, L., Hulo, N., Gattiker, A.,
(sequence for which the 3-D structure is known),
Falquet, L. Pagni, M., et al. (2002). PROSITE:
it is possible to predict the 3-D structure of a
A documented database using patterns and
protein or domain using publicly available
profiles as motif descriptors. Brief
sources (NCBI). Furthermore, methods like
Bioinformatics, 3, 265-274.
energy scoring functions, loop building, homol-
ogy modeling, and energy minimization can be
used. LINKS
http://en.wikipedia.org
REFERENCES
http://meme.sdsc.edu/meme/meme-intro.html
Actis, L. A., Rhodes, E. R., & Tomaras, A. P. http://pymol.sourceforge.net/
(2003). Genetic and molecular characterization
http://www.dhgp.de/info/lexica/dictionary.html
of a dental pathogen using genome-wide ap-
proaches. Advances in Dental Research, 17, http://evolution.genetics.washington.edu/phylip/
95-99. general.html
Altschul, S. F., Gish, W., Miller, W., Myers, E. http://lectures.molgen.mpg.de/Algorithmische_
W., & Lipman, D. J. (1990). Basic local align- Bioinformatik_WS0405/material/Steinke_
ment search tool. Journal of Molecular Biol- lecture_19_1.pdf
ogy, 215, 403-410.
http://www.ncbi.nlm.nih.gov/Education/
Bailey, T. L., & Elkan, C. (1995). The value of BLASTinfo/glossary2.html
prior knowledge in discovering motifs with
MEME. Proceedings of the International
Conference on Intelligent Systems and Mo- KEY TERMS
lecular Biology, 3, 21-29.
Alignment: The process of lining up two or
Krishnaraju, R. K., et al. (2003). Comparative more sequences to assess the degree of simi-
genomics and structure prediction of dental larity and homology.
matrix proteins. Adv Dent Res, 17, 100-103.
BLAST (Basic Local-Alignment Search
Moradian-Oldak, J. (2001). Amelogenins: As- Tool): A sequence-comparison algorithm opti-
sembly, processing and control of crystal mor- mized for speed and used to search sequence
phology. Matrix Biology, 20, 293-305. databases for optimal local alignments to a
query.
Rubin, G. M., Yandell, M. D., Wortman, J. R.,
Gabor Miklos, G. L., Nelson, C. R., Hariharan, Conservation: A high degree of similarity
I. K., et al. (2000). Comparative genomics of in the structure of homologous proteins amongst
the eukaryotes. Science, 287, 2204-2215. various phyla. This is seen as an indication of its
importance in cellular function.
Sebastiani, P. (2003). Bayesian machine learn-
ing and its potential applications to the genomic
355
Domains: Portions of a protein assumed to Nuclear Magnetic Resonance (NMR):

fold independently and possessing their own A physical phenomenon based upon the mag-
functions. netic property of an atom’s nucleus. NMR
spectroscopy can provide detailed information
Glycosylation: Process or result of the
on the exact three-dimensional structure of
addition of saccharides to proteins and lipids.
biological molecules in a solution.
The process is one of four principal posttrans-
lational modification steps in the synthesis of Phosphorylation: The addition of a phos-
membrane and secreted proteins. phate (PO4) group to a protein or a small
molecule.
Homologous: Two or more structures are
said to be homologous if they are alike because PHYLIP: The Phylogeny Inference Pack-
of shared ancestry from a common ancestor age is a package of programs for inferring
(evolutionary), or because they are from the phylogenies (evolutionary trees). Methods avail-
same tissue in embryonal development (devel- able in the package include parsimony, distance
opmental ancestry). matrix, likelihood methods, bootstrapping, and
consensus trees.
MEME: A software tool for discovering
motifs in a group of related DNA (deoxyribo- Posttranslational Modification: Chemi-
nucleic acid) or protein sequences. MEME cal modification of a protein after its transla-
uses statistical modeling techniques to auto- tion.
matically choose the best width, number of
Protein: A protein consists of amino-acid
occurrences, and description for each motif.
chains. Proteins play a key role in most biologic
Motif: A sequence pattern that occurs re- processes.
peatedly in a group of related proteins or genes.
PyMOL: A molecular visualization system.
Myristoylation: A posttranslational pro-
Sequence: A sequence defines the order
tein modification. It is catalyzed by the enzyme
of nucleotides in the DNA or RNA (ribonucleic
N-myristoyltransferase and occurs on glycine
acid), or the order of amino acids in a protein.
residues exposed during cotranslational N-ter-
minal methionine removal.
356
357
Chapter XLV
Genomic Databanks for
Biomedical Informatics
Andrea Maffezzoli
Politecnico di Milano, Italy
Marco Masseroli
Politecnico di Milano, Italy
ABSTRACT
In the area of medical informatics, the recent ICT (information and communication technology)
tools and systems supporting knowledge on sciences involved in the study of genes,
chromosomes, and protein’s expression level in various organisms, that is genomics and
proteomics, are becoming necessary to develop new prospects for the comprehension of
mechanisms lying at the base of biological processes which cause a disease. This can allow
more effective diagnostic and treatment methods and also personalized pharmacological
therapies. At this purpose, the mutual intervention of different sciences, such as biology,
medicine, engineering, informatics and mathematics, becomes an indispensable step: The
development of a science embracing all these fields is identified in bioinformatics, which was
conceived for the analysis, storage and processing of huge amount of biological data. The
achievement of all the aforementioned operations involves the creation of the so-called
genomic or proteomic databanks, which represent a major source of information on nucleotide
sequences, as well as biological, clinical, physiological and bibliographical annotations
related to singular sequences. There are different types of databanks based on their peculiar
characteristics and features (such as primary and derivative or specialized databanks), and
several ways to access data stored in these databanks; there are also specific bioinformatics
databank-based tools developed to perform searching operations and to extract significant
information, in order to summarize and compare gene annotations related to the causes of a
disease and finally to identify a list of the most significant genes as cause of disease.
Genomic Databanks for Biomedical Informatics
MEDICAL INFORMATICS AND For these purposes, new knowledge and

E-HEALTH IN THE skilled personnel are needed, as well as essen-
POST-GENOMIC ERA tial, in order to implement new technologies and
gain adequate financial support, from both the
The subject of medical informatics (MI) plays private and public sectors. This will carry very
a very important role among the several important consequences for the industries op-
branches of bioengineering because it deals erating in the field, the so-called e-health indus-
with the practical and theoretical issues of the tries, which represent today the third industrial
implementation of ICT in a wide range of force in the European health sector, and which
solutions in the medical field, as well as in the will bring new life to ICT industries.
more general health sector. Examples of the operational capabilities of
The constant evolution and improving of the MI are as follows:
technologies involved make it possible to guess
that medical informatics will become increas- • Communication networks in hospitals
ingly important in the years to come. Conse- • Medical databases
quently, in the last years, the complexity of • Telemedicine services
medical-informatics solutions causes them to • MI Web-based portals
always be referred to as belonging to the so- • ICT tools and systems dealing with the
called e-health field. This term hints at a wider patient
role and groups all the numerous activities • Diagnosis
arising from MI applications. • Monitoring
E-health responds to growing demands for • Treatment
quality health services, patient mobility, data re- • Prevention
cording and processing, and finally for the more
rational management of the economic resources In particular, if we refer to the ICT tools
and human efforts destined to these services. and systems (Lacroix , 2002), the related appli-
E-health tools and applications can provide cations can generate new and undiscovered
fast and easy access to electronic health records developments, especially if viewed from the
at the point of need: They can support diagnosis perspective of the latter discoveries carried out
by noninvasive imaging-based systems, they after human-genome sequencing and in the
support surgeons in planning clinical interven- post-genomic era, with the subsequent devel-
tions using patient-specific digital data, and opments of genomics, the systematic study of
they provide access to specialized resources genes, chromosomes, and nucleic acids in an
for education and training. Digital data transfer organism, and proteomics, which analyzes the
enables more effective networking among clini- proteins’ expression levels in the biological
cal institutions across the world and the cre- processes of an organism.
ation of a global network of centers of refer- In the postgenomic era, the complete se-
ence. Electronic health records also enable the quencing of the human genome has helped to
extraction of information for research, man- develop new prospects for the comprehension
agement, public-health, or other related statis- of the many mechanisms existing at the base of
tics of benefit to health professionals. biological processes that cause disease.
More generally, e-health aims to improve A new vision of pharmacology has also
the overall quality, productivity, and efficiency started, supporting so-called pharmacogenet-
of the sector. ics and pharmacogenomics, which study how
358
an individual’s genetic inheritance affects the The completion of the human genome project
body’s response to drugs. Pharmacogenomics has been greatly facilitated by MI tools and
focuses on a treatment that must be personal- modern computational capabilities. Currently,
ized according to the genetic profile of the BI is focused on the following three subjects:
patient: This could potentially reveal better and
safer drugs, undiscovered and related macro- 1. Structural bioinformatics, which studies
molecular targets, new therapies, and new and the prediction of molecular structure
more effective prevention methods. 2. Sequence analysis
It can also be noticed that this new era 3. Macromolecular databases
involves the intervention of different sciences,
such as biology, medicine, engineering, In bioinformatics, computers are used for all
informatics and mathematics, and different pro- types of data useful for biological and macro-
fessional figures corresponding to different molecular research, and consequently, the meth-
stages of analysis and study. ods peculiar to information and communication
technology are applied to forecast and analyze
the molecular structures taken into account.
BIOMEDICAL INFORMATICS: The ultimate goal of BI is to uncover the
A NEW APPROACH FOR richness of biological information hidden in the
GENOMIC RESEARCH mass of data and to obtain clearer insight into
the fundamental biology of organisms.
In this renewed scenario, the study of human From the synergy of BI and MI, biomedical
genomes has marked the need of synergy be- informatics (BMI) was created, born for the
tween two sciences that have stayed distinct development and sharing of biomedical knowl-
and unrelated until now: medical informatics edge coming from MI and BI, and for the
and bioinformatics (BI). support of computational molecular biology,
The two sciences have different origins: which is the combined performance of math-
ematical, statistical, informatics, and techno-
• MI was conceived with the introduction of logical techniques used in molecular-biology
computers in hospitals in order to perform research (Martin-Sanchez et al., 2004).
some special applications, such as those The union of the two sciences means a
for electronic data records, Bayesian sys- powerful research method that provides a sci-
tems, online bibliographic databases, and entific and technical framework supporting an
so forth. analysis that is totally personalized according to
• Bioinformatics was developed for the information available from each patient, con-
analysis, storage, and processing of huge sisting of data that were both clinical and ge-
amounts of biological data coming from nomic. This represents a more dynamic ap-
laboratories, and it is an interdisciplinary proach, based on the genotype of the patient
research area interfacing between the and able to carry out a predictive analysis,
biological and computational sciences, which is an analysis that can provide the appli-
resulting in a fundamental management cation of prevention measures.
information system for molecular biology. Biomedical informatics allows us to focus
on small DNA (deoxyribonucleic acid) changes
359
or protein syntheses, and to understand the consultable via the Web with different
functioning and regulation of genes in order to architectures or operating systems
discover new diagnosis techniques and to de- • Performing data-mining analysis includ-
velop specific products in pharmacogenomics, ing learning algorithms, methods for sig-
which fight the disease on a molecular level. nificant information extraction, decision-
BMI is intended to create a point of encoun- support systems, and queries into data-
ter between the genotypical and phenotypical bases (these methods should preferably
information of a patient. be accomplished online)
The purposes of BMI can be summed as • Performing text-mining elaboration, such
follows (Tavazoie, Hughes, Campbell, Cho, & as the extraction of data and information
Church, 1999): from text, thanks to the latest natural
language processing (NLP) capabilities
• Development of prevention measures in
disease treatment Nevertheless, some difficulties can arise in
• Discovery of new therapies and related such an integrated BMI approach due to differ-
biological targets to fight the disease ent learning and analysis strategies, or to the
• Efficiency, celerity, and cheapness of bi- absence of proper technologies useful to the
ology and pharmacogenomic research solutions.
The current state of BMI shows the need

for creating comprehensive databases of both GENOMIC DATA BANKS:
clinical data and genomic data, for having stan- FUNCTIONS, PURPOSES, AND
dards for the creation of databases and ontolo- THEIR EFFECTIVE USE
gies (Gene Ontology Consortium, 2000;
Martucci, Masseroli, & Pinciroli, 2004; Genomic data banks orderly store genomic and
Masseroli, Martucci, & Pinciroli, 2004), and for proteomic or molecular data, and provide an
analyses such as DNA microarrays for patient interface for their querying. Broadly referring
classifications. to the data they contain, biomolecular data
The first results of BMI are the so-called banks is another proper term for them. These
biomolecular data banks, which are databases data banks represent an integrated environ-
reporting clinical and genomic information to- ment to manage a great amount of data, and
gether with physiological and environmental they are able to report heterogeneous structural
data. There are also new informatics tools and and functional information on genes and pro-
Web applications, and even standards to obtain teins.
an effective interaction between the different Biomolecular data banks represent a major
phases of data processing, hardware compo- source of information on nucleotide sequences
nents, network architectures, and system secu- (DNA, RNA [ribonucleic acid], proteins), to-
rity technologies. gether with biological, clinical, physiological,
The following functionalities appear essen- and experimental data, and bibliographical an-
tial to create such a framework: notations related to the enquired genes. There-
fore, these data banks can be the information
• The storage of all data and information in source of biomedical-informatics analysis tools
special databases located at different sites, that extract significant information related to
360
the causes of a disease and on its possible Data Banks’ Accessibility

treatments.
To look for any piece of information in the Genomic data banks are accessible in different
large amount of data on molecular sequences ways, but at present, all these ways are not
and gene annotations collected in genomic data functional to efficiently use the provided anno-
banks, generally some keywords or identifiers tations for easily studying lists of genes.
have to be used. They enable one to identify The ways to access available data are the
stored biomolecular information by gene name following:
or identifier, GenBank accession number, or by
controlled terms describing their involvement in • Access through a Web Server (HTML
biological processes, molecular functions, bio- [HyperText Markup Language] or
chemical pathways, and so forth. XML [Extensible Markup Language]
Since 1994, every year the scientific journal Pages): This is the most common pro-
Nucleic Acids Research (http:// vided access, which usually presents un-
nar.oupjournals.org/) publishes a specific re- structured information and heterogeneous
view on molecular biology data banks, including Web interfaces, and the related query
a list of open-access data banks together with results on a single biomolecular sequence
brief descriptions and related URLs. The 2005 are mainly returned in HTML format; on
update quotes a list of 719 data banks, 171 more the other hand, it requires time to compre-
than the previous year (Galperin, 2005). hensively query multiple data banks.
As of February 2005, sequences for more • Access through an FTP (File Transfer
than 1,200 species, including 973 viruses, 197 Protocol) Server: It requires one to have
bacteria, and 39 eukaryotes, were known, and significant technological and human re-
207 genomes were completed. The most stud- sources for locally reimplementing the
ied genomes are those of the human, fruit fly data bank, and sometimes there are no
(Drosophila melanogaster), mouse, rat, zebra relations among provided data (ASCII
fish, thale cress (Arabidopsis thaliana), [American Standard Code for Informa-
escherichia coli, pea, maize, and wheat. tion Interchange] flat file format).
Most of the data stored in biomolecular data • Direct Access: It is rarely allowed due to
banks are public and freely accessible through security issues and because data-bank
the Internet, and they are submitted in the schemes are heterogeneous and unknown
following types: a priori, query languages differ among
data banks, and there is a lack of a com-
• Nucleotide sequences mon vocabulary.
• Genomic mapping data • Direct HTTP (HyperText Transfer
• Expression profiles (2D-SDS PAGE, DNA Protocol) Linking: It is generally avail-
chips) able if the data-bank entry identification
• Protein sequences code is known and each link returns a
• 3-D structures of nucleic acids and pro- Web page (usually in HTML format) with
teins all data available in the data bank for the
• Metabolic data considered entry.
• Functional annotations
• Bibliographic information Generally, biomedical researchers need to
have an aggregated form of the genomic data in
361
order to browse them easily and perform ar- • Automatically extract specific data of in-
ticulated queries on them. Despite efforts to terest in the HTML or XML pages of
integrate gene annotations, relevant gene data different data banks
are still sparsely stored among heterogeneous • Store in aggregate form the extracted
data banks. Consequently, the increasing amount data
of information available requires new ap- • Structure the aggregate data to enable the
proaches to summarize, visualize, and compare performing of subsequent, specific que-
the gene annotations in order to make it possible ries on them
to discover new knowledge.
The effective use of the huge amount of Biomolecular data banks can be classified in
data available in genomic data banks presents different ways according to specific character-
several difficulties because the storage is ac- istics.
complished in distinct data banks, and the data Above all, they can be subdivided into the
banks are heterogeneous in schema and con- two following classes:
tents, and generally can be interrogated only for
a single genomic sequence at a time. Moreover, • Primary data banks
they are mostly accessible for interrogation via • Derivative or specialized data banks
the Web only, and the data retrieved as interro-
gation results are usually available, not struc- Primary Data Banks
tured, on HTML pages only.
For the data banks with access through FTP The primary data banks regard nucleic acids
servers and the rare ones with direct access, and amino acids, and they only contain the
solutions to the interrogation difficulties can essential information to identify a sequence and
include the following: its main characteristics. Each sequence intro-
duced in a data bank with its annotation consti-
• Creating local databases (i.e., mirrors) tutes a so-called entry.
associated with the original data banks; DNA (nucleic acids) data banks include the
the related drawbacks are keeping the following:
mirrors updated and multiple-database is-
sues • GenBank at NCBI (USA, http://
• Designing and using special query lan- www.ncbi.nlm.nih.gov/)
guages to access and query data in mul- • EMBL at EBI (European Molecular Biol-
tiple databases of heterogeneous DBMS, ogy Laboratory, United Kingdom, http://
or through the definition and use of www.ebi.ac.uk/embl.html/)
metadata • DDBJ (Japan, http://www.ddbj.nig.ac.jp/)
• Automatic mapping of queries to answer
the need of performing the same query on Protein (amino acids) data banks include the
several databases following:
For data banks providing access through a • Swiss-Prot/TrEMBL (high level of anno-
Web server, solutions to the interrogation diffi- tation; http://www.expasy.org/sprot/)
culties reside in creating new tools allowing one • PIR (Protein Identification Resource,
to do the following: http://pir.georgetown.edu/)
362
• UniProt (Unified Protein Resource, http:/ with specific biological information inher-
/www.pir.uniprot.org/) ent to the considered subset. A good ex-
ample is the PIR Sequence-Structure data
The first DNA data bank, created in 1980, bank (PIR-NRL3D, http://www-
was the European Molecular Biology Labora- nbrf.georgetown.edu/; Pattabiraman et al.,
tory Data Library. In 1982, the American data 1990). PIR-NRL3D is a data bank of
bank GenBank was created, followed in 1986 proteins, derived from the Protein Infor-
by DDBJ, the Japanese DNA data bank. Data mation Resource databank, with a known
banks concerning proteins and carbohydrates are 3-D structure and whose atomic coordi-
Swiss-Prot/TrEMBL (http://www.expasy.org/ nates are recorded in the Protein Data
sprot/) and PIR (http://pir.georgetown.edu/). Bank (PDB).
The EMBL, GenBank, and DDBJ form the • A set of multi-aligned homologous se-
International Nucleotide Sequence Database quences, such as rRNA (Neefs et al.,
Collaboration, which uses the Taxonomy Project 1993; http://www.psb.ugent.be/rRNA/
in order to make available a unified taxonomy in index.html/) and tRNAC (Steinberg et al.,
all three data banks, and the Feature Table in 1993) data banks.
order to identify a set of information to associ- • A set of specific information complemen-
ate to each sequence and the mechanism of tary to those in the primary data banks,
data exchange. and specific for a well-defined class of
sequences. A good example for this class
Derivative Data Banks is the Eukaryotic Promoter Databank
(EPD, http://www.epd.isb-sib.ch/; Bucher
The derivative or specialized data banks con- et al., 1996).
tain both the genomic data of primary data • Genomic databanks, representative of the
banks and their taxonomic, biological, physi- whole set of information derived from
ological, or medical annotations; thus, they rep- mapping and sequencing projects of the
resent a very useful source of information human genome and of other genomes
complementary to primary data banks. selected as model organisms. A good ex-
The specialized data banks can be the fol- ample is the Genome Data Base (GDB,
lowing: http://gdbwww.gdb.org/).
• Integrational databanks, recently created
• Human curated (e.g., Entrez Gene, Swiss- to collect information dispersedly stored
Prot, NCBI RefSeq nRNA) in other specialized data banks. Good ex-
• Computationally derived (e.g., UniGene) amples are the GeneCards (http://
• A combination of both (e.g., NCBI Ge- bioinformatics.weizmann.ac.il/cards/) and
nome Assembly) SOURCE (http://source.stanford.edu/)
data banks.
Specialized databanks can be classified as
follows: Examples of other derivative data banks are
the following:
• A simple subset of the primary data bank,
homogeneous from the biological point of • Unigene (http://www.ncbi.nlm.nih.gov/
view, and accurately revised and enhanced UniGene/)
363
• Entrez Gene (http://www.ncbi.nih.gov/ above a defined threshold. There are three

entrez/query.fcgi?db=gene/) principal database search algorithms.
• Smith-Waterman algorithm (http://
The biomolecular data banks can be also decypher2.stanford.edu/), which uses
classified in other different and simpler ways, dynamic programming to compute the
such as in genome data banks or proteome data most sensitive similarity alignments
banks, depending on the type of information • FASTA (http://www-nbrf.georgetown.
they contain, that is, genomic or proteomic edu/pirwww/search/fasta.html/),
data. which is an approximate heuristic
algorithm used to compute subopti-
Bioinformatic Data-Bank-Based mal similarity comparisons
Tools • BLAST (http://www.ncbi.nlm.nih.gov/
BLAST/), which is another approxi-
Several bioinformatic software tools have been mate heuristic algorithm used to com-
developed to perform information extraction pute suboptimal similarity compari-
and searching operations on database-stored sons, but it is better and faster than
data by using different technologies (Masseroli, FASTA, giving a statistical evalua-
Stella, Meani, Alcalay, & Pinciroli, 2004). tion of the result’s significance
Factors that must be taken into consider- • Text-Based Tools: These include SRS6
ation when designing these tools are as follows: (http://srs6.ebi.ac.uk/srs6bin/cgi-bin/
wgetz?-page+top/), ENTREZ (http://
• The end user (the biologist) may not be a www.ncbi.nlm.nih.gov/Entrez/), and
frequent user of computer technology. DBGET/LinkDB (http://www.genome.jp/
• These software tools must be made avail- dbget-bin/www_bfind?linkdb/).
able over the Internet given the global
distribution of the scientific research com-
munity. CONCLUSION
A specific type of these bioinformatic tools In the last years, life sciences, such as biology
is represented by database search tools, which and medicine, and computational sciences, such
are software programs designed for extracting as informatics, statistics, and engineering, have
the meaningful information from the mass of been merging their efforts to try to reach a very
data and carrying out the analysis steps. difficult goal in genomics and proteomics: a
These database search tools are classified better knowledge of biological mechanisms and
as follows: processes, especially those that are the cause
of diseases.
• Sequence-Based Tools: These data- The development and application of biomedi-
base search algorithms are used to com- cal informatics in medicine, biology, and
pute comparisons between a candidate healthcare sectors have the objective of search-
query sequence and each of the sequences ing for new and more effective diagnostic, moni-
stored within a database in order to find all toring, and treatment methods. These are mainly
the pairs of sequences that have similarity focused on personalized pharmacological thera-
pies developed thanks to pharmacogenomics.
364
In order to achieve such aims, genomic data on Information Technology in Biomedicine,

banks are fundamental sources of structured 6(2), 123-128.
biomolecular data and biomedical knowledge,
Martin-Sanchez, F., Iakovidis, I., Norager,
including extremely valuable functional and clini-
S., Maojo, V., de Groen, P., Van der Lei, J.,
cal information.
et al. (2003). Synergy between medical
Recently developed bioinformatics algorithms
informatics and bioinformatics: Facilitating
and tools exploit such relevant information within
genomic medicine for future healthcare.
the heterogeneous and widely distributed data-
Journal of Biomedical Informatics, 37, 30-
bases of biomolecular data banks accessible
42.
through Web servers to extract lists of the most
significant genes as causes of disease, and to Martucci, D., Masseroli, M., & Pinciroli, F.
gather and evaluate the relevance of the gene (2004). Gene ontology application to genomic
annotations related to them. In the end, ge- functional annotation, statistical analysis and
nomic data banks and bioinformatic data-bank- knowledge mining. Studies in Health Tech-
based tools are useful to highlight significant nology Informatics, 102, 108-131.
biological characteristics and to support a glo-
Masseroli, M., Martucci, D., & Pinciroli, F.
bal approach in order to improve the under-
(2004). GFINDer: Genome Function INtegrated
standing of complex cellular mechanisms and
Discoverer through dynamic annotation, statis-
physiological knowledge.
tical analysis, and mining. Nucleic Acids Re-
search, 32, w293-w300.
REFERENCES Masseroli, M., Stella, A., Meani, N., Alcalay,
M., & Pinciroli, F. (2004). MyWest: My Web
Bucher, P., Karplus, K., Moeri, N., & Hofmann, extraction software tool for effective mining of
K., (1996). A flexible search technique based annotations from Web-based databanks.
on generalized profiles. Computers and Chem- Bioinformatics, 20, 3326-3335.
istry, 20, 3-24.
Neefs, J. M., Van de Peer, Y., De Rijk, P.,
Galperin, M. J. (2005). The molecular biology Chapelle, S., De Wachter, R. (1993). Compila-
database collection: 2005 update. Nucleic Ac- tion of small ribosomal subunit RNA structures.
ids Research, 33, D5-D24. Nucleic Acids Research, 21, 3025-3049.
Gene Ontology Consortium. (2000). Gene on- Pattabiraman, N., Namboodiri, K., Lowrey, A.,
tology: Tool for the unification of biology. Na- Gaver, B. P. (1990). NRL_3D: A sequence-
ture Genetics, 25, 25-29. environment. Protein Sequences Data Analy-
sis, 3, 387-405.
Glynn, D., Jr., Sherman, B. T., Hosack, D. A.,
Yang, J., Gao, W., & Lane, H. C. (2003). Steinberg, S., Misch, A., & Sprinzl, M. (1993).
DAVID: Database for annotation, visualiza- Compilation of tRNA sequences and sequences
tion, and integrated discovery. Genome Biol- of tRNA genes. Nucleic Acids Research, 21,
ogy, 4, R60. 3011-3015.
Lacroix, Z. (2002). Biological data integration: Tavazoie, S., Hughes, J. D., Campbell, M. J.,
Wrapping data and tools. IEEE Transactions Cho, R. J., & Church, G. M. (1999). Systematic
365
determination of genetic network architecture. statistical, experimental, and technological meth-

