Professional Documents
Culture Documents
Neuroinformatics For Neuropsychology
Neuroinformatics For Neuropsychology
Neuroinformatics For Neuropsychology
Vinoth Jagaroo
Neuroinformatics
for Neuropsychology
123
Vinoth Jagaroo
Department of Communication Sciences
& Disorders
Emerson College
120 Boylston Street
Boston, MA 02116
USA
vinoth_jagaroo@emerson.edu
and
The idea for this book was conceived over many years and through many influences.
The fields of neuropsychology, general neuroscience, and information technology
were certainly among the main influences. It was in particular an unusual context
in which I was on the one hand exposed to academic and clinical neuropsychology
and on the other to information technology that gave rise to the ideas that would
eventually lead to this work.
I began thinking about informatics for neuropsychology more than a decade ago
as a graduate student in behavioral neuroscience at Boston University School of
Medicine. My track in this broad interdisciplinary area cut across neuropsychology,
neuroanatomy and neurobiology, and my focus was visual cognitive neuroscience.
I had concurrently held a position in a large information technology unit at the
university where I gained experience in computer networks and database program-
ming. The neuropsychology component of my training involved neuropsychological
assessment, which was carried out at the Boston Veterans Administration Hospital,
one of the teaching hospitals of Boston University Medical School. It was at these
institutions that many legendary neuropsychologists had pioneered their craft and
where some famous assessment instruments were developed.
As I engaged in carrying out neuropsychological assessment, I could not help
being struck by how comfortably this subspecialty of neuropsychology had con-
tained critical problems tied to its origins and its development. Psychometric testing
had played a huge part in the shaping of neuropsychological batteries and in some
cases assessment batteries were nothing more than modified psychometric tests.
When tests were developed from scratch in clinical neuropsychology, they were
typically developed around symptom clusters or operational tasks. Assessment
tools bore little tie to highly defined neuroanatomic systems or to rich conceptual
frameworks of cognition. Where was the alignment between neuropsychological
assessment tools, which were developed in earlier generations, and that rich body
of theory on neurocognitive principles that had arisen through cognitive neuro-
science and cognitive neurobiology, in a more recent generation? The “decade of
the brain” had brought forth so many neural systems and modules that related, with
relative precision, cognitive processes to the brain. In comparison, neuropsycholog-
ical assessment tools and neuropsychological models of cognition appeared rather
unsophisticated. It would have been possible to strive for reconciliation between
vii
viii Preface
I am grateful to the many individuals who helped make this book possible.
The encouragement and support I received from my colleagues, Daniel Kempler,
Cynthia Bartlett, and David Maxwell in the Department of Communication Sciences
and Disorders at Emerson College, was simply invaluable. The very same must be
said of the many years of generous support that I have received from Marlene Oscar
Berman of the Behavioral Neuroscience Program and the Department of Anatomy
and Neurobiology at Boston University School of Medicine.
Research for a core section of the book, on neuroinformatics applications and
models for neuropsychology, could not have been completed were it not for the
cooperation that I received from twelve individuals – who graciously addressed my
inquiries and provided me valuable data on their work: Dennis Reeves and Joseph
Bleiberg (automated neuropsychological assessment); Mark Baggett, Mark Kelley,
and Daniel Christensen (Internet-enabled assessment systems pioneered in clinical
research programs of the US Army); Ho-Chuan Huang (computerized cancellation
test system); Frank Guenther (computational model of speech production); Huber-
tus Axer and Jan Jantzen (aphasia database); Carol Neidle (sign language database
project); Curtis Deutsch (behavioral phenomics and dysmorphology); and Robert
Bilder (cognitive and neuropsychiatric phenomics).
I also wish to thank Stephen Koslow, the former director of the Office of Neu-
roinformatics at the US National Institutes of Health, for reviewing my synthesis
of the Human Brain Project, the Neuroinformatics Program of the National Insti-
tutes of Health, and the Neuroinformatics Working Group of the Organization of
Economic Cooperation and Development.
Jane Emes, my diligent graduate student assistant deserves enormous credit for
her assistance with research and preparation of the manuscript.
Finally, I am indebted to my colleague and friend, Jon Hemperley, the former
manager of Information Technology at Office of the President at Boston University,
for years of mentorship on information technology systems.
xi
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 An Overview of Bioinformatics . . . . . . . . . . . . . . . . . 1
1.2 What Is Neuroinformatics? . . . . . . . . . . . . . . . . . . . . 2
1.3 Bringing Neuroinformatics to Neuropsychology . . . . . . . . . 3
2 Current Neuroinformatics Applications and Infrastructure . . . . 7
2.1 Brain Image Construction, Analysis, and Morphometric
Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Examples of Image Construction, Analysis,
and Morphometric Tools . . . . . . . . . . . . . . . . 8
2.2 Brain Image Atlases, Databases and Repositories . . . . . . . . 10
2.2.1 Examples of Image Databases . . . . . . . . . . . . . 10
2.3 Tools and Databases for Mapping Neural Structure
and Connectivity Patterns . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Examples of Tools and Databases for the Study
of Neural Architecture . . . . . . . . . . . . . . . . . 12
2.4 Tools and Methods for the Simulation of Neurons
and Neural Circuits . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Examples of Tools for the Simulation of Neurons . . . 14
2.5 Database and Knowledge Discovery Systems for
Clinical and Academic Research . . . . . . . . . . . . . . . . . 18
2.6 Neuroinformatics Management and Infrastructure . . . . . . . . 20
2.6.1 Examples of NI Organizations, Infrastructure,
and Management . . . . . . . . . . . . . . . . . . . . 21
3 Neuroinformatics for Neuropsychology . . . . . . . . . . . . . . . . 25
3.1 Differentiating Between the General Computer
Applications in Neuropsychology and Neuroinformatics
Systems for Neuropsychology . . . . . . . . . . . . . . . . . . 25
3.2 Defining Neuropsychology-Specific Neuroinformatics
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Neuroinformatics Applications and Models for
Neuropsychology . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.1 General Neuropsychological Assessment . . . . . . . . 32
xiii
xiv Contents
Over the past 25 years, the biomedical sciences have seen an unparalleled explosion
in the amount of research data generated. The mission of understanding complex
systems at discrete levels of analysis, be they protein structure or gene expression,
and the devising of complex tools to carry out such studies have inevitably led to
massive amounts of data and the need to process the data. Large-scale undertakings
in certain biomedical avenues, such as the mapping of the human genome, have
meant that large data sets would be generated by multiple research teams. Often
these teams would be working across disciplines and across international borders,
each team dealing with an aspect of the greater research project. In such contexts,
the need to rapidly code and index data, share the data in a common database, and
devise flexible cross-referencing and retrieval systems would be vital to the success
of the undertaking.
Occurring in parallel and serving as the critical enabler of biomedical research
have been the many strident advances in computer technology. Cheap and powerful
relational databases, Internet-accessible database systems, server–client computer
architecture, and visualization and simulation software have provided the essen-
tial tools that have greatly propelled biomedical research. So much of modern
biomedical research is so integrally woven with computational technology that the
computational environment is built into the very identity of biomedical research.
Bioinformatics is the name given to the discipline that has been formed around
this very specialized application of computer science and information technology
to conceptual, investigative, and cataloguing challenges in biology and medicine.
Bioinformatics, as a discipline, has carved a distinct identity between general bio-
logical science and computer science. It arises from both these disciplines but gains
its distinction from its very unique crafting and application of computational tech-
nologies to biological complexities that can be effectively investigated only with the
application of these computer-based methods. Numerous bioinformatics books and
journals have been published over the past two decades and many undergraduate
and graduate bioinformatics programs have emerged.
Genetics and molecular biology are the biomedical subdisciplines that first real-
ized a research dependence on informatics systems. Hence these disciplines were
the first to boast elaborate informatics systems and a very well developed bioin-
formatics infrastructure. The most well-known bioinformatics undertaking has been
the Human Genome Project (HGP), coordinated by the National Institutes of Health
and the US Department of Energy (see Collins et al. 2003; http://www.genome.gov/;
http://doegenomes.org/). The HGP ran from 1990 to 2003 and was a visionary
project. Its primary goals were to identify all the genes of human DNA (the human
genome) and the sequences of their base pairs. The project was officially completed
ahead of schedule, in 2003, when its central goal – the sequencing of the human
genome – was met but work continues on some remaining goals. Well represented
within the HGP were numerous aspects of a large-scale bioinformatics undertaking.
It grew out of the need to process massive amounts of data and elaborate data sets
in order to understand a complex biological system. It encouraged innovative com-
puterized tools designed to capture data and it shared these tools among the HGP
community. The data generated were fed into centralized databases, easily accessi-
ble to research teams internationally. The data had multiple uses. Analytic tools were
devised for the specialized assessment of the data (and some teams would focus only
on the assessment of captured data). The HGP involved biologists, mathematicians,
computer scientists, and IT specialists, melding their skills in various ways. It was
an international collaborative research effort with well-defined goals and it had a
centralized coordinating body. A clear understanding of the distribution of tasks
and the establishment of uniform research protocols and IT platforms made for a
streamlined global mission, maximizing resources while minimizing redundancy of
effort. The HGP had strong organizational leadership that guaranteed funding and
infrastructure for continuity of the project.
As indicated by the example of the HGP, there are many levels at which bioinfor-
matics can be practiced. Bioinformatics can refer simply to algorithms, databases,
and analytic tools used to further biomedical research. At broader levels, it may
encompass a large organizational system for project coordination and an IT infras-
tructure for integration of data sets. Genetics research and informatics systems have
defined such a powerful interface that the term genomics was coined to describe
this area. Similarly, the extent of informatics dependency in studying protein struc-
ture has given rise to the field of proteonomics. Numerous other subdisciplines in
biology and medicine have since inevitably been drawn into bioinformatics.
The broad field of neuroscience has clearly been a burgeoning field. It has continued
to face ever-complex research questions and large data volumes. It is not surprising
therefore that neuroscience has taken to bioinformatics. The many applications of
bioinformatics in the neurosciences have come to be labeled by the term "neuroin-
formatics" (Huerta and Koslow 1996; Shepherd et al. 1998). Neuroinformatics is in
itself a richly constituted subdiscipline with avenues that extend from neuroimaging
to neurobiology. Some respective examples are databases of fMRI images and sets
of algorithms to help connect functional patterns; computerized simulation tools
1.3 Bringing Neuroinformatics to Neuropsychology 3
for neuronal conduction and dendritic branching; and a data mining system that
can search published neuroscience data (electronic files) and then generate probable
relational patterns between data fragments.
Initially neuroscience was slow to adapt to a bioinformatics approach when com-
pared to genetics and molecular biology (Koslow 2000). In more recent years,
however, neuroinformatics has rapidly gained momentum, as will be illustrated
through the many examples described in Chapter 2. A major initiative in neuroin-
formatics has been the Human Brain Project (detailed in following chapters). It was
sponsored by the National Institutes of Health and a few other US federal agencies,
with the explicit aim of furthering brain research through application of informatics
(Huerta et al. 1993; Shepard et al. 1998). The project has been in essence a neu-
roinformatics initiative – to database the brain – and it has supported many types
of projects that have been drawn into this central goal. Its significance to the brain
sciences has been as great as the significance of the Human Genome Project to the
field of genetics. It has spawned a number of major neuroinformatics developments.
These developments together with many independently developed neuroinformat-
ics tools and networks have come to be spread over almost every subdiscipline of
neuroscience.
The Digitization of Neuroscience Data: Common to many of the facets of
neuroinformatics (NI) are a few essential features to do with the digitization of neu-
roscience data. With the aid of computers, other specialized hardware, and database
applications, data are first captured at very fine levels. The data may then be stored,
manipulated, analyzed, and reconstructed. This general digital format and the com-
putational platform offer a greatly enhanced degree of complexity in describing
structural or functional properties of the nervous system. The key features in the
NI approach to data are summarized in Table 1.1.
With the digital approach to data capture, the degree of data complexity can be
more closely in tune with the actual biological complexity of the object of function
being studied. The richness of the data capture is often enabled by specialized hard-
ware. These tools are quite unique as are the kinds of analyses and relationships that
can be derived from the data they generate. Conceptually and operationally, they are
on a very different level when compared to conventional tools or methods. These
are the features that make NI a very unique and powerful avenue in neuroscience
research.
Aspects of data
processing Key considerations
a. Data capture • Emphasis on digitizing the data – capturing or coding the data in ways
that can be stored and manipulated by computers.
• Capturing data at multiple levels of abstraction – data about all levels of
structure/function and data on as many variables as feasible.
• Indexing the data in ways that are optimal for database queries/searches
and pattern-seeking algorithms.
b. Data storage • Databases with data fields to accommodate multiplicity and breadth of
and sharing data to enable multiple levels of inference and analysis.
• Tagging or coding data; saving text, images, and data about spatial and
functional relationships.
• Web-accessible centralized databases or shared databases through
server networks.
• Data availability for current or future use – prospective and
retrospective study needs.
c. Data analysis, • Data mining – statistical techniques designed to mine the data for
modeling, patterns and relationships; search algorithms tuned to complex
manipulation, relationships among the data.
rendering, and • Data modeling – manipulate one or more variables; an application can
knowledge use real or hypothetical contingencies to model the effect of one set of
discovery data on another.
• Reconstruction – using the binary data to create realistic 3D spatial
renderings of structures, or probabilistic maps; semantic relationships
among text data can be expressed in spatial form.
• Ontology building – computer readable models of the properties and
relationships of elements of a data domain; aids the knowledge
discovery process
Society (INS) in St. Louis, Missouri. At the June 2008 INS meeting in Hawaii and
at the February 2009 meeting, Robert Bilder laid out compelling perspectives on
informatics frameworks for neuropsychology (these are discussed at length in later
chapters).
Both clinical and experimental neuropsychology have certainly built an increas-
ingly close relationship with specialty subdisciplines like neuroimaging, to which
NI is central. Yet, neuropsychology in itself has not yet realized the application of
NI to subject areas that are definitive to the field of neuropsychology. Traditional
neuropsychological research on the functional domains of language, visuospatial
function, memory, etc., tests and assessment batteries used in neuropsychologi-
cal assessment, and neuropsychological models for describing clinical syndromes
remain starkly unengaged with the activities of NI when compared to other subdis-
ciplines of the brain–behavioral sciences. Because neuropsychology is concerned
more with cognitive behavior and deals with the brain at a level that does not delve
into the minutiae of laboratory-style bench neuroscience, it could appear that it is
incongruous with the very technical stance of NI. This book attempts to demonstrate
1.3 Bringing Neuroinformatics to Neuropsychology 5
Functional magnetic resonance images are the “end products” of an elaborate and
complex sequence of experimental steps and computerized data processing. The
variables involved in the process are immense (see Buxton 2001; Huettel et al.
2004). The raw data that are captured are typically in the form of numerous graphed
signals tied to voxels (and they have the appearance of approximated sine wave
or time-series graphs). The raw data look nothing like the colorful brain images
that are synonymous with fMRI. The data are fed into computerized tools that fil-
ter and analyze the data in terms of signal strength, temporal properties, etc. They
then necessarily use a 3D brain coordinate system so that the data can be viewed
(transposed) as a neuroanatomic activation pattern in the brain. A suite of comput-
erized tools enable the transformations from raw data to 2D images, 3D volumetric
reconstructions, surface maps, etc.
The generation of 3D images from 2D image slices in a single subject typically
involves steps such as image slice assembly formatted for a particular computer
application for image analysis; image quality checks; realignment and smoothing
of the assembled slices; removal of non-brain imaged parts/structures; time-series
analysis stimulus and inter-stimulus activity; corrections for artifacts (e.g., head
movement); and segregation in differential patterns of activations (blobs and regions
of interest). In experiments involving multiple subjects the same steps apply to each
subject but the activations blobs are averaged across subjects.
Each and every one of these steps is dependent on highly specialized comput-
erized tool sets. These are powerful informatics tools without which fMRI images
would not be possible. A few examples of such tools are described below (and the
aim of this subsection is merely to describe some of these tools, not the specialized
domain of brain imaging).
1 Brain coordinate systems are used to define a “standard brain” against which to describe the locus
of neuroanatomic structures and activation patterns. They are necessary to imaging – data sets are
“normalized” to the space defined by a particular coordinate system (a process of stereotaxic trans-
formation) so that voxel-based computations can be carried out. One such system is the Talairach
system that derives from the Talairach and Tournoux Atlas, which was based on the brain of a
single individual (see Talairach and Tournoux 1988; 1993). Another is the MNI system developed
at the Montreal Neurological Institute in collaboration with the International Consortium for Brain
Mapping (see Section 2.2). This system was developed from a sample of 152 individuals and is
currently the most commonly used template for brain imaging. Some imaging software suites can
convert data between the Talairach and MNI systems.
2.1 Brain Image Construction, Analysis, and Morphometric Tools 9
a comprehensive suite for fMRI data processing. AFNI’s basic unit of data is a set
of 3D volumetric data – an array of voxels, for which AFNI codes various voxel
values. AFNI provides a number of ways of viewing the imaged data, for example,
in axial, sagittal, and coronal views; one slice at a time; many slices at once; with
functional overlay (merging two data sets); with anatomical overlay (merging a sub-
ject’s functional data with anatomic data from a database); as volumetric renderings;
as time-series graphs; etc.
AFNI has evolved into a very comprehensive image analysis package, currently
consisting of about 300 external programs and plug-ins that work with the core
application. Each deals with a very specific aspect of image analysis or rendering.
AFNI has been able to achieve this level of sophistication in part because it is also
an open source platform for development of new software. Users can download and
compile AFNI source code to program new features. Core developers of AFNI at
the NIH and the AFNI user community work together to iron out bugs and address
questions via the AFNI website. AFNI is a prime example of a dynamic platform
for neuroinformatics development. (URL: http://afni.nimh.nih.gov/afni)
Computerized Anatomical Reconstruction and Editing Toolkit (CARET): Devel-
oped at Washington University School of Medicine (St. Louis), CARET is a
powerful software application for generating and manipulating cortical surface
reconstructions. It is an open source software available to researchers and is accom-
panied by comprehensive online tutorials. CARET is designed to accommodate
various kinds of experimental data (neural connectivity patterns, anatomically or
functionally defined cortical geography, neuroimaging scans, etc.), which it can use
to produce surface reconstructions of cortical gyri. For example, CARET can use
a few neuronal connectivity maps of a gyral region or a few functional images
of the gyral region and from the data, it can project a gyral surface contour map.
It can also piece together individual image sections to create an entire gyral sur-
face. It can further manipulate the reconstruction to generate, for example, flat
maps (an unfolded gyrus) and inflated maps. Surface maps can be thought of as
a generic form of expressing brain images and neural organization. The advan-
tage of these maps is that they can represent the integration of various data types
arising from different studies. In accommodating different data types, CARET also
serves as a repository that successfully brings together and compares the data. (URL:
http://brainmap.wustl.edu/)
3D Slicer and FreeSurfer: Both 3D Slicer and FreeSurfer are software packages
freely downloadable from the Biomedical Informatics Research Network (BIRN),
an informatics consortium. The 3D Slicer utility has similarities with CARET in that
it enables neuroanatomic visualization and image segmentation. It can, for example,
process a segment from each of three brain scan slices (axial, sagittal, and coronal
planes) and then produce a 3D surface model highlighting the segmented region,
a function particularly important in the visualization of tumors or other lesions.
The development of this software is ongoing. It is the product of collaboration
between the Artificial Intelligence Laboratory at Massachusetts Institute of Tech-
nology and the Surgical Planning Laboratory at the Brigham & Women’s Hospital
(Boston). FreeSurfer is a set of software tools that can also reconstruct a cortical
10 2 Current Neuroinformatics Applications and Infrastructure
surface using imaging data. Using data about functional characteristics of a corti-
cal region it can then overlay these data on the reconstructed cortical surface. This
functionality is well geared to extract cortical regional correspondence across large
data sets. FreeSurfer has been developed at the Center for Biomedical Imaging at
Massachusetts General Hospital with support from the BIRN, the NIH, and other
agencies.
The examples described above are a sample of the range of tools avail-
able for imaging analysis. Without these tools, the elaborate forms of data
image manipulation and data modeling that stem from imaging data would
not be possible. Recognizing that “[I]nformatics tools are key at all stages
of neuroimaging, allowing scientists to control highly sophisticated imaging
instruments and to make sense of the vast amounts of complex data gen-
erated by them” (http://nifti.nimh.nih.gov/background), the NIH launched the
Neuroimaging Informatics Technology Initiative (NIfTI). Its mission is to help
researchers enhance the utility of such tools and to foster unification of
tool platforms between different research groups using a community-based
approach. (URLs: http://nbirn.net/research/morphometry/visualization_tools.shtm
http://surfer.nmr.mgh.harvard.edu/)
With the advent of computerized neuroimaging and the churning out of brain images
came the need to store the images. Brain image databases would serve as image
repositories, enable rapid retrieval of images (scans), and provide a means to com-
pare images. The emergence of functional neuroimaging only solidified the need
for imaging databases. Brain image databases have also meant that increased atten-
tion would be given to the processing of imaging data, post acquisition. As Wong
and Huang (1996) have highlighted, (a) this post-acquisition image analysis capa-
bility allows for the maximal analysis of the data, even long after the data have been
acquired, and (b) the informatics tools for managing and sharing these data have the
potential to reshape the clinical environment in the decades ahead.
particular axis (horizontal, coronal, and sagittal). Some of the images can be regen-
erated by changing the weighting parameters (e.g., T1 vs. T2), thereby reflecting
differences in tissue-image characteristics. A well-organized database of images
presented through a simple online interface makes the Whole Brain Atlas a com-
pact and powerful utility. The image data can be used simply to tour the brain
slice by slice or as a brain scan reference for normal and diseased brains. (URL:
http://www.med.harvard.edu/AANLIB/home.html)
The International Consortium for Brain Mapping (ICBM) Subject Database:
The ICBM Subject Database is one component of a multifaceted neuroinformat-
ics project based at the University of California, Los Angeles. It is a web-interfaced
database of images and is at the center of a much broader NI infrastructure. Among
the goals of this NI infrastructure is the fostering of unification in (a) the collection
of images from various sources and (b) the organization these images and their dis-
semination among researchers. The ICMB database is also a large and continually
growing image database. Its sophistication lies in its powerful functionality: Vari-
ables and attributes relating to the subject (individual) and to the actual scan can be
set in order to refine a query. The query interface offers, for example, search fields
for the sex and weight of the subject and image attribute search fields for modal-
ity (MRI, PET, etc.), weighting, pulse thickness, slice thickness, and acquisition
plane. A collaborator registered with the ICBM can use this unified set of data fields
to upload images and view customized representations of the larger data set. By
bringing organization and structure to image data set collection, the ICMB subject
database provides an effective and convenient system for collaborators to archive,
share, and retrieve image data. (URL: http://www.loni.ucla.edu/ICBM/Databases/)
The Database of the NIH MRI Study of Normal Brain Development: The NIH
MRI Study of Normal Brain Development is a large-scale, multi-centered study that
tracks normal brain development in healthy children and adolescents. The greater
goal of the study is to provide a large reference data set on normal brain maturation
that will aid in understanding a range of neurodevelopmental conditions. The study
has recorded MRI and cognitive/neuropsychological data from approximately 500
children, from newborns to age 18, across the United States. Data collection began
in 2001 and was completed in 2007. It is, to date, the most extensive MRI study
of normal brain development in children. The study is sponsored by the NIH and
involves seven participating pediatric centers in the United States and the Montreal
Neurological Institute (MNI) in Canada, which acts as the data coordinating center
(see Almli et al. 2007; Evans 2004; Evans and the Brain Development Coopera-
tive Group 2006). A study of this scale overcomes problems of small sample sizes
and problems of different data types being spread across different (incompatible)
databases.
The pediatric MRI data repository that has resulted from the study resides at
MNI and contains a unique collection of multi-spectral structural MRI images. The
images were fed into the databases via the web network linking the participating
centers. Since the database also contains behavioral data, it allows researchers to
relate cognitive and motor changes in growth to changes in brain anatomy. The
database is web accessible and is available internationally to qualified researchers.