Nature Genetics, 22, 281-285. ods that is energizing and accelerating the
development of new techniques and tools for
molecular biology.
KEY TERMS
E-Health: It refers to the application of
Bioinformatics: A branch of biology and information and communications technologies
informatics concerned with the development of to the whole range of activities related to the
techniques for the collection and manipulation health sector.
of biological data, and the use of such data to
Genomic Data Banks: They hold, treat,
make biological discoveries or predictions. It
and analyze genomic data, together with bio-
comprehends all computational methods and
logical, clinical, or experimental information.
theories applicable to molecular biology and
areas of computer-based techniques for solv- Genomics: The systematic identification
ing biological problems including the manipula- and study of genomes, each of which include all
tion of models and data sets. the genetic material of a living organism.
Biomedical Informatics: The discipline Medical Informatics: The field of infor-
that studies biomedical information and knowl- mation science concerned with the analysis and
edge, focusing in particular on their structure, dissemination of medical data through the appli-
acquisition, integration, management, and opti- cation of computer-science technologies to
mal use. It adopts and applies results from a various aspects of healthcare and medicine.
variety of other disciplines including informa-
Pharmacogenetics: A branch of genetics
tion science, computer science, cognitive sci-
that deals with the genetic components of vari-
ence, business and organization management,
ability in the individual responses to and me-
statistics and biometrics, mathematics, artifi-
tabolism of drugs.
cial intelligence, operations research, and basic
and clinical health sciences. Pharmacogenomics: An extension of the
established science of pharmacogenetics. The
Biomolecular Databanks: Structured re-
process of treatment is tailored to fit the precise
positories of biomolecular, genomic, or
makeup of each individual patient.
proteomic data and their related biological,
medical, and clinical information. Proteomics: The study of all possible pro-
teins (amino-acid sequences) of an organism,
Computational Molecular Biology: A
translated from different transcripts.
new discipline bringing together computational,
366
367
Chapter XLVI
Basic Principles and Applications
of Microarrays in Medicine
Andriani Daskalaki
Athina A. Lazakidou
ABSTRACT
The simultaneous expression of a large number of genes is a critical component of normal

growth and development, and the maintenance of health. Microarray technology is used to
understand fundamental aspects of growth and development, as well as to explore the
underlying genetic causes of many human diseases. Systematic analysis of microarray data
will yield insight into molecular biological processes and the functions of thousands of gene
products in parallel. This approach allows for better understanding in cellular signaling,
disease classification, diagnosis, and prognosis. Microarrays allow scientists to analyze the
expression of many genes in a single experiment quickly and efficiently. One important goal
of computational analysis of microarrays is to extract clues from microarray data and
translate the information into biological understanding diseases in medicine and dentistry.
There are different platforms or types of DNA microarrays that are commercially available:
Glass DNA microarrays and high-density oligonucleotide microarrays. DNA microarray
experiments generate large quantities of genome-wide data. To extract useful information
from expression profiles, computational tools that compute, statistically validate and display
data can be used. An important step in the computation of microarray data is normalization.
The purpose of the normalization prozess is to identify and remove the effects of systematic
variation in the measured fluorescence intensities other than differential expressions. There
are different methods for the normalization of data: total intensity normalization, regression
normalization, normalization using ratio statistics, and variance stabilization (VSN). A major
goal of microarray data analysis is to identify differentially expressed genes. Selecting
marker genes is an important issue for disease classification based on gene expression data.
Basic Principles and Applications of Microarrays in Medicine
The selection of marker genes is critical in tumor classification using gene expression data.
Many methods have been proposed to select differentially expressed genes, including
parametric and nonparametric tests, and others.
INTRODUCTION rate diagnosis, surgical treatment, and drug-

delivery therapy based on an individual patient’s
The proper and harmonious expression of a genetic profile.
large number of genes is a critical component of DNA microarray technology has been used
normal growth and development, and the main- for genome-wide gene-expression studies that
tenance of proper health. Disruptions or incorporate molecular-genetics and computer-
changes in gene expression are responsible for science skills on massive levels. This technol-
many diseases. ogy allows comparisons of gene-expression
Biomedical research evolves and advances levels in samples derived from normal and
not only through the compilation of knowledge, diseased tissues, treated and nontreated tis-
but also through the development of new tech- sues, and tissues in different stages of differen-
nologies. Using traditional methods to assay tiation or development. It uses nucleic-acid
gene expression, researchers were able to sur- hybridization techniques and computers to evalu-
vey a relatively small number of genes at a ate the mRNA (messenger ribonucleic acid)
time. The emergence of new tools enables expression profile of thousands of genes simul-
researchers to address previously intractable taneously for the purposes of gene discovery,
problems and to uncover novel potential targets disease diagnosis, improved drug development,
for therapies. Microarrays allow scientists to and therapeutics tailored to specific disease
analyze the expression of many genes in a processes.
single experiment quickly and efficiently. They DNA microarrays are miniature arrays con-
represent a major methodological advance and taining gene fragments that are either synthe-
illustrate how the advent of new technologies sized directly onto or spotted onto glass or other
provides powerful tools for researchers. Scien- substrates. Each spot serves as a highly spe-
tists are using microarray technology to try to cific and sensitive detector of the correspond-
understand fundamental aspects of growth and ing mRNA. Further computational analysis of
development, as well as to explore the underly- microarray data allows the classification of
ing genetic causes of many human diseases. genes by their mRNA expression patterns.
Since many genes contribute to normal func- Global gene-expression profiles in cells or tis-
tioning, research efforts are moving from the sues will provide us with a better understanding
search for a disease-specific gene to the under- of the molecular basis of a phenotype, pathol-
standing of the biochemical and molecular func- ogy, or treatment.
tioning of a variety of genes and how compli- Furthermore, there exists an increasing num-
cated networks of interaction can lead to a ber of applications for protein and antibody
disease state, such as oral cancer. With the microarrays (Feilner et al., 2004) in basic re-
DNA (deoxyribonucleic acid) microarray-based search diagnostics, drug discovery, and in vitro
research, we can look forward to more accu- risk assessment of nutrients.
368
TYPES OF DNA MICROARRAYS Figure 1. GeneChip System 3000Dx
There are currently two platforms or types of

DNA microarrays that are commercially avail-
able.
1. Glass DNA microarrays that involve the

microspotting of prefabricated cDNA
(complementary DNA) fragments on a
glass slide.
2. High-density oligonucleotide microarrays,
often referred to as chips, that involve in
situ oligonucleotide synthesis.
However, from a manufacturing point of

view, there are fundamental differences be-
tween the two platforms in regard to the sizes
of printed DNA fragments, the methods of expressed, or yellow when the level is the
printing the DNA spots on the slide or chip, and same in the two samples.
also the data images generated. • Oligonucleotide Microarrays: In these
arrays, DNA oligonucleotides are synthe-
• cDNA Microarrays: On a glass surface, sized in situ onto the DNA chip using
complementary DNA can be spotted. A photolabile protecting groups and photo-
high-speed robot is used to spot PCR- lithographic masks to add the selective
amplified cDNA onto a chemically modi- sequences of nucleotides. They offer a
fied (polylysine, aminosilane) glass slide. fast, high-throughput alternative for the
The DNA-arrayed slides are then hybrid- parallel detection of microbes from virtu-
ized with fluorescently labeled cDNA really any sample. The application potential
verse transcribed from mRNA popula- spreads across most sectors of life sci-
tions. During this process, the slide is ences, including environmental microbiol-
hybridized with two different cDNA ogy and microbial ecology. Each probe
samples labeled separately with two dis- consists of 25-base-pair (bp) oligonucle-
tinct fluorescent dyes, such as Cy3 (cya- otides (thus called 25-mers) where 20
nine 3, green dye) and Cy5 (cyanine 5, red different oligonucleotide pairs represent
dye; two-color hybridization). The rela- one gene. A pair consists of a perfect-
tive intensities of the two fluorescent dyes match and a mismatch oligonucleotide, in
within a spot represent the relative mRNA which the 13th nucleotide is verändert.
expression levels of the gene. For ex- The perfect-match signals are subtracted
ample, if fluorescent labels Cy3 (green) by the mismatch signals, and the net val-
and Cy5 (red) are used to make each ues are used for the comparisons. In the
sample’s cDNA probe, the expression Affymetrix GeneChip system (Figure 1),
level of a gene will be displayed as green in contrast to the cDNA or presynthesized
or red when the gene is differentially oligonucleotide deposited arrays, only one
369
Figure 2. Experimental design for a microarray experiment
Collection of clinical & experimental data (Biopsy, blood, cell lines)
RNA -- Isolation, Synthesing or "labeled targets"
Hybridisation of Microarrays
Scanning of Microarrays
Statistik Biological Verification

Datanormalisation
(Clustering, & Interpretation
Classification)
sample is hybridized on to one array (tar- brightness (foreground-background ratio) can

get), and comparisons can be made among vary within a spot. Beyond the spot quality
multiple arrays (one-color hybridization). (brightness, uniformity, morphology), the slide
quality (percentage of spots without signal,
range of intensities, distribution of spot signal
DATA ANALYSIS area) is of importance.
The raw data from a cDNA microarray experi- Normalization

ment consist of pairs of image files, 16-bit
TIFFs, one for each of the dyes. Image analysis The purpose of normalization is to identify and
is required to extract measures of the red and remove the effects of systematic variation in
green fluorescence intensities, R and G, for the measured fluorescence intensities other
each spot on the array. In the image analysis, than differential expressions, such as the fol-
the first step is to estimate the location of the lowing:
spot centers (addressing). Then pixels have to
be classified as foreground (signal) or back- • Expression, for example, different label-
ground (segmentation). ing efficiencies of the dyes.
At the end, we have to extract the available • Different amounts of Cy3- and Cy5-
information, foreground and background inten- labeled mRNA. In hybridizations, where
sities, and quality measures for each spot on the the same mRNA sample is labeled with
array and each dye. Spots usually vary in size the Cy3 and Cy5 dyes, the imbalance in
and shape (area, perimeter, circularity). Also, the red and green intensities is usually not
pixel and ratio intensities (uniformity) as well as constant across the spots within and be-
370
tween arrays, and can vary according to that a subset of genes, referred to as
overall spot intensity, location, and so housekeeping genes, do not change their
forth. profiles throughout the experiments. These
• Different scanning parameters (scanning genes are thought to be constantly ex-
artifacts). pressed across a wide range of biological
• Printing. samples (e.g., GAPDH). The normaliza-
tion factor calculated from this subset of
Normalization is needed to ensure that dif- housekeeping genes is used to adjust ex-
ferences in intensities are indeed due to differ- perimental variability in the samples being
ential expression. Normalization is necessary compared.
before any analysis that involves within- or
between-slide comparisons of intensities, for Alternatively, a set of exogenous controls
example, clustering and testing. can be spiked onto the arrays, and mRNA from
There are different methods for the normal- the set are equally added into the initial RNA
ization of data: samples before labeling. The average expres-
sion ratio from these controls should be equal to
• Total Intensity Normalization: This 1, and this factor is used to normalize the data
method stands on the assumption that the to identify differentially expressed genes.
total quantities of messages from both
samples are the same. Under this assump- • Variance Stabilization (VSN): This
tion, a normalization factor can be calcu- method builds upon the fact that the vari-
lated from the total integrated intensity (in ance of microarray data depends on the
one-color hybridization, for example, in signal intensity, and that a transformation
the Affymetrix GeneChip system) or from can be found after the variance is approxi-
the total average fold difference of the mately constant. VSN assumes that less
Cy3 and Cy5 channels (in two-color hy- than half of the genes on the arrays are
bridization, for example, in deposited cDNA differentially transcribed across the ex-
arrays) for all the elements in one array. periment.
This normalization factor is then used to
adjust the scale or fold for each gene in the Clustering
array.
• Regression Normalization: In a scatter DNA microarray experiments generate large
plot of both channels in two-color hybrid- quantities of genome-wide data. To extract
ization, the genes would scatter along a useful information from expression profiles,
straight diagonal line when two closely computational tools that cluster and display
related samples are compared. Normal- data can be used. Although there are many
ization of this data can be performed by ways to analyze gene-expression data, hierar-
calculating the best-fit slope and by apply- chical clustering (Eisen, Spellman, Brown, &
ing the regression to adjust the levels of all Botstein, 1998) and self-organizing map (SOM)
the genes. clustering (Tamayo et al., 1999) have been
• Normalization Using Ratio Statistics: widely used to display the data.
Using all spots on the array would be a Hierarchical clustering is simple and the
problem when many genes are differen- results are easily visualized. In hierarchical
tially expressed. This method assumes clustering, the distances between genes are
371
calculated for all of the genes based on their MAIN PROBLEMS OF

expression pattern, and the closer genes are MICROARRAYS
merged to produce a cluster. The distances
between these small clusters are calculated to • Complex Interpretation: Microarrays
produce a new cluster. Self-organizing map obtained from expression profiling are too
clustering assigns genes to a series of groups on complex to interpret. The problem of the
the basis of expression-pattern similarities. biological interpretation of gene-expres-
Random vectors are constructed for each group, sion data occurs when cellular events are
and a gene is assigned to the closest vector. mediated in protein levels. mRNA profil-
ing provides us with only the levels of
mRNA messages. In addition, the current
APPLICATIONS OF array data include the transcriptional be-
MICROARRAYS haviors of a large portion of, at the mo-
ment, uncharacterized genes.
Traditional molecular research tools for gene- • Expensive: Microarray technology is still
expression study are limited to dealing with a expensive and requires biological materi-
small group of genes at a time. Recent ad- als that may be difficult to collect. There-
vances in the microarray field have enabled the fore, most studies perform only a few
study of large numbers of genes in a single replicated microarray experiments.
experiment. DNA microarrays not only detect • Bias in Hybridization: We need to un-
global changes of gene expression, but also derstand array-specific effects regarding
have many other potential applications includ- the hybridization behavior of the
ing the identification of polymorphism (Wang et oligoprobes such as immobilization.
al., 1998), diagnostic tools for diseases, and • Sequence Databases of Less Con-
drug discovery. served Markers: Large sequence data-
The application of microarray technology bases of less conserved markers are
for microbial diagnostics is a field in the stage needed. Linking this sequence informa-
of dynamic development, with many options tion to phylogenetic traits (i.e., antibiotic
available and advantages and disadvantages resistance in clinical microbiology) will
associated with each option. One major goal of enable the prediction of these functions, at
microarray data analysis is to identify differen- least at a given level of certainty, from
tially expressed genes. Selecting marker genes microarray results.
for sample classification is also an important • Oligoprobe Design: Oligoprobe design
issue for disease classification based on gene- is often limited by the length, GC content,
expression data. The selection of marker genes number, and position of diagnostic resi-
is critical in tumor classification using gene- dues within a diagnostic region.
expression data. Many methods have been
proposed to select differentially expressed genes,
including the two-sample t -tests (Dudoit, Yang, CONCLUSION
Callow, & Speed, 2002), ANOVA (Kerr, Mar-
tin, & Churchill, 2000), SAM (Tusher, Tibshirani, One important goal of computational analysis is
& Chu, 2001), Wilcoxon nonparametric two- to extract clues from microarray data and
sample tests, and others. translate the information into biological under-
372
standing. Systematic analysis of microarray Dudoit, S., Gentleman, R., Irizarry, R., & Yang,
data will yield insight into molecular biological Y. H. (2002). Pre-processing in DNA
processes and the functions of thousands of microarray experiments: Bioconductor short
gene products in parallel. This approach allows course. Retrieved from http://www.bio
for better understanding in cellular signaling, conductor.org/workshops/WyethCourse
disease classification, diagnosis, and prognosis. 101702/PreProc/PreProc4.pdf
Proteome technologies for monitoring
Dudoit, S., Yang, Y. H., Callow, M. J., &
changes in protein abundance and protein modi-
Speed, T. P. (2002). Statistical methods for
fication are important because the correlation
identifying differentially expressed genes in
between gene and protein expression is vari-
replicated cDNA microarray experiments. Stat.
able, and the posttranslational protein modifica-
Sinica, 12, 111-139.
tions are responsible for realizing signaling and
information processing. Tissue microarrays Eisen, M. B., Spellman, P. T., Brown P. O., &
(Kononen et al., 1998) and protein microarrays Botstein, D. (1998). Cluster analysis and dis-
(MacBeath & Schreiber, 2000; Zhu et al., play of genome-wide expression patterns. Pro-
2001) have been developed in which samples of ceedings of the National Academy of Sci-
up to hundreds of tissues or proteins are ana- ences of the United States of Ameria, 95,
lyzed simultaneously on one glass slide. 14863-14868.
Unfortunately, microarray data are not eas-
Feilner, T., Kreutzberger, J., Niemann, B.,
ily shared due to the variation of standards
Kramer, A., Possling, A., Seitz, H., et al. (2004).
among experiments. The need for researchers
Proteomic studies using microarrays. Current
to agree on one particular standard, referred to
Proteomics, 1(4), 283-295.
as a universal standard, is very difficult to
achieve. Thus, ongoing efforts to find a com- Huber, W. (2003). Practical DNA microarray
mon standard sample for all experiments are in analysis. First Analysis Steps. Retrieved from
progress to facilitate widespread data sharing. http://www.bioconductor.org/workshops/
NGFN03/qcnorm.pdf
Kerr, M. K., Martin, M., & Churchill, G. A.
REFERENCES
(2000). Analysis of variance for gene ex-
pression microarray data [Tech. Rep.]. Bar
Bodrossy, L., & Sessitsch, A. (2004). Oligo-
Harbor, ME: The Jackson Laboratory.
nucleotide microarrays in microbial diagnos-
tics. Current Opinion in Microbiology, 7, Kononen, J., Bubendorf, L., Kallioniemi, A.,
245-254. Retrieved from http://www.diagnostic- Barlund, M., Schraml, P., Leighton, S., et al.
arrays.com/download/Bodrossy_and_ (1998). Tissue microarrays for high-throughput
Sessitsch_2004_preprint.pdf molecular profiling of tumor specimens. Nat
Med., 4, 844-847.
Buhler, J. (2002). Glossary of biotechnology
terms. Anatomy of a comparative gene ex- Kuo, W. P., Whipple, M. E., Sonis, S. T., Ohno-
pression study. Retrieved May 7, 2005, from Machado, L., & Jenssen, T. K. (2002). Gene
http://www.cs.wustl.edu/~jbuhler/research/ar- expression profiling by DNA microarrays and
ray/glossary.html#hybridize its application to dental research: Review. Oral
Oncol., 38(7), 650-656.
373
MacBeath, G., & Schreiber, S. (2000). Printing http:// www.diagnostic-arrays.com

proteins as microarrays for high-throughput
http://www.ima.umn.edu/talks/workshops/9-
function determination. Science, 289, 1760-
29-10-3.2003/huber/whuber-vsn-ima-
1763.
oct2003.htm
Tamayo, P., Slonim, D., Mesirov, J., Zhu, Q.,
http://www.lshtm.ac.uk/itd/grf/microarray
Kitareewan, S., Dmitrovsky, E., et al. (1999).
overview.htm
Interpreting patterns of gene expression with
self-organizing maps: Methods and application http://www.ncbi.nlm.nih.gov/About/primer/
to hematopoietic differentiation. Proceedings microarrays.html
of the National Academy of Sciences of the
United States of Ameria, 96, 2907-2912.
KEY TERMS
Tusher, V. G., Tibshirani, R., & Chu, G. (2001).
Significance analysis of microarrays applied to Channel: Data from one color (Cy3, cya-
the ionizing radiation response. Proc. Natl. nine 3, green dye; Cy5, cyanine 5, red dye).
Acad. Sci., 98, 5116–5121.
Hybridization: The process of joining two
Wang, D. G., Fan J.-B., Siao, C-J., Berno, A., complementary single-stranded nucleic acids
Young, P., Sapolsky, et al. (1998). Large-scale over their complementary bases (C-G and A-
identification, mapping, and genotyping of single- T). The microarray holds hundreds or thou-
nucleotide polymorphisms in the human ge- sands of spots, each of which contains a differ-
nome. Science, 280, 1077-1082. ent DNA sequence. If a probe contains cDNA
whose sequence is complementary to the DNA
Zhong, G., & Hongyu, Z. (2005). A
on a given spot, that cDNA will hybridize to the
semiparametric approach for marker gene se-
spot, where it will be detectable by its fluores-
lection based on gene expression data.
cence (Buhler, 2002).
Bioinformatics, 21(4), 529-536.
Microarray: A miniature array containing
Zhu, H., Bilgin, M., Bangham, R., Hall, D.,
gene fragments that are either synthesized
Casamayor, A., Bertone, P., et al. (2001).
directly onto or spotted onto glass or other
Global analysis of protein activities using
substrates.
proteome chips. Science, 293, 2101-2105.
Normalization: Data transformation. Its
purpose is to identify and remove the effects of
URL REFERENCES
systematic variation in the measured fluores-
cence intensities.
http://dial.liacs.nl/Courses/MicroArray
DataAnalysis/Exercises/Normalization_ Probe: Immobile substrate (DNA, protein)
VSN_R_Course_and_PB.pdf spotted on the array.
http://page.mi.fu-berlin.de/~fabioinf/internes/ Target: Mobile substrate (DNA, protein)
studium/vorlesungen/genetik/lh_genetik_ hybridized to the array.
ss02_Glossar1.pdf
374
375
Chapter XLVII
System Patterns of the Human
Organism and their Heredity
Manfred Doepp
Holistic DiagCenter, Germany
Gabriele Edelmann
Holistic DiagCenter, Germany
ABSTRACT
The frequency distribution analysis of biological data enables an insight into the regulatory
state of the organism. In case of strong or permanent deviations from the balance of the
systems (e.g. of the vegetative nervous system) abnormalities and/or diseases will result. They
are associated with a tendency either to chaos or to rigidity. We examined in this way families
over two or three generations. Similarities in their distribution histogram types are evaluated
which confirm a genetic disposition and a heritability of system patterns. Risk profiles are
resulting individually and concerning the descendants making possible a systemic prevention
therapy or a modification of the life style. The analysis method may be adapted to a lot of
medical examinations and represents an objective second opinion concerning health prevention.
INTRODUCTION closed-loop control-system patterns, the fre-