12 2 Current Neuroinformatics Applications and Infrastructure
As with any NI project of this type and scale, new NI methods are devel-
oped in the process. The NIH/MNI MRI data repository demonstrated new
methods for MRI segmentation, web-accessible image databasing, and a work-
able architecture for large-scale collaboration. (URL: http://www.bic.mni.mcgill.ca/
nihpd/info/index.html)
Just as the biophysical properties of neurons and neural conduction are of interest to
cell biologists, the dynamics of neural functioning and neural circuitry are of interest
to computational neuroscientists – those who deal with mathematical modeling of
14 2 Current Neuroinformatics Applications and Infrastructure
2 Computational modeling is the dominant method of simulating neuronal circuits and as such is
briefly covered in this subsection. The outline of main concepts given here will also aid in better
understanding a few of the NI examples that the reader will encounter in Section 3. A comprehen-
sive description of neural modeling is well outside the scope of this book. The interested reader is
referred to the original sources cited.
16 2 Current Neuroinformatics Applications and Infrastructure
processing. In such a system, the net function is the result of the entire network,
while each neuron is defined by connectional and functional properties that give it
a distinct role in the collective function. Modeling can be viewed as a way of draw-
ing abstractions of workings of a complex system such as the brain and thereby
approximate neural functioning. There are a few areas of concentration in the field.
There is a lot of focus on understanding actual neural systems in terms of their
functional neurobiology and their role in cognition and then attempting to model
these systems. On the application end, artificial neural networks are often the basis
for artificial intelligence; hence neural modeling is closely tied to practical applica-
tions of robotics and engineering initiatives to develop computers that closely mimic
human information processing.3 Though artificial neural networks (hereon referred
to as “neural networks”) are generally much more simplified than actual networks
in the brain, they help to guide research.
Structure of Neural Networks: An artificial neuron (or processing unit) is the
building block of a network. These units are interconnected such that they can influ-
ence each other and in this way modulate the larger outcome of the system. The term
node may describe a single neuronal unit or a cluster of them that serve a common
function in the network. Neurons can be arranged in various ways, for example, a
network can have 100 neurons divided into 10 clusters and each cluster can be linked
to each other, or a network may be built of multiple layers of neurons, each layer
carrying out a distinct function and feeding information to the next layer. A certain
pattern of input to the first layer of the network results in a differential pattern of
activation across its neurons. Different “afferent” signals are then propagated and
are received by the next layer. Successive stages of processing are carried across the
network until an output layer generates a final result.
There are many variables and functional parameters involved in neural networks
and many of these are specified when designing neural networks. They include a
host of neuronal properties such as pattern of conduction (influenced by ion channel
properties), types of dendritic formations, numbers and strengths of synapses, how
a neuron summates its signals, the rate of firing, and the action of neurotransmitters
and receptors. In general, the closer the neurons of a network simulate integration of
3 Some readers may have encountered the term “Cognitive Informatics,” a term coined by Wang
(2002). The claim has been made that cognitive informatics is a new, innovative, emerging dis-
cipline concerned with (a) mathematical and computational approaches to understanding neural
and cognitive systems, especially in terms of how they represent abstract knowledge, and (b) how
this understanding can potentially lead to new types of computer architecture (see Wang 2003a,b;
2006; 2007a,b). These two goals, however, have been at the core of cybernetics and neural model-
ing for more than five decades (see sources cited earlier on neural modeling). On closer inspection,
“cognitive informatics” is largely an approach to modeling influenced by a bit of philosophy
about knowledge representation. Specifically, it is an approach predicated on a set of descriptive
mathematical (algebraic) methods around which an elaborate set of propositions, corollaries, and
theorems have been built (see Wang 2003b; 2006; 2007b). This clarification is being made here as
a caution to the reader not to confuse “cognitive informatics” with neuroinformatics for neuropsy-
chology. “Cognitive informatics” is an approach to modeling, and the relevance of modeling to NI
in neuropsychology is articulated in this subsection.
2.4 Tools and Methods for the Simulation of Neurons and Neural Circuits 17
information from multiple synapses and at finer levels of neuronal detail, the more
the network approximates spatial and temporal dynamics of an actual neural system.
Neural networks are developed with computer programs, and a large variety of
software packages are available. The software enables the user to select types of net-
work topologies, types of neurons and various neuronal variables, different methods
to train the network, and ways of testing the network.
How Networks Learn: An important aspect of complex network is the bidirec-
tional flow of information – when information from later nodes can be fed back
to earlier nodes so as to adjust their properties. A network is trained by learning
rules (algorithms) that serve to adjust parameters (e.g., synaptic strength) of earlier
layers. Back-propagation is a commonly used training method by which the values
of data output from the output layer of a network are compared to expected val-
ues. Discrepancy values are propagated backwards in the network as feedback data
that the network uses to adjust its synaptic strengths. With continued training of the
network in this manner the network can achieve an output that is consistent with an
expectation. An adaptive network is one where connections and other network prop-
erties are automatically adjusted based on the input to the network and on how the
information flows through the network during the training/learning phase. In neural
modeling, neural networks are often configured as adaptive networks. They have a
powerful capacity to adjust to new data and to outcomes.
Core Application Functions of Neural Networks: Computer and mathematical
models based on neural networks have wide ranging applications beyond theoreti-
cal models of brain function. They may be applied, for example, in everything from
camera surveillance software to weather forecasting applications. In most cases, the
role of the neural network model is just one or more of a few common, sometimes
overlapping, functions. These have to do with algorithms that reduce or compress a
large mass of data to a manageable form and algorithms that mine the data for sim-
ilar temporal, spatial, or physical patterns or clusters; algorithms that match input
patterns (physical, spatial, temporal, numerical, etc.) and match them to stored data
so as to recognize the stimulus pattern or classify and store the pattern; algorithms
that extrapolate or approximate existing data to derive an approximate value for an
unknown value; and algorithms that do the same but on a temporal scale so as to
predict data sequences or event occurrences.
Relevance of Neural Modeling in Neuroinformatics: Neural networks are fun-
damentally methods of modeling data that have complex, non-linear relationships.
Neural modeling is by and large mathematical and statistical and can be described as
mathematical or statistical methods for the capture, classification, and matching of
data and predictions based on these data. Bioinformatics and NI have the same goals
relating to data but are not necessarily tied to statistical methods. Neural network
methods can be applied to NI when data volume or data complexity warrants the
application. They are hugely applicable in data mining – to help filter and condense
the data, find relationships between variables, etc. Essentially, modeling is a tool
that can directly tackle the larger NI goals of working out patterns and relationships
by clarifying the complex layers between input and output data patterns.
18 2 Current Neuroinformatics Applications and Infrastructure
Some of the NI examples surveyed above make evident that a database is often
the nucleus of the NI environment. More than a mere repository, a database allows
for selective querying and combining of data. When databases have to be designed
for clinical and research environments, they face some unique design challenges
brought about by the multi-tiered, multi-faceted nature of these environments. Large
clinical and research settings involve many data sources and data types and call
for numerous uses and applications of the data. A clinical setting typically also
includes multiple users with varying levels of data access privileges. In the academic
neuroscience research environment, massive volumes of data are churned out at rates
faster than any single scientist can reasonably integrate. The volume of published
literature is on a scale so enormous that it becomes impractical to grasp.
Spread throughout the strata of clinical and research environments are valuable,
often untapped data. Masses of data in databases cannot be manually examined
for relationships among the data. Special procedures are required to mine the data
for trends and patterns that may exist therein. The potential of the data is realized
through data mining and related methods of knowledge discovery.
Data Mining and Knowledge Discovery: Data mining refers to the process of
applying statistical, modeling, and visualization methods to large databases in
order to bring forth patterns and relationships that may lie obscured in the vastness
of the data (see Piatetsky-Shapiro and Frawley, 1991; Lavrač 1999; Prather et al.
1997; Zupan et al. 1999; Zhang et al. 2004). It serves to extract useful information
from the data by sifting out regularities or irregularities among the data. Since data
mining applies to databases and serves knowledge discovery, the process has come
to take on the name “Knowledge Discovery in Databases” (KDD). Data mining is
a crucial step in the overall process. Typically, before data mining procedures can
be applied, the data are cleaned and may be arranged in ways that facilitate specific
data mining techniques.
Data mining is ultimately concerned with drawing comparisons and viewing
interactions among the data (Kötter 2001) and using the data to discover hidden
correlations which may then be used to draw hypotheses (Wong and Huang 1996;
Wong et al. 2004). Data mining has therefore come to constitute an actual method of
research (Shepard et al. 1998). An entire branch of neuroscience research can be cre-
ated just by mining the data that lies within the multiple strata of its subdisciplines.
Ontology and Knowledge Discovery: In informatics, the term ontology refers to
a model that lays out the relationships and properties of the features that constitute
2.5 Database and Knowledge Discovery Systems for Clinical and Academic Research 19
a domain of data (see Gruber 1995; Maedche 2002; Staab and Studer 2004). The
model has a format that is readable by informatics tools and attempts to represent
concepts and their relationships in a manner that is approximate to how they occur in
reality. Ontologies attempt to conceptualize a data/knowledge domain by specifying
possible data sets, data classes, data concepts, and features, rules, and relationships
that possibly describe the data. By defining a common vocabulary and relational
structure, ontologies aid researchers in organizing the data of a knowledge domain,
which in turn makes for efficient ways of knowledge sharing.
Ontologies are used to guide data mining but they are also modified and adjusted
over time as more data are accumulated. Ontologies are crucial for knowledge
discovery in informatics because (a) they guide the choice and structure of the
data mining methods and (b) they critically influence results obtained through data
searches and data mining. In other words, relationships in the data as brought forth
by data mining are influenced by the ontology that guides the data mining pro-
cedure. Ontology that more accurately represents properties and relationships of a
domain as they occur in the real world will generate more accurate results through
KDD. This in turn can be used to refine the original ontology. Ontology develop-
ment in NI is a highly specialized area. Ontology building is essential to semantic
tools in knowledge discovery that enable the mapping of semantic relations among
large data sets. Semantic web technologies and a suite of “onto-tools” help provide
semantic frameworks for organizing the data and “engineering knowledge.”
There has been a clear trend in neuroscience toward centralized databases in
which the data are laid out along semantic categories (see Gorin et al.). Complex
programs using semantic criteria work to archive, index, or retrieve the data.
The NeuroScholar project (www.neuroscholar.org) is an example of database
system for the mining of neuroscience data for research purposes (see Burns
2001a, b; Burns et al. 2003). NeuroScholar is designed to index multiple sources of
neuroscience data such as published literature, websites, and laboratory-based data
files. It utilizes techniques from software engineering to construct its knowledge
model. Using NeuroScholar a user can, for example, extract the parts of pub-
lished papers that are most relevant to a question the user may seek to understand.
NeuroScholar terms these information parts “fragments,” which may be textual or
graphical in form. In a workspace provided by the system, the user can then decide
on the extent of linkage between the generated fragments. The user may also cre-
ate fragments. Once the user establishes a linkage between a set of fragments, the
relational pattern may then be stored on the system. This relational pattern may now
constitute an interpretation of data fragments, providing a model for knowledge
representation. NeuroScholar offers the facility to build complex neuroscientific
relationships using isolated fragments of neuroscience data. A host of support-
ing tools also enables the NeuroScholar user to carry out other operations such as
computational analysis of the knowledge representations.
The Neuroinformatics Database System (Wong et al. 2002; Wong et al. 2004)
is a database system for disease-oriented neuroimaging research, implemented at
the University of California, San Francisco. It is an example of a system tailored
to data integration within a clinical setting. In its current rendering, it contains data
20 2 Current Neuroinformatics Applications and Infrastructure
relating to patients with intractable temporal lobe epilepsy (TLE) – it serves and
is served by clinicians and scientists involved with TLE treatment or research. In a
large clinical setting, some of the departments that would typically be involved with
intractable TLE patients are neurology, radiology, and surgery. In a typical hospital’s
information system, each department inputs its data on a patient into the system. All
departments may have access to the patient file and can therefore view the patient’s
results from each department (presuming the hospital’s privacy policy makes the
allowance) and this is usually the extent of data interconnectedness. Hospital infor-
mation systems are designed to serve the essential functions of data input, storage,
and retrieval. The Neuroinformatics Database System (NIDS) is an integrated sys-
tem that builds on an existing hospital computer infrastructure. It is designed to
draw multiple data types from multiple sources (departments) and offers special-
ized analytic and statistical components for data meta-analysis. NIDS can extract
and replicate information from the hospital’s primary information system without
interfering with patient records. “Warehousing” the data as NIDS does and then
applying a range of analytic algorithms to the data enables a clinician or researcher
to search for correlations and patterns that otherwise could not be gleaned from the
standard database.
The emerging role of NI-oriented databases and data mining systems for neuro-
science cannot be underestimated. It is only through the NI-enabled integration of
diverse data types that many complex relationships in neuroscience can be viewed.
This idea has been succinctly expressed by Wong and Huang (1996): “[T]he motive
is to realize the greatest possible benefit from the data that already exist. The new
megachanges of the next 20 years will center around gathering, managing and using
multimedia clinical information.” (p. 285).
The neuroinformatics applications and projects sampled above often require a great
deal of management, funding, and coordination. Proper management infrastructure
is particularly important for an emerging multifaceted discipline like NI. The sup-
port structure can (a) provide an effective forum for exchange and communication of
ideas, (b) help formally identify areas of research and development, (c) coordinate
projects among research teams at national and international levels thereby maxi-
mizing research potential, (d) provide centralized databases and tool repositories,
and (e) provide funding sources for NI projects. The scale of a support structure for
NI can vary depending on the extent to which it serves one or more of the above
considerations.
A less obvious but equally crucial function that may be served by an organizing
body for NI has to do with formative decisions regarding tool and database designs,
modeling schema, and data-sharing protocols. These have the potential to influence
the entire NI landscape over the long term. This consideration addresses the fol-
lowing questions: What data entry fields and querying procedures should be created
when designing a database? What kinds of coding systems should be implemented
2.6 Neuroinformatics Management and Infrastructure 21
leaps and the promise of applied information technology was already being seen
in molecular biology projects such as the Human Genome Project and the C. ele-
gans research database (a database for mapping the genome of the roundworm,
Caenorhabditis elegans). Encouraged by these promising developments, the NIH
and NSF, in 1989, asked the National Academy of Sciences’ Institute of Medicine
(IOM) to create a working group of scientists that would explore the feasibility of
developing a national database resource for neural circuitry. The group engaged in
extensive discussion and consultations, and in 1991, published a report in which
the idea of integrating computational technology with neuroscience was firmly
endorsed (see Pechura and Martin 1991). Evolving from this initiative, the HBP was
launched in 1993, formalizing a structure for neuroinformatics. Incorporating many
of the recommendations made by the IOM’s working group in 1991, the HBP’s
coordinating committee shaped a guiding strategy for the project.
The HBP would aim to support research in all areas of neuroscience at both
federal and public institutions. It would call for the research to show a clear involve-
ment of both neuroscientists and technology specialists (computer/informational,
mathematical, engineering, etc.). Very importantly, the HBP would require a certain
degree of communication between the projects it would sponsor and it would require
that NI tools and technologies developed under its sponsorship be compatible across
research teams, and hence be easily shareable. The idea of the neuroscience com-
munity richly interlinked by a system of shared computer networks databases was
articulated in the IOM’s working group. It is one of the main ideas that have shaped
the HBP’s outlook.
The HBP has been conducted along three phases (see Huerta and Koslow 1996;
Huerta et al.; Gorin et al. 2001). In phase 1, it sought development of prototype
NI tools for capturing, analyzing, and manipulating information digitally. Much of
the emphasis in this phase was on tools and databases relating to (a) functional
neuroimaging and (b) neuronal structure and circuitry (see Shepard et al.). In phase
2, it encouraged the refinement and integration of the tools and systems developed
in phase 1. Attention was given to computerized methods of integrating graphical
data (of the type generated with phase 1 tools) and non-graphical (e.g., theoretically
described) neuroscience data. Phase 3 took the developments from phases 1 and 2
to shared platforms (e.g., Internet-accessible databases), making them available to
the broader neuroscience community.
High priority has been given to projects relating to data storage, data access, and
data mining; digital brain atlases and related tools; simulation and visualization soft-
ware; and Internet-based platforms for tool and data sharing and for collaboration
between research teams and sharing. After more than a decade of successful ini-
tiatives, the HBP has turned its focus to how these initiatives can be more broadly
applied in advancing research (HBP website). It is not, at the present time, soliciting
new funding applications but many other divisions of the NIH continue to call for
informatics-related projects.
The HBP has greatly propelled and shaped the field of NI. It started with a
visionary idea, which was translated into an effective plan of action. In addition
to providing management and funding, the HBP has placed emphasis on the
2.6 Neuroinformatics Management and Infrastructure 23
Conclusion
The rapid success of NI over the past years is closely linked to organizations like
those discussed above that have made NI imperative. An open but centralized sys-
tem for NI coordination, protocols, and standards for NI data and key funding have
been essential ingredients to the rise of NI. Stable, long-term funding, especially
when integrated with national science/research agendas are critical to large-scale NI
projects. It is still, however, necessary that independent or local NI initiatives con-
tinually sprout new ideas. Whether in the form of exploratory committees, interest
groups, or a small database on an academic department’s computer server it takes
these seeds to grow into larger NI projects which may then become the nodes of a
wider NI network. Small NI initiatives and the larger management structures that
may assist in coordinating these initiatives often co-evolve.
The various areas of NI sampled in this section highlight the definitive role of
databases, computational tools, analytic and modeling software, and frameworks
for data access in NI. Great emphasis is placed on the gathering, organization,
integration, and sharing of data. In some instances, approaches to a problem in neu-
roscience are built around informatics methods that are actively brought to bear on a
research problem. NI influences research questions as much as they influence it. NI
shares great credit in shaping the current era of neuroscience. Almost every avenue
in neuroscience has now incorporated NI.
Chapter 3
Neuroinformatics for Neuropsychology
There is clearly an enormous difference between NI applications and the general use
of computers in the neurosciences. NI is much more than a simple extension of gen-
eral computing. NI applications by definition constitute digital systems for active
information-seeking, problem-solving, and modeling, generally based on strong
theoretical frameworks.
Neuropsychology has always been quick to embrace general computing. Even
prior to the first generation of commercially available personal computers of the
early 1980 s, simple neuropsychological tests began to appear in computerized form
(see Adams and Heaton 1987; Chute 2002). Over the past two decades, arrays of
computerized neuropsychological tests have emerged and these tests have shown
increasing levels of complexity. However, even with the advances carried by these
applications, they still remain far short of incorporating NI functionality.
Neuropsychology features minimally in the landscape of NI. Similarly, NI has
only a small presence in neuropsychology. Yet this small presence is marked by
some powerful NI-oriented neuropsychological applications/models that are rela-
tively new and which have been developed in highly specialized clinical or research
environments. The two key factors marking these examples are that they are con-
cerned with neuropsychological topics and they are designed from an informatics
perspective.
This section begins by spelling out the differences between NI applications for
neuropsychology and the general use of computers in the field. Exactly what it
means to have neuropsychology-specific NI systems is then articulated – an attempt
is made to define core principles of NI approaches in neuropsychology. Following
this, nine NI-based applications or models for neuropsychology are discussed.
neuropsychologists analogous to the tools and databases such as CARAT and CoCo-
Mac that allow neurobiologists to create maps of cortical regions and circuits?
Do neuropsychologists have information sharing systems along the lines of the
ICBM Subject Database that gives users the facility to upload imaging data to
shared platforms and then reap the benefits of the collective data? Why have neu-
ropsychologists not developed a simulation environment for their work while other
neuroscientists have developed simulation environments like GENESIS and Neuron
for modeling neural systems?
The interface between NI and a definitive neuropsychological praxis is largely
an uncharted area. This, of course, does not apply to some subspecialties allied to
neuropsychology such as neuroimaging and clinical electrophysiology. Neuroimag-
ing (see Sections 2.1, 2.2) has certainly become the preferred research mode of
many experimental neuropsychologists but the contexts in which it is used and the
medical/research questions it addresses are far from unique to neuropsychology.
Clinical neurophysiology, especially to do with electroencephalography and evoked
potentials, is now conducted with fully computerized, informatics-heavy tools. Like
functional imaging, computerized electrophysiology is a specialized discipline that
happens to serve neuropsychology but is also applied to studying a range of func-
tions far outside the scope of neuropsychology. The NI seen within neuroimaging
and electrophysiology do not constitute NI applications that can be considered
unique to neuropsychology.
In neuropsychology, the areas of (a) head injury research and (b) computerized
neuropsychological assessment show levels of sophistication in computerized meth-
ods that are almost on the threshold of transitioning to NI methods. Informatics-
reliant functions are steadily filtering into computer applications in these areas
though they remain, for the large part, short of constituting NI functionality (auto-
mated assessment, discussed in Section 3.3, is a clear exception). Nonetheless, they
signal the value of databases and computational methods for neuropsychology.
The purpose of this subsection is to clarify aspects of computerized assessment
that border on NI and aspects that do not constitute NI. A review of computerized
neuropsychological assessment is not a goal of this subsection. Many such excellent
reviews have been conducted and are cited.
Head Injury Research: It has been suggested that a national traumatic brain injury
(TBI) database will help with uniformity in TBI assessment, better communication,
and data sharing among clinicians and improved prediction of clinical outcomes
(Hall 1997; Johnston 1997). These reports have centered more on the general util-
ity of such TBI databases. Hobbs (1999) mentioned “neuroinformatics” and the
Human Brain Project and then described some implications for TBI rehabilitation,
for example, NI-based brain atlases and 3-D reference systems for brain lesions can
be used to model TBI and help to better match treatment with type of injury. In 2002,
the Journal of Head Trauma Rehabilitation published a special edition on neuropsy-
chological technologies. While none of the articles concerned NI, a few touched
on issues that are also encountered in the broader discourse on NI. Schatz and
Browndyke (2002) described scenarios where neuropsychological assessment can
be beneficially carried out on portable computers or workstations and the data then
3.1 Differentiating Between the General Computer Applications 27
and scored with the aid of computers. The integration of statistical tools with the
assessment programs also accounts for a large part of the computerized functional-
ity conferred by these applications. Successive computerized versions of these tests
are more likely to add small increments of change rather than functionality that con-
fers a potential for research breakthroughs. A hard NI functionality remains absent.
In general, neuropsychology has been using a very simple rubric when structuring
computer applications.
In these scenarios, the computer platform has not been used to dynamically
modify the tests. In contrast, NI offers methods and tools that can be dynami-
cally tuned to the very complexity of neurocognition. The sophistication of its
methods is commensurate with the complexity of cognitive-behavioral patterns that
neuropsychological assessment seeks to describe.
• The application must render observations and findings that cannot be feasibly
achieved with conventional neuropsychological methods – there has to be a
very clear advance brought by the NI tool over and above the data rendered by
conventional tools.
• Where applicable, the structure of the NI system should be commensurate with
the complexity of neurocognitive mechanisms as described by current theory in
neuroscience, that is, the complexity of an NI system should, as far as possible,
be tuned to the complexity of the neuropsychological phenomenon at which it is
aimed.
3.3 Neuroinformatics Applications and Models for Neuropsychology 29
(a) Seeking a finer “grain” or increment of data, seeking to capture the data in
digital form or converting the data to digital form, and optimizing it for database
operations.
(b) Aiming to database the data with a breadth of data fields (the fine increments
of data), structuring the databases so that they are optimized for future data
manipulation, and where feasible, using web-accessible and shared-platform
databases.
(c) Applying data mining/pattern-seeking algorithms and data modeling methods
to elicit or test relationships between data sets.
Areas of
neuropsychology to
which the examples
relate Name of example Functions, innovations, and features
Areas of
neuropsychology to
which the examples
relate Name of example Functions, innovations, and features
and neuropsychology. For each of these areas, considerable background is first laid
out. Certain conceptual problems and methodological shortcomings in general neu-
ropsychological assessment, the visuospatial domain, and in the neuropsychology
of language, are described so as to highlight the rationale for NI. The examples per-
taining to cognitive phenomics are contextualized by an overview of phenomics
research that is pertinent to neuropsychology. Table 3.1 summarizes all of the
examples.
Many of the examples describe actual informatics applications and databases,
and a few describe theoretical models or integrative research models that are ori-
ented to NI approaches in neuropsychology. Earlier sections of this book illustrated
that the broader landscape of NI is characterized by a variety of NI applications and
systems. NI in neuropsychology is barely in its infancy and shows only a limited
number of developments but these developments are very significant and already
show some variety.
Information on the examples described was gathered using a number of sources:
published literature in neuropsychology and bioinformatics/software engineering,
various websites, and very importantly, from material obtained directly from the
few clinicians and researchers who happen to be working at the intersection of
32 3 Neuroinformatics for Neuropsychology
neuropsychology and NI. Since this field is so new, personal communication with
these individuals proved to be a very fruitful way of gathering pertinent data.1 A
lot of unpublished data, project proposals, and data on projects-in-progress, etc.,
were provided directed to the author in the course of 2006–2008 (see acknowl-
edgements). This material also gave broader perspective to the various researchers’
published data.
1 The author’s descriptions of all examples in this section have been reviewed by the primary
researcher/s associated with the examples. Appreciation is expressed in the Acknowledgements
3.3 Neuroinformatics Applications and Models for Neuropsychology 33
these benefits in turn translate to flexibility and practicality in the usage of tools,
improved organization of assessment protocols, and greater reproducibility of the
results generated.
NI also introduces into clinical assessment one particularly radical idea, an idea
that has been absent in the field of neuropsychology, if not rather alien to the
field, and almost never raised as a possibility: that is, the potential for large-scale
sharing of everyday assessment data. Conventional modes of assessment are such
that the results of tests are stored in office filing cabinets or on the hard drive of
a clinician’s computer. Valuable data go to waste and this has been the case for
decades, as forgotten boxes of paper would go through the process of degeneration
in moldy basements. One of the greatest impediments that the field of assessment
suffers (somewhat unknowingly to the field) is the lack of large, national or inter-
national databases of assessment data. Server–client computer platforms and online
databases make this sharing possible and the potential benefits are unparalleled. Cer-
tainly, the mere allusion to the possibility of large-scale sharing of assessment data
is bound to raise instinctual doubt and disbelief among neuropsychologists. Shar-
ing of patient data is often a violation of patient confidentiality – a code that is set
deep in clinical practice. The issue of data sharing is, of course, fundamental to any
discussion on shared NI systems for assessment. Not surprisingly, this is exactly
the kind of issue that has been at the forefront in the broader discussion of medical
informatics, especially in relation to genetic and imaging data, and many possible
solutions to the problem have been proposed. In view of the centrality of this issue, it
is given special attention within a different section of this book (Section 4.1) and the
current subsection will remain focused on NI systems for assessment. In a similar
vein, issues such as standardization in assessment are not washed over by an enthu-
siastic argument for NI-based assessment. Discussion is best given to such issues in
a separate section where they can be laid out systematically.
There have been some impressive NI developments in neuropsychological
assessment. They relate to such things as (a) computerized tests that tap discrete
cognitive operations; (b) the digitization of neuropsychological batteries so as to
produce sets of data files that can be automatically uploaded to databases; (c) the
structuring of databases for many types of neuropsychological data; and (d) the
use of a, b, and c in conjunction with the internet and highly portable computer
devices for large-scale assessment, data collection and analysis. The dimensions that
NI brings to assessment simply cannot be achieved with conventional assessment
methods.
Two informatics-based neuropsychological assessment initiatives are described
in this subsection. One describes an automated battery of tests interfaced with
a database and with portable assessment devices. The other describes a project
that was designed to test a possible structure for internet-based, database-linked
assessment.
It will be seen through the examples that great emphasis is placed on test mod-
ularity and on data capture. Both projects have served to establish the feasibility of
NI-based assessment. They demonstrate numerous crucial components that make
up a large-scale NI system for assessment.
3.3 Neuroinformatics Applications and Models for Neuropsychology 35
Data Capture and the ANAM Access Database (AADB): The AADB is a
Microsoft Access database designed to organize raw data files from ANAM tests,
create summaries of the data, and examine or combine the data in relation to one or
more variables of interest (see Bleiberg et al. 2004). ANAM test data files contain-
ing raw data and summary data can be automatically imported into the AADB. Data
import is made very simple, involving a few steps.
Data from ANAM subtests are fed into pre-existing tables in the database. Data
from a particular test are automatically fed into a table for that test (tied to the test’s
unique file extension). The database contains three main types of tables: Header
tables include data such as the subject ID, date and time of testing, and variables
selected for a test and the testing session (Fig. 3.1a). An entry (row) in the header
table corresponds to a single run of a test. If the test is administered twice in a
session, another row of data will be produced in the header table.
Fig. 3.1a Example of a header table for the Code Delayed subtest (showing one row without
data). From the ANAM Access Database User’s Manual (2004); printed with Permission, National
Rehabilitation Hospital.
Item tables contain data about the trials/stimuli of each subtest and the subject’s
responses – correct/incorrect, reaction time, and accuracy (Fig. 3.1b). In this table, a
row of data corresponds to a single trial or item of a subtest. A 15-item subtest will
therefore produce 15 rows of data in its corresponding table in the AADB.
Fig. 3.1b Example of an Item table for the Matching Grids subtest (showing one row without
data). From the ANAM Access Database User’s Manual (2004); printed with Permission, National
Rehabilitation Hospital.
Summary tables essentially summarize ANAM test data files to present informa-
tion such as mean and median scores on a test, number of items correct, number of
lapses, and throughput scores (Fig. 3.1c). A row in a summary table corresponds to
single run of a subtest. If a subtest was administered twice in a session, two rows of
summary data will be produced in the summary table.
Fig. 3.1c Example of a summary table for the Running Memory (continuous performance) subtest
(showing one row without data). From the ANAM Access Database User’s Manual (2004); printed
with Permission, National Rehabilitation Hospital.
38 3 Neuroinformatics for Neuropsychology
The AADB also contains two demographic tables. These tables draw data from
the demographics data file. They include information such as the file location (path)
of the data files, date and time of testing, the subject’s name, age, data of birth, etc.
Queries: Since the AADB is a Microsoft Access database, it is benefited by the
enormous power of querying functions. Queries allow the user to view the data in
various permutations. A query can be used to sift out or analyze data based on some
criteria and combine data from two or more tables. A user can, for example, utilize
a query to select out all subjects in the 22–23 year age range whose scores in one
subtest are lower than scores in another subtest and whose mean ANAM scores are
within a certain range.
A database need not necessarily allow the user to utilize or create queries. Often,
when a database is created for a specific purpose, the user access is limited to data
entry and a limited number of data views. The AADB does not impose such lim-
itations on the user. At the same time, its query process is designed to cater to a
range of users – from those with little database skills to those who can write com-
plex structured query statements. The AADB provides the user with a number of
automated (prewritten) summary queries. The use of summary queries does not
require any training in writing Microsoft Access queries. The user is taken through
a simple wizard that presents choices on tables or variables the user would like
to select, information on various aspects of the query, and instructions on how to
complete the query (see Fig. 3.1d). Traditional queries or manual queries usually
opted for by users familiar with Microsoft Access queries may be written by using
the applications default new query interface window. Here various types of queries
may be written and they may range widely in scope and complexity. Knowledge of
structured query language obviously facilitates this process.
Internet Enabled Assessment, Remote Assessment Devices, and Shared Soft-
ware: ANAM is currently Internet enabled using a “store-and-forward” method
(see Cernich et al.). Test modules can be downloaded from an ANAM server to
a resident computer. The tests can be run off the local computer without need
for continued link to the server (applications designed with platform independent
computer languages such as Java make this possible without too many download
compatibility obstacles). The data can then be uploaded to the server for storage and
analysis. Internet-enabled function is a vital feature of ANAM’s avenues of applica-
tion in remote cognitive assessment and monitoring. An Internet-accessible Oracle
database is currently being developed to store data files from individual ANAM
tests. It will be programmed with pattern-seeking algorithms, i.e., normative patterns
across tests, populations, and test conditions.
The AADB and ClinicView© , a software package for tabular and graphical dis-
plays of ANAM test results, are ANAM software utilities that have been developed
through the National Rehabilitation Hospital (Washington D.C.). These third party
software utilities are available at no cost to the user – see ANAM links above. They
are crucial to the digital capture and display of test data.
Portable Assessment Systems: Another significant ANAM related development
that is aligned with an informatics perspective is that of portable, handheld cog-
nitive screening systems (see Elsmore and Reeves 2004; Elsmore et al. 2007;
3.3
Fig. 3.1d One of the windows of the Query Wizard for automated summary queries in the AADB (the instructions in boxes also appear). From the ANAM
Access Database User’s Manual (2004); printed with Permission, National Rehabilitation Hospital
39
40 3 Neuroinformatics for Neuropsychology
Fig. 3.1e The BrainCheckers handheld device (PDA) displaying a trial of the ANAM Code Substi-
tution Test. From, BrainCheckers User’s Manual (2007); printed with permission, Dennis Reeves,
Behavioral Neuroscience Systems, LLC
Bleiberg et al. 2007; Reeves et al. 2007; Behavioral Neuroscience Systems 2008).
BrainCheckers is a handheld computerized neuropsychological assessment system,
and it is an advanced version of an earlier system called the ANAM Readiness Eval-
uation System (ARES). BrainCheckers runs on the Palm! R
PDA. A re-engineered
version of ANAM adapted to the Palm operating system can be self administered
with the system (Fig. 3.1e). This library of ANAM tests can be configured in many
ways as BrainCheckers is used for different screening, diagnostic, and assessment
purposes. The portability of the BrainCheckers system means that it can be utilized
in emergency rooms, battlefield environments, and other such settings where desk-
top computer-based assessment is impractical. It is also cost effective. Both of these
were major factors that drove its development.
The informatics functionality of this portable system is significant: The sys-
tem records subject information and test data in databases – a database record for
each subject. When the data are uploaded to a personal computer, Dataman, a data
manager and PC-PDA communication program, archives the data directly into a
Microsoft Access database. From the Access database, the data can be easily pasted
to other Windows applications for other kinds of viewing and analysis. A back-end
data viewer enables the examiner to view test results and summary data on the PDA
so that downloads to a personal computer are not required. As described in Part II,
Internet-enabled tools, databases, shared software and devices that help digitize data
are fundamental to the building of an NI environment.
From an informatics perspective, ANAM has been a milestone development in
computerized neuropsychological assessment. ANAM has taken a set of cognitive
tests and has adapted them to a framework that incorporates informatics functional-
ity. ANAM digitizes a wealth of raw data and creates data files that are compatible
with a powerful database application. ANAM’s database is the only database of
its kind available in automated assessment and is the first informatics-enabled
3.3 Neuroinformatics Applications and Models for Neuropsychology 41
problems facing any reviewer in any field, only perhaps to a greater degree. The
nature of the data, the absence of a clearly defined conceptual framework, and a his-
tory of confusing and inconsistent terminology all result in bias and lack of clarity”
(Newcombe and Ratcliff 1990, p. 333). A fundamental problem has been the lack
of agreement regarding the definition of visuospatial function. Further, descriptions
of what is visuospatial in neuropsychology have been crafted largely with reference
to operational definitions as carried by visuospatial tests:
object’s identity and location, respectively. The work provided the first systematic
account of visual-spatial processing beyond the visual cortex, the functional prop-
erties of which had been mapped earlier in cats and monkeys (Hubel and Wiesel
1962; 1978).
Through the 1980 s and 1990 s, neurobiological and neurophysiological stud-
ies on visual and spatial processing (largely in non-human primate brains) saw an
explosion in numbers. They aimed to draw clear associations between visual/spatial
processes and cortical systems. Noteworthy for neuropsychology is the precision
with which they defined visuospatial, perceptual, and cognitive operations. These
studies were concerned with such processes as the mnemonic coding of coordinates
involved with mapping spatial plans for eye movements in the dorsolateral pre-
frontal cortex (Funahashi et al. 1989); spatial frameworks for movement kinematics
and movement control coded by the area 5 of the parietal cortex (Kalaska 1991);
the representation of head-facial profiles in neurons around the superior temporal
sulcus (Perrett et al. 1992); and multimodal spatial coordinate representations and
coordinate transformations in the posterior parietal cortex (Andersen et al. 1993).
Synthesizing such data and the vast amount of data on cortical connections in neu-
robiology, Felleman and Van Essen (1991) compiled a map that indicated some 32
cortical areas involved with visual processing, including more than 300 connections
among these areas and a distinct neural processing hierarchy among these areas. In
visual neuroscience, this often cited map has served as a major reference for the
functional anatomy of visual and spatial processing.
Cognitive science has been equally focused on the “dissection” of visuospatial
processes: Earlier pioneering work laid out, for example, the variables of features
in visual attention and object recognition (Treisman, and Gelade 1980; Treis-
man 1982); graded differential hemispheric processing based on spatial frequency
(Sergent 1982; 1991) and categorical versus coordinate relations (Kosslyn 1987;
Kosslyn et al. 1992) in the visual scene; and the role of representational coordi-
nate systems in mental manipulation (Just and Carpenter 1985). Functional imaging
then signaled the possibility of localizing discrete spatial operations (see Roland
and Friberg 1985; Haxby et al. 1991) and soon the landscape of visual cognition
was transformed. Visuospatial processes would now be gleaned through a range
of cortical regional activation patterns, for example: The posterior parietal area in
relation to 3-dimensional mental rotation/manipulation (Cohen et al. 1996; Richter
et al. 1991); prefrontal, premotor, and occipito-temporal regions in relation to spatial
working memory tasks (Jonides et al. 1993); responses properties of a fusiform gyral
area (the fusiform face area) in relation to faces (Kanwisher et al. 1997; Kanwisher
2000) and its potential to be tuned to sets of objects that share similar looks and
spatial form (Tarr and Gauthier 2000); a parahippocampal gyral area (the “parahip-
pocampal place area”) in relation to visual scenes that convey relative layout of
objects (Epstein and Kanwisher 1998); and a region of the lateral occipitotemporal
cortex (the “extrastriate body area”) with regard to pictures of bodies and non-facial
parts of bodies (Downing et al. 2001). Numerous cortical modules for visuospatial
processing could then feasibly be laid out as functional-anatomic maps built from
functional imaging studies alone (see Grill-Spector and Malach 2004).
46 3 Neuroinformatics for Neuropsychology
(squares) to produce a grid. The fineness of the grid can also be varied from a few
large cells to thousands of small cells. The cell at the midpoint of the grid is given
the coordinates (0; 0). In one configuration of the grid, cells on the X-axis to the left
of the center are assigned negative coordinates, e.g., from –1 to –40. Cells to the
right are assigned positive coordinates, +1 to +40. Cells along the Y-axis, above cell
(0; 0) are assigned positive coordinates, +1 to +30. Cells below are assigned negative
coordinates –1 to –30. The resulting matrix amounts to a grid of uniform cells, each
with a unique address – a set of unique coordinates (see Fig. 3.2b). The default posi-
tion of the grid is such that coordinate (0; 0) defines the midpoint of the presented
visual field: The Y-axis (–30 to +30) defines the midline separating each hemifield
and the X-axis (–40 to +40) defines the horizontal separation between equal-sized
upper and lower fields. The grid acts as a reference system, i.e., an objective
Euclidean map.
When used in assessment, two computer screens are involved. The examiner
views one screen and the subject views another. Both screens are tied to the same
control system but the grid is invisible on the subject" s screen. On the subject’s
screen, the flower will appear on a plain background. The flower, however, occupies
certain cells on the invisible grid on which it is imposed.
Each cell in the grid is mapped to a database called the Coordinate Recording
Database (CRD). The position of a stimulus placed within the grid is also mapped to
the database. For each cell in the grid, the CRD has other data fields in addition to the
fields the code the cell’s position. The other fields can record attributes such as stim-
ulus color, percentage of the field being occupied, etc. There are a number of other
important programmed components that make up the system: The grid controller
is used to stipulate how many cells should make up the matrix, that is, the cellular
density or complexity of the matrix. It can also elongate cells and cell contents ver-
tically or horizontally to warp regions of space as may be required experimentally.
The option of varying the matrix configuration is an important one. A large-cell grid
configuration, for example, would be better suited to a patient exhibiting very gross
neglect. The coordinate computational kernel (CCK) is a computational engine that
is composed of algorithms and queries. This component essentially converts grid
data to numerical arrays and then compares the data. Computational criteria can be
programmed as needed. A graphics database provides a selection of basic shapes
and patterns and can also accommodate other figures and pictures that may be
inputted for use. A graphics kernel provides a simple graphics editor to alter the
shape, form, size, and color of a stimulus. A patient profile database is comprised
of a set of tables that store the patient’s performance data. It can be programmed
with analytic queries as needed. An analytic module is a set of strings, macros, and
queries that run on the CRD and Patient Profile Database to look for relationships
between stimulus/grid conditions and the patient’s performance and to carry out
inter-subject comparisons.
All the databases included in the system are designed using Microsoft Access;
other components are for the most part programmed with C++. The beta version
of the system employs only some of the components described: The CRD, a sim-
ple Grid Controller (offering 3 three-grid configurations), and the Patient Profile
50 3 Neuroinformatics for Neuropsychology
Fig. 3.2b Schematic illustration of the CVFN’s Coordinate Addressing System. From, Jaga-
roo (2002). Dynamic computational visual field matrices: A computerized mapping system for
the analysis of visual perception, spatial processing and featural recognition. In C.H. Dagli,
A.L. Buczak, J. Ghosh, M.J. Embrechts, O. Ersoy, & S.W. Kercel (Eds.) Smart Engineering System
Design, Neural Networks and Fuzzy Logic, vol. 12. Printed with permission, American Society of
Mechanical Engineers Press
Database. Other components are being worked on as the greater project progresses.
Figure 3.2c provides a schematic of the architecture of the CVFN system.
Functionality Gained Through a Computational Grid Matrix: An infinite number
of stimuli and visual field variables and permutations are enabled by this NI system
3.3 Neuroinformatics Applications and Models for Neuropsychology 51
GRID CONTROLLER
COORDINATE
VAR (1:8)
COMPUTATIONAL
Controls grid density Seq 1
KERNEL
and enables warping Σ((+3–5), (–5+6))
of grid sections Seq 2
Converts grid data
Σ ((+17–9), (–30+29))
into numerical arrays
If CRD(X,X) = open
Increment = (4) for data comparison
Enter (X,X)
GRAPHICS DATABASE Mark (–1+4)
table[gridpoint]
Begin level {23}
power =3
Offers a selection of Position if ƒ(+2–5)
graphical objects bit (0, α –1, 1, α –5)
(stimuli). Objects bit (3, δ +1, 4, δ –7)
can also be added to Stop Level {45}
this database. End
(+4+3) (+5+7)
(–3+5) (+4–5)
(–23+24) (–30+29)
(+13–14) (+27–18)
(–10+12) (–1–3)
COORDINATE
RECORDING
GRAPHICS KERNEL DATABASE
Fig. 3.2c Schematic illustration of the CVFN’s Architecture. From, Jagaroo (2002). Dynamic
computational visual field matrices: A computerized mapping system for the analysis of visual
perception, spatial processing and featural recognition. In C.H. Dagli, A.L. Buczak, J. Ghosh,
M.J. Embrechts, O. Ersoy, & S.W. Kercel (Eds.) Smart Engineering System Design, Neural
Networks and Fuzzy Logic, vol. 12. Printed with permission, American Society of Mechanical
Engineers Press
moved to the right of the visual field midline (Halligan and Marshall 1995).
Such discoveries have been made when both the line length and the point of
transection have been carefully measured (see Bisiach et al. 1990; Halligan and
Marshall 1991). Such data have been obtained manually, often requiring hundreds
of paper-illustrated variations of a stimulus. The tedium of such manipulations can
discourage further such analyses. With a computerized matrix system, all the follow-
ing data can be preset and/or referenced in seconds: (a) the coordinates occupied by
the line – indexing line length and displacement; (b) the coordinate corresponding to
the point of transection; (c) the relative degree (in numerical units) to which transec-
tion is displaced towards one hemifield; (d) the pattern of the perceiver’s regression
or improvement over a number of trials or over a period of time; (e) the aggregate
critical “sector” of the line where transection occurs (within and across patients);
(f) the average coordinate position of the patient’s (biased) midline based on his/her
line transections; and (g) transection points across varying line lengths.
Stimulus Displacement and Field Gradients: In neglect, a patient’s omissions are
not always limited to the left hemifield (see Small et al. 1994). Omissions may occur
along the midline and in some cases, in the right hemifield. This neglect “gradient”
often defines a patient’s unique neglect profile. Attempts have been made to map
this gradient along a simple 3-point scale (see Small et al.), a system that crudely
neglects large sectors in the hemifields. With a computerized matrix, the gradient
can be generated instantaneously and can be accurate to the point of a matrix cell.
All averaging functions and tracking functions (over time) can be applied to gradi-
ent assessment. The stimulus pattern can also be displaced by very fine increments
towards either hemifield in order to establish the X-axis point at which the gradient
is zero, that is, by displacing the stimulus by precise increments, one cell displace-
ment at a time, it will be possible to compute the patient’s perceived (displaced)
midline.
Masked Stimuli, Global-Local Effects, and Figural Modulation: The system can
be programmed to implement masking effects, obscuring effects, and global-local
effects on visual stimuli, for example, (a) stimuli in one hemifield can be masked
at varying degrees along a fine gradient; (b) featural or outline cues can be added
to one hemifield, again to assess how perception is modulated under these condi-
tions; (c) framing effects, shape modulation, and global cueing effects can also be
incorporated; and (d) the effects of masking or featural cueing on neglect can be
tracked.
Chimaeric Stimuli and Stimulus Compression: Chimaeric stimuli used in the
study of visuospatial processing have produced dramatic results but have been lim-
ited largely to pictures of faces and simple objects (see Young et al. 1992). Again,
practical constraints in manually generating and altering these figures prevent com-
plex permutations. With a matrix-controlled system, almost any aspect of chimaeric
stimuli can be easily adjusted, again, akin to the adjustments allowed by a sim-
ple computer graphics application: (a) the matrix of one hemifield governing the
chimaeric piece in that field will be able to adjust the dimensions and color of the
figure; (b) adjustments can be made to one half of the chimaeric figure so as to make
it incrementally similar or dissimilar to the other half. In this way, the degree of
3.3 Neuroinformatics Applications and Models for Neuropsychology 53
Blinking arrows in neglected field Center-to-left Arrow Train in neglected Center-to-left Color Wave in neglected
while displaying stimulus item field while displaying stimulus item field while displaying stimulus item
the National Cheng Kung University, respectively, in Taiwan. Its development has
been supported in part by the National Science Council of Taiwan.
Background: The development of CACTS was motivated by some key factors.
It was recognized that while visual search, scanning, and cancellation tasks have
been widely used in visuospatial assessment, inherent limitations are contained in
the conventional paper-and-pencil nature of these tasks (of the type described in the
preceding subsection). Again, the static nature of these conventional tasks makes
for a poor rendering of the overall visual search and attention process. Informa-
tion about onset time, search paths, search strategies, stops and pauses, etc., can be
captured only crudely. A computerized interface was sought to address these limi-
tations. A second consideration was the need to capture visual search and scanning
trajectories without necessitating the use of eye-tracking devices that usually require
more elaborate hardware and a specialized testing environment. A highly portable
system that would be capable of registering attentional search would be much more
preferable in, for example, a school setting where a clinician may need to assess
visual attention in a large population of children in an efficient manner. CACTS was
therefore developed to utilize only a personal computer and a monitor. The aim was
to create a computer-interfaced system to capture dynamics processes in a subject’s
visual search and attention during cancellation tasks and then apply algorithms to
analyze the attentional pattern.