quency distribution of biological events and/or
Analysis of the human genome goes forward dates enables an insight into the condition of the
continuously, and the genetic types of many organism (Popp, 1987; Rossmann & Popp,
organ functions, dysfunctions, and diseases are 1986; Zhang & Popp, 1996). The lognormal
meanwhile becoming known. Problems exist (LN) distribution is considered to represent
concerning genes directing regulatory systems health, and the normal (N) distribution (bell
or feedback mechanisms; as for complex pur- curve=random) is a suspicion of cancer. This is
poses, several genes are cooperating (Finch & the state of the art.
Tanzi, 1997). In order to study the heritability of
System Patterns of the Human Organism and their Heredity
The bell curve is the result of an accident ativity, whereas humans have low instinct and
and occurs in living beings with a deficit or an high creativity.
absence of regulatory networks and coher- This means that the lognormal distribution is
ence, what is to be referred to as chaos (Doepp typical for all living beings beside humans, and
& Edelmann, 2004; Zhang & Popp, 1996). In for humans the golden-section distribution is
the normality of a high-powered and controlled the typical one. This regulation type is situated
arrangement, which works in coherence and is at the laser threshold (Table 1), which is the last
in a steady state, an asymmetric=lognormal bastion of the order inside chaos (Type 3).
distribution (Gevelein & Heite, 1950; Sachs, Health is considered to be a wavelike motion, a
1969) seems to surrender. Recently, we found crest between two terminals: a gentle one and
that an exaggeration of this results in a ten- an extreme one.
dency of the regulations of the organism toward Order is necessary as a basis for continuity,
rigidity (delta distribution). It is accompanied by and lability is the basis for adaptations as well
rigidity biologically as, for example, with as for all charm-reaction courses of events.
Parkinson’s or arterial sclerosis (Doepp & These two soft polarities should be sufficient
Edelmann, 2003). for life; however, extreme terminals are usually
The basic principle is that the normal state taking place that lead to diseases and finally to
of all controlled systems, for example, the veg- death. Our analysis is able to detect the actual
etative nervous system, contains a certain varia- condition of the organism by distinguishing five
tion—an oscillation—around the middle line types.
concerning how it is usually at phase transi- There are two chaotic (1 + 2) and two rigid
tions, like the laser threshold (Haken, 1964). (4 + 5) types, and two soft (2 + 4) and two
The chaos theory supplies the reason that extreme (1 + 5) types. The soft types lead to
the best adaptation is guaranteed through con- dysfunctions and abnormal conditions that are
ditions changing continuously in the environ- usually reversible by lifestyle modifications.
ment due to the inherent order in chaos However, in the case of extreme types, more
(Feigenbaum, 1978; Prigogin & Stengers, 1981), serious consequences will follow like diseases
or through the deterministic chaos of nonlinear that need treatment in order to be cured. Table
and/or dissipative systems. Pure chaos is too 2 demonstrates examples of this.
confused and pure order too inflexible, so a
combination of both developed in the evolution
of living beings. PROBLEM FORMULATION
Synergy comes between all parts of the
organism enabled with coherence, with the Up to now, the theory of chaos and the methods
result of a highly organized entirety. This shows of statistical analysis seemed to be more impor-
itself equal to the demands and appropriately tant than appearances in nature. Previous re-
reactive to exterior stressors and dangers. sults (Doepp & Edelmann, 2003, 2004) had led
The distribution of accelerated electrons us to the assumption that analysis is not only a
and consequently entropy becomes keys for the question of distinguishing between lognormal
understanding of the organizational state of the and normal distribution, but of distinguishing
organism. However, there is a difference be- more patterns. The routine usage of distribution
tween animals and mankind: Animals are deter- analysis in more than 1,000 cases showed us
mined by their instinct. They do not own a that distributions with two, three, or more peaks
higher consciousness and they have low cre- representing the coincidence of different ab-
376
Table 1. Relationships between physics, system statistics, and medicine: five principally
different types
Frequency-Distribution Analysis of Biological Data: Five Types

open system health closed system
accelerated electrons < 50% dynamic equilibrium accelerated electrons > 50%
integration of influences metastable emission of influences
high entropy low entropy
life at the laser threshold
high adaption low adaption
high absorption low absorption
high-frequency modulation act low-frequency modulation
low coherence high coherence
of
predominantly coincidental predominantly ordered
intuition, flexibility ratio, reason
balance
Yin Yang
extremely: lability centered stability extremely:
catabolism wavelike motion, anabolism
chaos creativity clearness rigidity
sine curve,
vagotonic dynamics, evolution sympathetic
matrix: increased sol sol : gel = 50 : 50 matrix: increased gel
abnormal
abnormal diseases:
diseases: conditions:
order + coincidence = conditions:
auto-aggression
deterministic distress depression
cancer sensitivity
chaos overacidity Parkinson’s
fatigue syndr. neurosis
cramping sclerosis
allergy
tendency
bell curve:
toward bell golden section lognormal delta type
Gaussian
curve
a : b 1,0 a : b 0,8 a : b 0,618 a : b 0,45 a : b 0,3
Type 1 Type 2 Type 3 Type 4 Type 5
Notes: a and b are complementary distances on the abscissa in the frequency-distribution histogram. a =
distance from the left limit to the maximum, b = distance from the maximum to the right limit
normalities are not so rare. We investigated meridians within the meridian diagnostics and
several families concerning their frequency- achieve a number of mostly more than 1,600
distribution patterns in order to find out a pos- results per patient. As a result, good statistical
sible heritage. In this study, different patterns processing is guaranteed. The resistance val-
inside the frequency-distribution histograms ues are converted into their reciprocals of the
should be examined if certain types are to be conductivity since those correspond to the vi-
found in two to three generations of two fami- tality status.
lies, suggesting a heritability of system analysis The histograms receive an evaluation of
patterns. their dates retrospectively by means of chi-
square (chi) and Kolmogorov-Smirnov (KS)
Material and Methods analysis of their frequency distributions com-
pared with the adapted normal and lognormal
We use as qualified measured values the skin curves (Zhang & Popp, 1996). As discussed in
resistances in the original and/or final points of earlier publications (Doepp & Edelmann, 2004),
377
Figure 1. Examples for the five principal frequency distribution types of human beings
type 1 2 3 4 5
Table 2. Examples of abnormal conditions of the regulation state of the organism and typical
diseases
chaos: Types 1 or 2 rigidity: Types 4 or 5

2 and 4: allergy distress g cramping
abnormal neurosis overacidity
states low adrenalin High adrenalin
auto-aggression sclerosis g infarction
1 and 5: burnout syndrome multiple sclerosis
diseases cancer (entodermal) lymphatism
chronic fatigue syndrome Morbus Parkinson
we had found that the addition of the ratios of rigidity (Figure 3). According to the distribution
chi N/LN and KS N/LN revealed the best types of Family A, the following symptoms
diagnostic relevance: a sum index (SI; see exist: neurasthenia, different allergies, sensi-
Table 3). tivities, and dependencies on various alkaloids.
Obviously, a chaotic regulation means a predis-
position for two, three, or even five peaks. The
PROBLEM SOLUTION only person with a different pattern and pheno-
type is Son 1, who has another father with
Results certainly another genetic type. Impressive is
the similarity of Granddaughter 2 and her mother,
Out of several family results with close similari- showing two peaks beside a valley in the central
ties between relatives, we selected for demon- areas of their results. The father of this girl
stration two: one with lability-chaos images ought to have a chaotic genetic type, too.
(Figure 2) and one with a tendency toward According to the genetic type in Family B,
no chaos-typical symptoms exist; however, pre-
dominantly metabolic diseases like diabetes
mellitus, high blood pressure, arterial sclerosis,
Table 3. Ranges of the calculated sum index obesity, and lymphatism do exist. The mother
has a lognormal distribution, the father a delta-
1 - chaos < 1.9
2 - tendency toward 1.9-2.6
like distribution.
chaos The children are not yet in the extreme
3 - normal range 2.7-5.2 rigidity phase but show clear tendencies. They
4 - tendency toward 5.3-8.2
rigidity now have the chance to work against their
5 - rigidity > 8.2 genetic material by bringing flexibility into their
378
Figure 2. Family A’s frequency-distribution histograms of 12 relatives from three generations:

Chaotic
mother daughter 1 daughter 2 daughter 3
3 peaks 5 peaks 3 peaks 3 peaks
son 1 (with other father) daughter 4 son 2 grandson 1
1 peak (type 4) 2 peaks 3 peaks 2 peaks
grandson 2 granddaughter 1 grandd.s’ mother granddaughter 2
3 peaks 1 peak (bell) 2 peaks 2 peaks
Figure 3. Family B’s frequency distribution histograms of five relatives from two generations:
Tendency to order and rigidity
mother father
son 1 son 2 daughter
lifestyle in order to keep the genotype from regulatory status within the range between the
becoming the phenotype as a whole. extreme types 1 and 5. This corresponds to a
genetic disposition for the regulation variability
Discussion of the vegetative nervous system. Not one of
the histograms is contradictory to the assump-
The two examples demonstrated here let us tion.
assume that there is a genetic disposition for a
379
Figure 4. Frequency-distribution histogram (total analysis) of a patient with cardiac

dysrhythmia (chaos peak, right) and a cerebrovascular accident (rigidity peak, left)
In the last years, classical distribution pat- that the apoplectic insult is not the effect of a
terns showing one peak have been becoming general arterial sclerosis.
rare more and more. However, the tendency It is to be considered, thus, that with the
toward two or more peaks is higher in persons existence of two or more different affecting
having a chaotic distribution pattern. It seems diagnoses, the frequency distribution in fact
that chaos-determining genetic material is a shows a superposition of two or more typical
predisposition for bi- or multipolarity of the peaks. Here, pattern-recognition analysis, which
organism regulation, which is usually acceler- can indicate distribution models characteristic
ated by external influences. Sometimes the of every illness, should be performed in addition
psyche and the personality are incorporated by to the mathematical calculation of the SI. Soft
fragmentation or dissociation. Then dependen- regulation abnormalities may exist temporarily
cies and addictions are found according to the and are reversible: Distress leads to rigidity as
genetic disposition (Doepp & Edelmann, 2002). well as neurosis to chaos. Such functional-
A chaotic tendency usually causes hypotonic abnormal discoveries are not to be interpreted,
blood pressure (low adrenalin), while a rigid thus, as illnesses, although they deviate from
tendency causes hypertonic blood pressure (high the normal.
adrenalin).
In Figure 4, we demonstrate the histogram
of a patient with two peaks resulting from a CONCLUSION
chaos disease (heart arrhythmia) and a rigidity
disease (cerebrovascular infarction caused by The analysis of the frequency distribution of
a thromboembolism). His sum index is 1.0, biological data (here, the skin resistance values
which means that the chaos is predominant and in the original and final points of the meridians)
facilitates a diagnostic insight into the control
380
systems of the organism to be gained easily. Doepp, M., & Edelmann, G. (2004). Chaos in
The regulatory type can be classified and quan- human being: An aid for medical decision mak-
tified. With a chaotic or rigid inclination go both ing. WSEAS Transactions on Biology and
conditions and illnesses. In this study, we dem- Biomedicine, 4(1), 403-409.
onstrate that these results can be used for the
Feigenbaum, M. J. (1978). Quantitative univer-
recognition of genetic disposition and heritabil-
sality for a class of nonlinear transformations.
ity.
Journal of Statistical Physics, 19, 25.
The presented analysis not only has an
objective character and an evidence basis, but Finch, C. E., & Tanzi, R. E. (1997). Genetics of
it can give the therapist information about the aging. Science, 278, 407-410.
existing risks and possible illnesses in the future
Gebelein, H., & Heite, H.-J. (1950). About the
for the individual and the descendants. In knowl-
imbalance of biological frequency distributions.
edge of this, prevention can be performed
Klin. Wochenschrift, 28, 41.
successfully. Approximately 80% of the popu-
lation dies of one of two illnesses: arterial Haken, H. (1964). A nonlinear theory of laser
sclerosis (rigid) with the result of an infarct, and noise and coherence. Zeitschrift für Physik,
cancer (partly chaotic). 181, 96-124.
Frequency-distribution analysis is not a
Popp, F.-A. (1987). New horizons in medicine
method on its own, but is a by-product of a lot
(2 nd ed.). Heidelberg, Germany: Haug.
of clinical investigations of feedback-controlled
systems creating adequate numbers of biologi- Prigogine, I., & Stengers, I. (1981). Dialog
cal data (>500). As a computer program, it may with the nature: New ways of scientific think-
be implemented into several medical routine ing. Munich, Germany: Piper.
activities. In each case, the state of the regula-
Rossmann, H., & Popp, F.-A. (1986). Statistics
tion network is determined and the patient may
of the electro-acupuncture according to Voll.
receive well-based recommendations for his or
Ärztezeitschrift für Naturheilverfahren, 1,
her lifestyle. In many cases, a second opinion is
51-59.
derived in the field of medical decision making.
This mathematical-statistical procedure as a Sachs, L. (1969). Statistical evaluation meth-
system analysis of the human organism has ods. Berlin, Germany: Springer.
proven to be of high relevance in healthcare.
Zhang, C., & Popp, F.-A. (1996). Log-normal
distribution measure of physiological param-
REFERENCES eters and the coherence of biological systems.
In C. Zhang, F.-A. Popp, & M. Bischof (Eds.),
Current developments of biophysics (pp. 102-
Doepp, M., & Edelmann, G. (2002). The switch-
111). Hangzhou University Press.
ing of the brain as an addiction-like condition:
Interference of the meridians by electrosmog?
AKU, 30(3), 133-139. KEY TERMS
Doepp, M., & Edelmann, G. (2003). Clinical
evaluation of the frequency distribution analy- Biometric Medicine: The human organ-
sis of biological data. Erfahrungsheilkunde, ism allows the detection of huge amounts of
12, 818-824. data describing the function of organs, cell
381
populations, and systems. For each of those independent procedure producing a second opin-
data collections, mathematical and statistical ion. An agreement of both enhances the prob-
evaluations may be applied in order to find ability of the diagnosis (and/or a therapy rec-
normal ranges and to control the agreement of ommendation) in the run of the decision mak-
a person’s results with those of a normal popu- ing.
lation. Significances are calculated.
Meridian Diagnostics: The net of merid-
Frequency Distribution: Biological data ians is one of the body’s regulation systems.
are collected that are generated by a random- According to traditional Chinese medicine
ized procedure of a measuring principle con- (TCM), 12 paired meridians exist in man. In
cerning any system of the organism. There order to determine their functional states, the
should be more than 500 data. They are plotted electrical resistance in the beginning and the
in a graph with the measured values on the x- end points (Ting points) of those meridians is
axis and the frequencies on the y-axis, thereby measured (24 points). The important conduc-
producing a distribution pattern. tance shows that the energetic state is recipro-
cal to the resistance. The test runs are per-
Frequency-Distribution Analysis: The
formed 20 to 100 times per patient under differ-
distribution pattern is compared with a normal
ent and randomized circumstances.
curve (Gaussian) and a lognormal curve. Two
mathematical methods are used for this pur- Prevention Therapy: Four hundred years
pose: the chi-square analysis and the ago, Sir Walter Raleigh said, “Prevention is the
Kolmogorov Smirnov analysis. Both are able to daughter of intelligence.” This means it is bet-
analyse the adaption of a person’s value distri- ter to stop a negative development from hap-
bution by the two curves. pening rather than try to deal with the problems
after it has happened. However, a prevention
Heredity: It is the process by which mental
therapy must be based on the knowledge of an
and physical characteristics are passed by par-
undesirable development. For that purpose,
ents to their children. One’s genetic type can-
existing functional and regulatory problems may
not be changed later (up to now); however,
be detected long before a disease becomes
one’s lifestyle often decides whether or not
obvious. The solution is to use an early detec-
changes become reality in the phenotype. So, a
tion method that evaluates the state of the
hereditary risk is not an unchangeable fate but
regulatory systems of the organism and finds
may be compensated for if the person gets to
out the individual risk factors.
know it early enough and consequently can
avoid certain lifestyle mistakes. Systems of the Human Organism: Inside
the human organism, several systems exist that
Medical Decision Making: Clinical medi-
are all regulated by feedback mechanisms in
cine is not a science and does not produce any
order to maintain parameters concerned around
truth. Every diagnosis and therapy is the result
its mean value. Some of those controlled sys-
of a probability calculation. For that purpose, all
tems are the heart rate, blood pressure, perfu-
symptoms and digital results are emphasized
sion rate, immune system, lymphatic system,
according to their evidence abilities. The diag-
hormonal system, breath rate, oxygen supply,
nosis with the highest probability is used as a
meridian system, skin temperature, and so forth.
hypothesis. It is also called the first-opinion
All those systems may be used in order to
diagnosis and should be controlled by another
generate accumulations of biological data.
382
383
Chapter XLVIII
Evaluation Methods for
Biomedical Technology
Maria Sevdali
Scientific Collaborator of Technological Educational Institution Kalamata, Greece
ABSTRACT
Over the past few years specialized tools for the measurement of the health level have been
developed related to the quality of life (health-related quality of life—HRQOL) and in general
they include both the objective and subjective criteria of human operation as it is illustrated
through a person’s individual and social activities. These psychometric tools are addressed
either to adults or to children and actually elevate the health services user to a basic assessor
of the effectiveness of medical interventions (medication, modern surgical techniques,
biomedical equipment, etc.) and generally of the entire health system.
INTRODUCTION status of an individual. Today, several ways of

measuring health exist that reflect the variety
The definition of health has been approached of perceptions of health. Thus, if somebody
by different points of view. However, the pre- believes that health is good physical condition,
vailing definition is the one adopted by the it is natural to use indicators of physical condi-
World Health Organization (1948), which de- tion. If, however, social or sentimental sides of
scribes health as a “[s]ituation of complete the health definition are also considered, indica-
bodily, mental and social well-being and not tors that include also these sides of health
simply the absence of illness or infirmity.” would be used.
Due to the particularity of health, objective The process of health measurement is as-
difficulties exist in efforts to estimate the health sisted by indicators of mortality, sickliness, and
Evaluation Methods for Biomedical Technology
positive health. The latter can be expressed numerical scale from 0 to 1, with death being 0
itself as a subjective behavior (indicators of and complete health being 1. Thus, death and
functional ability, general health profiles, indi- the quality of life are combined in a single
cators of good psychological condition, indica- number that can be used at the same time with
tors of social balance, indicators of quality of other methods such as QALYs (qualitatively
life) but also as an objective reality (general parked years of life), which can be used for
indicators, indicators of general behavior, envi- cost-benefit analyses (Bennett & Torrance,
ronmental indicators, socioeconomic indicators). 1996).
The utilisation of the above-mentioned indi- The HRQOL evaluation methods are subdi-
cators reflects the perception that the benefit of vided into two categories: (a) adult HRQOL
medical care aims at the improvement of qual- tools and (b) underage HRQOL tools (Tables 1
ity of life. For that reason, these indicators are and 2). However, the sample’s age-related
also named indicators of quality of life. Conse- composition is not the only application criterion
quently, the degree of improvement of life for a medical result evaluation method. The
quality for a social team constitutes the control selection of the assessment methods depends
criterion of provided medical care. Thus, such on a number of factors such as (a) the process
indicators constitute basic elements of the evalu- means of the method (i.e., the time required to
ation of health services, different treatments, extract the results), (b) the evaluation method’s
medical operations, and so forth. reliability, possible modifications in time, and
Several standardised models (question- process changes, (c) the application features of
naires) exist for the measurement of the health the tool, and (d) the applicability of the tool
status of a population concerning the quality of content in the particular study (http://ntl.bts.gov/
life (health-related quality of life, HRQOL). lib/11000/11400/11433/keynote_3.htm).
These tools produce a depicted plan of prosper-
ity and portray changes of levels of natural and
mental health before and after a patient’s intro- EVALUATION OF BIOMEDICAL
duction to the health system. TECHNOLOGY
The presence or absence of health symptoms