Design and Architecture of CACTS: The system runs on a personal computer
using Microsoft’s Windows XP platform. The software is programmed in C++. In
the latest setup of CACTS, the hardware consists simply of a tablet computer with
a touch-sensitive screen and a cordless screen pen. When the user applies the pen to
the screen, its pressure-sensitive tip transmits signals to the tablet computer at about
the point of contact. Euclidean (x, y) coordinates together with corresponding time
of the event are automatically recorded in a database.
CACTS involves four software modules (see Fig. 3.3a). A profile module serves
to hold a subject’s general and demographic data such as age, sex, and education
level. A test module is designed to contain various forms of attention/cancellation
tests. An outcome analysis module records and analyses a subject’s attentional pat-
tern and provides the examiner with graphical tools for the analysis. A print module
allows the examiner to be selective in terms of data to be printed. A Microsoft
Access database is used to store the subject’s demographic and test data.
Kinematic and Metric Functionality: One component of the CACTS database,
the process table, stores data on kinematic aspects of the subject’s visual scan path.
General attentional dynamics can subsequently be pictured using this data. Kine-
matic data include response coordinates, distances between consecutive response
coordinates, and pause and completion times between consecutive response coor-
dinates. Total times for movement, pauses, and completion can also be recorded.
The system’s analysis module can use this data together with a pattern identifica-
tion algorithm to construct a scaled figural representation of the subject’s search
sequence and overall scanning strategy (see Fig. 3.3b). Sequential nodes along
the visual scan path can be highlighted. The coordinates marking the subject’s
attentional center can also be highlighted. This convenient function can effectively
3.3 Neuroinformatics Applications and Models for Neuropsychology 55
Participant Examiner
TEST MODULE
Analysis of Analysis of
Search Patterns Scanning Trend
Test Stimuli Type of Presentation
1, 2, 3 A, B, C
(Alphabets)
Structured
(Numbers) PRINTING MODULE
,•
(Symbols)
" and "
Random
(Radicals) Data Printing Graphic Printing
Databases
Fig. 3.3a CACTS System Architecture. Adapted from Wang et al. (2006) and printed with
permission, Elsevier
summarize an individual’s search strategy; it can also sort out different search
patterns among a group of individuals.
In the current version of CACTS, the test module contains four forms of tests
stimuli – numerical forms (0–9); alphabetical forms (A–Z) in upper case; 50 two
symbol forms (example as “ ” and “•”); and 26 basic components of Chinese
alphabetical characters, called radicals (example, “ ” and “ ”). The system also
allows the option of adding new stimulus items to test module. The stimuli may be
presented in random or structured arrays.
As is the case with the CVFN, the center point in the CACTS screen is the 0;0
coordinate point. All other points are referenced in terms of x and y coordinates
relative to the x and y axes (this is a standard method in many image analysis
applications). To aid the examiner with analysis of the subject’s performance, the
coordinates in the CACTS screen can also be grouped into nine regions – a central
circle surrounded by eight circular demarcations. Dividing the screen in this way
allows for a simpler delineation of the response pattern, such as “center,” “left,”
“far left,” etc.
The examiner can view test results with a screen interface of the Outcome Anal-
ysis Module (Fig. 3.3b,c). Different outcome screens offer different data sets or
different levels of detail for a data set. Figure 3.3b is a snapshot of one outcome
screen. It includes some of the subject’s personal data such as name, education
A pie chart showing
56
Records of each
cancellation by
orders
Fig. 3.3b Snapshot of CACTS outcome analysis screen 1. From, Wang et al. (2006); printed with permission, Elsevier
3 Neuroinformatics for Neuropsychology
3.3 Neuroinformatics Applications and Models for Neuropsychology 57
Fig. 3.3c Snapshot of CACTS Outcome Analysis Screen 2. From, Huang and Wang (2005).
Toward a graphical analysis tool for computer-assisted assessment of visual search patterns. Paper
presented at the Fifth IEEE International Conference on Advanced Learning Technologies; printed
with permission, IEEE (©2005 IEEE)
level, and date of testing. It also shows the number of cancellations made by
the individual in the left and right half-fields; the number of correct responses
in each half-field; total time spent in each half-field, and a graphical depiction
of the subject’s scan path. Another outcome screen, shown in Fig. 3.3c, includes
some of the above data and offers more detailed temporal data tied to the subject’s
scan path.
Cancellation tasks constitute a special category of tests in the assessment of atten-
tion and visuospatial function. Their simplicity, ease of administration, and often
clear results have given them a proven utility in neuropsychology. These factors
have also resulted in the tool remaining essentially unchanged in more than three
decades. So much more data can come from cancellation tests, data that are not fea-
sibly realized by the conventional paper-and-pencil versions of these tests. CACTS
is a system that changes the landscape of cancellation tasks. It adds a powerful
informatics dimension to these tasks. In so doing, valuable hidden information about
attentional strategies and visual field biases is brought to the fore.
58 3 Neuroinformatics for Neuropsychology
The broad functional domains of speech and language are ripe and ready for neu-
roinformatics. Production and perception of speech and language involve complex
neurocognitive mechanisms. These mechanisms continue to be investigated at an
exciting pace only to unveil an even more multifaceted puzzle. In formulating mod-
els or dealing with assessment of speech and language, neuropsychology is called to
(a) parse out contributing cognitive variables such as phonology, syntax, semantics,
discourse, working memory, auditory and orthographic perception, and sequencing
of cognitive and motor programs, (b) match these functions to their corresponding
neural systems, and (c) determine the nature of disruption to these functions follow-
ing damage to a neural system. The difficulty of such tasks need not be emphasized.
Neuropsychological Models of Language: The evolution of neuropsychologi-
cal models of language, including neuropsychological speech mechanisms, while
marked by numerous epochs, can be viewed in terms of the degree of emphasis
placed on functional-anatomic systems, sensory and motor processes, cognitive pro-
cesses, or neural networks involved with speech and language (Arbib and Caplan
1979; Benson 1993). Neuroanatomic localization of psycholinguistic and speech
sensorimotor processes has been an underlying theme.
The Broca-Wernicke model of language has been a canon in neuropsychol-
ogy. Key events around this model are common historical knowledge in the field
(and even though this model has been described endless number of times in
neuropsychology, a few highlights are called for here to serve the discussion
building up to the need for NI). Broca’s (1861/1865) reports associated motor
speech production with the frontal opercular region. Wernicke’s (1874) extensive
monograph served to delineate processes of sensory (auditory) language compre-
hension tied to the posterior aspect of the superior temporal gyrus and to highlight
the fiber tract linking this area to Broca’s speech motor area. Lichteim’s (1884)
elaboration on Wernicke’s model (the Wernicke-Lichteim model) then consolidated
3.3 Neuroinformatics Applications and Models for Neuropsychology 59
2 The model has also been advocated (Pinker 1994) as the neural basis of Chomsky’s (1957)
Generative-Transformational model of language. In this context, its shortcomings have been
described from the standpoint of evolutionary neurobiology of language (Deacon 1997; Lieberman
2002, and see Aboitiz and Garcia 1997).
60 3 Neuroinformatics for Neuropsychology
It can flexibly incorporate new research data and allow for greater precision in
associating symptoms of aphasia with neuroanatomic systems.
Limitations of Box-and-Arrow models: A neuropsychological working model of
language that describes wide spread neuroanatomic loci may represent an achieve-
ment over the classical Broca-Wernicke model in that it draws a larger network. It
may recognize that interaction and feedback can occur between the loci. Yet, how-
ever elaborate this kind of “box-and-arrow” model may be, it will always have some
unfortunate limitations: It has neither a way of systematically exploring the relation-
ships between the boxes (nodes) nor a way of dynamically testing the model. The
dynamics of the network cannot easily be studied or rendered with this kind of
model. The model becomes proliferated with more and more boxes-and-arrows (see
Marshall 1986)
Assessment of Language: The assessment of language in clinical neuropsychol-
ogy has hardly been deterred by the difficulty of formulating comprehensive models
of this functional domain. Certainly with regard to aphasia, assessment tools such
as the Boston Aphasia Diagnostic Exam, the Western Aphasia Battery Tap are
well-delineated behavioral features. They strive for well-defined symptomological
categories. As neurocognitive models of speech-language evolve, some challenges
for neuropsychology will be to further clarify symptoms and to match them more
closely with neural systems.
Role of NI: Arbib and Bota (2003), writing about neurolinguistics and neuroinfor-
matics, stated that “ . . . we need neuroinformatics methods to develop in a fashion
which more tightly integrates modeling with databases . . . to allow the more effec-
tive integration of data from neuroanatomy, neurophysiology, brain imaging . . . ”
(p. 1257). Neuroinformatics is integral to the process of clarifying speech and lan-
guage mechanisms in neuropsychology. Coding, sorting, and connecting the myriad
of cognitive and neural variables involved with speech and language is dependent
on NI methods. NI can structure databases of symptoms such that syndromes of
language are indentified. NI can help model discrete neural operations in speech,
which in turn can contribute to complex neural models of speech-language produc-
tion. To further improve models of speech and language, neuropsychology must
expand its scope. It has to borrow concepts from computational neuroscience that
relate to neural coding and representations of data, neural transformations of the
coded data, and feedforward and feedback mechanisms. Models are needed that take
into account the dynamics of speech-language circuitry and NI methods can help
to parse out perceptual and cognitive speech-language processes that result from
these interactive dynamics. A related model is described in this Section (3.3.3.1)
and its significance in fostering NI approaches in the neuropsychology of speech
and language is emphasized.
Sign and Gestural Language: The study of language in neuropsychology has
not been limited to spoken language. Descriptions of the complexity of structure
of sign language (Bellugi and Klima 1976; Klima and Bellugi 1979) prompted
many neuropsychological reports on the neural basis of sign language and gestu-
ral communication (e.g., Hickok et al. 1996; McNeill and Pedelty 1995; Petitto and
Bellugi 1988). Spatial relations and spatial processing involved with sign language
syntax (Poizner et al. 1984; Bellugi et al. 1988), linguistic versus affective facial
3.3 Neuroinformatics Applications and Models for Neuropsychology 61
expressions in sign language (Corina et al. 1999), and their differential lateraliza-
tion have been among the major topics in this area. More recently, neuroimaging
studies have advanced finer distinctions in the neural mechanisms of sign language,
e.g., for different types of facial expressions, aspects of phonological processing,
and in dissociations between motor and linguistic processes (see Emmorey 2002;
Emmorey et al. 2007; McCullough et al. 2005).
In sign language research, the transcription and analysis of signs and gestures
are fundamental steps, but this is complicated by the fact that signed languages
are visual languages with no standard written form. Thus, transcription from video
involves making annotations of some kind, frequently glossing signs by the nearest
spoken-language equivalents, despite the fact that there is no one-to-one corre-
spondence between a sign in, e.g., ASL, and a word in, e.g., English. Moreover,
it is difficult to capture essential information about non-manual gestures – move-
ments of the head (including periodic movements such as nods or shakes) and upper
body, and facial expressions (e.g., raised or lowered eyebrows, eye aperture, mouth
movements, etc.) that carry important grammatical information – which occur over
phrases (frequently sequences of individual manually produced signs).
Somewhat at odds with sophistication of topics in sign language research has
been a long standing lack of consensus on how to annotate signs and gestures and a
lack of appropriate tools to aid with the transcription (see Miller 2001; Pizzuto and
Pietrandria 2001). One of the reasons for the lack of a standard method has to do
with theoretical and analytical assumptions reflected in symbols and notation sys-
tems (Emmorey and Reilly 1995; Miller 2001). “One of the most difficult issues
for sign language is how to represent space itself and the linguistic distinctions
conveyed by space . . . ” (Emmorey and Reilly, pp. 14–15).
Role of NI: It is precisely on issues such as the annotation and spatial frameworks
for coding sign language that NI offers refreshingly new and efficient options. NI
lends the methods of digital recording and computerized motion analysis in the
capture and analysis of gestures. It also offers ways to database a range of gestures
and signs and to mine this data. The third example in this Section (3.3.3.3) illustrates
the case.
The potential of NI in the neuropsychological study of speech and language in
general is illustrated below through three examples. The first is a computer model
of motor control in speech production, but one that relates to the cortical circuitry
for language, and accounts for various phenomena in speech and language. The
second is a web-based aphasia database designed to help clarify aphasic syndromes
by using computational methods to cluster symptoms. The third is a computerized
tool for digitally annotating and analyzing gestural language.
out. It is in this context that attention is given here to DIVA (and this relevance
elaborated later in this subsection).
DIVA is both a neural and computational model of SPA. It is neural in that all its
components are tied to centers of the cortex or cerebellum – it posits a broad neural
network that is informed greatly by neurobiological and neurophysiological data. It
is computational in that (a) many aspects of the model can be simulated in computer
modeling, which includes the learnt control of an artificial voice tract (synthesizer)
and (b) data from other avenues of SPA research, especially fMRI research, can be
modeled in the DIVA framework for further interpretation.
Detailed descriptions of DIVA can be found in the above references.3 A synopsis
is provided below.
Synopsis of DIVA: The DIVA model involves four major classes of information
or reference frames: (a) an acoustic frame that deals with how sounds are coded
as auditory maps; (b) a phonetic frame constituted by sets of speech sounds learnt
through the neural network; (c) a somatosensory (orosensory) frame that receives
sensory/proprioceptive information from the vocal tract to monitor the sound being
produced, and (d) a motor (articulatory) frame that controls the speech musculature.
DIVA also specifies the nature of the relationships or mappings between the four
frames as well as feedforward and feedback systems. Mapping refers to the trans-
formation of a neural representation that occurs between one node of the neural
network and another.
Figure 3.4 is a schematic of the DIVA model. The neural systems underlying the
framework can be grouped as three interacting subsystems involved with the con-
trol of speech production. These are (1) a feedforward motor system and a feedback
system that involves (2) a subsystem for auditory feedback and (3) a subsystem
for somatosensory feedback. A region in the left ventral premotor cortex, within
Broadmann’s areas 6 and 44, is hypothesized to be a starting point in speech sound
production. This area contains speech sound maps, i.e., higher-level motor repre-
sentations of speech sounds. A speech sound refers to any single speech sound unit
such as a phoneme or syllable. When a speech sound map is activated, the three
subsystems are involved in the relay of afferent signals to the motor cortex to drive
and control the production of speech:
The Feedforward Control Subsystem: The feedforward system is involved in the
actual, primary production of a speech sound. Feedforward signals (A) from the pre-
motor speech area have two components, one going directly to areas of the primary
motor cortex and the other going first to the cerebellum and then to motor cortex
speech areas. Target regions are all areas of the motor cortex involved with speech
sounds (e.g., areas controlling tongue, lips, mandible). A region may code the posi-
tion and velocity of a part of the speech apparatus in directing a motor movement.
3 DIVA has been described with immense mathematical and neuroanatomical substantiation and is
best appreciated when viewed against its mathematical and theoretical backing. For these details,
see especially Guenther (1995), Guenther et al. (1998), and Guenther et al. (2006).
64 3 Neuroinformatics for Neuropsychology
Fig. 3.4 Schematic of neural processing stages in speech production and acquisition as described
by the DIVA Model (From, http://speechlab.bu.edu/diva.php; adapted with permission, Frank
Guenther)
The likely cerebellar regions are the anterior parvermal and superior lateral areas as
target input areas and the medial subcortical cerebellum as the output area.
The Feedback Control Subsystem: This subsystem involves auditory and
somatosensory feedback systems.
A set of premotor signals (B) generated by a speech sound map targets the
auditory area/superior temporal cortex. The pathway of this signal encodes the
spatiotemporal auditory properties of the speech sound. When the auditory area
receives feedback (C) via subcortical nuclei (the sound of the produced speech), a
comparison is made between this sound and the expected sound. Two auditory maps
are involved, a state map and an error map. In neural terms, coordinates of activa-
tion in the auditory area (state map) are compared with the expected coordinates of
activation based on previous learning. If a discrepancy occurs between expected and
activated coordinates, an error map is generated. A corrective signal (D) is then sent
to the motor cortex where required motor adjustments are made. The hypothesized
loci of state map cells are the medial aspect of Heschl’s gyrus (primary auditory)
and the anterior planum temporale (association auditory). Auditory error map cells
are thought to lie at the temporoparietal juncture in the left Sylvian fissure or in the
lateral aspect of the posterosuperior temporal gyrus.
3.3 Neuroinformatics Applications and Models for Neuropsychology 65
Another set of premotor signals (E) generated by a speech sound map targets the
speech areas of the somatosensory cortex. This pathway encodes the spatiotemporal
somatosensory properties of the sound. The somatosensory areas also receive tactile
and proprioceptive feedback signals (F) from the speech articulators during produc-
tion. As with the auditory feedback system, two maps are involved. The occurring
somatosensory activation pattern is compared with the expected pattern (state map)
coded through previous productions. Discrepancies are coded in an error map. It in
turn sends corrective signals (G) to the motor cortex. Somatosensory state maps are
hypothesized to lie in the inferior postcentral regions for the speech; lips, tongue,
larynx, etc. Error maps are thought to lie along the supramarginal gyrus, in the
inferoposterior parietal cortex.
Feedback and Feedforword Process in Speech Acquisition: In the DIVA model,
attempts at production of a speech sound can begin after the acoustic proper-
ties of the sound have been coded, i.e., auditory coding in neural terms has
been established. The first attempt at producing a particular sound will typi-
cally be crude. The feedforward neural circuitry will be operating for the first
time and hence will be greatly reliant on the auditory feedback system for feed-
back of the sound errors. With subsequent attempts, feedback is more smoothly
integrated, the feedforward system becomes smoothly conditioned, now produc-
ing fewer errors and becoming less dependent on auditory feedback. The same
principles would apply to the somatosensory aspects – reliance on propriocep-
tive input is lessened with practice as the sound’s proprioceptive “signature”
gets well coded and the burden of the somatosensory error to state check is
minimized.
With the above described representational and sensorimotor processes, DIVA
provides an account of how individual speech sounds such as phonemes are strung
together in speech acquisition and how neural signals are transformed in produc-
ing fluently sequenced movements of the speech articulators. The framework also
enables DIVA to account for a variety of speech-language phenomena or provide a
unique investigative avenue into these phenomena. Three examples are very briefly
conveyed here:
Infant Babbling: According to the model, it is during the babbling phase that crit-
ical aspects of sensorimotor parameters of speech sounds are tuned. Feedforward
and feedback signals involved with babbling allow the infant to mark the canonical
range or signature of sensory-acoustic, motor, and sensorimotor states tied to par-
ticular speech sounds. This example very well conveys the self-organizing system
described by DIVA where learning is based entirely on sensory and motor neural
information.
Phonetic Categories: The neural dynamics of acoustic representational (neural)
space for phonemic categories are richly detailed in the DIVA model. Phonemic
encoding occurs through a consolidation of neural representational matrices – as a
sound is learnt and successfully produced, the neural cluster encoding the sound
in the auditory/association auditory cortex is tuned to rely on fewer neurons with
a tighter relationship between them. Prototypical sounds learnt in one’s native lan-
guage are, for example, coded in this manner. Later learning of a non-prototypical
66 3 Neuroinformatics for Neuropsychology
4 The case of Erik Ramsey and the DIVA-inspired electrode signal decoding of his premotor speech
neural signals has already received much attention in the popular media (for example, New Scien-
tist, July 9, 2008; CNN, December 14, 2007). A major collaborator on the project has been Dr.
Philip Kennedy at Neural Signals, Inc. in Atlanta, Georgia.
3.3 Neuroinformatics Applications and Models for Neuropsychology 67
applied to the data could potentially yield fine tuned relationships between the data,
advancing significantly over conventional methods of diagnosis. In addition, any
success with a database and computational techniques relating to classification of
such a complex set of syndromes (aphasia) would provide an informatics model for
classification of problems in general medicine.
Structure and Contents of the AD: The AD system uses a server-client set up.
Data for the AD were originally collected in the format of a Microsoft Excel file. The
data were then exported to a Notes/Domino database residing on a Lotus Domino
server5 (in Denmark). Pages of the server website are accessed in database views
in any browser. Data collection for the AD began in 1986 in the Department of
Neurology at the RWTH Aachen. The database currently has 265 records.
Two main types of data are included in the AD, aphasia test profiles and neu-
roanatomic lesion data. Language profiles for the AD are generated by the Aachen
Aphasia Test (AAT), a test battery commonly used in German speaking countries.
The AAT contains six subtests: spontaneous speech, token test, repetition, written
language, confrontation naming, and comprehension, each designed to tap a partic-
ular symptom. All subtests have multiple subcategorical measures, for example, the
repetition subtest, involves tests of single phonemes, monosyllabic nouns, local and
foreign words, compound words, and sentences. Database entries are the numerical
scores of the subtests.
The neuroanatomic data in the AD are lesion profiles obtained through CT scans.
Neuroanatomic data are included for only a subset of 146 patients in which the
lesion sizes were found to be relatively stable. The AD holds a lesion profile as a
set of 3D coordinates of the border lesion. These coordinates are derived from the
Aachen Voxel Model (see Section 2.2) that can be used to describe the locus of a
lesion using standardized 3-dimensional model of a normal brain.
For the 265 records in the database, many types of (aphasia) diagnosis are
included but the most frequently occurring types are Broca’s, Wernicke’s, Global,
and Anomic aphasia.
Neural Modeling and Diagnoses with the AD: The AD and its related modeling
system are geared to standardization and consistency in relating aphasic symptoms
to aphasic syndromes. Tied to the AD is a neural network modeling program. The
point of this artificial neural network is the association of symptoms with diagnoses,
computationally. The neural network is programmed to examine the various symp-
toms, i.e., test scores in the AD. The multilayered network selects and classifies
particular features of the symptoms and outputs diagnoses with a specified degree of
certainty of Broca’s, Wernicke’s, Global, or Anomic aphasia. Figure 3.5 represents
5 The Notes database is a non-relational database. A data object or “note” may consist of any
number or type of data fields, and the database can hold any number of these notes. Because the
database header and note header contain certain kinds of information, a user can search multiple
Notes databases at the same time. This structure is advantageous for the efficient storage of diverse
data types in the same database. Lotus Domino is a very flexible server application that can be used
with a variety of Lotus Notes applications.
3.3 Neuroinformatics Applications and Models for Neuropsychology 69
articulation Artificial
Neural
automatized language Network Anomic Aphasia
repetition
reading aloud
Fig. 3.5 Schematic of neural network diagnosis by classification of AD Test Scores. (Adapted
from Axer et al. 2000a,c; and Jantzen 1999a; printed with permission, Elsevier)
this process. The key aspect of the neural network is the classifier. It is capable of
classifying symptoms (score patterns) even if it has not previously encountered these
symptoms. Using a back propagation learning method, the network was trained on
a test set drawn from the database. For example, high scores in measures a, b, and c
combined with low scores in measures x and y and an average score in measure z,
may amount to the profile of a certain type of aphasia. When the network encounters
this pattern consistently, it reinforces its association between the particular symptom
profile and the diagnosis. The classifier can be adjusted in terms of its complexity
and the number of input data. In experimental trials with AD data, one classifier
produced correct diagnosis in 92.4% of test cases (Jantzen 1999).
The AD’s modeling component can also use another approach in searching
for similarities in the language assessment and neuroanatomic data. The nearest-
neighbor approach measures an abstract (standardized) distance between an object
in the database and all other objects across all records or selected records. The
objects most similar to a given object can thus be found. This method can, for exam-
ple, be applied to the lesion data to gauge the average anterior border (coordinates)
among all cases of Broca’s aphasia.
The AD stands as an embodiment of the potential that neuroinformatics holds for
research in language and aphasia. With a relatively small but carefully categorized
data set, it has demonstrated the advantages of digitizing aphasia test data and the
utility of neural network models in analyzing the data. The AD initiative relates
to the need for standardizing aphasia test data, the need to objectively diagnose
a syndrome using a certain data set, and the need to develop ways to accurately
associate the data with neuroanatomic loci. In this regard, AD, a neuroinformatics
system, is an outstanding development in aphasia research and provides a strong
foundation for other NI initiatives in aphasia.
70 3 Neuroinformatics for Neuropsychology
individuals. SignStream! R
, as a digital, computational, analytic tool, was developed
in direct response to the problem.6
The General Structure and Functionality of SignStream! R
: At the National Cen-
ter for Sign Language and Gesture Resources (NCSLGR) at Boston University,
a subject’s gestural communication is captured through digital video recordings.
Multiple synchronized digital cameras simultaneously record the subject’s gestures
from different angles. Similar hardware arrangements can be set up at different
research centers with relative ease. The digital video recordings are made publicly
available in multiple movie file formats; SignStream! R
requires formats compati-
TM
ble with QuickTime . Segments are marked that correspond to different gestural
utterances. These individual “units” of utterances are the data units that populate
the SignStream! R
database – which is therefore a collection of utterances. Since an
utterance is usually captured from various angles, more than one video file may be
associated with the utterance. The user first annotates a segment of video by enter-
ing data about the utterance. This produces a corresponding transcription for that
utterance. The SignStream program provides numerous data fields through which
an utterance may be coded. The three main types of data fields are non-manual
fields, gloss fields, and text fields. Types of non-manual gestures (e.g., raised eye
eyebrows, rapid headshake, or squinted eyes) are coded through non-manual fields.
For each event, annotation includes identification of start and end frame, in addi-
tion, potentially, to onsets, offsets, or holds of particular gestures. The ability for
part of speech tagging is included and there is also an English translation field avail-
able. The new version of the program will provide additional tools for fine-grained
phonological analysis (e.g., specification of hand shapes and movement types).
The data fields that are provided (see above snapshot) also allow for distinctions
in the intensity of specific gestures. Non-manual gestures are coded through multiple
fields relating to form, position, and movement of the eyes, head, etc. Head behavior,
for example, is coded through 8 fields that amount to a rich 3-dimensional coding
frame. They address head position at the starting (base) point and then in relation to
x, y, and z coordinates.
SignStream! R
offers some intuitive ways for data entry such as manual typing
into the fields or selections from drop-down menus. It also allows the user to create
new fields that the user may wish to define – enabling the description of new types
of data. Numerous other options are available for screen display parameters and for
audio and video controls.
On the SignStream! R
program’s screen, a video segment and its corresponding
transcription can be viewed in separate adjacent windows at the same time. Up to
four video files may be displayed for each utterance. As a video segment plays, a
media alignment indicator (a vertical bar) moves from left to right in the transcrip-
tion window. This provides for the temporal alignment of the linguistic features
6 There are now also other computer-based tools available for transcription of multimedia data
(e.g., Elan; URL: http://www.lat-mpi.eu/tools/elan/).
72 3 Neuroinformatics for Neuropsychology
Fig. 3.6 Snapshot of SignStream video and gloss windows showing annotated ASL data. (Figure
provided by Carol Neidle and printed with permission. © Boston University/National Center for
Sign Language and Gesture Resources)
described in the transcription window with the visual images in the video window.
Hence, the program offers dynamic linkage of the coded linguistic data with the
visual data.
The SignStream! R
Database: The database links transcription data for an utter-
ance with its respective video files. A sample database accompanies the application.
It contains 10 short American Sign Language (ASL) utterances. It provides the start-
ing point for users to develop their databases. In addition, users may sign on to an
online database ( http://ling.bu.edu/asllrpdata/queryPages /) from which data can
be drawn for analysis. The online database therefore acts both as a distribution point
for SignStream! R
encoded data as well as a data repository serving the ongoing
effort of building large shared corpora of sign and gestural language. At present,
the data in the online database is from native signers of ASL. The online database
now contains over 1300 utterances gathered primarily from four subjects since data
collection began in 1999. It is a constantly expanding database and the brand new
web-based Database Access Interface (DAI) greatly facilitates searching through
the data and downloading subsets of videos and annotations that may be of interest
to particular researchers. The DAI provides an array of search options to the user, as
does the search capability within SignStream! R
itself. Many types of queries rang-
ing from simple to complex may be formulated. Search criteria may be combined, to
search, for example, a specific sign only when it co-occurs with a specified gesture
3.3 Neuroinformatics Applications and Models for Neuropsychology 73
such as a head tilt to the right. Temporal search criteria may be applied such that a
gesture may be located only when it occurs with, before, or after another gesture.
A script window within SignStream! R
allows the user to select a set of utterances
and position them to be viewed in a particular order, a function that is particularly
useful in teaching and research. Queries and scripts can be saved for future use.
The NCSLGR shared corpus of data has achieved a size and scale such that
subsets of data can be extracted to study a particular feature of gestural commu-
nication. These subsets of data can then be grouped and benchmarked to form
smaller, specialized databases that relate to discrete features. Dreuw et al. (2008)
have reported on three data subsets that relate to language recognition in ASL.
These have been created at Rheinisch-Westphalian Technical University (RWTH) in
Aachen, Germany, drawing from the central database at Boston University. One of
the newly created databases at RWTH is the RWTH-Boston-400, a database of 843
sign language sentences. This is currently the largest publicly available benchmark
corpus for video-based continuous sign language recognition.
Computational Analysis of Gestures: A major aspect of the collaborative research
involving linguists (esp. Carol Neidle) and computer scientists (esp. Stan Sclaroff
at BU and Dimitris Metaxas at Rutgers) has to do with developing algorithms for
the automatic analysis and recognition of gestures and motion in communication.
Algorithms used in computer-based recognition/machine vision are applied in cod-
ing the forms, positions, orientations, and trajectories of the eyes, face, head, hands,
upper body, etc. Because these body parts move simultaneously and often subtly
in communication, it is an enormous challenge to classify their motions and their
relationships with regard to each gesture. Computational algorithms enable this kind
of analysis, and with such a fine-grained analysis comes the potential to sift out rich
yet often-elusive phonological details contained in gestural communication. The
project has given much attention to algorithms for coding head motion in view the
linguistic significance of head/face movements.
In the context of neuroinformatics, SignStream! R
is particularly remarkable. As
a unified platform, it seamlessly integrates so many facets of NI – it offers a com-
puter application geared to the analysis and databasing of gestural language. To aid
the analysis, it develops computational methods for finely deconstructing this com-
plex behavior. It opens itself to the research community worldwide. SignStream! R
This subsection has described three major developments that serve NI approaches
to speech-language research. Clearly, there are many ways of applying NI to
neuropsychological research on speech and language.
There are other NI initiatives in this area, either smaller projects or projects indi-
rectly related to neuropsychology. The Aphasia Modeling Project (also known as
74 3 Neuroinformatics for Neuropsychology
Webfit)7 is an online tool for the analysis of aphasia picture naming and repetition
data, using the two-step model of lexical access in language production (see Dell
et al. 1997; Dell et al. 2007; Foygel and Dell 2000). A patient’s picture naming test
data are entered into a program and are mapped to the model’s parameters that fit
normal error patterns. The model can then be “lesioned” by adjusting phonologi-
cal and semantic variables to produce various error patterns. There are also tools
geared to linguistics and normal language. WordNet8 is an online, downloadable
lexical database of English. It contains open class words (nouns, verbs, adjectives,
and adverbs), which it organizes as synonym sets. Lexical, semantic, and conceptual
relationships link these synonym sets such that a word search generates numerous
class and contextual attributes. Such a database of the English lexicon and rela-
tionships of its elements makes for a useful tool in cognitive and computational
linguistics. Latent Semantic Analysis (LSA) is a statistical method for extracting
and representing semantic relationships between words in phrases that are contextu-
ally used in written documents across a body of knowledge (Landauer and Dumais
1997; Landauer et al. 1998).9 The LSA method mathematically decomposes text
into a matrix, which allows for the analysis of each work or passage in relation to
other words and passages. The representation of a particular body of text (e.g. text on
social psychology, cardiac research, French literature) is defined LSA as a semantic
space. Words within this space (i.e., corpus of texts) are mathematically coded in
terms of particular vectorial dimensions and weightings based on their occurrence
in relationship to other words. Using a technique akin to factor analysis, a partic-
ular usage of a word is weighted in terms of the average meaning derived from its
usage throughout the semantic space. The meaning of a passage of text is coded
in relation to the average value/meaning of all its constituent words. LSA therefore
extracts relationships written discourse using a statistical technique as opposed to
using a dictionary, knowledge base, or artificial intelligence program. Its frame of
reference is generated by the semantic space in question and so offers a unique way
of analyzing patterns in the discourse of the texts. LSA is of special relevance to
neuropsychological assessment of language in that its methods can be adapted in
analyzing word usage and discourse in aphasia.
NI brings promising new avenues to the neuropsychology of speech and lan-
guage. The sophistication in the methods and tools of NI are invaluable in matching
up to the cognitive and neural complexity of speech and language.
7 Webfit is a project of the Language Production Laboratory at the University of Illinois Urbana-
Champaign. URL: http://langprod.cogsci.uiuc.edu/cgi-bin/webfit.cgi
8 WordNet is based at the Cognitive Science Laboratory at Princeton University. URL:
http://wordnet.princeton.edu WordNet: An Eletronic Lexical Database (Fellbaum 1998) is a
written guide to the online tool.
9 The Latent Semantic Analysis project is based at the University of Colorado at Boulder URL:
http://lsa.colorado.edu / See also the Handbook of Latent Semantic Analysis (Landauer et al. 2007)
3.3 Neuroinformatics Applications and Models for Neuropsychology 75
2003). Because they are associated with causes rather than effects of a condition,
they are viewed as more reliable diagnostic indicators of syndromes (Zobel and
Maier 2004). In contrast, the traditional methods of neuropsychiatric classification
and diagnosis in psychiatry, as embodied by the DSM IV and its predecessors, are
based on general consensus around observed clinical features and are theoretical.
(Clinicians need only be reminded of the vigorous discussion and debates regarding
classification of mental disorders that played out in medical journals over the last
three decades.10 ) Using neurocognitive endophenotypes as markers of neuropsychi-
atric conditions may bring a more scientific, objective dimension to neuropsychiatric
classification and diagnosis.
The role of neuropsychology: Neuropsychology has a crucial role in phenomics
research. Neuropsychological constructs and tools are relied upon to describe cog-
nitive and neuropsychiatric endophenotypes such as response inhibition, attentional
control, and verbal working memory. Consider how performance on the Wiscon-
sin Card Sorting Test is viewed as an index of executive function/set switching and
prefrontal functioning; neuropsychological phenotypes are about such processes but
with a need for more precision in their definition (see Bearden and Freimer 2006;
Kremen et al. 2007).
In the context of phenomics research, neuropsychology also finds itself oper-
ating a sophisticated web of neuroinformatics. Neuropsychology is challenged to
describe cognitive-behavioral phenotypes in highly refined ways and in a manner
that can be interfaced with other levels in inquiry (see Bilder 2007b; Bilder et al.
a, in press). To accomplish this, neuropsychology has to be informatics compatible.
How neuropsychology is integrated in current phenomics projects provides a broad
and compelling illustration of what it means to have the discipline of neuropsychol-
ogy geared for NI, why this should be the case, and ways to go about the process
(Section 3.2 stressed the importance of defining NI systems that can be considered
unique to neuropsychology. It also indicated that NI systems for neuropsychology
should be designed with the view of potentially integrating neuropsychological data
with data from other disciplines. The examples of phenomics research described in
this section, especially the second example, demonstrate particularly well the latter
aspect.)
Two examples of phenomics research are described below. The first concerns a
body of work that seeks relationships between craniofacial dysmorphology, neu-
roanatomy, and neurosychological variables in neurodevelopmental and neuropsy-
chiatric disorders, using informatics methods. The second describes a large-scale,
NIH-funded research consortium for neuropsychiatric phenomics, through which
numerous novel and powerful NI developments in neuropsychology are being pio-
neered. It is a unique example of the kind of informatics infrastructure that serves
knowledge building in neuropsychology.
10 See for example, Spitzer et al. (1975); Spitzer et al. (1980); Williams & Spitzer (1982); Klerman
et al. (1984); Sadler et al. (1994).
3.3 Neuroinformatics Applications and Models for Neuropsychology 77
a large, normative database (Farkas and Deutsch). The database permits quantifi-
cation of craniofacial dysmorphology, in, for example, autism and schizophrenia.
Phase 1 of the database development is complete. It involved the creation of a sim-
ple PC-based tabled database. The database consists of numerous anthropometric
measurements, each entered in separate column and subjects are entered by rows.
Data continue to be added as studies progress. Phase 1 also involved the overall plan
for the incorporation of a neuropsychological data component. Having data on both
physical and neuropsychological features (a) will better enable the search for possi-
ble links between different classes of phenotypes in autism and (b) could possibly
point to etiological subtypes in autism. Deutsch and Joseph (2003) demonstrated this
kind of convergence in a study that linked uneven development of non-verbal/spatial
skills in a subgroup of autism patients who also happened to manifest unusually
large head circumferences (macrocephaly being a frequently seen feature in autism).
In Phase 2, the researchers are now developing database functions that introduce
descriptive statistics and data visualization functions. The aim is to work towards a
larger normative database that will be referenced against a craniofacial atlas.
The studies of craniofacial and brain dysmorphology described above help
to map out complex etiologies in neurocognitive disorders by examining con-
vergences of certain physical and behavioral traits. Matthysse et al. (1992), in
describing a few ways to clarify the complex mix of genetic factors and phe-
notypes in neuropsychiatric conditions, stated that “[i]n the long run, the most
successful method is likely to be direct refinement of non-Mendelian behav-
ioral and physiological traits into more fundamental components” (p. 461). The
behavioral component, to do with defining cognitive-behavioral phenotypes, is
essentially neuropsychological. In such an intensely multivariate research context,
NI offers the enabling framework for the interface between neuropsychological and
physiological data.
different levels of phenotypic data (cellular, neuronal, anatomic, etc.). As was the
case with the HGP, the proposed HPP will require new technologies to automate
complex processes of data analysis, comparison, and modeling. The proposal also
emphasized that novel collaborations and research methods will be essential to the
project.
The Consortium for Neuropsychiatric Phenomics (CNP) is a large, new consor-
tium that is focused on neurobehavioral phenotypes. It can be viewed as the first
large-scale phenomics project/component to be embarked upon that fits in with the
proposed HPP. Thus all that is described in the above paragraph is consistent with
the CNP. The CNP is based at the University of California at Los Angeles and
involves 52 researchers and clinicians from across the institution. Earlier work at
the UCLA Center for Cognitive Phenomics laid the platform for the CNP.
The CNP is funded by 5-year, a 22 million dollar grant from the National Insti-
tutes of Health. Among its strategic initiatives laid down in its 2006 Reform Act (see
Krensky 2008), the NIH established a fund to support transdisciplinary research. It
set forth a series of “Roadmap Initiatives” to fund highly integrative, transformative
biomedical and behavioral research. It stressed research opportunities in emerg-
ing interdisciplinary junctures in medical science and the potential therein to solve
complex medical/health problems. It also stressed the role of bioinformatics as new
pathways to discovery. Nine grants were awarded to research teams around the US.
One was a grant to establish the CNP, which would lead numerous initiatives in
behavioral phenomics.
Of the projects funded by the NIH Roadmap Initiatives program, CNP is the
only one that has a distinct neuropsychological component. Robert Bilder, a clinical
neuropsychologist by training, is the primary investigator and director of CNP. It
should be noteworthy to neuropsychologists, that this groundbreaking, informatics-
infused initiative is led by a neuropsychologist. (At the February 2008 meeting of the
International Neuropsychological Society, in Waikoloa, Hawaii, a continuing edu-
cation course titled “Phenomics for Neuropsychology: Transdisciplinary Strategies
for Research on Brain-Behavior Relations in the Post-Genomic Era” was given by
Robert Bilder.11 )
The Mission and Strategy of CNP: Simply stated, the CNP seeks to identity
neuropsychiatric phenotypes tied to perception, cognition, and emotion. This is
motivated by the complete lack of a knowledge base on cognitive/neuropsychiatric
phenotypes – “there is no coherent knowledge base representing cognitive con-
cepts and cognitive measurements” (Bilder 2008c). Building a knowledge base of
behavioral phenotypes can potentially serve to unravel the mechanisms of neuropsy-
chiatric syndromes and aid in more accurate classification of clinical disorders (see
discussion on endophenotyopes above): It is often the case with neuropsychiatric
disorders that a set of endophenotypes may manifest in more than one disorder, i.e.,
similar behavioral features mediated by common neural mechanisms may manifest
11 The references cited, Bilder 2008b,c, refer to Robert Bilder’s 97 slides presented at the
conference and the audio recording of the lecture, respectively.
80 3 Neuroinformatics for Neuropsychology
in two or more different disorders. This can blur both the scientific understanding of
a disorder and its diagnosis. Of greater value are certain “signature” endophenotypes
or combinations of endophenotypes that may be associated with a neuropsychiatric
condition. Also of interest is how shared endophenotypes may vary in subtle ways
across disorders.
Understanding the behavioral/neural endophenotypes specific to a disorder leads
to the possibility that disorders can be accurately diagnosed by sampling of tar-
get behavioral phenotypes. The idea comes from genome-wide associations studies
(GWAS), where a rapid scanning of genetic markers as opposed to a scan of the
entire genome can reliably signal risk or presence of a condition. If one attempts
to link the rather precise genomics data to neuropsychiatric data, one is forced to
use conventional psychiatric clusters (syndromes) as the neuropsychiatric data set.
There is wide discordance between the two data types – symptom clusters are not
compatible with a scientific structure that seeks to mechanistically link syndromes to
genes. This seriously inhibits linkage between genotypic and phenotypic data in try-
ing to understand neuropsychiatric mechanisms. Hence at one level (“horizontal”),
the CNP seeks a mapping of neuropsychiatric phenotypes to better characterize
neuropsychiatric disorders and link them to underlying genotypes.
The CNP also seeks to describe data such that they can be related to levels of
analysis above and below the level of cognitive phenotypes. Refer to Fig. 3.7a.
The levels above are symptoms and syndromes (and various clinicians and clini-
cal researchers are involved at these levels). The levels below cognitive phenotypes
are those of neural systems, cellular systems, proteins, and genes (each involving
different specialists in biology/neuroscience). These levels of analysis define the
“vertical” structure of the CNP – researchers at each of these levels aim to generate
data relating to the same set of neuropsychiatric disorders. The greater goal of the
Syndromes
Symptoms
Cognitive
Cognitive
Phenotypes
Phenotypes
Neural Systems
Fig. 3.7a Schematic of the
multi-leveled (vertical Cellular Systems/
structure) of the each level, Signalling Pathways
and the position of cognitive
informatics in CNP, the focus Proteins
within relation to the other
levels. From,
Genes
www.phenomics.ucla.edu/index.asp;
printed with permission, Environment
Robert Bilder
3.3 Neuroinformatics Applications and Models for Neuropsychology 81
CNP is to draw mechanistic links between the gene level and the syndrome level,
through all the other levels. With such a multileveled investigative paradigm and an
emphasis on phenotypes, “we may start to see phenotypes quite different from what
we are used to as the genotype data sets start to come in” (Bilder 2008c).
Neuropsychology and Neuroinformatics in the CNP: Of the CNP’s five interre-
lated research projects, three are concerned with schizophrenia, bipolar disorder,
and ADHD. They presently focus on neurocognitive phenotypes relating to mem-
ory mechanisms and response inhibition, which cut across all three disorders. (Since
the phenome is so vast, a manageable study must limit the number of phenotypes
on which it focuses. The proof-of-concept that is established paves the way for
expansion.)
A current study aims to examine the phenotypes (and their corresponding geno-
types) in a sample of 2000 normal controls and a sample of 300 individuals who
suffer one of the three disorders of interest. In both groups, genotypes and physical
and functional neuroanatomy are examined and cognitive/neuropsychological func-
tions (working memory and response inhibition/cognitive flexibility) are assessed.
For levels of analysis (molecular and cellular) that cannot feasibly be carried out on
human subjects, transgenic mouse models are developed to examine very specific
effects.
In the context of the CNP’s work, neuropsychology bears a very sophisticated
role. This involves defining cognitive/behavioral endophenotypes, which in turn
involves thoroughly examining various cognitive constructs. Existing neuropsycho-
logical tests that may tap an endophenotype has to be selected or new tasks have to
be tailored to an endophenotype. All of this has to be done with the consideration
that the data have to be meaningfully integrated with other levels of analysis and
that comparable tasks have to be constructed when devising animal models.
One of the goals of the CNP is to develop statistical and informatics tools for
analyzing the data across the many research levels of the consortium. The tools
relate to data visualization, modeling, and hypothesis generation (see Bilder et al.
a, in press; Sabb et al. 2008). PubGraph, PubBrain, and Phenowiki are three novel
tools that have been developed by CNP and available for public use:
URLs: http://www.pubgraph.org/
http://www.pubbrain.org/
http://www.phenowiki.org/
PubGraph is a tool for literature mining and for the visualization of the associa-
tions between a set of results. It allows the user to enter one or more search terms,
which it uses to search PubMed/MEDLINE. It then generates a graphical view of the
results, representing resulting terms and concepts as nodes. Lines or “edges” link-
ing the nodes convey coefficient and correlational strengths of associations between
the terms, as statistically inferred through literature mining. Figure 3.7b shows the
resulting graph for the search of the term “schizophrenia.” One way by which the
tool can be utilized is in aiding neuropsychological refinement of a cognitive con-
struct. Strengths of association between the nodes of the graph ultimately reflect
factors and variables common to the nodes. The construct of “cognitive control” as
82 3 Neuroinformatics for Neuropsychology
Fig. 3.7b A PubGraph representation of the literature mining results for the term “schizophrenia”
(results on cognitive/behavioral studies are clustered on the left, and results on genetic/cellular
studies are clustered on the right). From, www.pubgraph.org/; printed with permission, Stott Parker
related in PubMed indexed literature, for example, is closely tied to the constructs
(nodes) of response inhibition, response selection, task/set switching, and working
memory (Bilder 2008b; Sabb et al. 2008).
PubBrain works in a similar fashion – the user inputs search terms, which are
queried in PubMed. The difference is that the results are displayed as a list of search
3.3 Neuroinformatics Applications and Models for Neuropsychology 83
terms that are extracted from the most frequently occurring search results. Simpli-
fied brain activation maps are generated based on the list of anatomical terms in the
returned list.
Phenowiki is a database-enabled wiki platform for the collaborative building of
cognitive phenotypes. It enables users to explore a concept (e.g., working memory)
by laying out definitions and citations, by describing related concepts, by annotating
the existing content on the wiki, and by linking the descriptions to other resources
such as PubMed, PubGraph, PubBrain, and Wikipedia. Phenowiki provides a space
that promotes incubation, development, and modeling of neurocognitive concepts.
Researchers are given a common medium in which their collective insights can be
brought to bear on ideas and constructs and this is interfaced with other knowledge
building resources.
There are other related developments at CNP: A Cognitive Atlas has being
built for illustrating maps/relationships between cognitive concepts. Cognitive
Ontologies that reflect the sophistication of cognitive concepts and their complex
associations are also being developed. Sophisticated search ontologies (see Section
2.5) are crucial for accurate data retrieval and for the building of detailed knowledge
bases.
These informatics-based tools developed by the CNP are more than just novel.
They are quantitative tools geared for the association of complex data and they are
a vital compliment to large and complex agenda of the CNP.
In the context of the CNP, neuropsychology has to operate in an environment
far different from that in which it evolved. In this kind of research and clinical
environment, the fine granularity on genomic and cellular data pertaining to a dis-
order sets the bar that shapes the entire working environment. Neuropsychology,
as a discipline, is required to apply the data in sharpening cognitive and behav-
ioral constructs. It in turn is relied upon to precisely identify and investigate
cognitive/behavioral phenotypes. It has to deliver neuropsychological data that are
meaningful in the larger scheme of an explanatory framework that stretches from
genomes to phenomes.