MODELS OF HRQOL does not constitute the unique criterion for
conclusions with regard to the health level of an
Basic notions that can be processed by the individual. The multidimensional tools of the
HRQOL tools are natural and mental health evaluation of health-related quality of life de-
through the evaluation of individual health no- pict the sense of prosperity in a patient through
tions. Health bodies assemble the evaluations the recording of his or her psychological, social,
of restrictions in individual activities (that is to and natural activity.
say, difficulties of implementation and output in In case of the comparison of different sur-
a specific activity) but also in social actions gical methods (a simple day operation com-
(that is to say, insufficient correspondence in pared to a traditionally longer one), a day opera-
the implementation of common activities) with tion has a limited influence on the mental health
a tool or a health plan (World Health Organiza- of a patient provided that he or she is allowed to
tion, 1997). More specifically, all the psycho- return shortly to his or her way of life. It
metric tools portray health status through a immediately exempts the patient from senti-
384
Table 1. Examples of generic health status and HRQOL measures for use in biomedical
technology (adults)
Number of
Selected
Measure Abbreviation Items or Domains
References
Questions
Medical SF-36 Sevdali and 36/12 Physical functioning, physical
Outcomes 36- Petropoulou role, bodily pain, general
Item Short Form (2004) health, vitality, social
Health Survey functioning, emotional role,
mental health, physical
summary score, mental
summary score
Quality of Well- QWB Kaplan and 107 Mobility, physical activity,
Being Scale Bush (1982) social activity, symptom and
problem complexes
European Quality EuroQoL EuroQoL 15 Mobility, self-care, usual
of Life Scale Group activities, pain or discomfort,
(1990) anxiety or depression
Functional FCI MacKenzie, 64 Excretion, eating, sexual
Capacity Index Dainiano, function, ambulation, bending
Miller, and and lifting, hand and arm
Luchter movement, visual function,
(1996) auditory function, speech,
cognitive function
Health Utilities HUI:3 Feeney, 45 Vision, hearing, speech,
Index Mark III Torrance, and ambulation, dexterity, emotion,
Furlong cognition, pain
(1996)
Barthel Index Barthel Mahoney and 10 Bladder and bowel control,
Barthel grooming, toilet use, feeding,
(1965) transfer, mobility, dressing,
climbing stairs, bathing
Functional FIM Linacare et al. 18 Self-care, sphincter control,
Independence (1994) mobility, locomotion,
Measure communication, social
cognition
ments of pain and probable infections, de- ward, if the health benefit to a patient is ensured
creases the cost of hospital care because of his at the highest extent, the one with smaller cost
or her short stay, and helps in the decongestion is selected so that resources are guaranteed
and saving of hospital beds through the benefit both at the individual and the sanitary level.
of qualitative utility. As far as the evaluation of Additionally, health plans can be very useful in
the contribution of medicines of the same type the drawing of objective conclusions in cases of
(with the same degree of effectiveness for different estimates for the result of an illness.
certain diseases; e.g., antibiotics) is concerned, For example, in incidents of the hospitalisation
the first step is to check the medicines’ likely of individuals with cerebral episodes that came
short-term or long-term side effects. After- back from comatose situations with residues of
385
Table 2. Examples of generic health status and HRQOL measures for use in biomedical
technology (children and adolescents)
Number of
Selected
Measure Abbreviation Items or Domains
References
Questions
Functional FSII(R) Stein and 43/14 Physical,
Status II- Jessop psychological,
Revised (1990) intellectual, and social
behavioral,
manifestations of
illness that interfere
with age-appropriate
activities
Child Health CHIP Starfield et 175 Activity, comfort,
and Illness al. (1993) satisfaction with
Profile health (perceived well-
being), disorders,
achievement,
resilience
Pediatric PedsQL Varni, Seid, 30 Physical functioning,
Quality of Life and Rode emotional functioning,
Inventory (1999) social functioning,
school functioning,
well-being, global
perception of overall
health status
Child Health CHQ Landgraf 50/28 Physical functioning,
Questionnaire and Abetz self-esteem, mental
(1996) health, general health
perceptions, behavior,
bodily pain, social-
physical role, social-
emotional-behavioral
role, parental impact
and time, parental
impact and emotion,
family activities,
family cohesion
infirmity, this situation is recorded as improve- psychometrics indicators investigates the pro-
ments in their medical files; however, the pa- motion of health as good and contributes to the
tients consider that they henceforth experience objective evaluation of sanitary programs and
lower levels of quality of life, with impacts on health systems. HRQOL tools can cover cases
their natural and mental health (Sevdali and in which conflicts exist between the patient and
Petropoulou, 2004). medical science. They give multifaceted di-
mension in health and bring the patient as a
basic factor in the collection of information
CONCLUSION about the degree of the effectiveness of each
therapeutic method, pharmaceutical education
Therefore, deductively we would say that the program, and piece of biomedical equipment in
evaluation of health levels with the use of the quality of life. Their usefulness is unques-
386
tioned since they provide possibilities for appli- Landgraf, J. M., & Abetz, L. N. (1996). Mea-
cations at various levels, such as (a) micro- suring health outcomes in pediatric populations:
scopically at a clinic or hospital level, as in a Issues in psychometrics and application. In B.
database with elements that concern the natu- Spilker (Ed.), Quality of life and
ral, psychological, and mental health of patients pharmacoeconomics in clinical trials (2 nd
with particular diseases before and after their ed.) (pp. 793-802). Philadelphia: Lippincott-
submission in therapies, and (b) macroscopi- Raven Publishers.
cally (e.g., evaluation of the utility of patients in
Linacre, J. M., Heinemann, A. W., Wright, B.
two or more stages due to the issuing of medi-
D., Granger, C. V., & Hamilton B. B. (1994).
cines for the confrontation of certain illnesses
The structure and stability of the functional
in combination with their cost) for the mapping
independence measure. Arch Phys Med
out of effective health policy.
Rehabil, 75, 127-132.
MacKenzie, E. J., Dainiano, A., Miller, T., &
REFERENCES Luchter, S. (1996). The development of the
functional capacity index. Journal of Trauma,
Bennett, K. J., & Torrance, G. W. (1996). 41, 799.
Measuring health state preferences and utili-
Mahoney, F. l., & Barthel, D. W. (1965).
ties: Rating scale, time trade-off, and standard
Functional evaluation: The Barthel index. Mary-
gamble techniques. In B. Spilker (Ed.), Quality
land State Medical Journal, 14, 56-61.
of life and pharmacoeconomics in clinical
trials (2 nd ed.) (pp. 253-265). Philadelphia: Sevdali, M., & Petropoulou, M. (2004). Mea-
Lippincott-Raven Publishers. surement of health levels using the evalua-
tion method sf36: Comparative study of 2
DeBruin, A. F., De Witte, L. P., Stevens, F., &
hospital units (pp. 11-15). Patra, Greece:
Diederiks, J. P. M. (1992). Sickness impact
Helenic Open University.
profile: The state of the art of a generic func-
tional status measure. Social Science Medi- Staffield, B., Reiley, A., Green, B., Ensminger,
cine, 35(8), 1003-1014. M., Ryan S., Kelleher, K., et al. (1997). The
adolescent child health and illness profile: A
EuroQol Group. (1990). EuroQol: A new facil-
population-based measure of health. Medical
ity for the measurement of health-related qual-
Care, 33, 553-566.
ity of life. Health Policy, 16, 199-208.
Stein, R. E. K., & Jessop, D. J. (1990). Func-
Feeny, D. H., Torrance, G. W., & Furlong, W.
tional status II(R): A measure of child health
J. (1996). Health utilities index. In B. Spilker
status. Medical Care, 28, 1041-1055.
(Ed.), Quality of life and pharmacoeconomics
in clinical trials (2nd ed.) (pp. 239-252). Phila- Varni, J. W., Seid, M., & Rode, C. A. (1999).
delphia: Lippincott-Raven Publishers. The PedsQL: Measurement model of the pedi-
atric quality of life inventory. Medical Care,
Kaplan, R. M., & Bush, J. W. (1982). Health-
37, 126.
related quality of life measurement for evalua-
tion research and policy analysis. Health Psy- World Health Organization (1948). World
chology, 1, 61-80. Health Organization constitution. Geneva,
Switzerland: Author.
387
World Health Organization. (1997). The inter-

national classification of impairments, ac-
tivities and participation: A manual of di-
mensions of disablement and functioning.
Beta-1 draft for field trials. Geneva, Switzer-
land: Author.
388
390
Chapter XLIX
Ergonomic User Interface
Design in Computerized
Medical Equipment
D. John Doyle
Cleveland Clinic Foundation, USA
ABSTRACT
Current statistics suggest that preventable medical error is a common cause of patient
morbidity and mortality, being responsible for between 44,000 and 98,000 deaths annually,
and resulting in injuries that cost between $17 billion and $29 billion annually. An important
approach to tackling this problem is to apply system design principles from human factors
engineering (ergonomics). By doing so, systems and equipment become easier for people to
work with, ultimately reducing the frequency of errors. In particular, in the case of medical
equipment, the design of the user interface can impact enormously on its successful use. In this
chapter we consider some of the elements of good and bad medical equipment design, using
examples drawn from the literature and elsewhere. The concept of ecological interface design
is also discussed, and some practical design guidelines are provided.
INTRODUCTION tant approach to improving patient safety is to

apply system design principles from human-
American statistics suggest that preventable factors engineering (ergonomics; Kohn et al.;
medical error is the eighth leading cause of Leape, 1994). Human-factors engineering is a
death, being responsible for between 44,000 relatively young scientific discipline that focuses
and 98,000 deaths annually, and resulting in on those factors that affect the performance of
injuries that cost between $17 billion and $29 individuals using systems or equipment (Kroemer,
billion annually (Kohn, Corrigan, & Donaldson, 2001). The product may be as simple as a spoon
1999). Experts have often stated that an impor- or an office chair, or as complex as an aircraft
Ergonomic User Interface Design in Computerized Medical Equipment
carrier, but in all cases the goal is to design www.fda.gov), a patient was overdosed after a
products to conform to human nature rather than nurse read the number 7 as a 1 in the drug-
merely expect people to adapt to technology. By infusion pump display. Because the flow-rate
doing so, systems and equipment become easier display was recessed, the top of the 7 was
for people to work with, ultimately reducing the blocked from view at many viewing angles.
frequency of errors. In another case report from the same source,
In the case of medical equipment, the design a physician treating a patient with oxygen set
can impact enormously on its successful use. In the flow-control knob between 1 and 2 liters per
particular, errors in operating such equipment minute, not realizing that the scale numbers
are often caused, at least in part, by the design represented discrete, rather than continuous,
of the user interface. Of course, such errors settings. Unbeknownst to the physician, there
can not only hamper patient care, but in some was no oxygen flow between the settings, even
cases can even lead to injury or death. It is though the knob rotated smoothly, implying that
obviously important that medical equipment be intermediate settings were available. The pa-
designed with special consideration given the tient, an infant, became hypoxic before the
impact of design on safe operation. Thus, the error was discovered. One solution could have
user interface for medical equipment should be been a rotary control that snaps into discrete
straightforward and intuitive: If its operation is settings.
excessively complex or counterintuitive, safety In yet another case, a patient on a ventilator
can be compromised. Human-factors techniques died following the accidental detachment of the
have been applied to other industries, such as ventilator tubing from the humidifier. Unfortu-
nuclear power and aviation, and have been very nately, an alarm did not sound because the
successful in reducing error and improving pressure-limit alarm setting had been set so low
safety in these contexts. Note also that in that it was essentially nonfunctional.
addition to increasing safety, an added benefit Finally, Figure 1 illustrates a less hazardous
of using good ergonomic design practices is the example drawn from personal experience.
likelihood that training costs will be reduced. A series of reports from the laboratory of
Dr. Kim Vicente of the University of Toronto
have looked at user-interface issues for pa-
BAD DESIGN EXAMPLES tient-controlled analgesia (PCA) equipment
(Lin, Isla, Doniz, Harkness, Vicente, & Doyle,
Examples of perplexing, arcane, and hazardous 1998; Lin, Vicente, & Doyle, 2001; Vicente,
designs produced in violation of ergonomic Kada-Bekhaled, Hillel, Cassano, & Orser,
principles are not hard to find. For instance, 2003). PCA is a computer-based medical tech-
Michael J. Darnell’s Web site nology used to treat severe pain via the self-
www.baddesigns.com offers a collection of administration of analgesic agents such as mor-
frequently humorous examples. But when bad phine. Potential benefits include superior pain
designs in medical equipment lead to injury or control, automatic documentation, and improved
death, the situation can be far from amusing. utilization of nursing resources.
This is sometimes the case for computerized Unfortunately, however, one of these units
medical equipment. (Abbott Lifecare 4100 PCA Plus II) has been
In one case reported on the U.S. Food and linked to a number of overdose deaths. This
Drug Administration Web site (http:// machine is easily set up incorrectly by
391
Figure 1. The user interface for medical equipment should be straightforward, friendly, and
intuitive. Also, rarely is the operating manual available to the end user, which makes the
labeling of the controls especially important. Consider then the user controls shown for this
operating-room table in use at the author’s facility. The top and bottom left-hand controls
lower the head and feet, respectively, while the right-hand controls raise the head and feet.
But what if the entire operating table is to be raised or lowered, which is by far the most
common request from the surgeon? It turns out that the entire table is raised by pushing both
right-hand buttons, while the entire table is lowered by pushing both left-hand buttons. This
arrangement makes sense if one thinks about it for a while, but an intuitive interface should
not require a lot of thinking. Furthermore, there is plenty of space available on the control
panel to add two extra buttons.
caregivers, who must manually enter the PCA this initially displayed value or modify it using
parameters, and a number of patients have the arrow controls. The critical flaw in the
received drug overdoses as a result of user design is that in this situation, the Lifecare
errors when using this product: the insertion of 4100 offers the minimal drug concentration as
a 5 mg/mL morphine cartridge when the ma- the initial choice. If nurses mistakenly accept
chine is expecting a 1 mg/mL concentration, or the initially displayed minimal value (e.g., 0.1
the acceptance of the default (initial) drug mg/mL) instead of changing it to the correct
concentration when the correct action is to (and higher) value (e.g., 2.0 mg/mL), the
scroll up to the correct value, among other machine will run as if the drug is less concen-
errors. In the latter case, when nurses pro- trated than it really is. As a result, it will pump
gram the drug concentration, the Lifecare more liquid, and thus more of the drug, into the
4100 display shows a particular concentration patient than is desired.
(e.g., 0.1 mg/mL). Nurses can either accept
392
Aware of the dangers of the Lifecare 4100, • Ensure that the intensity and pitch of
Lin, Isla, et al. (1998) and Lin, Vicente, et al. auditory signals and the brightness of vi-
(2001) studied the unit using cognitive task- sual signals allow them to be perceived by
analysis techniques. Based on this analysis, the users working under real-world condi-
interface was then redesigned “to include a tions.
dialog structure with fewer steps, a dialog • Make labels and displays so that they can
overview showing the user’s location in the be easily read from typical viewing angles
programming sequence, better command feed- and distances.
back, easier error recovery, and clearer labels • Use color and shape coding to facilitate
and messages ” (Lin, Isla, et al., 1998, p. 253). the identification of controls and displays.
Studies of the new interface showed signifi- Colors and codes should not conflict with
cantly faster programming times, lower mental industry conventions.
workload, and fewer errors compared to the
manufacturer’s original interface. Regrettably,
the improved interface design was not used by ECOLOGICAL INTERFACE
the manufacturer. DESIGN
Ecological interface design (EID) is a concep-

DESIGN GUIDELINES tual framework developed by Vicente and
Rasmussen (1990, 1992) for designing human-
The U.S. Food and Drug Administration has machine interfaces for complex systems such
offered a number of guidelines to help with the as are often found in process-control applica-
design of medical equipment, such as the fol- tions or in computer-based medical equipment.
lowing (adapted from http://www.fda.gov): The primary goal of EID is to aid operators,
especially knowledge workers, in handling novel
• Make the design consistent with user ex- or unusual situations. Studies suggest that the
pectations; both the user’s prior experi- use of EID methods can improve operator
ence with medical devices and well-es- performance when compared with classical
tablished conventions are important. design approaches (Vicente, 2002). The basis
• Design workstations, controls, and dis- for EID is Rasmussen’s (1986) skills-rules-
plays around the basic capabilities of the knowledge model of cognitive control, and in
user, such as strength, dexterity, memory, this model, critical incidents can result from
reach, vision, and hearing. errors at any of the skills-rules-knowledge lev-
• Design well-organized and uncluttered els of human cognition.
control and display arrangements. Keys, The first form of incident includes skill-
switches, and control knobs should be based errors involving the faulty execution of
sufficiently apart for easy manipulation an otherwise correct plan. Here, behavior is
and placed in a way that reduces the unconscious, nonverbal, and automatic. An
chance of inadvertent activation. example would be inadvertently turning on the
• Ensure that the association between con- wrong switch. Even the most experienced cli-
trols and displays is obvious. This facili- nicians are prone to skill-based errors as they
tates proper identification and reduces the often occur during highly routine procedures
user’s memory load.
393
such as in reading a drug label or adjusting a eling the system to allow one to determine the
control. consequences of an open valve.
A second category is rule-based errors and What does all this have to do with avoiding
involves the failure to apply a rule, such as human error when operating complex medical
stopping at a stop sign when driving a car, or not equipment? The answer lies in the following.
administering a drug to which the patient is Rasmussen’s (1986) three levels of cognition
allergic. At this level, one step up in Rasmussen’s can be grouped into two broader categories: (a)
(1986) cognitive model, people use stored (or analytical-based behavior (knowledge-based
precompiled) rules acquired with training or behavior) and (b) perceptual-based behavior
with experience on the job. (rule and skill based). Such a categorization is
Lastly, there are knowledge-based errors in helpful because perceptual processing has im-
which the initial intention is itself wrong, often portant advantages over analytical-based be-
due to inadequate knowledge or experience. A havior: Analytical behavior is slow, demanding,
clinical example would be administering 10 mg of and serial in nature, whereas perceptual behav-
morphine to an infant. This corresponds to ior is fast, effortless, parallel, and less error
Rasmussen’s (1986) highest cognitive level, and prone. Thus, the goal of design should be to help
is most suited for unfamiliar environments where people avoid situations requiring them to work
prior experience is unavailable to provide a at the knowledge-based level, while supporting
system of rules, such as troubleshooting a new the use of analytical problem solving for use in
piece of medical equipment for the first time. unfamiliar situations. And, as emphasized above,
Under Rasmussen’s (1986) skills-rules- design guidelines that match the environment to
knowledge model, human behavior moves along the people involved is known as ecological
a ladder as experience increases. Early on, interface design.
when one is placed in an unfamiliar environ-
ment, problem-solving behavior will be at the
knowledge level. As experience is gained so REFERENCES
that rules can be formed, the rules level takes
over. In some situations, further experience Kohn, L. T, Corrigan, J. M., & Donaldson, M.
may lead to even further automation (skills S. (1999). To err is human: Building a safer
level). health system. Washington, DC: National Acad-
For each of the three cognitive levels, the emy Press.
way in which information and environmental
Kroemer, K. H. E. (2001). Ergonomics: How
cues are perceived differs. Signals guide skill-
to design for ease and efficiency (2nd ed.).
based behavior, while symbols apply to knowl-
Upper Saddle River, NJ: Prentice Hall.
edge-based behavior. Signals supply time-space
information only; they have no meaning at Leape, L. L. (1994). Error in medicine. The
higher levels, and they cannot be verbalized. Journal of the American Medical Associa-
Signs may trigger rules (stop, start, etc.) or may tion, 272, 1851-1857.
indicate the state of the system (valve open or
Lin, L., Isla, R., Doniz, K., Harkness, H., Vicente,
closed), but they do not express functional
K. J., & Doyle, D. J. (1998). Applying human
relationships (e.g., the consequences of an
factors to the design of medical equipment:
open valve). Finally, symbols refer to concepts
Patient-controlled analgesia. Journal of Clini-
that support analytical reasoning, such as mod-
cal Monitoring and Computing, 14, 253-263.
394
Lin, L., Vicente, K. J., & Doyle, D. J. (2001). edge elicitation (the process of obtaining infor-
Patient safety, potential adverse drug events, mation through in-depth interviews and by other
and medical device design: A human factors means) and knowledge representation (the pro-
engineering approach. Journal of Biomedical cess of concisely displaying data, depicting
Informatics, 34, 274-284. relationships, etc.).
Rasmussen, J. (1986). Information process- Ecological Interface Design (EID): A
ing and human-machine interaction: An ap- conceptual framework for designing human-
proach to cognitive engineering. New York: machine interfaces for complex systems such
North-Holland. as computer-based medical equipment. The
primary goal of EID is to aid operators, espe-
Vicente, K. J. (2002). Ecological interface
cially knowledge workers, in handling novel or
design: Progress and challenges. Human Fac-
unanticipated situations.
tors, 44, 62-78.
Ergonomics: A synonym for human-fac-
Vicente, K. J., Kada-Bekhaled, K., Hillel, G.,
tors engineering, especially in the European
Cassano, A., & Orser, B. A. (2003). Program-
literature.
ming errors contribute to death from patient-
controlled analgesia: Case report and estimate Human-Factors Engineering (HFE): The
of probability. Canadian Journal of Anaes- branch of engineering devoted to the study of
thesia, 50, 328-332. the interactions between humans and systems,
especially complex systems, with the goal of
Vicente, K. J., & Rasmussen, J. (1990). The
designing systems that are safe, comfortable,
ecology of human-machine systems II: Mediat-
effective, and easy to use.
ing “direct perception” in complex work do-
mains. Ecological Psychology, 2, 207-250. Patient-Controlled Analgesia (PCA): A
computer-based medical technology used to
Vicente, K. J., & Rasmussen, J. (1992). Eco-
treat severe pain via the self-administration of
logical interface design: Theoretical founda-
analgesic agents such as morphine.
tions. IEEE Transactions on Systems, Man
and Cybernetics, 22, 589-596. Skills-Rules-Knowledge Model: A mul-
tilevel model of cognitive control developed by
Rasmussen that is especially helpful in examin-
KEY TERMS
ing human error and critical incidents.
Cognitive Task Analysis (CTA): A fam- User Interface: The components of a sys-
ily of methods and tools for understanding the tem that the operator uses to interact with that
mental processes central to observable behav- system. For example, in the case of a computer,
ior, especially those cognitive processes funda- the operator interacts with it using the monitor,
mental to task performance in complex set- keyboard, mouse, and so forth.
tings. Methods used in CTA may include knowl-
395
396
Chapter L
Safety Issues in Computerized
Medical Equipment
D. John Doyle
ABSTRACT
Computers now are being used increasingly in safety-critical systems like nuclear power
plants and aircraft and, as a consequence, have occasionally been involved in deadly
mishaps. As microcomputer technology continues to proliferate, computers are also now
increasingly being used in medical equipment such as ventilators and pacemakers, sometimes
with safety-critical results. This chapter discusses some of the special concerns that arise when
computer technology is introduced into medical equipment, using two case studies as
examples: the Therac-25 radiation therapy unit and Abbott’s patient controlled analgesia
machine. Also discussed are some of the regulations that have been proposed by the
(American) Food and Drug Administration (FDA) to help tackle the special problems that can
arise when developing software-based medical equipment.
INTRODUCTION pacemakers, sometimes with safety-critical