The study of cognitive and neuropsychiatric phenotypes is a rapidly emerg-
ing, leading-edge discipline that involves neuropsychology. Once again, being part
of such an enterprise requires that neuropsychology assimilate a rich variety of
informatics methods and tools.
The nine examples of NI for neuropsychology described in this section embody the
ideas and goals of the NI approach to research and assessment. In relation to their
respective neuropsychological domains, all the examples aspire to “capture, relate
and analyze diverse types of data at multiple levels of abstraction” – the goals of NI
as described by Wong and Koslow (2001, p. 104). All the examples can be described
84 3 Neuroinformatics for Neuropsychology
institution, etc. Many supported the idea, citing the potential opportunity to review
the work of others; facility to compare similar scans from different studies; and the
building of a large image database – the debate continues. The controversy, said
Gazzaniga, did a service to the field by engaging imaging researchers in an exami-
nation of the data sharing (see Aldhous 2000). “Sharing is justified because analysis
of the same data sets by different scientists might yield a different interpretation and
would not require someone to repeat the whole experiment to query the data with
a different question. Sharing would also provide a rich data source for theoretical
neuroscientists” (Koslow 2002).
Clinical neuropsychologists may sooner or later be engaged in similar debates,
as NI systems become part of their discipline. Some early lessons can be gained by
looking at how issues around data sharing have been mediated in neuroscience. Van
Horn and Gazzaniga (2002) described measures taken by the Dartmouth fMRI Data
Center to ensure subject anonymity and protection:
In compliance with US federal regulation 45 CFR 46 on human subject protection, we
had to address the requirement that any and all information that can be used to identify
a subject must be removed from the data, while maintaining its experimental integrity.
This is accomplished in a two-pronged effort. First, contributing authors remove identi-
fiers before submitting study information the Data Center. Then the Data Center checks for
identifiers that might have been missed by the authors, while removing other potential iden-
tifying subject descriptors To remove [the possibility facial contour reconstruction from
high resolution head images], the high resolution images are stripped of facial features.”
Many proposed methods for data sharing involve design elements in the NI system
to do with controlled access to information (see Lee et al. 2000). Some examples are
(a) A database that holds detailed raw data from multiple researchers but only gives a
researcher access to an index of the data; the index contains the contact information
for the contributor of each data set. With additional permission protocols, the data
set can presumably be obtained from the contributor in a ready-to-use, standard
format. (b) Data can be stored at progressive levels or stages, ranging, for example,
from specific personal descriptors to the broad diagnostic categories and at each
level, it can be divided in sets. Particular configurations of the data grid can be made
available to a researcher based on the level of permission he/she holds (IT network
and personal computer operating systems involve different levels of user rights).
The greater point in this discussion concerns not the exact details of data access
but the need for neuropsychology to recognize that engaging with NI also requires
engaging the question of access to primary data. Stephen Koslow, the former direc-
tor of the Office of Neuroinformatics at the NIH, has argued passionately on this
cause (for data sharing) in articles such as “Should the neuroscience community
make a paradigm shift to sharing of primary data?” (Koslow 2000) and “Sharing
primary data: a threat or asset to discovery” (Koslow 2002). In the current time,
these debates in NI have less to do with the question of whether data should be
shared; that is largely accepted. The questions are about the rules and guidelines
for data sharing that are needed. The need has come about suddenly with a new
enabling technology showing enormous prospects for knowledge discovery through
data sharing.
88 4 Obstacles and Aids to Neuroinformatics in Neuropsychology
Increasing throughput may involve among other things a great reliance on auto-
mated and online assessment tools: “Another possibility for increasing throughput
is to adopt less labor-intensive procedures [f]or example, it is possible to collect
information of certain behavioral features through assessment tools that can be
self-administered over the Internet by individuals throughout the world.” (Freimer
and Sabatti 2003). How open will the field clinical neuropsychology be to the
introduction of large-scale automation in such a core component of its charge?
How will these issues play out if neuropsychologists embark on developing an
open-platform online database of results from assessment batteries? What kind of
neuropsychological battery should be used? Would it be an adaptation of the widely
used WAIS? Would it be more akin to ANAM? What if one or the other is considered
inappropriate in terms of the precision with which a perceptual or cognitive feature
is parsed out? If one embarks on a very ambitious NI project such as that conducted
by the Consortium for Cognitive and Neuropsychiatric Phenomics, then it becomes
imperative that the data fields in the database be very finely tuned to a cognitive
phenotype.
Usage of any neuropsychological test or battery presupposes a definition of the
functions being tested (an issue raised many times in Chapter 3). Neuropsychol-
ogy has always been using taxonomies, both implicit and explicit. The big problem
here is that neuropsychology cannot optimally engage with NI without first con-
fronting some of its taxonomies. Some may be simply inaccurate in the light in
newer data and some may be so implicitly held in assessment tools that they have
never been questioned (see Chapter 3). With refined taxonomies, the NI systems
for neuropsychology can build ontologies that are more suited to the classification
of complex data (see Bilder et al. c, in press). Taxonomy and ontology develop-
ment influence each other. Data mining functions require some level of specification
of how groups of data are related and if these relations are inaccurately specified,
subsequent computed relations will be inaccurate (see Sections 2.5).
Almost certainly, the talk of shared databases, open-ware neuropsychological
batteries that will upload to these databases, and the radical reinvention of assess-
ment tools to make them NI capable will raise the ire of certain commercial entities.
These are the entities that have for many decades been pumping out spring-bound
cardboard booklets and painted puzzles, nicely contained in immaculate black brief-
cases. Neuropsychologists, generally well intentioned, have aided in the process by
authoring revisions of the batteries, and the field of neuropsychology has partaken
in establishing assessment norms for various tests. Altogether, the arrangement
has been that neuropsychologists have been dependent on these batteries (there
have been no viable alternatives); this has legitimized the batteries even though
they are based on questionable constructs and have designs that are antithetical to
shared electronic data building; and in return, neuropsychologists have paid pre-
mium prices for copyrighted sheets of paper containing some rows and lines, on
which to score results! In this era of NI, much superior alternatives are afforded.
In fact, if these alternatives are developed under federal funding programs, they
are likely to be freely available to neuropsychologists. How can neuropsycholo-
gists on a global scale achieve the development of assessment tools that are truly
90 4 Obstacles and Aids to Neuroinformatics in Neuropsychology
be required to be deposited following the rules of some federated ontology, so if you have
an index of working memory, you will be able to choose from a menu of what that is . . .
otherwise justify exactly how what you are doing is different and why you are not using
existing techniques. I think that’s where it’s going.” (Bilder 2008c)
Indeed that is where it’s going! All the signs and signals on NI in neuropsychology,
as laid out in Section 3.3 clearly say so. An urgent focus for neuropsychology is a
broad-based examination of how to make this transition as smooth as possible.
1A commercial entity that produces a NI tool that facilitates information sharing in neuropsychol-
ogy and does not impose numerous limits that serve primarily a profit making interest, cannot be
viewed as an entity that impedes NI in neuropsychology.
92 4 Obstacles and Aids to Neuroinformatics in Neuropsychology
on analysis of performance patterns on the Wechsler-Bellevue, which was hardly the opti-
mal instrument for the purpose, or on the first version of the Halstead-Reitan battery.”
(Benton 1992, p. 410)
Benton went on to suggest that assessment in the current era needs to be informed
by contemporary theory just as assessment decades ago needed to be informed by
theory of the time: “Given our present knowledge of brain-behavior relationships,
of cognitive processes . . . are we doing as well as we can? . . . In short, we should
take a hard look at the neuropsychological examination. It deserves its own critical
examination.” (p. 415)
As the discipline faced such cautions and calls for changes, it did apparently
respond – neuropsychology today is not what it was in its infancy and is marked
with many successful efforts towards formalization. What then is different about the
present call for neuropsychology to adopt NI? Will the field just adapt as it faces
the evolutionary pressures on NI? As already conveyed, NI is not just a technol-
ogy that can be tacked on nor does it amount to a retrofitting project. Unlike other
adjustments that neuropsychology may have made in the course of its development
(a tweak here and a recalibration there), NI implies deeply systemic changes. Nev-
ertheless, as the benefits of NI in neuropsychology begin to emerge, the system can
be expected to reorganize.
To aid the process, many simple steps can be taken within the field: Professional
bodies can appoint task forces to examine possible NI approaches and make rec-
ommendations. Conference organizers can aim to build NI themes into conference
agendas. Clinicians and researchers can seek alliances with informatics special-
ists to help design NI systems. Perhaps the most effective (conventional) way to
spread NI consciousness is through neuropsychology journals – by adding a com-
ponent/section devoted to NI or at least encouraging manuscripts on the topic. An
old problem in neuropsychology is that when relevant new technologies emerge
they are often outside the view of the neuropsychologists. Kane and Kay (1992)
noted this problem and Kane (2007) summed it up (referring here to computerized
assessment):
“Information about these efforts was available; however, it was found mostly in technical
reports or in journals not typically monitored by neuropsychologists. Kane and Kay (1992)
noted that the amount, quality and importance of the work being done in the field of com-
puterized assessment was not fully appreciated and was in fact occurring under the radar of
most neuropsychologists. (Kane 2007, p. S3).
who typically handle the more technical aspects of NI. It is imperative though
that the neuropsychologists familiarize themselves with the field of NI in general.
Neuropsychologists should have a good sense of NI potential – its functions and
capabilities – so that they can at least conceptualize different NI methods when
thinking of clinical and research problems. Of course, keeping abreast of cognitive
neuroscience will also aid the formulation of these ideas. Writing in the context
of automated assessment (see Section 3.1), Schlegel and Gilliland (2007) stated,
“Given the rapidly expanding frontiers of neuroscience, test developers will increas-
ingly have to keep one eye on test programming and one eye on neuroscience
research to remain abreast of construct validity advances.” (p. S54).
Conferring NI perspectives on neuropsychology trainees should be a high prior-
ity. This implies changes to the syllabi and requirements of clinical, doctoral, and
postdoctoral programs. Just as training in psychology generally requires courses
in statistics, budding neuropsychologists should be empowered by courses in NI
for neuropsychology. In making such curricular adjustments, a good resource may
again be federal funding agencies. The NIH, for example, has a funding program
for clinical research curriculum development; it supports the development of formal
courses in many biomedical areas including medical technology.
The more neuropsychology is informed about NI, the smoother its course of
engagement will be. Exposure to existing projects (Section 3.3), understanding
their rationale and their gains, and the emergence of a few more could help cat-
alyze change with the field. Bringing NI to neuropsychology will undoubtedly
be challenging but all that has occurred in NI thus far should be inspiring and
motivating.
Chapter 5
An International Society for Neuroinformatics
in Neuropsychology
URL: www.scnn.org
Neuroinformatics interest groups have steadily emerged within research orga-
nizations and academic societies. The NIH has one, as does the Society for
Neuroscience. These interest groups can vary in scope, from offering platforms for
discussion to acting as central coordinating facilities for global research in NI. The
neuroinformatics interest group of the OECD (see Section 2.6) is an example of
the latter. The importance of NI interest groups has been conveyed throughout this
book – they foster collaboration and promote NI initiatives.
In the course of writing the book, the author developed a collaborative network
with many clinicians and researchers whose projects are described in Section 3.3.
Much of the collaboration involved exchange of information, updates on NI projects,
reviews of data, etc. As noted in the introduction to Section 3.3, essential material
for the book was obtained through this collaboration. By the end of writing the book,
the author had networked a group of researchers with common interests.
A logical step was to form an organization to promote these interests. An inter-
national organization devoted to neuroinformatics for neuropsychology was hence
recently formed. It was named the Society for Neuroinformatics in Neuropsychology
(SCNN). A website for the organization has been launched (see above URL). The
website may be of benefit to researchers interested in NI for neuropsychology.
The SCNN will serve three broad functions: (a) foster discussion on the topic,
(b) compile and communicate resources, and (c) assist in coordination of projects.
These functions will to some degree overlap with each other. Emphasis on each
function will vary at different times depending on the needs expressed by members
and the specific priorities of the organization at a given point. However, the organi-
zation will remain structured around the three main functions. These functions are
summarized below (see Fig. 5.1).
Discussion Forum: A discussion forum will offer researchers the space to explore
ideas for NI in neuropsychology. It will also help to focus discussion on a number
of relevant topics such as
• Database platforms and distributed architecture for large-scale NI systems for
neuropsychological assessment data
• NI systems for the study of specific cognitive functions
• Synchronization of databases, computer platforms, and protocols between multi-
ples teams
Examples of Discussion
Topics
-General design of NI systems
Examples of Working Group Projects
-Database/network architecture
-Ontology for data mining
-Synchronization of systems
-NI in neuropsychology
-Theoretical frameworks curricula
-Data privacy -SCNN conferences
-Neuropsychological
PROJECT NI in National Health
DISCUSSION COORDINATION Information Networks
FORUM & WORKING
-Coordination of
GROUPS multiple small
SCNN scale projects
Examples
-Relevant NI projects
-Books, websites, etc
COMMUNICATION
& RESOURCES
-Funding programs
-Meetings, conferences, etc.
-Resources for software/hardware development
Fig. 5.1 Summary illustration of the major functions of the society for neuroinformatics in
neuropsychology
Project Coordination and Working Groups: As needs and projects arise, the orga-
nization will be able to draw on its network to form working groups and committees.
These may be set up to address, for example, some of the topics in the discussion
forum that may be actively discussed and thereby call for special focus committees.
Working groups could be set up to explore for example, ways of introducing NI in
neuropsychology training curricula, and the place of neuropsychological informat-
ics in the context of national health information networks. Another possible area for
a coordinating committee is the tracking and coordination of many small-scale NI-
neuropsychological efforts around the world so that research groups can consider
ways of making their data sets and NI systems interoperable.
The SCNN is a very new organization and is currently led by the author in col-
laboration with some of the researchers whose work has been described in earlier.
It is hoped that as the group gains a broader membership, more individuals can
contribute to its structure and mission. The SCNN website will provide updates
periodically.
Chapter 6
Concluding Remarks
Neuropsychology has a long and proud history. It grew out of behavioral neurology
and psychology and developed into a full-fledged discipline. So much of cognition
and behavior is meaningful only when studied in relation to the brain – and this
is where neuropsychology has successfully demonstrated its unique focus. Clinical
neuropsychology and experimental neuropsychology have been the tracks that have
formerly defined the discipline for more than half a century.
Since the emergence of “cognitive neuroscience” about three decades ago, there
has been a noticeable “shrinking” of the field of experimental neuropsychology.
Many topics and researchers that once belonged to experimental neuropsychology
have sought a more fitting (and perhaps more prestigious) home in cognitive neu-
roscience. One can debate endlessly the degree of overlap or difference between
experimental neuropsychology and cognitive neuroscience or even whether there
is any difference at all. What is less arguable is that the playing field of the
discipline of neuropsychology changed drastically in the 1980 s and 1990 s as
brain-behavioral research rapidly accelerated in disciplines like neurobiology and
as new imaging methods came onto the scene. The traditional neuropsychological
methods of deconstructing cognitive behavior and seeking relationships between
cognition and neural systems had by this point reached a plateau. Assessment
tests and lesion models in neuropsychology would remain valuable as always
but they had reached fundamental limits. Much clearer, discrete brain-behavioral
associations were being generated through behavioral-neurophysiologic investiga-
tions in primates and through cognitive functional imaging in humans. What was
once a clear transitional zone between neuropsychology and neuroscience started
to see some complex demarcations – making for the emergence of “cognitive
neuroscience.”
Clinical neuropsychology, though influenced by this change in the landscape,
remained largely as it was. Since it was a thriving discipline, established in its own
right, there was no pressure for fundamental shifts from its well-codified tools and
procedures. It continued along with its assessment batteries that had their roots in
psychometry and intelligence testing. It continued with its armament of individual
tests that were tuned to gross cognitive/perceptual complexes. A strong neural-
theoretical basis for many of these tools would remain unclear though the need
was acknowledged. It had revised and formed these tools over many generations
and they influenced the definition of the functional domains on which clinical neu-
ropsychology traditionally focused. For the most part, the data generated by these
tools were in physical form, viewed by the clinician and then stored away.
As was the case in almost all other disciplines, knowledge building in neuropsy-
chology was a slow and manual process. Data would be published in journal articles,
and over the years, patterns would be observed and models could be built.
A Major Change in the Neuropsychological Landscape: The landscape on which
neuropsychology operates is once again changing. The change began towards the
end of the 20th century. It continues into the beginning of the 21st century and all
indications are that it will transform the way many disciplines are practiced. This
change is all about data. It has to do with how data are extracted and dealt with in
this greater digital informational era. Once upon a time, clinicians and researchers
were relatively disconnected and only got to share their ideas and data in publica-
tions and at conferences. Now, through the Internet, through databases, and a host
of informatics technologies, new and previously unimaginable methods for collab-
oration have been minted. These methods are rapidly becoming the currency for
discovery, effectiveness, and innovation in clinical, research, and academic worlds.
The vast amount of data on the Internet is only one of two key factors in this revo-
lution. The other factor centers on a powerful set of next-generation Internet tools.
These tools facilitate the knowledge building process by enabling individual sets
of data across the web to be integrated and by enabling researchers to dynamically
participate in the process. These changes in how knowledge is built are influencing
all disciplines including the likes of literature and philosophy.
Neuropsychology is at a critical threshold. It faces entry into an altogether new
way of transacting its business. Yet, ironically, the identity that neuropsychology
has so carefully built over the decades seems to have locked in a consciousness that
prevents it from recognizing this change. How it crosses this threshold will have far
flung ramifications. It is imperative that neuropsychology perceives very clearly the
new informational platform that has grown all around it.
knowledge models: It was characterized by a large number of servers that held infor-
mation; users went to these servers and got the information. The sheer number of
servers and users and the resulting global flow of information made for an infor-
mation revolution. Yet the server (vendor) and client (receiver) model was in some
ways reminiscent of information dissemination during the pre-digital age.
In 2004, formal recognition was made of a unique set of practices that had begun
to evolve over the Internet. These practices altogether established ways of collec-
tive knowledge building in which large groups of users became the contributors
and co-developers. “Web 2.0” was the term coined to describe these practices, the
principles behind the practices, the specialized web tools involved, and the radical
shift to a model of networked communities.1 Web 2.0 captures a trend in open plat-
form Internet use that is geared to user-driven online networks and knowledge bases
that constantly change and grow. It is a flexible, user-driven approach to building a
product. “. . . [A] single monolithic approach, controlled by a single vendor, is no
longer a solution, it’s a problem” (O’Reilly 2005, p. 2). Wiki platforms are one of
many tools that enable collaborative knowledge building. The online encyclopedia
wikipedia.org is among the best known examples of a product derived with Web 2.0
principles. Web 2.0 in essence is about (1) harnessing interactive user networks that
stem from the Internet, to (b) collectively develop and customize ideas and products
and (c) refine these ideas or products as more users contribute to them. The result
is a knowledge-oriented participatory environment in which great amounts of data
are integrated and novel products are molded. Dynamic forms of collaboration and
engagement are hallmarks of Web 2.0, and the power of the phenomenon has been
referenced by numerous terms over the past few years, such as digital democracy,
collective intelligence, wisdom of groups, architecture of participation, and dynamic
content creation.
At the February 2009 meeting of the International Neuropsychological Society
in Atlanta, Georgia, Robert Bilder presented a paper titled “Neuropsychology 2.0:
Leveraging Genomics, Phenomics, and the World Wide Web”. Bilder described how
existing technologies (of the Web 2.0 variety) can be used to re-energize the field
in view of the limitations brought about by its outmoded neurobehavioral models
and “instrument inertia.” Bilder proposed an action plan involving web collabo-
rations to build cognitive ontologies, knowledge bases, and automated assessment
instruments.
It is inevitable that such developments will transform neuropsychology, mak-
ing for “Neuropsychology 2.0”. The field is destined to reach a critical point of
departure from its now maladapted tools and traditions. Informatics tools will enable
new modes of data gathering and knowledge building. With NI, neuropsychology
1 An authoritative commentator on the Internet, Tim O’ Reilly, is widely credited with having
coined the term “Web 2.0” in 2004. O’Reilly’s 2005 article, “What Is Web 2.0: Design Patterns
and Business Models for the Next Generation of Software” is considered a landmark refer-
ence on the topic (http://www.oreillynet.com). Numerous informative books on Web 2.0 have
since been written. See for example Jones (2008) and Vossen and Hagerman (2007). See also
http://en.wikipedia.org/wiki/Web_2.0
102 6 Concluding Remarks
“In addition to connecting our libraries and schools to the internet, we must also ensure that
our hospitals are connected to each other through the internet. That is why the economic
recovery plan I’m proposing will help modernize our health care system – and that won’t
just save jobs, it will save lives. We will makes sure that every doctor’s office and hospital in
this country is using cutting edge technology and electronic medical records so that we can
cut red tape, prevent medical mistakes, and help save billions of dollars each year” (Obama
2008; excerpt from a radio address delivered on Dec. 6).
The promise drew applause but also words of caution. Some specialists on the topic
responded in an open letter, highlighting among many points, the lack of interoper-
ability among current systems, and the need for some basic steps such as extending
high speed internet access to all practitioners.2 Undaunted by debates on the value
of health information networks or the difficulty of the task, about $20 billion dollars
have been allocated toward health information technology (HIT) through the Amer-
ican Recovery and Reinvestment Act of 2009. The initiative will be overseen by the
newly created Office of the National Coordinator of Health Information Technology
(see Blumenthal, 2009; www.recovery.gov). Discussions and debates around health
information technology and electronic medical records continue and have broadened
into the public domain.3
The issue is very pertinent to neuropsychology as neuropsychology considers
large-scale informatics systems for communication and data sharing. A national
medical informatics system that carries comprehensive health records could likely
hold neuropsychological data as well. Whether or not it does have a place for neu-
ropsychological data may depend on how quickly neuropsychologists ride into this
frontier of health information networks.
To be tuned to such considerations, the discussion on NI for neuropsychology
needs to be all encompassing. It has to consider the broader infrastructure and when
there are planned developments within that can potentially serve neuropsychology,
the opportunity should be promptly examined.
In the course of its historical development, neuropsychology has been very open to
new technology – from brass instruments to EEG, from microcomputers to fMRI.
The discipline has been very good at incorporating new technology into its existing
2 David Kibbe, a senior advisor to the American Academy of Family Physicians, and Bruce
Klepper, a healthcare market analyst, sent an open letter to Barack Obama in which they
outlined key problems and made suggestions regarding electronic health records in the USA.
(www.thehealthcareblog.com/the_health_care_blog/2008/12/where-should-fe.html)
3 For example, the National Public Radio program, All Things Considered, interviewed Dr. Ashish
Jha (Harvard School of Public Health) on March 25, 2009, on the topic on HIN implementa-
tion in USA: “Electronic Medical Record Change Not Easy”. (http://www.npr.org/templates/story/
story.php?storyId=102360638)
104 6 Concluding Remarks
structure and then adjusting its focus or developing subspecialties based on insights
gained through the technology.
As great an impact that various technologies have had on the field, none has
implied or required that the discipline of neuropsychology undergo an evaluation
of so many of its fundamental operating principles – relating to its constructs of
cognition, how it assesses these constructs, how its data are gathered and protected,
how inferences are drawn from the data, how models of behavior are built, and how
neuropsychologists communicate as a community. NI has fundamentally different
implications in this regard. NI is much more than just a new technology. It is about
a major change of paradigm on the treatment of neuroscience data at all levels. The
implication for neuropsychology, especially clinical neuropsychology, is that its tra-
ditions and tool sets are going to be widely challenged. From its clinical assessment
methods to its modes of training, clinical neuropsychology is at risk of severe incon-
gruence with a surrounding mass of fine-grained neuroscience data. NI is the great
unifying force for data in the brain sciences. It is undoubtedly a powerful, transfor-
mative force that has reoriented much of neuroscience. This is the brave new world
in which neuropsychology has to function. The effect of NI on neuropsychology
will not be ameliorating but instead greatly enlivening.
It is because of the rich history of neuropsychology and its long line of
remarkable pioneers that the discipline has reached such a high level of distinc-
tion. To continue this legacy, it must embark on entirely new trajectories set
forth by technologies of our time. It should do so with vision and firm embrace.
Neuropsychology in the 21st century will be defined by neuroinformatics.
References
Blumenthal, D. (2009). Stimulating the adoption of health information technology. The New
England Journal of Medicine, 360(15), 1477–1479.
Blumenthal, D. & Glaser, J.P. (2007). Information technology comes to medicine. The New
England Journal of Medicine, 356(24), 2527–2534.
Bower, J.M. & Beeman, D. (1998). The book of GENESIS: Exploring realistic models with the
GEneral NEural SImulation System (2nd ed.). New York: Springer-Verlag New York, Inc.
Bower, J.M., Beeman, D., & Hucka, M. (2003). The GENESIS simulation System. In M.A. Arbib
(Ed.) The Handbook of Brain Theory and Neural Networks (2nd ed.). Cambridge, MA: The
MIT Press
Broca, P. (1861). Remarques sur le siege de la faculte de langage articule, suivie d’une observation
d’aphemie (perte de la parole). Bulletins de la Societe Anatomigue, 6, 330–357.
Broca, P. (1865). Sur la faculte du langage articule. Bulletin de la Societe d’Anthropologie, 6,
337–393.
Burnett-Stuart, G., Halligan, P.W., & Marshall, J.C. (1991). A newtonian model of perceptual
distortion in visuo-spacial neglect. Neuroreport, 2, 255–257.
Burns, G. (2001a). Knowledge mechanics and the neuroscholar project: A new approach to neu-
roscientific theory. In M. Arbib & J. Grethe (Eds.) Computing the Brain, 1 (pp. 319–336). San
Diego, CA: Academic Press.
Burns, G.A. (2001b). Knowledge management of the neuroscientific literature: The data model
and underlying strategy of the NeuroScholar system. Philosophical transactions of the Royal
Society of London. Series B, Biological sciences, 356(1412), 1187–208.
Burns, G.A.P.C., Khan, A.M., Ghandeharizadeh, S., O’Neill, M.A., & Chen, Y. (2003). Tools
and approaches for the construction of knowledge models from the neuroscientific literature.
Neuroinformatics, 1(1), 81–109.
Bush, S., Naugle, R., & Johnson-Greene, D. (2002). Interface of information technology and
neuropsychology: Ethical issues and recommendations. Clinical Neuropsychology, 16(4),
536–547.
Buxton, R.B. (2001). Introduction to Functional Magnetic Resonance Imaging: Principles and
Techniques. Oxford: Cambridge University Press.
Cernich, A.N., Brennana, D.M., Barker, L.M., & Bleiberg, J. (2007). Sources of error in
computerized neuropsychological assessment. Archives of Clinical Neuropsychology, 22S,
S39–S48.
Chalupa, L.M. & Werner, J.S. (2004). The Visual Neurosciences. Cambridge, MA: The MIT Press.
Chomsky, N. (1957). Syntactic Structures.The Netherlands: Mouton & Co., N.V., Publishers, The
Hague.
Churchland, P. & Sejnowski, T.J. (1994).The Computational Brain (Computational Neuro-
science).Cambridge, MA: The MIT Press.
Chute, D.L. (2002). Neuropsychological technologies in rehabilitation. Journal of Head Trauma
Rehabilitation, 17(5), 369–377.
Cohen, M.S., Kosslyn, S.M., Breiter, H.C., DiGirolamo, G.J., Thompson, W.L., Anderson, A.K.,
Brookheimer, S.Y., Rosen, B.R., & Belliveau, J.W. (1996). Changes in cortical activity during
mental rotation. A mapping study using functional MRI. Brain, 119, 89–100.
Collins, F.S., Morgan, M., & Patrinos, A. (2003). The human genome project: Lessons from large
scale biology. Science, 300 (5617), 286–290.
Corina, D.P., Bellugi, U., & Reilly, J. (1999). Neuropsychological studies of linguistic and affective
facial expressions in deaf signers. Language and Speech, 42, 307–331.
Damasio, A. (1990). Synchronous activation in multiple cortical regions: A mechanism for recall.
Seminars in the Neurosciences, 2, 287–296.
Dayan, P. & Abbott, L.F. (2005). Theoretical neuroscience: Computational and Mathematical
Modeling of Neural Systems. Cambridge, MA: The MIT Press.
Deacon, T.W. (1997). The Symbolic Species: The Co-evolution of Language and the Brain. New
York: W.W. Norton & Company, Inc.
108 References
Dell, G. S., Burger, L.K., & Svec, W.R. (1997). Language production and serial order: A functional
analysis and a model. Psychological Review, 104(1), 123–147.
Dell, G.S., Martin, N., & Schwartz, M.F. (2007). A case-series test of the interactive two-step
model of lexical access: Predicting word repetition from picture naming. Journal of Memory
and Language, 56(4), 490–520.
Deutsch, C.K., Hobbs, K., Price, S.F., & Gordon-Vaugh, K. (2000). Skewing of the brain midline
in schizophrenia. Neuroreport, 11(18), 3985–3988.
Deutsch, C.K. & Joseph, R.M. (2003) Brief report: Cognitive correlates of enlarged head cir-
cumference in children with autism. Journal of Autism and Developmental Disorders, 33(2),
209–215.
Donnelly, N., Guest, R., Fairhurst, M., Potter, J., Deighton, A., & Patel, M. (1999). Developing
algorithms to enhance the sensitivity of cancellation tests of visuospatial neglect. Behavior
Research Methods, Instruments, & Computers, 31(4), 668–673.
Downing, P.E., Jiang, Y., Shuman, M., & Kanwisher, N. (2001). A cortical area selective for visual
processing of the human body. Science, 293(5539), 2470–2473.
Dreuw, P., Neidle, C., Athitsos, V., Sclaroff, S. & Ney, H. (2008, May). Benchmark databases for
video-based automatic sign language recognition. Paper presented at the Sixth International
Conference on Language Resources and Evaluation in Marrakech, Morocco.
Duhamel, J.R., Colby, C.L., & Goldberg, M.E. (1991). Congruent representations of visual and
somatosensory space in single neurons of monkey ventral intra-parietal cortex (area VIP). In J.
Paillard (Ed.) Brain and Space (pp. 223–236). New York: Oxford University Press.
Editorial (2000, September). “A debate over fMRI data sharing.” Nature Neuroscience, p. 845.
Elsmore, T. & Reeves, D. (2004). ANAM readiness evaluation system (ARES): User’s Guide.
Activity Research Services.
Elsmore, T.F., Reeves, D.L., & Reeves, A.N. (2007). The ARES! R test system for palm OS
handheld computers. Archives of Clinical Neuropsychology, 22S, S135–S144.
Emmorey, K. (2002). Language, Cognition, and the Brain. New Jersey: Lawrence Erlbaum
Associates.
Emmorey, K., Mehta, S., & Grabowski, T.J. (2007). The neural correlates of sign versus word
production. Neuroimage, 36(1), 202–208.
Emmorey, K. & Reilly, J.S. (1995). Theoretical issues relating language, gesture, and space: An
overview. In K. Emmorey & J.S. Reilly (Eds.) Language, Gesture and Space (pp. 1–19).
Hillsdale, New Jersey: Lawrence Erlbaum Associates, Publishers.
Epstein, R. & Kanwisher, N. (1998). A cortical representation of the local visual environment.
Nature, 9(392), 598–601.
Evans, A.C. & The Brain Development Cooperative Group (2004). NIH MRI study of normal
brain development. Presented at Pediatric Functional Neuroimaging: A Trans-NIH Workshop,
Bethesda, Maryland, May, 25.
Evans, A.C. & the Brain Development Cooperative Group (2006). The NIH MRI study of normal
brain development. NeuroImage, 30(1), 184–202.
Farah, M.J. (2000). The cognitive Neuroscience of Vision. Oxford: Blackwell Publishers.
Farkas, L.G. & Deutsch, C.K. (1996). Anthropometric determination of craniofacial morphology.
American Journal of Medical Genetics, 2(65), 1–4.
Fellbaum, C. (1998). WordNet: An Electronic Lexical Database. Cambridge, MA: The MIT Press.
Felleman, D.J. & Van Essen, D.C. (1991). Distributed hierarchical processing in the primate
cerebral cortex. Cerebral Cortex, 1(1), 1–47.
Flint, J. & Munafò, M.R. (2007). The endophenotype concept in psychiatric genetics. Psychologi-
cal Medicine, 37, 163–180.
Foygel, D., & Dell, G.S. (2000). Models of impaired lexical access in speech production. Journal
of Memory and Language, 43(2), 182–216.
Friedl, K.E., Grate, S.J., Proctor, S.P., Ness, J.W., Lukey, B.J., & Kane, R.L. (2007). Army research
needs for automated neuropsychological tests: Monitoring soldier health and performance
status. Archives of Clinical Neuropsychology, 22S, S7–S14.
References 109
Freimer, N. & Sabatti, C. (2003). The human phenome project. Nature Genetics, 34, 15–21.
Funahashi, S., Bruce, C.J., & Goldman-Rakic, P.S. (1989). Mnemonic coding of visual space in
the monkey! s dorsolateral prefrontal cortex. Journal of Neurophysiology, 61(2), 331–349.
Geschwind, N. (1965). Disconnexion syndromes in animals and man. Brain, 88, 237–294 and
585–644.
Goldstein, G. (1997). The clinical utility of standardized or flexible battery approaches to neuropsy-
chological assessment. In G. Goldstein & T.M. Incagnoli (Eds.)Contemporary Approaches to
Neuropsychological Assessment (pp.67–92). New York: Plenum Press.
Goldstein, G. (1998). Introduction to neuropsychological assessment. In G. Goldstein, P.D.
Nussbaum & S. R. Beers (Eds.) Neuropsychology (pp. 1–8). New York: Plenum Press.
Goldstein, G., Nussbaum, P.D., & Beers, S.R. (Eds.) (1998). Neuropsychology. New York: Plenum
Press.
Goodglass, H. (1993). Understanding Aphasia. San Diego, California: Academic Press, Inc.
Goodglass, H. & Kaplan, E. (1979). Assessment of cognitive deficit in the brain-injured patient.
In M.S. Gazzaniga (Ed.) Handbook of Behavioral Neurobiology: Volume 2, Neuropsychology
(pp. 3–21). New York: Plenum Press.
Gorin, F., Hogarth, M., & Gertz, M. (2001). The challenges and rewards of integrating diverse
neuroscience information. The Neuroscientist, 7(1), 18–27.
Gottesman, I.I. & Gould, T.D. (2003). The endophenotype concept in psychiatry: Etymology and
strategic intentions. American Journal of Psychiatry, 160(4), 636–645.
Grill-Spector, K. & Malach, R. (2004). The human visual cortex. Annual Review of Neuroscience,
27, 649–677.
Gruber, T.R. (1995). Toward principles for the design of ontologies used for knowledge sharing.
International Journal of Human-Computer Studies, 43, 907–928.
Guenther, F.H. (1994). A neural network model of speech acquisition and motor equivalent speech
production. Biological Cybernetics, 72, 43–53.
Guenther, F.H. (1995). Speech sound acquisition, coarticulation, and rate effects in a neural
network model of speech production. Psychological Review, 102, 594–621.
Guenther, F.H. (2006). Cortical interactions underlying the production of speech sounds. Journal
of Communication Disorders, 39, 350–365.
Guenther, F.H., Ghosh, S.S., & Tourville, J.A. (2006). Neural modeling and imaging of the cortical
interactions underlying syllable production. Brain and Language, 96(3), 280–301.
Guenther, F.H., Hampson, M., & Johnson, D. (1998). A theoretical investigation of ref-
erence frames for the planning of speech movements. Psychological Review, 105(4),
611–33.
Gur, R.C., Ragland, J.D., Moberg, P.J., Turner, T.H., Bilker, W.B., Kohler, C., Siegel, S.J., & Gur,
R.E. (2001). Computerized neurocognitive scanning: I. Methodology and validation in healthy
people. Neuropsychopharmacology, 25, 766–776.
Hall, K.M. (1997). Establishing a national traumatic brain injury information system based upon a
unified data set. Archives of Physical Medicine and Rehabilitation, 78, S5–S11
Halligan, P.W. & Marshall, J.C. (1991). Figural modulation of visuo-spatial neglect: A case study.
Neuropsychologia, 29(7), 619–628.
Halligan, P.W. & Marshall, J.C. (1995). Lateral and radial neglect as a function of spatial position:
A case study. Neuropsychologia, 33(12), 1697–702.
Halligan, P.W. & Marshall, J.C. (1994). Figural perception and parsing in visuo-spatial neglect.
Neuroreport, 5, 537–539.
Halligan, P.W. & Marshall, J.C. (1998). Visuospatial neglect: The ultimate deconstruction. Brain
and Cognition, 37(3), 419–438.
Heilman, K.M., Watson, R.T., Valenstein, E. (1993). Neglect and related disorders. In K. Heil-
man, M. Kenneth, & E. Valenstein (Eds.) Clinical Neuropsychology (3rd ed.) (pp. 279–336).
New York: Oxford University Press.
Heuttel, S.A., Song, A.W., & McCarthy, G. (2004). Functional Magnetic Resonance Imaging.
Sunderland, MA: Sinauer Associates.
110 References
Hickok, G., Bellugi, U., & Klima, E.S. (1996). The neurobiology of sign language and its
implications for the neural basis of language. Nature, 381(6584), 699–702.
Hines, M. (1994). The NEURON simulation program. In J. Skrzypek (Ed.) Neural Network
Simulation Environments (pp. 147–163). Norwell, MA: Kluwer.
Hines, M.L. (1998). The neurosimulator NEURON. In C. Koch & I. Segev (Eds.) Methods in
Neuronal Modeling (pp. 129–136). Cambridge, MA: The MIT Press.
Hines, M.L. & Carnevale, N.T. (2001). NEURON: A tool for neuroscientists. The Neuroscientist,
7(2), 123–135.
Hines, M.L. & Carnevale, N.T. (2003). The NEURON simulation environment. In M.A. Arbib
(Ed.) The Handbook of Brain Theory and Neural Networks (2nd ed.) (pp. 769–773). Cam-
bridge, MA: The MIT Press.
Hobbs, A. (1999). Mapping variation in brain structure and function: Implications for rehabilita-
tion. The Journal of Head Trauma Rehabilitation, 14(6), 616–621.
Hubel, D.H. & Wiesel, T.N. (1962). Receptive fields, binocular interaction and functional
architecture in the cat! s visual cortex. Journal of Physiology, 160(1), 106–154.
Hubel, D.W. and Wiesel, T.N. (1978) Anatomical demonstration of orientation columns in
macaque monkey. Journal of Comparative Neurology, 177, 361–379.
Huang, H.C. & Wang, T.Y. (2005). Toward a graphical analysis tool for computer-assisted assess-
ment of visual search patterns. Paper presented at the Fifth IEEE International Conference on
Advanced Learning Technologies, Kaohsiung, Taiwan.
Huang H.C. & Wang T.Y. (2008). Visualized representation of visual search patterns for a
visuospatial attention test. Behavioral Research Methods, 40(2), 383–90.
Huerta, M.F. & Koslow, S.H. (1996). Neuroinformatics: Opportunities across disciplinary and
national borders. Neuroimage, 4, S4–S6.
Huerta, M.F., Koslow, S.H., & Leshner, A.I. (1993). The human brain project: An international
resource. Trends in Neuroscience, 16(11), 436–438.
Jagaroo, V. (1998). Allocentric and representational subprocesses in visuospatial function: Four
focal lesion groups. Unpublished doctoral dissertation, Boston University School of Medicine,
Boston.
Jagaroo, V. (1999). Towards an analytic framework for the visuospatial domain: Spatial refer-
ence frames, cognitive operations and neural systems.The Journal of Clinical and Experimental
Neuropsychology 21(1): 134–146.
Jagaroo, V. (2002). Dynamic computational visual field matrices: A computerized mapping system
for the analysis of visual perception, spatial processing and featural recognition. In C.H. Dagli,
A.L. Buczak, J. Ghosh, M.J. Embrechts, O. Ersoy, & S.W. Kercel (Eds.) Smart Engineering
System Design, Neural Networks and Fuzzy Logic,vol. 12 (pp. 81–90). New York: American
Society of Mechanical Engineers Press.
Jagaroo, V. (2004). Mental rotation and the parietal question in functional neuroimaging: A
discussion of two views. European Journal of Cognitive Psychology, 16(5), 717–728.
Jagaroo, V. & Wilkinson, K. (2008). Further considerations of visual cognitive neuroscience in
aided AAC: The potential role of motion perception systems in maximizing design display.
Augmentative and Alternative Communication, 24(1), 29–42.
Jensen M. (2007). The new metrics of scholarly authority. The chronicle of higher education (The
chronicle review), 53(41), B6 (June 15).
Jha, A.K., Ferris, T.G., Donelan, K., DesRoches, C., Shields, A., Rosenbaum, S. & Blumenthal, D.
(2006). How common are electronic health records in the United States? A summary of the
evidence. Health Affairs, 25, 496–507
Jha, A.K., DesRoches, C., Campbell, E.G., Donelan, K., Rao, S.R., Ferris, T.G., Shields, A.,
Rosenbaum, S., & Blumenthal, D. (2009). Use of electronic health records in U.S. hospitals.
The New England Journal of Medicine, 360(16), 1628–1638.
Johnston, M.V. (1997). Toward a national information system for traumatic brain injury: Foreword
and overview. Archives of Physical Medicine and Rehabilitation, 78(8), S1–S4.
References 111
Jones, B.L. (2008). Web 2.0 heroes: Interviews with 20 web 2.0 influencers. Indianapolis, IN: Wiley
Publishing, Inc.
Jonides J., Smith E.E., Koeppe R.A., Awh E., Minoshima S., & Mintun M.A. (1993). Spatial
working memory in humans as revealed by PET. Nature,363(6430), 623–625.
Just, M.A. & Carpenter, P.A. (1985). Cognitive coordinate systems: Accounts of mental rotation
and individual differences in spatial ability. Psychological Review, 92(2), 137–172.
Kalaska, J.F. (1991). Parietal cortex area 5: A neuronal representation of movement kinematics for
kinaesthetic perception and movement control? In J. Paillard (Ed.) Brain and Space (pp. 133–
146). New York: Oxford University Press.
Kane (2007). Editorial. Archives of Clinical Neuropsychology, 22S: S3–S5.
Kane, R.L. & Kay, G.G. (1992). Computerized assessment in neuropsychology: A review of tests
and test batteries. Neuropsychology Review, 3(1), 1–117.
Kane, R.L. & Kay, G.G. (1997). Computer applications in neuropsychological assessment. In
G. Goldstein & T.M. Incagnoli (Eds.) Contemporary Approaches to Neuropsychological
Assessment: Critical Issues in Neuropsychology (pp. 359–392). New York: Plenum Press.
Kane, R.L. & Reeves, D.L. (1997). Computerized test batteries. In A.M. Horton, D. Wedding & J.
Webster (Eds.) The Neuropsychology Handbook (pp. 423–467). New York (NY): Springer.
Kanwisher, N. (2000). Domain specificity in face perception. Nature Neuroscience, 3(8), 759–763.
Kanwisher, N., McDermott, J., & Chun, M. (1997). The fusiform face area: A module in human
extrastriate cortex specialized for face perception. Journal of Neuroscience, 17(11), 4302–4311.
Kaplan, E. (1988). A process approach to neuropsychological assessment. In T. Boll & B.K. Bryant
(Eds.) Clinical Neuropsychology and Brain Functioning: Research, Measurement, and Practice
(pp. 129–167). Washington, DC: American Psychological Association.
Klerman, G.L., Vaillant, G.E., Spitzer, R.L., & Michels, R. (1984). A debate on DSM-III. American
Journal of Psychiatry, 141(4), 539–553.
Klima, E.S. & Bellugi, U. (1979). The Signs of Language. Cambridge, MA: Harvard University
Press.
Kolb, B. & Whishaw, I.Q. (2003). Fundamentals of Human Neuropsychology. New York: Worth
Publishers.
Koslow, S.H. (2000). Should the neuroscience community make a paradigm shift to sharing
primary data? Nature Neuroscience, 3(9), 863–865.
Koslow, S.H. (2002). Sharing primary data: A threat of asset to discovery? Nature Reviews:
Neuroscience, 3, 311–313.
Kosslyn, S.M. (1987). Seeing and imagining in the cerebral hemispheres: A computational
approach. Psychological Review, 94(2), 148–175.
Kosslyn, S.M., Chabris, C.F., Marsolek, C.J., & Koenig, O. (1992). Categorical versus coordinate
spatial relations: Computational analysis and computer simulations. Journal of Experimental
Psychology: Human Perception and Performance, 18(2), 562–577.
Kötter, R. (2001). Neuroscience database: Tools for exploring brain structure-function relation-
ships. Philosophical Transactions of the Royal Society of London, 356(1412), 1111–1120.
Kremen, W.S., Eisen , S.A., Tsuang, M.T., & Lyons, M.J. (2007). Is the wisconsin card sorting
test a useful neurocognitive endophenotype? American Journal of Medical Genetics. Part B,
Neuropsychiatric Genetics, 5(144B), 403–406.
Lainhart, J.E., Bigler, E.D., Bocian, M., Coon, H., Dinh, E., Dawson, G., Deutsch, C.K., Dunn, M.,
Estes, A., Tager-Flusberg, H., Folstein, S., Hepburn, S., Hyman, S., McMahon, W., Minshew,
N., Munson, J., Osann, K., Ozonoff, S., Rodier, P., Rogers, S., Sigman, M., Spence, M.A.,
Stodgell, C.J., & Volkmar, F. (2006). Head circumference and height in autism: A study by the
collaborative program of excellence in autism. American Journal of Medical Genetics 1A(140),
21:2257–2274.
Landauer, T.K. & Dumais, S.T. (1997). A solution to plato’s problem: The latent semantic analysis
theory of acquisition, induction, and representation of knowledge. Psychological Review, 104,
211–240.
112 References
Landauer, T.K., Foltz, P.W., & Laham, D. (1998). An introduction to latent semantic analysis.
Discourse Processes, 25, 259–284.
Landauer, T.K., McNamara, D.S., Dennis, S., & Kintsch, W. (2007). Handbook of Latent Semantic
Analysis. Mahwah, NJ: Lawrence Erlbaum Associates Publishers.
Lavrač, N. (1999). Selected techniques for data mining in medicine. Artificial Intelligence in
Medicine, 16(1), 3–23.
Lee, A. Thompson, P. & Toga, A. (2000). Databasing the brain. Nature, 406, 822–825.
Levin, B.L. (1990). Spatial cognition in parkinson disease. Alzheimer’s Disease and Associated
Disorders, 4(3), 161–170.
Lezak, M.D., Howieson, D.B. & Loring, D.W. (2004). Neuropsychological assessment 4th edition.
New York: Oxford University Press.
Lezak, M.D. (1995). Neuropsychological assessment. 3rd edition. New York: Oxford University
Press.
Lichtheim, L. (1885). On aphasia. Brain, 7, 433–484.
Lieberman, P. (2002). Human Language and our Reptilian Brain. Cambridge, MA: Harvard
University Press.
Luciana, M. (2003). Practitioner review: Computerized assessment of neuropsychological func-
tion in children: Clinical and research applications of the Cambridge neuropsychological
testing automated battery (CANTAB).Journal of Child Psychology and Psychiatry, 44(5),
649–663.
MacLaughlin, D., Neidle, C., & Greenfield, D. (2000). SignStreamTM User’s Guide, Version 2.0.
Boston, MA: American Sign Language Linguistic Research Project No. 9, Boston University.
Mahner, M. & Kary, M. (1997). What exactly are genomes, genotypes and phenotypes? And what
about phenomes? Journal of Theoretical Biology, 7(186), 55–63.