results. This article illustrates some of the
Computers are now increasingly being intro- special concerns that arise when computer
duced into safety-critical systems like nuclear technology is introduced into medical equip-
power plants and aircraft, and, as a conse- ment using two case studies as examples. Also
quence, have occasionally been involved in discussed are some of the regulations that have
deadly mishaps. As the cost of microcomputer been proposed by the (American) Food and
technology continues to drop, computers are Drug Administration (FDA) to help tackle the
also now increasingly being used in medical special problems that can arise when develop-
systems and equipment such as ventilators or ing software-based medical equipment.
Safety Issues in Computerized Medical Equipment
THE CASE OF THE THERAC-25 the machine about the radiation parameters to
RADIATION THERAPY UNIT be used. Srpecifically, if an extremely fast-
typing operator inadvertently selected the X-
In 1986, two cancer patients died when they ray mode instead of the electron-beam mode,
received lethal doses of radiation from a Therac- and then used an editing key to correct the
25 radiotherapy unit. An investigation revealed command to select the electron-beam mode
that one contributor to this catastrophe was the instead, it was possible for the computer to lag
failure of the design team to recognize a race behind the orders. The result was that the
condition: a miscoordination between concur- device appeared to have made the correction
rent tasks. This error resulted in individuals but in fact still had incorrect settings.
being overradiated (to death in two cases) in How could this happen? Experts speculate
Texas and Georgia while receiving cancer that the software developer might not have
therapy using the Therac-25 system (Leveson considered it necessary to guard against this
& Turner, 1993). failure mode or might not have even imagined it
Although the technical details of the failure since radiation-therapy designers have tradi-
remain secret as a result of a legal settlement, tionally used electromechanical interlocks to
experts have come up with the following ac- ensure safety in this setting. Also, analysts
count as the most likely accident scenario. A reviewing the case noted that the unit should
modern radiation-therapy machine is based on have been programmed to discard unreason-
a linear accelerator that produces a high-enable readings, as the injurious setting presum-
ergy electron beam. One may direct the elec- ably would have been. Finally, there should
trons directly into the patient, or, to produce X- have been no way for the computer’s verifica-
rays, one places a heavy metal target in the tions on the video screen to become
electron beam so that when the electron beam unsynchronized from the keyboard commands.
hits the target, X-rays come out from the other
end. The target is moved in and out of the beam
automatically under software control, depend- ERGONOMIC ISSUES
ing on whether an electron beam or an X-ray
beam is selected to treat the patient. Also, the Ergonomics is the art and science of matching
current in the electron beam is programmed to equipment design and job procedures to the
be much greater in the X-ray mode because of worker, usually with a view to reducing error
energy losses that result when the target is used and improving productivity. Ergonomics may
in making X-rays. additionally be defined as the study of the
However, in the overdose cases, because of interaction between people and machinery, and
a software-design error, the computer ran as if the factors that influence that interaction. Also
it was in X-ray made rather than in electron known as human factors, ergonomics is a rela-
mode, resulting in excessive radiation. The tively new discipline, but one that has led to
problem was a subtle error that no one had enormous improvements in equipment design
detected during the extensive testing the ma- as the principles of good ergonomic design have
chine had undergone before being introduced become discovered and refined over time
into clinical use. In fact, the error surfaced only (Kroemer, 2001; Vicente, 2004). As a result,
when an operator happened to use a specific, equipment ranging from automobiles and pho-
unusual combination of keystrokes to instruct tocopiers to nuclear power plants have all seen
397
improvements in design from the application of set for 5 to 10 minutes) has passed. In addition,
ergonomic principles. 1-hour or 4-hour cumulative limits also are
Still, examples of confusing, baffling, and available in some models.
even dangerous designs produced in violation Despite such safety features, numerous re-
of ergonomic principles are not hard to find. For ports of respiratory depression and death asso-
instance, Michael J. Darnell’s Web site ciated with PCA pumps have appeared ( Baxter,
www.baddesigns.com offers a collection of 1994; Etches, 1994; Geller, 1993; Kwan, 1995).
frequently amusing examples. But when bad One particularly notorious unit is the Abbott
designs lead to injury or death, the situation can Lifecare 4100 PCA Plus II machine. This
be far from entertaining. This is sometimes the machine is easily misprogrammed by caregivers,
case for computerized medical equipment. who must manually enter the PCA parameters,
and is in need of a more sensible and forgiving
user interface (Lin, Isla, Doniz, Harkness,
THE CASE OF THE ABBOTT Vicente, & Doyle, 1998). A number of patients
PATIENT-CONTROLLED have received drug overdoses as a result of
ANALGESIA MACHINE PCA errors when using this product: the inser-
tion of a 5 mg/mL morphine cartridge when the
In the case of the Therac-25 system, the unit machine is expecting a 1 mg/mL concentration,
did not operate as the designers intended. How- or the acceptance of the default (initial) drug
ever, the poor ergonomic design of complex concentration when the correct action is to
medical equipment can lead to patient morbidity scroll up to the correct value, among other
and mortality even while still operating cor- errors (Doyle, 2003; Lin, Vicente, & Doyle,
rectly. The following case study illustrates this 2001; Vicente, Kada-Bekhaled, Hillel, Cassano,
situation. & Orser, 2003) .
Patient-controlled analgesia (PCA) is a com- In 1997, ECRI (Emergency Care Research
puter-based medical technology now used ex- Institute, 1997) documented three deaths that
tensively to treat postoperative pain via the occurred while patients were connected to the
self-administration of analgesic agents such as Lifecare 4100. In at least two of the cases, the
morphine (Ferrante, Ostheimer, & Covino, alleged reasons for the deaths were the same.
1990). Potential benefits include superior pain In the mode of operation in use, when nurses
control, automatic electronic documentation, program the drug concentration, the Lifecare
and improved utilization of nursing resources. 4100 display shows a particular concentration
Unfortunately, however, analgesics are also a (e.g., 0.1 mg/mL). Nurses can either accept
frequent cause of adverse drug events, usually this initially displayed value or modify it using
related to respiratory depression (Bates et al., the arrow controls. The critical flaw in the
1995; Classen, Pesotnik, Evans, & Burke, 1991). design is that in this situation, the Lifecare 4100
A typical PCA machine contains an embed- offers the minimal drug concentration as the
ded computer programmed to give, for in- initial choice. If nurses mistakenly accept the
stance, 1 mg of morphine intravenously every initially displayed minimal value (e.g., 0.1 mg/
time the patient pushes a button on the end of a mL) instead of changing it to the correct (and
cable. To help prevent excessive drug adminis- higher) value (e.g., 2.0 mg/mL), the machine
tration, the onboard computer ignores further will run as if the drug is less concentrated than
patient demands until a lockout period (usually it really is. As a result, it will pump more liquid,
398
and thus more of the drug, into the patient than Food and Drug Act have led to significant
is desired. changes in the regulation of medical software.
One potentially important consideration in This act now places special emphasis on quality
the case of the Abbott PCA machine is that the issues and the need to incorporate validation
device operates exactly as specified in the criteria in design from the very beginning of
technical and operations manuals: The problem system development. The act also replaces the
is primarily one of the unwise selection of user prior emphasis on a premarket approval pro-
defaults. This makes the situation different cess with a focus on postmarket surveillance,
from design flaws such as those in the Therac- and users are now required to report to the
25 where the flawed unit operates in a manner FDA and the manufacturer any defects that
differently than the design specifications re- cause injuries or death. A number of reports
quire. While some individuals might argue that dealing with these and related issues have been
there is no design flaw present in the Abbott published (Coppess, Miller, Zipes, & Groh,
PCA unit in the sense that it operates exactly as 1999; Crumpler & Rudolph, 1997; Martensson,
the designers intended, it should be clear that 1993; Murfitt, 1990; Schnider, 1996; Trimbach,
where an alternative design exists that is safer 1995). In addition, the FDA itself has placed
and no more costly or difficult to implement, the some excellent didactic materials on the Web
original design must be considered to be infe- (FDA, 2005a, 2005b, 2005c). These resources
rior, flawed, or defective. would be expected to be valuable to all workers
in mission-critical domains.
Medical software can be defined as a set of
FDA REGULATIONS FOR instructions that enables a computer to monitor
MEDICAL SOFTWARE or control a medical device. Their regulations
DEVELOPMENT for medical software developers would require
a software developer to show that the algo-
Concerns such as those identified above raise rithm, or mathematical technique, used in the
the issue of whether a certification process for computer program has been correctly imple-
the development and testing of medical soft- mented in software. The FDA also requires
ware should be in place. Interested parties assurance through a risk analysis that any
include all medical software developers, medical software failure could not injure a patient. How
equipment manufacturers, the Food and Drug that assurance can be provided in general is still
Administration, Canada’s Health Protection unclear since any risk analysis is necessarily
Branch, and, of course, a variety of standards linked to the specifics of the application. For
agencies (ANSI, ASTM, CEN, CSA, IEEE, instance, clinicians will instantly recognize that
ISO, UL). Most efforts emphasize good soft- software errors in controlling a sodium nitro-
ware-development practices and special safety prusside infusion are far more likely to result in
measures appropriate to the clinical setting. patient injury than, say, a real-time phonocar-
In recent years, at least partly as a result of diographic monitoring system designed to de-
patients injured or killed due to medical soft- tect the new onset of cardiac murmurs in
ware defects, the FDA has taken a special coronary care unit patients (for example, with
interest in computer and software issues as suspected ischemia of the mitral valve muscu-
they relate to the clinical domain. In particular, lature). Techniques for evaluating software
the 1990 Medical Device Amendments to the safety are relatively new, drawn in part from
399
the aviation and nuclear industries as well as tive and injures a consumer or a patient, then
from academia. Who does the checking, how the manufacturer may be liable. To many medi-
much evidence is enough, and whether the cal-device producers, the threat of litigation
FDA or other authority can perform an inde- may be even more effective than government
pendent check of the software are important regulations for assuring the quality of medical
regulatory and cost issues. Furthermore, soft- software products.
ware developers may be wary of submitting
complete listings of their computer programs
because of concerns that competitors might get REFERENCES
a look at the source code via a request under the
(American) Freedom of Information Act. Bates, D. W., Cullen, D. J., Laird, N., Petersen,
One advocated approach to this challenge is L. A., Small, S. D., Servi, D., et al. (1995).
to begin with the idea that not all computer Incidence of adverse drug events and potential
errors are equally serious, as with the example adverse drug events: Implication for preven-
given above where drug-controller systems are tion. The Journal of the American Medical
seen to be riskier than, say, a simple monitor. Association, 274, 29-34.
Thus, risk-analysis techniques would require
Baxter, A. D. (1994). Respiratory depression
more exhaustive safeguards in software devel-
with patient-controlled analgesia. Canadian
oped for risky applications as compared to
Journal of Anaesthesia, 41, 87-90.
applications that are inherently safer. Software
developers must also continue to improve the Classen, D. C., Pesotnik, S. L., Evans, R. S., &
methods they use for documenting, writing, Burke, J. P. (1991). Computerized surveillance
testing, and maintaining medical computer proof adverse drug events in hospital patients. The
grams. The need for software-development Journal of the American Medical Associa-
methods that insure that programs are written in tion, 266, 2847-2851.
a consistent, easily understood, and reliable way
Coppess, M. A., Miller, J. M., Zipes, D. P., &
must be seen to be paramount. Too often, pro-
Groh, W. J. (1999). Software error resulting in
grammers include a description of what each
malfunction of an implantable cardioverter
part of their program does (the documentation)
defibrillator. Journal of Cardiovascular Elec-
only as an afterthought rather than starting their
trophysiology, 10, 871-873.
software writing from specifications provided in
a carefully developed design document. Crumpler, E. S., & Rudolph, H. (1997). FDA
As with buggy commercial software, in the software policy and regulation of medical de-
rush to the marketplace, when delays can put a vice software. Food and Drug Law Journal,
company at a competitive disadvantage, soft- 52, 511-516.
ware testing may lose out. This can happen
Doyle, D. J. (2003). Programming errors from
even in the medical environment. Delays in
patient-controlled analgesia. Canadian Jour-
releasing a software package to allow addi-
nal of Anaesthesia, 50, 855-856.
tional testing must be balanced against the
possibility of failing to detect any errors. When Emergency Care Research Institute (ECRI).
financial losses or injury result from software (1997). Abbott PCA Plus II patient-controlled
defects, the situation can lead to lawsuits. analgesia pumps prone to misprogramming re-
Software is clearly a product, and if it is defec-
400
sulting in narcotic overinfusions. Health De- Lin, L., Vicente, K. J., & Doyle, D. J. (2001).
vices, 26, 389-391. Patient safety, potential adverse drug events,
and medical device design: A human factors
Etches, R. C. (1994). Respiratory depression
engineering approach. Journal of Biomedical
associated patient-controlled analgesia: A re-
view of eight cases. Canadian Journal of
Anaesthesia, 41, 125-132. Martensson, K. (1993). Prevalidation of com-
puter systems regulating medical device manu-
Ferrante, F. M., Ostheimer, G. W., & Covino,
facturing processes. Medical Device Tech-
B. G. (1990). Patient-controlled analgesia.
nology, 4, 22-25.
Boston: Blackwell Scientific Publications.
Murfitt, R. R. (1990). United States govern-
Food and Drug Administration (FDA). (2005a).
ment regulation of medical device software: A
Design control guidance for medical device
review. Journal of Medical Engineering and
manufacturers. Retrieved from http://
Technology, 14, 111-113.
www.fda.gov/cdrh/comp/designgd.html
Schnider, P. (1996). FDA & clinical software
Food and Drug Administration (FDA). (2005b).
vendors: A line in the sand? Healthcare
Do it by design: An introduction to human
factors in medical devices. Retrieved from
http://www.fda.gov/cdrh/humfac/doit.html Trimbach, J. (1995). FDA regulation of clinical
software: What does this mean for the indus-
Food and Drug Administration (FDA). (2005c).
try? Healthcare Informatics, 12, 10-14.
Guidance for industry: General principle of
software validation. Retrieved from http:// Vicente, K. J. (2004). The human factor:
www.fda.gov/cdrh/comp/guidance/938.html Revolutionizing the way people live with
technology. New York: Routledge.
Geller, R. J. (1993). Meperidine in patient-
controlled analgesia: A near-fatal mishap. An- Vicente, K. J., Kada-Bekhaled, K., Hillel, G.,
esthesia and Analgesia, 76, 655-657. Cassano, A., & Orser, B. A. (2003). Program-
ming errors contribute to death from patient-
Kroemer, K. H. E. (2001). Ergonomics: How
controlled analgesia: Case report and estimate
to design for ease and efficiency (2nd ed.).
of probability. Canadian Journal of Anaes-
Upper Saddle River, NJ: Prentice Hall.
thesia, 50, 328-332.
Kwan, A. (1995). Overdose of morphine during
PCA. Anaesthesia, 50, 919.
KEY TERMS
Leveson, N., & Turner, C. (1993). An investi-
gation of the Therac-25 accidents. Computer, ECRI: Formerly the Emergency Care Re-
26, 18-41. search Institute, ECRI is an independent, non-
profit health-services research agency that
Lin, L., Isla, R., Doniz, K., Harkness, H., Vicente,
conducts research on patient safety issues.
K. J., & Doyle, D. J. (1998). Applying human
They maintain an online presence at http://
factors to the design of medical equipment:
www.ecri.org.
Patient-controlled analgesia. Journal of Clini-
cal Monitoring and Computing, 14, 253-263.
401
Ergonomics: A synonym for human-fac- designing systems that are safe, comfortable,
tors engineering, especially in the European effective, and easy to use.
literature.
Medical Device Software: Software in-
Food and Drug Administration (FDA): ternal to medical devices such as pacemakers,
A branch of the American government con- ventilators, anesthesia machines, and so forth.
cerned with safety in medical equipment (among Software written for medical devices should be
other things). Their Web site (http:// written to higher standards of safety and reli-
www.fda.gov) provides a variety of useful ability than ordinary software.
resources for equipment designers.
Patient-Controlled Analgesia (PCA): A
Human-Factors Engineering (HFE): The computer-based medical technology used to
branch of engineering devoted to the study of treat severe pain via the self-administration of
the interactions between humans and systems, analgesic agents such as morphine.
especially complex systems, with the goal of
402
403
Chapter LI
Alarm Design in Computerized
Medical Equipment
D. John Doyle
ABSTRACT
Alarms are frequently employed in safety-critical environments such as in aviation and

nuclear power plants. Now that microcomputer technology has revolutionized the design of
patient monitors for use in modern hospital operating rooms (ORs) and intensive care units
(ICUs), alarms are used in countless medical products ranging from infusion pumps to
ventilators. This is especially true in anesthesia/surgical and critical care environments. In
this chapter we examine the use of alarms in the acute care clinical environment, focusing on
their strengths and limitations in the setting of patient monitoring equipment.
INTRODUCTION brief review, we examine the use of alarms in

the acute-care clinical environment, focusing
Alarms are frequently employed in safety- on its strengths and limitations in the setting of
critical environments such as in aviation and patient-monitoring equipment.
nuclear power plants. Now that microcomputer
technology has revolutionized the design of
patient monitors for use in modern hospital ALARM DESIGN
operating rooms (ORs) and intensive-care units
(ICUs), alarms are used in countless medical For a clinical alarm system to be helpful when
products ranging from infusion pumps to venti- an adverse clinical situation occurs, an alarm
lators. This is especially true in anesthesia or must be sounded, the problem identified and
surgical and critical-care environments. In this corrected, and the patient treated. The earlier
Alarm Design in Computerized Medical Equipment
that an alarm occurs, the easier it is for the numbers, the clinician can be informed that
clinician to take actions to prevent patient injury there is a data-quality problem, such as might
(Schreiber & Schreiber, 1989). result from movement artifacts or from other
Alarm systems in medical equipment should causes.)
also be easy to (temporarily) silence, should New developments in intelligent or knowl-
offer power-on default settings to prevent the edge-based alarm technology have also been
inadvertent use of settings meant for a previous introduced commercially and experimentally
patient, and should incorporate a display that with the hope of improving patient safety (Koski,
enables the operator to detect problems or Sukuvaara, Makivirta, & Kari, 1994;
trends in its early stages. Also, the physical Westenskow, Orr, Simon, Bender, &
composition of auditory alarms should be de- Frankenberger, 1992). Such smart alarms com-
signed to convey a sense of urgency that bine expert-system techniques with alarm tech-
matches the actual urgency of the triggering nology either to provide more informative alarms,
clinical situation. Finally, the alarm should be to reduce the frequency of false alarms, or to
nonstartling, instantly recognizable to the trained provide initial suggestions about how to deal
respondent, and designed such that it would not with the problem that triggered the alarm. These
generally lead to anxiety in others, such as systems may offer the ability to change alarm
patients and their families (Quinn, 1989). priority with elapsed time (cascading alarms),
Alarms in medical equipment may be as or suppress secondary alarms that are the
straightforward as a simple threshold alarm consequence of a primary alarm condition.
such as an alarm that is activated when a Other smart alarm designs may suggest either
patient’s heart rate exceeds 120 beats per a diagnosis or an operator intervention to tell the
minute. Such simple alarm designs have the user more about how to handle the situation, or
potential to be enhanced in several interesting may offer a context-sensitive information dis-
ways. First, a duration condition might be added, play with specific clinical suggestions.
such as the requirement that a patient’s heart For instance, imagine a clinical monitoring
rate exceed 120 beats per minute for a period system that was aware that the patient being
of, say, 30 seconds before alarm activation monitored had severe coronary artery disease.
occurs. Note that such an arrangement has the Knowing that high heart rates are likely to
potential to reduce the frequency of false alarms, produce coronary ischemia in such individuals,
but may also delay the detection of some impor- the monitor might offer a default high-heart-
tant clinical events. rate alarm of 85 beats per minute in such cases
Second, the heart-rate data used for such an instead of a higher default heart-rate alarm
alarm system might be drawn from multiple (say, 120 beats per minute) that it might offer
sources, again in an effort to reduce the fre- for patients with normal hearts. Similarly, alarms
quency of false alarms (sensor fusion). As an can be annunciated according to the urgency of
example, an alarm system might require that the required response. Thus, a high-priority
both the heart rate obtained from the electro- alarm requiring immediate response by a clini-
cardiogram and the heart rate obtained from an cian might use a red indicator with a flashing
alternate data source (e.g., pulse oximeter de- frequency of 2 Hz, while a medium-priority
vice, arterial catheter) exceed (or fall below) a alarm requiring prompt response to deal with a
particular number. (As a bonus, when the condition might use a yellow indicator with a
sources of information fail to provide similar flashing frequency of 1 Hz, and a low-priority
404
alarm, one simply requiring awareness of a similar piezoelectric devices to provide audio
condition or indicating a change of status, might warning signals. The result was an unintegrated,
use a constant yellow indicator. awkward system of monitors often designed
Consider also the following hypothetical such that when an alarm sounded, the user had
example. Imagine a system in which if a patient’s to visually scan all the monitors to establish the
heart rate goes too low, say, under 30 beats per source of the alarm. Aware of these issues,
minute, it would suggest the administration of engineers, manufactures, and standards bodies
intravenous atropine, calling for the crash cart, set about to design integrated alarm systems in
and checking for low oxygen levels in the which all alarms are routed through a common
patient (using a pulse oximeter). It could even operator interface in order to facilitate alarm
alert the responsible physician or operating- recognition and management. In particular, the
room coordinator by pager to inform him or her International Standards Organization (ISO) has
that one of the patients is in trouble. developed ISO standard ISO/IEC 60601-1-8 to
A more mundane aspect of alarm design specify in detail how medical equipment alarms
concerns which alarms can be disabled and for should be implemented. Such developments
what period of time. As an example, a clinician offer considerable potential to improve patient
may want to be able to silence or disable a high- safety.
heart-rate alarm for 30 seconds or a minute in A final aspect of this topic concerns the
order to concentrate on treating the patient. eventual interconnection and computer control
Thus, an alarm policy engine may exist within of medical devices in the OR or ICU, such as
the software to ensure that the alarm limits the anesthesia workstation, patient cardiac
chosen are sensible and that important alarms monitors, and even drug-infusion pumps. Some
are not disabled or silenced for too long a time. of these efforts center on the medical informa-
In fact, some alarms should arguably not be tion bus (IEEE Standard 1073, http://
disabled under any circumstances. A good ex- www.ieee1073.org). The ability to regulate
ample is an alarm signaling a low concentration drug-infusion pumps from a central control
of oxygen in the gas mixture with which the point offers obvious advantages when many
patient’s lungs are being ventilated. This would infusion pumps are used, as is common during
be an appropriate alarm policy since false alarms cardiac surgery. A less appreciated benefit of
are vary rare for the oxygen-concentration such an arrangement is that in principle it allows
signal and low oxygen levels can quickly lead to for the automatic initial management of certain
brain damage. Similarly, an alarm policy engine hazardous clinical conditions. For example, it is
might indicate what alarms are to be escalated commonplace to use infusions of dopamine (a
to higher levels (or de-escalated to lower lev- drug that raises blood pressure) and sodium
els) as clinical circumstances evolve. nitroprusside (a drug that lowers blood pres-
Another aspect of alarm design involves sure) for cardiac surgery.
alarm integration. Until recent years, separate However, a clinician may sometimes mo-
monitors existed for tracking blood pressure, mentarily forget that a drug infusion is running
the electrocardiogram, arterial oxygen levels, and forget to turn off an existing nitroprusside
and so forth, and each monitor had its own infusion as the first response to the manage-
alarm system with its own default alarm con- ment of a hypotensive (low blood pressure)
ventions. Furthermore, all alarms tended to episode, or forget to discontinue an existing
sound the same as manufacturers all used dopamine infusion in order to treat an episode
405
of hypertension (excessive blood pressure). where a physician will be caring for many
The ability to control infusion pumps from a patients.)
central controlling station offers the potential to The potential value of alarms not withstand-
provide early warning about such mishaps. It ing, a high level of frustration exists among
would even be possible to automatically discon- healthcare workers about clinical alarm de-
tinue the offending infusion if no manual re- signs. Of all the complaints presented by clini-
sponse were detected within a given time pe- cians about current alarm technology, unques-
riod. Indeed, the concept of smart alarm sys- tionably the high rate of false alarms would be
tems can even be extended to provide for first on their list. For example, even small
automatic-initiation drugs to support blood pres- amounts of patient movement can introduce
sure should the crisis not be resolved in a timely artifacts into the patient’s electrocardiogram,
manner. Finally, as suggested above, smart pulse-oximeter signal, blood-pressure signal,
alarms can even be designed to automatically and other monitored variables. Quite often the
call for help should certain clinical crises not be artifact is not recognized to be garbage by the
resolved in an acceptable time period. alarm-management software, resulting in the
false annunciation of some alarm condition. So
common and so frustrating is this situation that
CLINICAL REALITIES many clinicians globally disable all alarms to
allow them to focus on caring for the patient
Despite all these developments, however, many rather than dealing with false alarms.
physicians working in the clinical trenches have Edworthy and Meredith (1994) reviewed
become quite cynical about many of the devel- many of the issues of alarm design from the
opments in medical alarm technology, regard- perspective of cognitive psychology, with an
ing them as more nuisances to deal with rather emphasis on the construction of effective alarm
than as contributions to patient care. For ex- sounds. They point out that there may be cir-
ample, many anesthesiologists note that be- cumstances where the excessive use of audi-
cause they are usually more or less perma- tory warnings may be counterproductive, while
nently situated near the patient, they are able to the principle of urgency mapping (involving a
keep an eye on the patient on a moment-to- graded series of alarms with increasing per-
moment basis so that they are usually aware of ceived urgency levels) may be helpful to pro-
any clinical deterioration before an alarm sounds. duce ergonomically sensible alarm systems.
When the alarm does sound, dealing with the In a study by Kestin, Miller, and Lockhart
alarm (e.g., silencing it, etc.) may distract the (1988), 50 patients undergoing anesthesia were
anesthesiologist from his or her efforts to monitored to determine the frequency of false
treat the patient (e.g., administering drugs to alarms. They found that 75% of all the alarms
restore the patient’s blood pressure into the overall were spurious and 22% represented a
normal range). (Of course, an important change above the upper alarm limits, but only
counterargument is that clinicians do not al- 3% corresponded to patient risk situations.
ways maintain 100% vigilance, especially With electrocardiogram alarms the situation
when fatigued or distracted, so alarms may was worse, with 81% of the alarms being
notify them of any life-threatening conditions spurious, 19% representing a change above the
that may have escaped their notice. This argu- upper alarm limits, and 0% representing a pa-
ment may be especially true in the ICU setting, tient risk situation. Similar results were ob-
406
tained with blood-pressure alarms and heart- Kam, P. C. A., Kam, A. C., & Thompson, J. F.
rate alarms from pulse oximeters. (1994). Noise pollution in the anaesthetic and
A similar study by Lawless (1994) reviewed intensive care environment. Anaesthesia, 49,
the false-alarm problem in a pediatric ICU. 982-986.
During a 7-day period, ICU staff recorded the
Kestin, I. G., Miller, B. R., & Lockhart, C. H.
type and number of alarms, categorizing them
(1988). Auditory alarms during anesthesia moni-
as false, significant, or due to staff manipula-
toring. Anesthesiology, 69, 106-109.
tions. Of 2,176 alarm soundings, 68% were
false and 26.5% were due to staff manipula- Koski, E. M., Sukuvaara, T., Makivirta, A., &
tions, while only 5.5% were significant. Of Kari, A. (1994). A knowledge-based alarm
interest, the pulse oximeter was the largest system for monitoring cardiac operated pa-
alarm source (44%), with ventilators (31%) tients: Assessment of clinical performance.
and electrocardiograms (24%) being other com- International Journal of Clinical Monitor-
mon alarm sources. In contrast, only 1% of ing and Computing, 11, 79-83.
alarms were from the capnometer, a device
Lawless, S. T. (1994). Crying wolf: False alarms
that measures expired carbon-dioxide levels
in pediatric intensive care unit. Critical Care
from the lung.
Medicine, 22, 981-985.
As already noted, not infrequently, anesthe-
siologists disable alarms at the beginning of a McIntyre, J. W. (1985). Ergonomics:
case to avoid false alarms and other alarm- Anaesthetist’s use of auditory alarms in the
related difficulties. McIntyre (1985) conducted operating room. International Journal of
a retrospective study in which he asked the Clinical Monitoring and Computing, 2, 47-
question, “Have you ever deliberately deacti- 55.
vated an audible alarm device at the start of a
Quinn, M. L. (1989). Semipractical alarms: A
case?” A majority (57%) of respondents re-
parable. Journal of Clinical Monitoring, 5,
plied, “Yes.”
196-200.
These and other studies tend to support two
points. First, most alarms (fortunately) do not Schreiber, P. J., & Schreiber, J. (1989). Struc-
signify a potentially critical medical event. Sec- tured alarm systems for the operating room.
ond, poorly performing alarm systems may Journal of Clinical Monitoring, 5, 201-204.
hinder, rather than help, the delivery of clinical
Westenskow, D. R., Orr, J. A., Simon, F. H.,
care, especially in environments where noise
Bender, H. J., & Frankenberger, H. (1992).
pollution may be problematic (Kam, Kam, &
Intelligent alarms reduce anesthesiologist’s re-
Thompson, 1994).
sponse time to critical faults. Anesthesiology,
77, 1074-1079.
REFERENCES
KEY TERMS
Edworthy, J., & Meredith, C. S. (1994). Cogni-
tive psychology and the design of alarm sounds. Alarm Integration: A design approach in
Medical Engineering and Physics, 16, 445- which all alarms are routed through a common
449. operator interface in order to facilitate alarm
recognition and management.
407
Alarm Policy Engine: Software structures ventional alarm technology either to provide
to ensure that the alarm limits chosen are more informative alarms, to reduce the fre-
clinically sensible and that important alarms are quency of false alarms, or to provide initial
not disabled or silenced for an excessive period suggestions about how to deal with the problem
of time. that triggered the alarm. They are also known
as smart alarms or knowledge-based alarms.
Cascading Alarms: Alarm systems that
have the ability to change alarm priority with Sensor Fusion: Employing data from mul-
elapsed time, such as a medium-priority apnea tiple sensors with the aim of improving system
alarm changing to a high-priority alarm when no reliability.
breaths are detected for an additional 30 sec-
Threshold Alarm: An alarm that is acti-
onds.
vated when a monitored parameter exceeds or
Intelligent Alarms: An alarm design that goes below set thresholds.
combines expert-system techniques with con-
408
410
Chapter LII
Organizational Factors in
Health Informatics
Michelle Brear
ABSTRACT
There is a general recognition that numerous organizational factors will influence the success
of an informatics intervention. This is supported by a body of evidence from multi-disciplinary
and health-specific research. Organizational factors are highly interrelated and the exact
nature and contribution of each to the success of an intervention is not clear. A health-specific
understanding and recognition of these factors is necessary if informatics applications are to
reach their potential in healthcare settings.
INTRODUCTION However, it remains rare for organizational

factors to be explicitly addressed in the imple-
The influence of organizational factors on the mentation process. As such, their contribution
success of informatics interventions in to the success or failure of informatics applica-
healthcare has been clearly demonstrated. This tions is not properly understood. This has impli-
health-specific research, informed by a larger cations for future interventions. Applications
body of evidence emerging from interdiscipli- that were not utilized or did not perform ad-
nary organizational, psychological, and socio- equately in a particular setting may be dis-
logical research, has confirmed the view that missed, while other less appropriate systems
organizational factors can be the decisive fac- may be adopted because organizational factors
tor in the success of an intervention (Lorenzi, influenced their success. Explicit study of the
Riley, Blythe, Southon, & Dixon, 1997). role of organizational factors on the implemen-
Organizational Factors in Health Informatics
tation of health-informatics interventions is nec- change, to identify a particular course through

essary to develop an understanding of their which the change can occur, and to take actions
influence in the healthcare context. to make it happen (Lorenzi, 2004). Resistance
Healthcare organizations tend to be highly to change occurs if users are not aware of the
task oriented, labor intensive, and dependent on need for change, are not convinced of the
interdisciplinary teamwork, so the influence of course of action set out, or are unable to carry
organizational factors within them may differ out the necessary action. It is the users, not the
considerably from the business settings in which technology, that should be the centre of the
they have traditionally been studied (Chau, change process as the decision to utilize the
2001). Health organizations are also increas- system is ultimately theirs (Berg).
ingly underresourced due to the global down- Even the best designed and well-intentioned
turn in government social spending, to health- informatics interventions are likely to lead to
sector privatization, and to aging populations. It productivity losses in the early stages and cre-
is these characteristics that necessitate the ate major changes (Lorenzi, 2004). The timely
rapid uptake of informatics applications, ca- and effective training of users can reduce the
pable of automating aspects of healthcare pro- disruption; however, it is not enough to ensure
vision and reducing labor intensity (Coiera, success as even a correctly used system can
2004). have far-reaching effects. Informaticians taking
From a technical perspective, rapid and a sociotechnical approach view the application
fundamental transformation of the healthcare as one component of a complex system (the
sector through informatics is achievable. How- health organization) whose introduction will dis-
ever, without a clear understanding of, and rupt other components of the system (e.g., pa-
ability to, manage organizational factors, it is tients and clinicians). They advocate design
unlikely that informatics applications will real- approaches that aim to create technology that
ize their potential in the health sector. This short fits within the complex system (Kaplan, 2001).
review provides an overview of the key organi- The multidisciplinary nature of health-sec-
zational factors influencing the success of tor organizations makes finding the correct fit
informatics interventions. It begins by position- challenging (Kaplan, 2001). A range of profes-
ing informatics interventions in the broader sionals with different needs, expectations, and
context of organizational change before dis- work norms are likely to use an application, and
cussing the current understanding of selected each will expect it to fit with his or her work
factors. practice. When an application does not fit,
resistance will increase. This is often due to
valid concerns about increased workload or the
INFORMATICS IMPLEMENTATION ability to care for patients (Timmons, 2003).
AS ORGANIZATIONAL CHANGE When systems do not fit, the best way to
overcome resistance is to change them. How-
Implementing informatics applications is es- ever, when they are essentially effective, resis-
sentially “a politically textured process of orga- tance can be overcome by changing people’s
nizational change” (Berg, 1999, p. 87), aimed at opinions or work norms. Organizational culture
achieving user acceptance and the utilization of and social networks, from which many of these
informatics applications. Organizational change norms and opinions arise, need to be understood
requires people to be aware of a need for and managed.
411
ORGANIZATIONAL CULTURE that do not threaten the values they are most
passionate about. The structure of an organiza-
Organizational culture is the set of shared norms, tion, and work patterns and roles of individuals
values, and tacit rules within which members of within it are influential and may be appropriate
an organization function (Lorenzi & Riley, 2000). areas to encourage change.
“Every culture supports a political and social
values system” (Lorenzi et al., 1997, p. 85) that Organizational Structure
will influence the reaction to an informatics
application. Healthcare settings often involve a The structure of an organization will affect the
professional hierarchy between doctors and way in which decisions are made, the type of
nurses, are characterized by high levels of leadership that emerges, and the way resis-
informal and disruptive communication, and tance is dealt with in the implementation pro-
place value on clinician-patient relationships cess. Flatter organizational structures tend to
and patient care. encourage the sharing of ideas, the emergence
It is necessary to identify and target the of innovation, and broader involvement in deci-
aspects of organizational culture presenting sion making (Leonard, Graham, & Bonacum,
opportunities for and barriers to success when 2004). In these types of organizations, manage-
changing the organization through an informatics ment tends to adopt a collaborative approach,
intervention. Managing change requires medi- working alongside, listening to, and involving
ating the influence of culture on events rather those working on the ground rather than making
than necessarily aiming to change it (Demeester, decisions on their behalf and communicating
1999). Where organizational culture and orders. Management is supportive, approach-
informatics applications appear incompatible, able, and accountable, and shows dedication to
adaptation of the application should be consid- continuous learning (Zimmerman et al., 1993).
ered. These types of organizations are more likely to
If it is not possible to modify the system, include practicing clinicians in formal decision-
success is dependent on changing the organiza- making bodies such as management commit-
tional culture to make it compatible. Cultural tees. They are also more likely to recognize,
change directly targeted at the strongly held encourage, and legitimize the role of grassroots
values of users may only increase resistance. If leaders with clinical credibility and presence,
the organizational culture supports a belief that and have a commitment to involving informatics
informatics applications undermine good clini- users in the implementation process.
cian-patient relationships, attempting to con- User involvement throughout the process
vince clinicians that good relationships with “leads to increased user acceptance and use by
their patients are not important is unlikely to be encouraging realistic expectations, facilitating
a successful strategy for winning acceptance the user’s system ownership, decreasing resis-
of the application. However, it may be possible, tance to change, and committing users to the
through an educational process, to convince system” (Lorenzi et al., 1997, p. 86). It also
clinicians that informatics applications do not allows a better definition of problems and solu-
necessarily undermine good relationships and tions from the user’s perspective, and develops
in the right conditions can even enhance them. a better understanding amongst users of the
Users may already be convinced of the need to application (Lorenzi et al.). Involving users in
change some aspects of organizational culture the design and implementation of a system is
412
more likely to result in applications suited to the Clinicians tend to be patient focused, re-
current work patterns of the intended users. quire an ability to maintain control of patient
care, and make decisions specific to individu-
Work Patterns and als, which computers are not capable of. They
Roles of Clinicians must be convinced that applications will not
jeopardize their ability to care for patients
Any informatics application must be compat- (Timmons, 2003). Overly prescriptive systems,
ible with the current work practices and values or those that attempt to take on the uniquely
of the organization (Greenhalgh, Robert, human quality of thinking, are unlikely to be
MacFarlane, Bate, & Kyriakidou, 2004; Kaplan, successful. Rouseau et al. (2004) noted inappli-
2001; Lorenzi & Riley, 2000). Compatibility will cable reminders were a barrier to effective use
differ between organizations and cultures, so and found that practitioners formed a habit of
applications require the capacity to be tailored ignoring all reminders. In a study of adherence
to the needs of individual organizations. Take to electronic HIV treatment reminders,
an electronic prescribing system, for example. Patterson, Nguyen, Halloran, and Asch (2004)
In one hospital, it may be used to enter prescrip- found that the inapplicability of reminders to
tion orders during ward rounds via a laptop many patients’ specific situations and the time
computer; however, a different hospital (or taken to document why the reminders were not
even another ward within that hospital) may adhered to were significant barriers to effec-
find it more appropriate to install the system on tive system use.
a computer terminal at the nurses’ stations so All applications will create some change to
that orders can be entered retrospectively. normal work patterns and roles. That is essen-
Bearing in mind the necessity to consider the tially what their implementation is intended to
unique context of individual organizations, it is do (Berg & Toussaint, 2003). Users need to be
possible to make some generalizations regard- realistically informed of and prepared for
ing the work patterns and roles of clinicians to changes to normal work practice. It is inevi-
broadly inform the design of informatics inter- table that users will expect some future benefit
ventions. from adapting their behaviors, and realistic
Doctors have traditionally worked with a communication of the likely benefits, particu-
high degree of professional autonomy and sta- larly if they are indirect or not clearly visible,
tus (Gagnon et al., 2003). Applications per- should form an integral part of the communica-
ceived to undermine their autonomy and status tion strategy.
as professionals or “subvert the art of medical
practice” (Kaplan, 2001, p. 4) are more likely to Communication
meet with resistance. In a qualitative study
examining factors influencing the adoption of a Communication binds individuals together and
CDSS (clinical decision-support system) in- is integral to the implementation process. With-
volving automatic clinical reminders, Rousseau, out effective communication, it is impossible to
McColl, Newton, Grimshaw, and Eccles (2004) lead, learn, make decisions, prepare individuals
found clinicians favored on-demand evidence for an intervention, or use the intervention
systems to automatically generated reminders. effectively (Zimmerman et al., 1993). The lack
The latter were perceived to be intrusive, often of communication, or ineffective communica-
inapplicable, and not particularly useful for tion that lacks trust, can negatively influence
making patient-management decisions.
413
the uptake of a technology (Ash, 1997). There should be actively elicited from users. Magrabi,
is no magic formula for effective communica- Westbrook, Coiera, and Gosling (2004) incor-
tion; however, there are some key principles porated two mechanisms for feedback into an
that should be applied to enhance the effective- online DSS. Users could volunteer feedback at
ness of communication. any time; however, it was also actively elicited
Communication must be timely. People re- by randomly prompting users. Once received,
quire time to digest information and prepare for feedback should be acted upon to adapt the
changes (Tilley & Chambers, 2004). However, application to meet user needs (Greenhalgh et
they also forget. Users who are trained to use al., 2004). For example, in response to clinician
a system months before its implementation may feedback, the patterns of automatically gener-
not remember how to use it when it finally ated reminders in a CDSS were altered to limit
arrives. Information must also be communi- the number of reminders perceived by clini-
cated at an appropriate time in the day. Disrupt- cians to be inapplicable. Communication through
ing a lunch break or patient care may not be the informal structures (e.g., gossiping in the tea
most appropriate way to inform users about a room) is inevitable, and the ability to manage it
new application. directly is limited. However, the less effective
Communication must be sincere and truth- formal communication mechanisms are, the
ful. Users must be offered an honest and real- more likely it is that communication will take
istic assessment of the potential negative con- place through informal channels in the social
sequences and expected benefits of an applica- network.
tion. For example, if an electronic medical-
record application is introduced, it may be real- SOCIAL NETWORKS
istic to expect its use to be more time consum-
ing in the initial stages while users master the A social network consists of the individuals,
system. However, in time, its ability to provide groups, and organizations with whom, and pat-
comprehensive patient information at the click terns of communication by which, individuals in
of a button may save time and provide better an organization interact. It is through social
quality information. It is also important to ac- networks that organizational culture and be-
knowledge unexpected benefits and problems haviors are reinforced and adapted (Lorenzi et
if they occur. al., 1997). The culture represented within a
Sources of communication have an immense social network can be influential. For example,
impact on the perceived message, its credibil- Gagnon et al. (2003) found that physicians who
ity, and its influence (Kaplan, 2001). A perceived social and professional responsibility
manufacturer’s leaflet declaring the new sys- from others in their social network to adopt
tem as easy to use is unlikely to carry much telemedicine applications had a stronger inten-
weight with users; however, a respected clini- tion to do so. The culture represented in each
cal peer conveying this information may be individual’s social network will differ, and each
quite influential. individual within will interact and be influenced
It is essential that the communication be differently.
recognized as multidimensional rather than a To properly understand a social network, it
one-way channel from management to users. is necessary to examine the interactions, not
Mechanisms for receiving feedback must be the individuals. The frequency of interaction is
created, and where it is not forthcoming, it important but should not be viewed in isolation
414
from the style of communication (e.g., formal Consideration should also be given to the
or informal), the type of communication (e.g., size and composition of teams. Teams of more
synchronous or asynchronous), the strength of than 15 tend to fragment into subteams, while
ties between the participants in an interaction, very small teams have a tendency to become
and the power relations involved (Katz, Lazer, cliquish, so ideally they should consist of 10 to
Arrow, Contractor, 2004). When implementing 15 members (Gosling et al., 2003). The work
an informatics application, interactions that environment may largely dictate a team’s com-
occur in clinical teams and with respected position. In healthcare settings, teams are usu-
opinion leaders are particularly influential in ally multidisciplinary. Individuals, however, are
relation to individuals’ decisions to utilize appli- more likely to create ties with those they per-
cations. ceive to be similar, so identifying similarities
amongst multidisciplinary teams is pertinent
Effective Teamwork (Katz et al., 2004). Identifying and utilizing the
influence of respected clinicians within the
The clinical team has been identified as the team can also be useful.
organizational unit most influential in the diffu-
sion of innovation (Gosling, Westbrook, & Clinical Champions and
Braithwaite, 2003). Well-functioning teams Opinion Leaders
facilitate effective communication, encourage
continuous learning, and offer a trusting envi- Respected clinicians with influence amongst
ronment in which ideas and issues can be their peers and who support the intervention
raised. Through these interactions, teams de- play an important role in convincing others of an
velop shared visions and common goals that application’s worth, as do those who oppose the
support the introduction of new innovations to intervention. They are commonly referred to as
fulfill these goals. As work in health organiza- clinical champions or opinion leaders. As the
tions is highly dependent on teamwork, and it is name champion tends to imply a positive influ-
teams, not individuals, that must adopt informatics ence and does not necessarily imply influence
applications, well-functioning teams are a pre- amongst peers (Lolock, Dopson, Chambers, &
requisite for successful informatics implemen- Gabbay, 2001), opinion leader will be used in
tation (Goldstein et al., 2004). this article.
In organizations where well-functioning Opinion leaders are individuals with the abil-
teams do not exist, an informatics intervention ity to influence others in the social network and
may be an opportunity to develop teams by who make a major personal commitment to
uniting individuals around a shared vision and diffusing information about an informatics ap-
common goals that the application can fulfill. plication (Lorenzi & Riley, 2000). Such diffu-
Doctors and nurses take on different roles and sion may have a negative or positive influence
responsibilities in the process of caring for and can discourage or encourage the adoption
patients, so a shared vision may not be immedi- of the application. For example, Ash (1997)
ately apparent. However, both groups ultimately identified the presence of champions as a sig-
work toward the goal of providing optimal nificant factor in the diffusion of e-mail in
patient care, so incorporating an application academic health-science centres. Conversely,
into a vision of improved patient care may be a Timmons (2003) discusses a “strong and ar-
way to unite users. ticulate” ward sister whose resistance to using
415
an electronic system in the ward was success- In particular, research has noted the influence
ful in preventing its implementation. It is not of organizational culture and social networks.
unusual to have both negative and positive Organizational culture, the shared norms and
opinion leaders within a network. values within which members of an organiza-
Whatever their persuasion, they tend to be tion function, influences the organization’s struc-
charismatic individuals with good interpersonal ture and patterns of communication. Social
relationships based on trust and understanding. networks, the individuals and groups with whom
They act through clinical conviction, generally one interacts and the interactions that occur
outside of the formal structures, and give appli- between them, are the social space in which
cations credibility at a local level. Their role is teams are formed and work, and in which
essentially informal, and, on the proviso their individuals are influenced, particularly by those
colleagues respect them, they tend to be self- individuals known as opinion leaders or clinical
appointed (Lolock et al., 2001). The role is champions. Organizational factors are highly
largely dependent on personal motivation and interrelated, and the exact nature and contribu-
conviction, and therefore it is difficult to for- tion of each to the success of an intervention is
malize. The potential alienating effect of being not clear. A health-specific understanding and
an opinion leader means that those without recognition of these factors is necessary if
sufficient commitment are likely to be reluctant informatics applications are to reach their po-
to take on such a role (Lolock et al.) tential in healthcare settings.
Despite a general consensus that successful
interventions are more likely when champions
are present, there is a lack of understanding REFERENCES
about exactly what it is they do, the circum-
stances in which they will be influential, and
Berg, M. (1999). Patient care information sys-
how best to describe them. As with each of the
tems and healthcare work: A sociotechnical
factors mentioned here, opinion leaders are one
approach. International Journal of Medical
small component of a large and complex system
(the health organization), and it is difficult to
isolate their effect. There is also considerable Berg, M., & Toussaint, P. (2003). The mantra
evidence to suggest that other factors, in par- of modeling and the forgotten powers of paper:
ticular the suitability of the application, influ- A sociotechnical view on the development of
ence the emergence of opinion leaders. Further process-oriented ICT in health care. Interna-
research is needed to identify how opinion tional Journal of Medical Informatics, 69,
leaders influence, and how the champions 223-234.
amongst them can be encouraged.
Chau, P. Y. K., & Hu, P. J.-H. (2002). Inves-
tigating healthcare professionals’ decisions to
accept telemedicine technology: An empirical
CONCLUSION
test of competing theories. Information and
Management, 39, 297-311.
There is a general recognition that numerous
organizational factors will influence the suc- Coiera, E. (2004). Four rules for the reinvention
cess of an informatics intervention. This is of healthcare. British Medical Journal, 328,
supported by a body of evidence from 1197-1199.
multidisciplinary and health-specific research.
416
Demeester, M. (1999). Cultural aspects of tion in providing safe care, quality and safety in
information technology implementation. Inter- health care, 13(Suppl. 1), i85-i90.
national Journal of Medical Informatics,
Lolock, L., Dopson, S., Chambers, D., & Gabbay,
56, 25-41.
J. (2001). Understanding the role of opinion
Gagnon, M.-P., Godin, G., Gagne, C., Fortin, J.- leaders in improving clinical effectiveness. So-
P., Lamothe, L., Reinharz, D., et al. (2003). An cial Science and Medicine, 53, 745-757.
adaption of the theory of interpersonal behaviour
Lorenzi, N. M. (2004). Beyond the gadgets:
to the study of telemedicine adoption by physi-
Non-technological barriers to information sys-
cians. International Journal of Medical
tems need to be overcome. British Medical
Journal, 328, 1146-1147.
Goldstein, M. K., Coleman, R. W., Tu, S. W.,
Lorenzi, N. M., & Riley, R. T. (2000). Manag-
Shankar, R. D., O’Connor, M. J., Musen, M.
ing change: An overview. Journal of the Ameri-
A., et al. (2004). Translating research into
can Medical Informatics Association, 7(2),
practice: Organisational issues in implementing
116-124.
automated decision support for hypertension in
three medical centers. Journal of the Ameri- Lorenzi, N. M., Riley, R. T., Blythe, A. J. C.,
can Medical Informatics Association, 11(5), Southon, G., & Dixon, B. J. (1997). Anteced-
368-376. ents of the people and organizational aspects of
medical informatics: Review of the literature.
Gosling, A. S., Westbrook, J. I., & Braithwaite,
Journal of the American Medical Informatics
J. (2003). Clinical team functioning and IT
Association, 4(2), 79-93.
innovation: A study of the diffusion of a point-
of care online evidence system. Journal of the Magrabi, F., Westbrook, J. I., Coiera, E. W., &
American Medical Informatics Association, Gosling, A. S. (2004). Clinicians’ assessments
10(3), 244-251. of the usefulness of online evidence to answer
clinical questions. Proceedings of the 11th
Greenhalgh, T., Robert, G., MacFarlane, F.,
World Congress on Medical Informatics (pp.
Bate, P., & Kyriakidou, O. (2004). Diffusion of
297-300).
innovations in service organisations: System-
atic review and recommendations. The Patterson, E. S., Nguyen, A. D., Halloran, J. P.,
Millbank Quarterly, 82(4), 581-629. & Asch, S. M. (2004). Human factors barriers
to the effective use of ten HIV clinical remind-
Kaplan, B. (2001). Evaluating informatics ap-
ers. Journal of the American Informatics
plications. Some alternative approaches: Theory,
Association, 11(1), 50-59.
social interactionism, and call for methodologi-
cal pluralism. International Journal of Medi- Rousseau, N., McColl, E., Newton, J.,
cal Informatics, 64, 39-56. Grimshaw, J., & Eccles, M. (2003). Practice-
based, longitudinal, qualitative interview study
Katz, N., Lazer, D., Arrow, H., & Contractor,
of computerised evidence based guidelines in
N. (2004). Network theory and small groups.
primary care. British Medical Journal, 326,
Small Group Research, 35(3), 307-332.
314-321.
Leonard, M., Graham, S., & Bonacum, D.
Tilley, S., & Chambers, M. (2004). The process
(2004). The human factor: The critical impor-
of implementing evidence-based practice: The
tance of effective teamwork and communica-
417
curates egg. Journal of Psychiatric and Men- Organizational Culture: Organizational