Marshall, J.C. (1986). The description and interpretation of aphasic language disorder. Neuropsy-
chologia, 24(1), 5–24.
Matthysse, S., Levy, D.L., Kinney, D., Deutsch, C., Lajonchere, C., Yurgelun-Todd, D., Woods,
B., & Holzman, P.S. (1992). Gene expression in mental illness: A navigation chart to future
progress. Journal of Psychiatric Research, 26(4), 461–473.
McClelland, J.D., Rumelhart, & the PDP Research Group (1986). Parallel Distributed Processing:
Explorations in the Microstructure of Cognition, Vol. 1: Foundations. Cambridge, MA: The
MIT Press.
McCullough S., Emmorey K., & Sereno M. (2005). Neural organization for recognition of gram-
matical and emotional facial expressions in deaf ASL signers and hearing nonsigners. Brain
Research: Cognitive Brain Research, 22(2), 193–203.
McNeill, D. & Pedelty, L.L. (1995). Right brain and gesture. In K. Emmorey & J. Reilly (Eds.),
Language, Gesture and Space (pp. 63–87). New Jersey: Lawrence Erlbaum Associates.
Meadche, A. (2002). Ontology learning for the semantic web. Boston: Kluwer Academic
Publishers.
Mesulam, M.M. (1990). Large-scale neurocognitive networks and distributed processing for
attention, language, and memory. Annals of Neurology, 28, 597–613.
Miller, C. (2001). Some reflections on the need for a common sign notation. Sign Language &
Statistics, 4, 11–28.
National Academy Press (1999). A Question of Balance: Privacy Rights and the Public Interest in
Scientific and Technical Databases. National Research Council, Washington, D.C.
Neidle, C. (2001). SignStreamTM : A database tool for research on visual-gestural language. Sign
Language & Linguistics, 4, 203–214.
Neidle, C. (2002b). SignStreamTM Annotation: Conventions used for the American sign language
linguistic research project. Boston, MA: American Sign Language Linguistic Research Project
Report No. 11, Boston University.
Neidle, C. (2002a). SignStreamTM : A Database Tool for Research on Visual-Gestural Language.
Journal of Sign Language and Linguistics, 4, 203–214.
References 113
Neidle, C., Sclaroff, S., & Athitsos, V. (2001). SignStreamTM : A tool for linguistic and com-
puter vision research on visual-gestural language data. Behavior Research Methods and
Instrumentation, 33(3), 311–320.
Newcombe, F. and Ratcliff, G. (1990) Disorders of visuospatial analysis. In F. Boller & J. Grafman
(Eds.) Handbook of Neuropsychology, Vol. 2. (pp. 333–356). Amsterdam: Elsevier.
Nichelli, P., Rinaldi, M., & Cubelli, R. (1989). Selective spatial attention and length representation
in normal subjects and in patients with unilateral spatial neglect. Brain and Cognition, 9(1),
57–70.
O’Reilly, R.C., Munakata, Y., & McClelland, J.L. (2000). Computational Explorations in Cog-
nitive Neuroscience: Understanding the Mind by Simulating the Brain. Cambridge, MA: The
MIT Press.
Organization for Economic Co-operation and Development (2002). Report on Neuroinformatics
by The Global Science Forum Neuroinformatics Working Group (June).
Pechura, C.M. & Martin, J.B. (Eds.) (1991). Mapping the brain and its functions: Integrating
enabling technologies into neuroscience. Institute of Medicine (U.S.). Committee on a National
Neural Circuitry Database. Washington, D.C.: National Academy Press.
Perrett, D.I., Hietanen, J.K., Oram, M.W., Benson, P.J., & Rolls, E.T. (1992). Organization and
functions of cells responsive to faces in the temporal cortex. Philosophical Transactions of the
Royal Society of Biological Sciences, 335(1273), 23–30.
Petersen, S.E., Fox, P.T., Posner, M.I., Mintum, M., & Raichle, M.E. (1988). Positron emis-
sion tomographic studies of the cortical anatomy of single-word processing. Nature, 331,
585–589.
Petitto, L.A. & Bellugi, U. (1988). Spatial cognition and brain organization: Clues from the
acquisition of a language in space. In J. Stiles-Davis, M. Kritchevsky, & U. Bellugi (Eds.) Spa-
tial Cognition: Brain Bases and Development (pp.299–327). New Jersey: Lawrence Erlbaum
Associates.
Piatetsky-Shapiro, G., & Frawley, W.J. (Eds.) (1991). Knowledge Discovery in Databases (AAAI
press copublication). Cambridge, MA: The MIT Press.
Pinker, S. (1994). The Language Instinct: How the Mind Creates Language. New York: William
Morrow and Company.
Pizzuto, E., & Pietrandrea, P. (2001). The notation of signed texts: Open questions and indications
for further research. Sign Language & Linguistics, 4, 29–45.
Potter, J., Deighton, T., Patel, M., Fairhurst M, Guest R, & Donnelly N. (2000). Computer
recording of standard tests of visual neglect in stroke patients. Clinical Rehabilitation, 14,
441–446.
Poizner, H., Kaplan, E., Bellugi, U., & Padden, C.A. (1984). Visual-spatial processing in deaf
brain-damaged signers. Brain and Cognition, 3(3), 281–306.
Prather, J.C., Loback, D.F., Goodwin, L.K., Hales, J.W., Hage, M., & Hammond, W.E. (1997).
Medical data mining: Knowledge discovery in a clinical data warehouse. Proceedings:
A Conference of the American Medical Informatics Association. AMIA Fall Symposium,
101–105.
Reeves, D. L., Winter, K. P., Bleiberg, J., & Kane, R. L. (2007). ANAM! R Genogram: Historical
perspectives, description, and current endeavors. Archives of Clinical Neuropsychology, 22S,
S15–S37.
Reeves, D., Elsmore, T., Wiederhold, M.D., Wood, D., Center, C., Spira, J., & Wiederhold,
B.K. (2007). Brain checkers: Handheld computerized neuropsychological assessment in a
virtual reality treatment protocol for combat PTSD. Annual Review of CyberTherapy and
Telemedicine: Transforming Healthcare Through Technology, 5: 151–156.
Reitan, R.M. (1964). Psychological deficits resulting from cerebral lesions in man. In J.M. War-
ren & K. Akert (Ed.), The Frontal Granular Cortex and Behavior (pp. 295–312). New York:
McGraw-Hill Book Company.
Richter, W., Ugurbil, K., Georgopoulos, A., Kim, S-G. (1991) Time-resolved fMRI of mental
rotation. Neuroreport, 8(17), 3697–3702.
Roland, P.E. & Friberg, L. (1985). Localization of cortical areas activated by thinking. Journal of
Neurophysiology, 53(5), 1219–1243.
114 References
Rowe, P.M. (1995). Neural networks: A bridge between neuroscience and psychology. Molecular
Medicine Today 1(4), 168–173.
Sabb, F.W., Bearden, C.E., Glahn, D.C., Parker, D.S., Freimer, N., & Bilder, R.M. (2008).
A collaborative knowledge base for cognitive phenomics. Molecular Psychiatry, 13,
350–360.
Sadler, T.W. (2004). Langman’s Medical Embryology, 9th Ed. Philadelphia, PA: Lippincott,
Williams & Wilkins.
Sadler, J.Z., Hulgus, Y.F. & Agich, G.J. (1994). On values in recent American psychiatric
classification. Journal of Medical Philosophy 19(3), 261–277.
Schatz, P., Browndyke, J. (2002). Applications of computer-based neuropsychological assessment.
The Journal of Head Trauma Rehabilitation, 17(5), 395–410.
Schlegel, R.E., & Gilliland, K. (2007). Development and quality assurance of computer-based
assessment batteries. Archives of Clinical Neuropsychology, 22S, S49–S61.
Schultheis, M.T., Himelstein, J., & Rizzo, A.A. (2002). Virtual reality and neuropsychology:
Upgrading the current tools. The Journal of Head Trauma Rehabilitation, 17(5), 378–394.
Sergent, J. (1982). The cerebral balance of power: Confrontation or cooperation? Journal of
Experimental Psychology: Human Perception and Performance, 8(2), 253–272.
Sergent, J. (1991). Judgments of relative position and distance on representations of spatial
relations. Journal of Experimental Psychology, Human Perception and Performance, 17(3),
762–780.
Shepherd, G.M., Mirsky, J.S., Healy, M.D. et al. (1998). The human brain project: Neuroinformat-
ics tools for integrating, searching and modeling multidisciplinary neuroscience data. Trends in
Neuroscience, 21(11), 460–468.
Small, M., Cowey, A., & Ellis, S. (1994). How lateralised is visuospatial neglect? Neuropsycholo-
gia, 32(4), 449–464.
Smith, E.E. & Jonides, J. (1998). Neuroimaging analyses of human working memory. Proceedings
of the National Academy of Science, 95, 12061–12068.
Spitzer, R.L., Endicott, J., & Robins, E. (1975). Clinical criteria for psychiatric diagnosis and
DSM-III.American Journal of Psychiatry, 132(11), 1187–1192.
Spitzer, R.L., Williams, J.B., & Skodol, A.E. (1980). DSM-III: The major achievements and an
overview. American Journal of Psychiatry, 137(2), 151–164.
Spreen, O. & Strauss, E. (Eds.). (1998). A Compendium of Neuropsychological Tests: Administra-
tion, norms, and commentary. New York: Oxford University Press.
Staab, S. & Studer, R. (2009). Handbook on Ontologies. New York: Springer.
Stein, J.F. (1991). Space and the parietal association areas. In J. Paillard (Ed.) Brain and Space
(pp. 185–222). New York: Oxford University Press.
Sternberg, S. (1966). High-speed scanning in human memory. Science, 153(736), 652–654.
Strauss, E., Sherman, M.S., & Spreen, O. (2006). A Compendium of Neuropsychological Tests:
Administration, Norms, and Commentary. New York: Oxford University Press.
Talairach, J. & Tournoux, P. (1993). Referentially Oriented Cerebral MRI Anatomy: An Atlas of
Stereotaxic Anatomical Correlations for Gray and White Matter. New York: Thieme Medical
Publishers.
Talairach, J. & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain: 3-dimensional
proportional system – an approach to cerebral imaging. New York: Thieme Medical Publishers.
Tarr, M.J. & Gauthier, I. (2000). FFA: A flexible fusiform area for subordinate-level visual
processing automatized by expertise. Nature Neuroscience, 3(8), 764–769.
Thorne, D.R. (2006). Throughput: A simple performance index with desirable characteristics.
Behavior Research Methods, 38(4), 569–573.
Thorne, D.R., Genser, S.G., Sing, H.C., & Hegge, F.W. (1985). The Walter Reed performance
assessment battery. Neurobehavioral Toxicology & Teratology, 7, 415–418.
Treisman, A.M. (1982). Perceptual grouping and attention in visual search for features and
for objects. Journal of Experimental Psychology: Human Perception and Performance, 8(2),
194–214.
References 115
Treisman, A.M. & Gelade, G. (1980). A Feature Integration Theory of Attention. Cognitive
Psychology, 12, 97–136.
Ungerleider, L.G. & Mishkin, M. (1982). Two Cortical Visual Systems. In D.J. Ingle, M.A.
Goodale, & R.J.W. Mansfield (Eds.) Analysis of Visual Behavior (pp. 549–586). Cambridge:
The MIT Press.
Valler, G., Guariglia, C., Nico, D., & Bisiach, E. (1995). Spatial hemineglect in back space. Brain,
118, 467–472.
Van Essen, D.C., Felleman, D.J., DeYoe, E.A., Olavarria, J., & Knierim, J. (1990). Modular and
Hierarchical Organization of Extrastriate Visual Cortex in the Macaque Monkey. Cold Spring
Harbor Symposia on Quantitative Biology, Vol. LV: 679–696. Cold Spring Harbor, NY: Cold
Spring Harbor Laboratory Press.
Van Horn, J.D. & Gazzaniga, M.S. (2002). Databasing fMRI studies–towards a ‘discovery science’
of brain function. Nature Reviews, Neuroscience, 3, 314–318.
Vossen, G. & Hagemann, S. (2007) Unleashing web 2.0: From Concepts to Creativity. Burlington,
MA: Morgan Kaufmann Publishers
Walsh, K. (1987). Neuropsychology: A Clinical Approach. New York: Churchill Livingstone.
Wang, T.Y., Huang, H.C., & Huang, H.S. (2006). Design and implementation of cancellation tasks
for visual search strategies and visual attention in school children. Computers & Education, 47,
1–16.
Wang, Y. (2002). On Cognitive Informatics (Keynote Speech). Proceedings 1st IEEE International
Conference on Cognitive Informatics (ICCI ’02). Calgary, Canada: IEEE CS Press, pp. 34–42.
Wang, Y. (2003a). Cognitive informatics: A new transdisciplinary research field. Brain and Mind:
A Transdisciplinary Journal of Neuroscience and Neurophilosophy, 4(2), 115–127.
Wang, Y. (2003b). On cognitive informatics. Brain and Mind: A Transdisciplinary Journal of
Neuroscience and Neurophilosophy, 4(2), 151–167.
Wang, Y. (2006). Cognitive informatics towards the future generation computers that think and feel.
In Proceedings of the 5th IEEE International Conference on Cognitive Informatics (ICCI’06)
(pp. 3–7). Beijing, China: IEEE CS Press.
Wang, Y. (2007a). Cognitive informatics: Exploring the theoretical foundations for natural intelli-
gence, neural informatics, autonomic computing, and agent systems. International Journal of
Cognitive Informatics and Natural Intelligence, 1(1), i–x.
Wang, Y. (2007b). The theoretical framework of cognitive informatics. International Journal of
Cognitive Informatics and Natural Intelligence, 1(1), 1–27.
Wernicke, C. (1874). Der Aphasische Symptomencomplex. Breslau: Cohn & Weigert.
Williams, J.B. & Spitzer, R.L. (1982). Research diagnostic criteria and DSM-III: An annotated
comparison. Archives of General Psychiatry, 39(11), 1283–1289.
Wilson, S.L., & McMillan, T.M. (1992). Computer-based assessment in neuropsychology. In J.
R. Crawford, D.M. Parker, & W. W. McKinlay (Eds.) A Handbook of Neuropsychological
Assessment (pp. 413–431). Hove, UK: Lawrence Erlbaum Associates,
Wong, S.T.C. & Koslow, S.H. (2001). Human brain program research progress in bioinfor-
matics/neuroinformatics. Journal of the American Medical Informatics Association, 8(5),
510–511.
Wong, S.T.C., Hoo, K.S., Cao, X., Tjandra, D., Fu, J.C., & Dillon, W.P. (2004). A neuroinformat-
ics database system for disease-oriented neuroimaging research. Academic Radiology, 11(3),
345–358.
Wong, S.T.C., Hoo, K.S., Knowlton, R.C., Laxer, K.D., Cao, X., Hawkins, R.A., et al. (2002).
Design and applications of a multimodality image data warehouse framework. Journal of
American Medical Informatics Association, 9(3), 239–254.
Wong, S.T.C. & Huang, H.K. (1996). Design methods and architectural issues of integrated
medical image data base systems. Computerized Medical Imaging and Graphics, 20(4),
285–299.
Young, A.W., Hellawell, D.J., & Welch, J. (1992). Neglect and visual recognition. Brain, 115 Pt 1,
51–71.
116 References
Zhang, S., Zhang, C., & Wu, X. (2004). Knowledge Discovery in Multiple Databases. New York:
Springer.
Zobel, A. & Maier, W. (2004). Endophänotypen-ein neues Konzept zur biologischen character-
isierung psychischer störungen. Der Nervenarzt, 75(3), 205–214.
Zupan, B., Lavrač, N., & Keravnou, E. (1999). Data mining techniques and applications in
medicine. Artificial Intelligence in Medicine, 16, 1–2.
American Recovery and Reinvestment Act (2009). http://www.recovery.gov. Cited 30 Mar 2009
Athinoula A. Martinos Center for Biomedical Imaging. (2009). FreeSurfer. http://www.nmr.mgh.
harvard.edu. Cited 15 May 2009.
Barras, C. (2008). Brain implant helps stroke victim speak again. In: New Scientist (Tech). Brain
implant helps stroke victim speak again. http://www.newscientist.com/article/dn14277-brain-
implant-helps-stroke-victim-speak-again.html. Cited 11 Feb 2009.
Biomedical Informatics Research Network. (2005). http://www.nbirn.net. Cited 22 Nov 2005.
Boston University American Sign Language Linguistic Research Project. (2008). Database Access
Interface. http://ling.bu.edu/asllrpdata/queryPages/) Cited 26 Sept 2008.
Boston University American Sign Language Linguistic Research Project. (2008). SignStream.
http://www.bu.edu/asllrp/SignStream/. Cited 26 Sept 2008.
Boston University Department of Cognitive & Neural Systems, CNS Speech Lab. (2008). The
DIVA Model. http://speechlab.bu.edu/diva.php. Cited 9 Jul 2008.
Computational Systems Neuroscience Group. (2005). CoCoMac (Collations of Connectivity data
on the Macaque brain). http://cocomac.org/home.asp. Cited 30 Nov 2005
Deutsch, C. (2008). Collaborative biological research in schizophrenia: Dysmorphology in
schizophrenia. http://www.umassmed.edu/shriver/research/psychological/projects/cbrsds.aspx.
Cited 8 May 2008.
Duke University. (2005). NEURON for computer simulations of neurons and neural networks.
http://neuron.duke.edu/. Cited 6 Dec 2005
ERUDIT. (2008). ERUDIT Aphasia tutorial. http://fuzzy.iau.dtu.dk/aphasia.nsf/htmlmedia/
index.html. Cited 21 Jul 2008.
GENESIS. (2005). General neural simulation system. http://www.genesis- sim.org/GENESIS/.
Cited Dec 6 2005
GENI. (2008). Global Environment for Network Innovations. http://www.geni.net/. Cited 9
Feb 2009
Information Sciences Institute, University of Southern California. (2009). Neuroscholar.
http://www.neuroscholar.org/. Cited Jan 15 2009
International Consortium for Brain Mapping. (2005). http://www.loni.ucla.edu/ICBM/. Cited
Nov 16 2005
International Neuroinformatics Coordinating Facility. (2006). The neuroinformatics site.org.
http://www.incf.org. Cited 8 Dec 2006
International Neuroinformatics Coordinating Facility. (2006). The neuroinformatics site.org.
http://www.neuroinf.org. Cited 8 Dec 2006
International Neuroinformatics Coordinating Facility. (2006). G-Node (German Neuroinformatics
Node). http://www.neuroinf.de. Cited 8 Dec 2006
Jantzen J. (1999a). Aphasia and classification methods [PowerPoint slides]. Cited from
http://fuzzy.iau. dtu.dk/tutor/apha/lecture1/index.htm. Cited 21 Jul 2008
Jantzen J. (1999b). Second test: Other features [PowerPoint slides]. Cited from http://fuzzy.iau.
dtu.dk/tutor/apha/lecture3/index.htm. Cited 21 Jul 2008
Johnson, K.A. & Becker J.A. (1999). The whole brain atlas. http://www.med.harvard.edu/
AANLIB/home.html. Cited 1 Dec 2006
References 117
Kelly, M.P. Walter Reed Army Medical Center (2001). Internet Enabled Aviation Neuropsy-
chological Assessment. http://www.stormingmedia.us/59/5911/A591104.html. Cited 15 May
2009
Kelly, M.P. Walter Reed Army Medical Center (2001). Internet Enabled Aviation Neu-
ropsychological Assessment. http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=
html&identifier=ADA401195 Cited 15 May 2009
Kibbe, D. & Klepper, B. (2008). An open letter to the Obama health team. http://www.
thehealthcareblog.com/the_health_care_blog/2008/12/where-should-fe.html. Cited 2 Jan 2009
Lee, Y.S. (2007). Scientists seek to help ‘locked-in’ man speak. In: CNN Health.
CNN. http://www.cnn.com/2007/HEALTH/conditions/12/14/locked.in/index.html. Cited 11
Feb 2009
McLean Hospital Mailman Research Center Psychology Research Lab. (2008).
http://www.mclean.harvard.edu/research/mrc/psychlab.php. Cited 8 May 2008.
National Institutes of Health. (2005). National Human Genome Research Institute.
http://www.genome. gov/. Cited 9 Nov 2005
National Institutes of Health MRI Study of Normal Brain Development. http://www.bic.mni.
mcgill.ca/nihpd/info/index.html. Cited 26 Jan 2009
National Institute of Mental Health. (2008). Analysis of functional neuroimages (AFNI).
http://afni.nimh.nih.gov/. Cited 10 June 2008
National Institute of Mental Health. (2008). Neuroimaging informatics technology initiative.
http://nifti.nimh.nih.gov/ Cited 10 June 2008
National Institute of Mental Health. (2005). Neuroinformatics: The human brain project.
http://www.nimh.nih.gov/neuroinformatics/index.cfm. Cited 10 Nov 2005
National Public Radio (2009). All Things Considered: “Electronic Medical Record Change Not
Easy” http://www.npr.org/templates/story/story.php?storyId=102360638 Cited Mar 25 2009
National Rehabilitation Hospital (2009) http://www.nrhrehab.org/Research/Projects/ATRC+2000/
ANAM/default.aspx. Cited 15 May 2009
Obama B. (2008). President-elect Barack Obama lays out key parts of economic recovery plan.
http://change.gov/newsroom/entry/the_key_parts_of_the_jobs_plan/. Cited 14 Jan 2009.
O’Reilly, T. (2005). What is web 2.0: Designing patterns and business models for the next
generation of software. http://oreillynet.com/pub/a/oreilly/tim/news/2005/09/30/what-is-web-
20.html. Cited 16 Jan 2009
Organization for Economic Co-operation and Development. (2006). OECD Global science forum,
neuroinformatics working group. http://www.neuroinformatics.nl/. Cited 1 Dec 2006
Press, B. & Olshausen, B. (2005). Xanat 2.0: A Graphical anatomical database. http://redwood.
psych.cornell.edu/bruno/xanat/xanat.html. Cited 30 Nov 2005.
Society of Neuroinformatics for Neuropsychology. http://www.scnn.org. Cited 15 Jun 2009.
University of California-Davis. (2005). BrainMaps.org. http://brainmaps.org/ Cited 29 Nov 2005
University of California Los Angeles Center for Cognitive Phenomics. (2008). PubBrain.
http://www.pubbrain.org/ Cited 6 Oct 2008
University of California Los Angeles Center for Cognitive Phenomics. (2008). PubGraph.
http://www.pubgraph.org/ Cited 6 Oct 2008
University of California Los Angeles Semel Institute Consortium for Neuropsychiatric Phenomics.
(2008). http://phenomics.ucla.edu. Cited 6 Oct 2008
University of California Los Angeles Semel Institute Consortium for Neuropsychiatric Phenomics.
(2008). http://www.phenowiki.org/index.php5/Main_Page. Cited 6 Oct 2008
University of Texas at Austin Center for Learning and Memory. (2009). Synapse Web.
http://synapses.clm.utexas.edu/learn/filo3D/howto.stm. Cited Jan 15 2009
US Army. (2009). Automated Neuropsychological Assessment Metrics. http://www.army
medicine.army.mil/prr/anam.html. Cited 15 May 2009
U.S. Department of Energy Office of Science. (2005). Genomics programs of the U.S. Department
of Energy Office of Science. http://genomics.energy.gov/ Cited 9 Nov 2005
118 References
Washington University in St. Louis. (2008). Computerized Anatomical Reconstruction and Editing
Toolkit (CARET). http://brainmap.wustl.edu/. 6 June 2008
Wikipedia. (2009) Web 2.0. http://en.wikipedia.org/wiki/Web_2.0. Cited 16 Jan 2009
Yale University. (2008). NEURON for empirically-based computer simulations of neurons and
neural networks. http://www.neuron.yale.edu/neuron/. Cited Dec 6 2005
Index
119
120 Index