tal Health Nursing, 11, 117-119. culture is the set of shared norms, values, and
tacit rules within which members of an organi-
Timmons, S. (2003). Nurses resisting informa-
zation function.
tion technology. Nursing Inquiry, 10(4), 257-
269. Organizational Factors: In an informatics
context, organizational factors are factors re-
Zimmerman, J. E., Shortell, S. M., Rousseau,
lating to the culture and functioning of an orga-
D. M., Duffy, J., Gillies, R. R., Knaus, W. A.,
nization that, negatively or positively, influence
et al. (1993). Improving intensive care. Obser-
its ability to adapt to an informatics interven-
vations based on organisational case studies in
tion.
nine intensive care units: A prospective
multicenter study. Critical Care Medicine, Social Network: A social network consists
21(10), 1143-1451. of the individuals, groups, and organizations
with whom an individual interacts, and the
interactions that take place between the indi-
KEY TERMS
vidual and other components of his or her social
network.
Clinical Champion: Clinical champions are
opinion leaders who champion or encourage the Sociotechnical Approach: A
uptake of an application. sociotechnical approach is one that views
informatics applications as part of the broader
Opinion Leader: An opinion leader is an
social and political context within which they
individual, respected amongst his or her peers,
are implemented.
who acts out of clinical conviction to influence
the opinions of others vis-à-vis an informatics Teamwork: The cooperative effort of a
application. small group to achieve a specified outcome.
418
419
Chapter LIII
Measurement of
Cost and Economic Efficiency
in Healthcare
Panagiotis Danilakis
Pericles Robolas
ABSTRACT
The developments in new sectors, as in medical physics and biomedical technology, contributed
a lot to the progress and especially to the medical practice. However, these developments
created a complicated and costly environment in which health services have to take place. This
chapter reports on the sensitive sector of finances of health and particularly the ways of cost
assessment. Furthermore, it presents the significance of efficiency and effectiveness of
technology that is used in the area of health.
INTRODUCTION that influence the growth of expenses for hos-

pital care.
During the last 30 years, the financial situation In the economic bibliography, there is a
of the health industry has known swift growth. wealth of theoretical and econometric ap-
However, the limited resources that are being proaches that aim to analyze the cost of hospital
disposed internationally for the sector of health care. Naturally, the factors that determine the
point out the need of some concern for the cost operation of hospital systems differ from coun-
of provided services. This was the main reason try to country, as well as the structure and the
that turned the interest of economists of health organization of the private sector of care. Con-
from very early to the investigation of factors stants, functions, means, and the marginal cost
Measurement of Cost and Economic Efficiency in Healthcare
exist as the objects of study, and it is supported medicines, and diagnostic examinations
that not only the objectives but also the more constitute variable costs.
general system of the market and operation of • Total Cost: The total cost of a hospital
the hospital unit is taken into consideration and unit, which can the total monthly or annu-
is analyzed. One of the sectors that affects ally, is composed of these two main com-
considerably the increase or the reduction of ponents. This relation in simple form is
cost in the care of health is biomedical technol- expressed as:
ogy (biomedical engineering), which consti-
tutes a very important tool for hospital organi- Total Cost = Constant Cost + Variable Cost.
zations. Biomedical technology is developed
continuously, encouraging the creation of new Many times in economic analysis, be-
sectors of medical specialization, influencing cause the total costs do not provide enough
decisively hospital structures, and simultaneously information, we seek the determination of
puzzling its right use with particular accent in average or marginal costs.
the economic dimension. • Average Cost: With the term this we
mean his reason total cost to the number
of patients. The average (or medium) cost
COST ESTIMATION shows the cost per unit, and in a hospital it
can be expressed by a lot of indicators,
The operation of each hospital unit aims at the such as (a) the cost per patient, (b) the
offering of services of hospital care. Medical, cost per day of hospitalization, (c) the
nursing, administrative, and auxiliary personnel cost bed, and (d) the cost per doctor. The
as well as certain diagnostic and therapeutic significance of the medium cost is impor-
means are used for the offering of these ser- tant for the exercise of rational economic
vices. The personnel and the means, depending policy because it provides useful informa-
on the way and degree of utilization, have a tion on the change of costs concerning the
certain cost. This cost is constant or variable. productive activity of the hospital.
• Marginal Cost: With this term, we mean
• Constant Cost: Remains immutable and the change that befalls the costs of the
includes various aspects such as rents and hospital from a small change in patient
the cost of buildings and instruments. It is number. The marginal cost constitutes a
independent, therefore, of the number of useful significance in the economic analy-
patients or beds in the hospital. A constant sis because it shows how much an in-
cost is considered a datum and is pre- crease in the number of patients will af-
sented diagrammatically as a horizontal fect the medium cost.
line.
• Variable Cost: Influenced by the level of
the operation of a hospital and increases EVALUATION OF BIOMEDICAL
proportionally with the number of patients TECHNOLOGY
in the hospital, with the duration of their
hospitalization, and with the more general As we reported previously, one of the more
use of pharmaceutical products and diag- important sectors that present particular inter-
nostic examinations. Sanitary material, est in analysis and more generally in the inves-
420
tigation of the relation of cost and effectiveness namic models of systems (Forrester),
is the evaluation of biomedical technology. It analyses of cost and utilities, analyses of
was built in the ’70s by the intense efforts in the cost and effectiveness, meta-analyses
search of evaluation methodologies of tech-
nologies, and was shaped by the incorporation According to the initial objective of the
of the special analyses of cost and effective- evaluation of biomedical technology, the collec-
ness, and cost and utility. The evaluation of tion and treatment of data so that the essential
biomedical technology constitutes a sector of information is produced is required in decision
the social evaluation of technologies (or social making with regard to investments in new mainly
technology assessment, STA), for which the medical technologies. For this aim, a mesh of
following definition was in effect in the ’70s. points of evaluation is proposed that is shaped
by considered dimensions and levels, and that
• The (social) evaluation of technologies is fixes the systematic structure of information
the dynamic process of the multidimen- and also criteria for evaluation. For the con-
sional analysis, forecast, and estimate of frontation of each question, a particular mesh
all the short-term and long-term effects of of criteria is used that is determined by consid-
technology in the individual, in society, and ered dimensions, the levels, and the time hori-
in the environment. zon.
• The objective of the evaluation of tech-
nologies is the support of decision making • Dimensions: The evaluation of biomedi-
(in government, in the hospital, in the cal technology is not only carried out in the
industry, and so forth) for the wider com- medicine and economic dimensions, but
prehension of activities. Society produces, also in the technique dimension. For the
influences, and uses technologies so that indirect cost of medical technology, vari-
they achieve its objectives in the best ous technical criteria are proposed in re-
possible way. gard to the level of the development of the
• The aim of the evaluation of technologies technology, the type used, safety, the ef-
is toward the preparation of policies and fects of the way of work, the level of the
plans of action that will be useful as points development and support of software, and
of departure for decision making, via which so forth. Depending on the examined prob-
the negative social repercussions of tech- lem of evaluation, additional dimensions
nologies will be minimized. are also selected such as further training.
• Organization Levels: The evaluation of
The methods of the evaluation of technolo- the social, ethics, law, and education di-
gies are the following: mensions is carried out on the following
levels: (a) the patient (the individual pa-
• Heuristic Methods: Historical retro- tient and his or her family), (b) the popu-
spections, consensus conferences, hear- lation of the service, (c) the hospital be-
ings, questionnaires, Delphi-analysis cause the technology will alter its struc-
scripts ture and way of work according to the
• Concise Methods: Analyses of sys- population of the service (point of depar-
tems, morphological method (Zwicky), ture for the measurement needs and the
analyses of surge flows (Leontieff), dy- demand), and (d) the system of hospitals
421
that serves the population of the prefec- which, however, in practice, are proved
ture or sanitary region. inapplicable (or at least applicable with
• Time Horizon: The time horizon of the difficulty). It is presupposed that the com-
evaluation should harmonize itself with parison concerns therapeutic medical in-
the practical requirements: the budget (1 struments and groups as far as for diag-
year), the plan for investments (e.g., 5 nostics should they be taken into consider-
years), and medium duration of the effi- ation as other criteria of evaluation.
cient operation of medical instruments (5 • Optimization of Effectiveness: For
to 10 years). whom will the medical method (or instru-
• Evaluation Steps: In each point of evalu- ment) be more effective so that some
ation (section of dimensions and levels of predetermined objective is achieved within
evaluation), the following correspond: (a) the given amount of investment? This
the placed questions, (b) the methods of question is placed for the choice of medi-
collection and treatment essential for the cal instruments for the achievement of
evaluation of data, and (c) the criteria, concrete aims when there are limited eco-
according to which the data are evaluated nomic resources. Examples are the choice
so that the information that is required for of equipment for arthroscopy and
decision making is produced. Most ques- arthroscopy interventions.
tions that are placed (or should be placed) • Minimization of Cost: For whom is a
to be answered for the evaluation of biomedical instrument more economical when
medical technology at the staff level in a given the achievement objective? The ef-
hospital, that is to say, at the processes of fectiveness under examination for medi-
the syntax of the budget for biomedical cal instruments can be given or even non-
technology, belong to one of the following existent. This question is usually exam-
categories. ined for the choice of diagnostic instru-
• Optimization of the Relation of Cost ments. The answer should promote the
and Effectiveness: For the achievement choice of the more economically accept-
of different (and desirable) medical objec- able and widespread instrument. Often,
tives, to whom does medical technology the choice of the medical instrument should
present the better relation of cost and correspond with a better way to meet
effectiveness? This question is placed needs. Examples are the choice of appli-
when, for example, it should be decided if ances for the automation of biochemical
investments should be made for the cre- laboratories, diagnostic technology and
ation of a new radiological clinic or for the appliances for the provision of physio-
automation of a biochemical laboratory, or therapy, and electrocardiographs, moni-
when deciding on the disposal of existing tors, and automated systems for the ex-
economic resources for equipment for amination of neurology or ophthalmology.
either a (new) unit for artificial kidneys or
for a unit for the intensive hospitalization
of nurslings. Still, for the decision-making CONCLUSION
if is in my interest the supply of chirurgical
instrument Laser opposite of some other. The big growth of science and technology
Basically, such types of questions are during the last decades of the 20th century
faced with analyses of cost and utilities, resulted in the rapid development of medical
422
science and its benefits, but also of the models Lovel, C. (1995). Econometric efficiency analy-
of the demand for services of health. These sis: A policy oriented review. European Jour-
developments, which influenced all the profes- nal of Operational Research, 452-461.
sions of health, helped so that the benefits of
services are carried out henceforth in the frames
KEY TERMS
of a completely organized system. Such a sys-
tem, therefore, will be aim to combine human,
Biomedical Technology: The sector that,
technological, and natural resources so that the
with the application of scientific and techno-
exercise of modern medical practice is fea-
logical methods, aims toward the manufacture
sible. It is exceptionally critical for profession-
of medical technological equipment, for ex-
als and future professionals of health to com-
ample, monitors and many other appliances. In
prehend not only the sectors of their profes-
a lot of ways, this sector is the same as medical
sions, but also the nature, organzation, and
information technology. The two sectors use
operation of the systems inside that make their
the PC (personal computer), which also has a
professions, as well how important it is to
secondary role in the two sectors. Thus, medi-
evaluate the sector of health and its economic
cal biotechnology focuses on the appliances,
output generally.
and medical information technology focuses on
the knowledge and information, and on the
management of these with a PC.
REFERENCES
Delphi Analysis: A method of the achieve-
Abel-Smith, B. (1976). Value for money in ment of consent between scientists of prestige
health services. London: Heinmann. on some difficult question.
Abel-Smith, B. (1986). The world economic Effectiveness: A metre that estimates how
crisis, part 2: Health manpower out of balance. much a product, service, or program achieves
Health Policy and Planning, 309-316. its objectives under real conditions. As the
effectiveness of medical equipment is improved,
Butler, J. R. (1995). Hospital cost analysis.
this contribution is in the treating of illness.
Dordrecht, The Netherlands: Kluwer Academic
Publishers. Efficiency: The achievement of concrete
results with the minimal possible cost of the
Evans, R. G. (1971). Behavioural cost func-
occasion, or the maximization of results with a
tions for hospitals. Canadian Journal of Eco-
given cost for the occasion.
nomics, 198-215.
Evaluation of Technology: The evalua-
Evans, R. G. (1972). Information theory and
tion of technology is a dynamic process that
the analysis of hospital cost structure. Cana-
examines short-term and long-term social, eco-
dian Journal of Economics, 398-418.
nomic, and legal repercussions from the use of
Feeny, D. (1986). New health technologies: this technology.
Their effect on health and cost of health care.
Medical Information Technology: It is
In Health care technology: Effectiveness,
the way that medical personnel collect, orga-
efficiency and pubic policy (pp. 5-24). Insti-
nize, and use data, information, and knowledge
tute for Research and Public Policy.
in decision making, and the export of conclu-
423
sions contributing to the quality care of the Reliability of Technical Equipment: It is

patient. the faculty of an element, appliance, or techno-
logical system to satisfy those requirements
Meta-Analysis: A method of evaluating
that are dictated from the aim for which it was
many clashing studies around some question,
manufactured.
and the export of useful conclusions.
424
425
Chapter LIV
Understanding Telemedicine
with Innovative Systems
Irene Berikou
Athens University of Economics and Business, Greece
Athina A. Lazakidou
ABSTRACT
A system of innovation (SI) is a new approach for the study of innovations as an endogenous
part of the economy. An SI can be defined as encompassing all the important factors that
influence the development, diffusion, and use of innovations as well as the relations between
these factors. For example, the SI approach is also used as a framework for designing
innovation policy at the national level in some EU (European Union) member countries such
as Finland and Ireland. It is simply at the center of modern thinking about innovation and its
relation to economic growth, competitiveness, and employment. In this chapter, the adaptation
of this framework for telemedicine applications is simply presented.
INTRODUCTION facts. The successful implementation of

telemedicine services requires its underlying
The healthcare sector is in need for innovation technologies to be configured in a way that
during the process of delivering services, and meets the particular needs of the healthcare
ICT has the potential to change the organization and social care providers, the individuals re-
and delivery of health and social services in ceiving care, and all other stakeholders. Tech-
Europe. The slow rate of telemedicine adoption nical success alone will not result in the wide-
in a country’s healthcare system is not only spread diffusion of telecare technologies
caused by technology factors related to the (Barlow, Bayer, & Curry, 2003).
speed, security, and capacity of the national Academic researchers have clarified little
infrastructure or by inferior technological arti- in understanding the phenomenon of
Understanding Telemedicine with Innovative Systems
telemedicine in its wider economic and social a much broader way than technical orientation.
performance. The complexity of ICT in He specifically mentions new forms of organi-
healthcare has also confused the definition of zation and new combinations.
telemedicine. Many terms have been used by Innovation is the new creation of economic
academics, researchers, healthcare profession- significance of either a material or intangible
als, informaticians, the ICT industry, and gov- kind (Edquist, 2001). Innovations may be brand
ernmental and nongovernmental organizations new, but often they are new combinations of
to describe the phenomenon. existing elements. A useful taxonomy is to
This chapter is an attempt to use a rather divide innovations into new products and new
new economic framework in order to define the processes. Product innovations may be goods
value and the placement of ICT within the or services. It is a matter of what is being
healthcare sector: the way that it transforms produced. Process innovations may be techno-
previous healthcare delivery into a coproduced logical or organizational. They concern how
activity. Therefore, this effort might prove to be goods and services are produced.
helpful to facilitate implementation issues and Today, it is widely recognized that techno-
to diffuse telemedicine in each healthcare sys- logical change is the primary engine for eco-
tem, no matter if we see it from either a micro nomic development. Innovation, at the heart of
or macro perspective. technological change, is essentially the innova-
tion process that depends upon the accumula-
tion and development of relevant knowledge of
DEFINITION OF TELEMEDICINE a wide variety. Certainly, individual firms play
a crucial role in the development of specific
Telemedicine is an information technology- innovations, but the process that nurtures and
driven application that provides the means of disseminates technological change involves a
delivering healthcare services at a distance. complex web of interactions among a range of
The term telemedicine is constituted by two other firms, organizations, and institutions
words, tele and medicine. The first component (Fisher, 2001).
is the Greek word τηλε that means from far A system of innovation (SI) is a new ap-
away or at a distance, and the second compo- proach for the study of innovations as an endog-
nent is of Latin origin from the word mederi, enous part of the economy. The SI has emerged
which means healing. It involves not only medi- only during the last decade or so. An SI can be
cal activities for ill patients, but also public- defined as encompassing all the important fac-
health activities involving well people. In other tors that influence the development, diffusion,
words, telehealth is a process and not a technol- and use of innovations as well as the relations
ogy, including many different healthcare activi- between these factors. These factors can be
ties carried out at a distance (Riva, 2000). studied in a national, regional, or sectoral con-
text; therefore, national, regional, and sectoral
Origins of Systems-of-Innovation systems of innovation coexist and complement
Approach in Economist Cycles each other. The SI has diffused surprisingly
fast in the academic world as well as in the
Joseph Schumpeter, a classical economist who realms of public innovation making and firm
wrote his famous theory for entrepreneurship, innovation strategy formulation. That is, SI
in 1939 was the first to conceive innovation in provides a framework, not a theory, of analysis
426
for identifying specific policy issues to under- Tether and Metcalfe (2001) made a major
stand differences between national and re- contribution recently to develop the framework
gional economies and various ways to support for services, shifting the interest from the manu-
technological change. For example, the SI ap- facturer. The adaptation of this framework for
proach is also used as a framework for design- telemedicine applications is the following.
ing innovation policy at the national level in
some EU (European Union) member countries
such as Finland and Ireland. It is simply at the TELEMEDICINE AS A SYSTEM
center of modern thinking about innovation and OF INNOVATION IN SERVICES
its relation to economic growth, competitive-
ness, and employment. In an SI approach, there It is known that the sector of healthcare is a
are at least three main categories of system service sector. Usually, the definition of a
failure (Edquist, 2001), which are different service is embedded with intangibility as it lacks
from market failures as system failures are an autonomous physical existence contrary to
identified through comparisons between exist- products.
ing systems. This sense has reason as the service of
healthcare can change our physical being, and
• Organizations in the system of innovation in order to offer health services, the process is
may be inappropriate or missing. often indistinguishable from the product, while
• Institutions or codification may be inap- in the long term the outcome is intangible. All
propriate or missing. transformation processes transform combina-
• Interactions or links of the knowledge tions of material, energy, and information into
flow between these elements in the SI new, more highly valued combinations of these
might be inappropriate or missing. elements (Tether & Metcalfe, 2001). The dif-
ferentiated nature of services and the multiple
The development of SI approaches has been ways that service activities can be defined, or
influenced by different theories of innovation the way telemedicine is defined in the healthcare
such as interactive learning theories and evolu- sphere, are central to an understanding of the
tionary theories. The first two books exclu- complexities of innovation systems of services.
sively devoted to the analysis of national sys- The nature of services and the transformations
tems of innovation were Lundvall (1992) and they provide also tend to have a significant
Nelson (1993). However, Chris Freeman (1987) bearing on their organizational form. We con-
first used the expression in published form. sider that telemedicine offers an innovation
Regional systems of innovation have been ad- process into the chain of healthcare delivery
dressed, for example, in Braczyk, Cooke, and (Berikou, 2004).
Heidenreich (1998). Sectoral SIs have been Another characteristic of services is that
analysed in Carlsson (1995), Breschi and they are coproduced, involving simultaneous
Malerba (1997), and Nelson and Mowery relationships between producers-physicians and
(1999). All these books and others are re- consumers-patients. With telemedicine, we have
viewed in the introduction to Edquist and one more new relationship to consider: that
McKelvey (2000), which is a collection of 43 between back office experts and front-office
central articles on systems of innovation. physicians in remote areas. The service of
427
healthcare with telemedicine brings to the fore making adjustments for telemedicine, the fol-
the interactive aspects of a dual character. lowing apply.
These complex interactions constitute multiple
systems of innovation (den Hertog, in press). • Services are not normally engaged in the
As telemedicine may transform previous en- production of tangible products, but cover
gagement with healthcare, we may consider a huge range of diverse activities associ-
that the core function of services in health ated with various types of transformation,
today is the service delivery of better quality that is, the transformation of people (physi-
and equity faster than before, and the periph- cally and mentally), things, and informa-
eral form of the service can vary enormously tion. There are important connections be-
from teleconsulting to telemonitoring at home tween service innovation (rearrangement
(Berikou, 2004). in the delivery of healthcare) and artifact
As information technologies and networks innovation (telemedicine applications) as
have developed, so have new forms of coordi- normally defined.
nation and delivery. Technological develop- • The study of services brings to the fore to
ments have reduced the power of location for a greater extent the interrelationships be-
the service of healthcare. Where traditionally tween business models, organizational
most services were provided locally, with con- forms, technology, and outputs. The sig-
sumers often coming to the service provider, nificance of market knowledge and pro-
now many services are provided at arm’s length, cedural knowledge is highlighted.
for example, through the Internet. Arm’s-length • Healthcare with telemedicine shows high
provision typically allows the exploitation of degrees of interaction and interdepen-
economies of scale, which provide advantages dency between the service provider (phy-
over traditional, local provision (Tether & sician) and the service user (patient), be-
Metcalfe, 2004). This encourages the develop- tween the service provider (back-office
ment of cooperation between the front office expert) and the service user (physician in
(which deals directly with patients) and the remote areas, front office), as well as
back office that carries out the service pro- between the provider (front office or back
cesses. It can bring a significant impact on the office) and equipment suppliers. Such in-
organization in terms of the size of the healthcare teraction and interdependency is a central
unit, the number of sites, and the location of feature of systems of innovation. The
functions. system involves a wide range of agents
As Gallouij and Weinstein mention, the analy- from many different sectors (IT suppliers,
sis of innovation is difficult because of the university researchers, research and de-
“fuzzy” nature of service outputs, in which it velopment companies and institutes, pub-
can be difficult to distinguish the service prod- lic sector, healthcare levels, physicians).
uct from the background process or the organi- An interesting feature of these systems is
zation of provision. Miozo and Soete have done that the agents involved and the interrela-
some attempts at the taxonomy of innovation in tionships between these agents can change
services in order to adapt the Pavitt taxonomy over time; thus, the boundaries of the
into services. system are not fixed but are dynamic and
From a system’s perspective on the service evolve.
innovation of Tether and Metcalfe (2001), and
428
• We consider that systems of innovation electronic applications are implemented. Elec-

often develop around identifiable se- tronic cooperation with all partners in the sector
quences of problems and opportunities, and with other areas such as pharmacies, rela-
which are themselves framed by a num- tives, and patients is required. The boundaries
ber of contingencies including the regula- of the system are changing as new agents are
tory, cultural, and technological context. included, including common standards, the free
In this way, the problem or opportunity flow of information between different applica-
(telemedicine in remote areas) at the heart tions, and the national implementation of impor-
of the system of innovation becomes the tant applications. More institutions playing by
focusing device around which it is devel- the rules of the game can constrain, coordinate,
oped. As the problem or opportunity and enable activity. ICT development must
changes or is redefined, so the system of follow organizational development, changes in
telemedicine can change, changing the work processes, and new forms of cooperation
agents involved and the relations between and divisions of labour. This means that learn-
these agents: It is a dynamic, distributed ing processes and knowledge bases are going
process (Coombs, Harvey, & Tether, to be established.
2001). One important aspect of this view
is that, individually and collectively, firms
and communities of practitioners (provid- REFERENCES
ers and users) take a leading role in as-
sembling innovation systems in the pursuit Barlow, J., Bayer, S., & Curry, D. (2003).
of a competitive advantage. Integrating telecare into mainstream care de-
livery. The IPTS Report.
Berikou, E. (2004). Telemedicine systems: A
CONCLUSION
comparison with Norway for innovation and
implementation in Greece. Unpublished
Such a framework may help us describe the
master’s thesis, Linkoping University, Tema-T
context of innovation in services for ICT in the
Department, Sweden.
healthcare sector, but the field is still frag-
mented (den Hertog, in press). The problem or Braczyk, H.-J., Cooke, P., & Heidenreich, M.
opportunity framework must be seen as a con- (1998). Regional innovation systems: The
text-specific policy tool and not be restricted role of governances in a globalized world.
into a solution for all patterns. London: UCL Press.
Generalizations under the influence of this
Breschi, S., & Malerba, F. (1997). Sectoral
framework in telemedicine practice can be
innovation systems: Technological regimes,
given as such: There must be greater capacity
Scumpeterian dynamics and spatial boundaries.
and broadband development in between hospi-
In C. Edquist (Ed.), Systems of innovation:
tals, and between hospitals and primary health
Technologies, institutions and organizations.
services in order to enable the use of telemedical
London: Pinter/Cassel Academic.
and IT applications. The clarification of re-
sponsibility, rules, guidelines, and rates in con- Coombs, R., Harvey, M., & Tether, B. S.
nection with telemedical consultations must be (2001). Analysing distributed innovation-
a prerequisite. Papers must be removed when
429
processes: A CRIC position paper. UK: CRIC, KEY TERMS

University of Manchester, & UMIST.
Adoption: Rogers (2003) differentiates the
Den Hertog, F. (in press). Mapping health care
adoption process from the diffusion process in
innovation: Tracing walls & ceilings. MERIT.
that the diffusion process occurs within society
Edquist, C. (2001). Innovation policy: A sys- as a group process, whereas the adoption pro-
temic approach. In B. A. Lundvall & Archingi cess pertains to an individual. Rogers defines
(Eds.), The globalizing learning economy. the adoption process as “the mental process
Oxford University Press. through which an individual passes from first
hearing about an innovation to final adoption.”
Edquist, C., & McKelvey, M. (Eds.). (2000).
Systems of innovation: Growth, competitive- Codification: A process of knowledge con-
ness and employment. Gheltenham: Edward version between the two forms: tacit and ex-
Elgar. plicit.
Fischer, M. M. (2001). Innovation, knowledge Diffusion: The process in which an innova-
creation and systems of innovation. Annals of tion is communicated through certain channels
Regional Science, 35, 199-216. over time among the members of a social
system (Rogers, 2003).
Freeman, C. (1987). Technology policy and
economic performance: Lessons from Ja- Evolutionary Economics: This science
pan. London: Pinter. describes the unleashing of a process of tech-
nological and institutional innovation that dis-
Lundvall, B.-A. (1992). National systems of
covers more survival value for the costs in-
innovation: Towards a theory of innovation
curred than competing alternatives. It explains
and interactive learning. London: Pinter.
the evidence and suggests that economies grow
Nelson, R. R. (1993). National innovation faster when there is diverse participation, com-
systems: A comparative study. Oxford: Ox- petitive survival, and the replication of success.
ford University Press.
Implementation: The process of putting
Nonaka, I. (1994). A dynamic theory of organi- all program functions and activities into place;
zational knowledge creation. Organization it is part of the innovation-decision process
Science, 5(1). quoted by Rogers (2003).
Riva, G. (2000). From telehealth to e-health: Interactive Learning Theories: Learn-
Internet and distributed virtual reality in health ing by interacting between users and produc-
care. Cyberpsychology & Behavior, 3(6). ers, embodying knowledge in new services and
products.
Rogers, E. M. (2003). Diffusion of innova-
tions (5th ed.). New York: Free Press. Knowledge Flow: The way knowledge
travels and grows within an organization.
Tether, B. S., & Metcalfe, J. S. (2001). Ser-
vices and systems of innovation. Paper pre-
sented at the DRUID Academy Winter 2002
PHP Conference.
430
431
Chapter LV
A Capacity Building Approach to
Health Literacy through ICTs
Lyn Simpson
Queensland University of Technology, Australia
Melinda Stockwell
Susan Leggett
Leanne Wood
Danielle Penn
ABSTRACT
There has been substantial interest in delivering ICT training options to rural and remote
areas of Queensland, Australia, in order to bridge the rural-urban divide. But there is more
than just education and training going on: Participants are being empowered to gain new
skills and confidence, form new networks, become active in the community, and be proactive
in addressing their own health and well-being needs.
THE RURAL AND & Seinen, 2001; Wagenfeld, Murray, Mohatt,

REMOTE CONTEXT & DeBruyn, 1997). This applies to essential
health services, to services that enable indi-
Rural and remote populations often experience viduals and communities to gain the skills nec-
poor access to services (Simpson, Wood, Daws, essary to participate in the social changes af-
A Capacity Building Approach to Health Literacy through ICTs
fecting the population, and to the peer-support literacy skills. A socially inclusive society re-
services needed by isolated professionals. In quires informed communities that have the
Australia, long-standing factors in service de- means, skills, and opportunities to communicate
livery to rural communities, such as sparse (IBM, 1997). For those unable to meet these
population, distance, and limited availability of criteria because of age, ethnicity, disability,
public transport, are being exacerbated by the income, or circumstance, difficulties associ-
reduction in and withdrawal of existing face-to- ated with the acquisition of everyday informa-
face services. In terms of health services, tion via the Internet can potentially create
access constitutes a significant issue for rural considerable frustration and distress, increas-
communities. At the same time, the traditional ing the degree to which these people are
jobs base for these communities in primary marginalized within their community, and im-
production is shrinking, resulting in psychologi- pacting their health and well-being.
cal pressures and the need for the re-skilling of
many of those people formerly employed in
such industries. THE CASE-STUDY PROJECTS
Health literacy, understood in its broadest
sense, is a key issue for these communities. The projects build on earlier work that identi-
Parker (2000, p. 280) notes, “… for those with fied the difficulty in accessing accurate current
limited health literacy, as health care is becom- information and in obtaining appropriate health
ing increasingly complex and health informa- and well-being support for these populations.
tion is becoming more diffuse in the public Interviewees in fieldwork for Creating Rural
domain, there is more reliance on written mate- Connections (Simpson et al., 2001) reported a
rials to educate and inform people about their variety of information needs, including more
health.” As governments at all levels seek timely access to a wider range of information,
ways to simplify, and reduce the costs of, the and the desire for specific information in re-
task of meeting health-service needs, the at- sponse to an identified need (such as to address
tractions of e-government for service delivery a health problem), to locate employment, or to
to a receptive “wired” community are strong improve the family business.
(and ICT vendors have encouraged that attrac- Community members identified two levels
tion). However, for those community members of specific need. First is the need for access to
used to the supportive environment of face-to- specialist services, including medical services
face service delivery and unfamiliar with ICTs, and counseling; ongoing access to help, com-
the focus on service provision via the Internet panionship, and mentoring; community-service
creates new challenges. The incentive for com- databases to facilitate better networking and
puter literacy (Hamm as cited in Loader & referral services; improved access to services
Keeble, 2004) is strengthening. for disability groups, particularly services that
The problems associated with negotiating have the potential to overcome the effects of
the changed rural social and service environ- communication limitations and personal isola-
ment impact particularly those people who have tion; and opportunities to identify and reinforce
characteristics that may intensify their isolation existing support networks. They also identified
and lack of access to information, including the need for re-skilling regarding computing
people of cultural and linguistic diversity skills as a necessary tool in the changing work
(CALD), disabled people, and people with low and social environment. The projects devel-
432
oped to help address these needs have provided marginalized groups that will assist them
an inclusive and supportive learning environ- in conducting training that is appropriate
ment for individuals who had experienced for their client base, taking into account
marginalization in their communities. The focus the specific needs arising from age, gen-
in each project has been on Bella and Bishop’s der, rurality, ethnicity, literacy, and/or dis-
(2004) “building on” community capacity: “help- ability.
ing people believe in their own skills” (p. 13).
While these aims focus on the technical and
training needs of people in rural and remote
HOW THE PROJECTS OPERATE communities, it is how these goals are enacted
that creates the powerful social and health
The projects are built on the recognition that outcomes. Projects intended to deliver ICT
social isolation and access to information are training to marginalized groups have often failed.
significant issues for marginalized people in When asked what the problems were with
far-north Queensland and, in particular, for computer training programs they had attempted,
people who speak little or no English, for indig- participants identified the following:
enous peoples, and for disabled people. Draw-
ing on research that has shown the Internet to • Modules not relevant to their needs
be an effective, appropriate means of meeting • Training moved too quickly for them to
such needs for a diverse range of people (Com- keep up with the group
munication Centre, 2001; Simpson et al., 2001), • Could not understand the trainer
the projects sought to: • Class sizes too big
• Felt alienated from the other students
• enhance the skills base in small communi- • Felt alienated from the teacher
ties; • Too old
• help create a more informed community • Left it too late
and a more equitable society in part via
access to information by marginalized As Feinstein, Hammond, Woods, Preston,
groups; and Bynner (2003) found, for those who have
• facilitate the sharing of skills through the been away from learning for some time, there
development and creation of online and are particular qualities in the learning environ-
face-to-face social networks; ment that enable the wider benefits to flow.
• develop an innovative and transferable Such programs need to “ensure engagement …
process that will be relevant to other [C]lasses that are not sensitive to learners’
groups and other communities; needs are not necessarily going to generate the
• increase awareness of the potential of wider benefits. The provision of facilities and
online technologies for contributing to the encouragement for interaction would also ap-
social connectedness and overall well- pear to be essential components” (p. 74). The
being of potentially marginalized groups, projects described emphasize the need to sur-
particularly but not exclusively in rural and mount these issues, and the need for building a
remote areas; and supporting and sustaining social infrastructure.
• provide “train the trainer” instruction for The e-life cycle is the methodology common to
volunteer community members from all these projects, 1 revolving around a five-
433
Figure 1. Inclusion, connectedness, and information sharing: Making connections face-to-

face and online
Formulate Deliver
content to suit content—
participants grow skills,
empower
individuals
Networks of
Listen and Evaluate and Networks
connected of
identify modify content connected
individuals
needs and delivery individuals sharing
sharing and
information
information
supportand
support
Engage
new
participant
stage approach (illustrated in Figure 1): Engage Simpson et al., 2001). Providing an environ-
the marginalized, listen to the learner, identify ment in which participants feel they are valued
the need(s), formulate effective programs, and and are treated as capable equals is crucial to
evaluate and modify. the e-life cycle’s listening stage; participants
here voice their needs explicitly, and those
Engage the Marginalized needs are addressed. Empowerment is also
important for, as Feinstein (2002, p. 9) found,
Engaging the community and ascertaining their “Psychological well being, which encompasses
training needs is the crucial first step. Tutors a sense of personal control, and freedom from
publicize their programs through community stress and hostility, in turn appears to lead to
places like hotels, community centers, and shops, better health outcomes.”
as well as through recruiting clients via other
community organizations, which refer people Identify
with literacy and/or numeracy needs to the
program, and also offer services to learners In this environment, specific training needs can
taking part in the program. be accurately identified since participants gain
a sense that they are valued, that their needs
Listen are unlikely to be “way out” or in some way
unacceptable, and that their contributions and
Developing effective strategies for access and concerns will be treated with respect. Tutors
participation that take differences in commu- encourage participants to use the technology to
nity needs and the whole range of local social, follow their own interests in the belief that
economic, cultural, and technological factors something that has direct relevance to their
into account can provide more equitable access own lives will be more enjoyable and more
to ICTs while increasing the overall success of relevant, and will enhance the internalization of
community informatics initiatives (Rural Women the learning process.
and ICTs Research Team, 1999; Simpson, 2004;
434
Formulate comes. Four crucial aspects of the e-life-cycle

approach are the following.
Formulating effective programs, then, draws
on the expressed interests of the participants. Participation and Inclusion
For example, participants might express a de-
sire to be able to e-mail a friend or relative The equitable and inclusive participation of
overseas, or to access a particular Web site or people in planning, evaluation, and decision
support group via the Internet. The learning is making is vital to sustainable community and
structured around this need while following a economic development. The programs are
series of specified guidelines, and it is intended specifically targeted at marginalized individu-
to provide all participants with skills in basic als, with the trainer taking a proactive approach
computing and connectivity. to inclusion. The existing capacities and expe-
There are both accredited and nonaccred- riences of participants are overtly valued and
ited schemes in operation. In the nonaccredited celebrated, with trainers and other class mem-
training programs, material is taught in a flex- bers supporting and empowering one another in
ible, self-paced delivery mode. Participants are an encouraging, inclusive environment.
encouraged to attend each session but are not
penalized for not doing so. Emphasis is on Content
meeting the learner’s needs. Feinstein et al.
(2003) found that the particular value of It is crucial that content is individually tailored
unaccredited courses is that they “may equip in a group setting to the needs and interests of
adults with the personal and social confidence the clients, is culturally sensitive and gender
as well as other necessary skills to progress to inclusive, and is made accessible by attention to
more challenging accredited courses, espe- the language and literacy needs of specific
cially if appropriate guidance is available” (pp. groups within the community. Hence, courses
76-77). are designed around what participants want,
using real-life learning needs or desires to
Evaluate shape learning delivery.
In the projects in north Queensland, opportuni- Delivery

ties are provided for trying out new skills,
sometimes in real-world situations, enabling the Delivery must be flexible and responsive to
teacher to evaluate and modify the training. Is learners’ needs, incorporate design features
it working? Are participants moving toward that capitalise on the tools made available by
confidence and facility? What adjustments can evolving communication technologies, and sup-
be made? port access by people with varying levels of
ability, including those with mobility, vision,
PRINCIPLES FOR hearing, and cognitive difficulties. In designing
COMMUNITY WORKERS training-delivery projects, differences between
people therefore need to be taken into account,
These projects have demonstrated a clear link including gender, age, ethnicity, occupation,
between increased social cohesion and con- and level of knowledge of new technologies.
nectedness, health literacy, and health out-
435
Time literacy and health. Paper presented at the

Canadian Conference on Literacy and Health,
Allowing sufficient time for the development of CPHA, Ottawa, Ontario.
the positive attitudes and skills necessary for
Communication Centre. (2001). Developing
the effective adoption and utilization of online
indicators to determine the effectiveness of
technology is important. For example, the need
Web-based service delivery: An holistic ap-
for “hastening slowly” is evident in each of the
proach. Brisbane, Australia: Author, QUT.
human factors affecting the acceptance and
value of Web-based services (Communication Feinstein, L. (2002). Wider benefits of learn-
Centre, 2001). Taking this into consideration, ing research report no. 6: Quantitative esti-
some courses are self-paced, allowing partici- mates of the social benefits of learning, 2:
pants to work through individual modules at Health (depression and obesity). London:
their own pace and with the support of peers, Centre for Research on the Wider Benefits of
tutors, family, or friends. This flexibility allows Learning.
for work and family commitments, and allows
Feinstein, L., Hammond, C., Woods, L., Preston,
participants to work around farm and seasonal
J., & Bynner, J. (2003). Wider benefits of
commitments.
learning research report no. 8: The contri-
bution of adult learning to health and social
capital. London: Centre for Research on the
FUTURE TRENDS
Wider Benefits of Learning.
AND CONCLUSION
IBM. (1997). The net result: Social inclusion
Advances in information technology and tele- in the information society. Report of the
communications may offer many potential ben- National Working Party on Social Inclu-
efits to Australia’s underserved communities sion. London: IBM UK.
by reducing the barriers of distance and space
Loader, B., & Keeble, L. (2004). Challenging
that disadvantage rural areas; this can only
the digital divide? A literature review of
happen where projects enable the participation
community informatics initiatives. York, UK:
of all groups in the community. Improvements
Joseph Rowntree Foundation & York Publish-
in the knowledge and use of such technologies
ing Services.
have been identified as being beneficial in im-
proving health literacy in rural communities, Parker, R. (2000). Health literacy: A challenge
and they will continue to be an area of attention for American patients and their health care
to reduce health inequities in Australia. providers. Health Promotion International,
15(4), 277-283.
Rural Women and ICTs Research Team. (1999).
REFERENCES
The new pioneers: Women in rural
Queensland collaboratively exploring the
Bella, L., & Bishop, R. (2004, October 17-19).
potential of communication and information
Community capacity development: A frame-
technologies for personal, business and
work for understanding the contribution of
community development. Brisbane, Australia:
community access computers at MacMorran
The Communications Centre, QUT.
Community Centre to community capacity,
436
Simpson, L. (2004). Community informatics KEY TERMS

and sustainability: Why social capital mat-
ters. Paper presented at the Community Capacity Building: The development of
Informatics Research Networks Conference, sustainable skills, organizational structures, re-
Prato, Italy. sources, and commitment to health improve-
ment in health and other sectors.
Simpson, L., Wood, L., Daws, L., & Seinen, A.
(2001). Creating rural connections. Brisbane, Health Literacy: The capacity of an indi-
Australia: The Communication Centre, QUT. vidual to obtain, interpret, and understand basic
health information and services, and the com-
Wagenfeld, M. O., Murray, J. D., Mohatt, D.
petence to use such information and services in
F., & DeBruyn, J. C. (1997). Mental health
ways that are health enhancing.
service delivery in rural areas: Organizational
and clinical issues. NIDA Research Mono- Social Connectedness: The relationships
graph, 168, 418-437. people have with others and the benefits these
bring to individuals and society.
ENDNOTE
1
Workers delivering training refer to the
process as “normal mode,” a label that
downplays its exceptional and innovative
qualities.
437

Healthcare &amp; Biomedicine Informatics Handbook

Uploaded by

Copyright:

Available Formats

You might also like

Healthcare &amp; Biomedicine Informatics Handbook

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Healthcare &amp; Biomedicine Informatics Handbook

Uploaded by

Copyright:

Available Formats

Information Science Reference

IGI Online Bookstore Reference Exact Search

About This Book

• Topics Covered Edited By: Athina A. Lazakidou, University of Peloponnese, Greece

• Table of Contents Chapter I

http://www.igi-global.com/reference/details.asp?ID=5571&v=tableOfContents (1 of 5)7/4/2009 11:15:46 πμ

SECTION III: VIRTUAL REALITY APPLICATIONS IN MEDICINE

SECTION IV: VIRTUAL LEARNING ENVIRONMENTS IN HEALTHCARE AND BIOMEDICINE

SECTION V: COMPUTER-ASSISTED DIAGNOSIS

SECTION VI: DATA MINING AND MEDICAL DECISION MAKING

SECTION VII: CURRENT ASPECTS OF KNOWLEDGE MANAGEMENT IN MEDICINE

SECTION VIII: TELEMEDICINE AND E-HEALTH SERVICES

http://www.igi-global.com/reference/details.asp?ID=5571&v=tableOfContents (3 of 5)7/4/2009 11:15:46 πμ

Starida, and Dimitrios I. Fotiadis

SECTION IX: IMAGE PROCESSING AND ARCHIVING SYSTEMS

SECTION X: SIGNAL PROCESSING TECHNIQUES

SECTION XI: USE OF NEW TECHNOLOGIES IN BIOMEDICINE

http://www.igi-global.com/reference/details.asp?ID=5571&v=tableOfContents (4 of 5)7/4/2009 11:15:46 πμ

SECTION XII: ERGONOMIC AND SAFETY ISSUES IN COMPUTERIZED MEDICAL EQUIPMENT

SECTION XIII: HEALTH ECONOMICS AND HEALTH SERVICES RESEARCH

http://www.igi-global.com/reference/details.asp?ID=5571&v=tableOfContents (5 of 5)7/4/2009 11:15:46 πμ

IGI Online Bookstore Reference Exact Search

About This Book

• Topics Covered Edited By: Athina A. Lazakidou, University of Peloponnese, Greece

The application of computer technology to education often refers to computer-assisted learning

Athina A. Lazakidou, Ph.D.

http://www.igi-global.com/reference/details.asp?ID=5571&v=preface (3 of 4)7/4/2009 11:17:31 πμ

IGI Online Bookstore Reference Exact Search

About This Book

• Topics Covered Edited By: Athina A. Lazakidou, University of Peloponnese, Greece

• Table of Contents Arvanitis, Theodoros N. / University of Birmingham, UK

http://www.igi-global.com/reference/details.asp?ID=5571&v=contributors (1 of 5)7/4/2009 11:19:50 πμ

Chandrashekar, Sharath T. / Sarayu Softech Pvt Ltd., India

Damigou, Dionisia / National and Kapodistrian University of Athens, Greece

Danilakis, Panagiotis / National and Kapodistrian University of Athens, Greece

Daskalaki, Andriani / Max Planck Institute of Molecular Genetics, Germany

Davies, G. Powell / University of New South Wales, Australia

Della Mea, Vincenzo / University of Udine, Italy

Diaz, Javier / University of Houston, USA

Doepp, Manfred / Holistic DiagCenter, Germany

Doumouchtsis, Evangelos K. / National and Kapodistrian University of Athens, Greece

Doyle, D. John / Cleveland Clinic Foundation, USA

Drougas, Bill Ag. / ATEI Education Institute of Epirus, Greece

Economopoulos, Georgios / National and Kapodistrian University of Athens, Greece

Edelmann, Gabriele / Holistic DiagCenter, Germany

Esslinger, A. Susanne / Friedrich-Alexander-University, Germany

Exarchos, Themis P. / University of Ioannina, Greece

Fotiadis, Dimitrios I. / University of Ioannina, Biomedical Research Institute—FORTH, and

Frize, Monique / Carleton University and University of Ottawa, Canada

Galani, Olga / National and Kapodistrian University of Athens, Greece

Ganiatsas, G. / University of Ioannina, Greece

Georgiou, Andrew / University of New South Wales, Australia

Giannakakis, Nikolaos / National and Kapodistrian University of Athens, Greece

Goletsis, Yorgos / University of Ioannina, Greece

Harris, M. F. / University of New South Wales, Australia

http://www.igi-global.com/reference/details.asp?ID=5571&v=contributors (2 of 5)7/4/2009 11:19:50 πμ

Hirakis, Odysseas / National and Kapodistrian University of Athens, Greece

Holzinger, Andreas / Medical University Graz (MUG), Austria

Hussain, Kashif / University of Valenciennes, France

Healthcare & Biomedicine Informatics Handbook

Healthcare & Biomedicine Informatics Handbook

Healthcare & Biomedicine Informatics Handbook