Module II : Healthcare

❏ Managing Medical Records

❏ Doing Repetitive Jobs
❏ Treatment Design
❏ Digital Consultation
❏ Virtual Nurses
❏ Medication Management

1
 Managing Medical Records
Benefits of applying AI to medical records?

❏ Improve productivity
❏ Accelerate digital health
❏ Improve personalized care
❏ Decision support

2
 Electronic Health Record (EHR)
An electronic health record (EHR) is a digital version of a patient’s
paper chart. EHRs are real-time, patient-centered records that make
information available instantly and securely to authorized users.
While an EHR does contain the medical and treatment histories of
patients, an EHR system is built to go beyond standard clinical data
collected in a provider’s office and can be inclusive of a broader view
of a patient’s care.

3
 Improve Productivity
AI tools have recently been developed that can help healthcare
providers (HCPs) extract clinically-relevant insights from free text
housed in medical records or insurance claims.

4
 Improve Productivity
One such tool, Healthcare Natural Language API, released by Google
Cloud, generates a structured data representation of the medical
knowledge stored in these data sources for downstream analysis and
automation.

5
 Example from Real World

Quotient Health
Location: Denver, Colorado

How it’s using machine learning in healthcare: With the help of

machine learning, Quotient Health developed software that aims
to “reduce the cost of supporting EMR [electronic medical records]
systems” by optimizing and standardizing the way those systems
are designed. The ultimate goal is improved care at a lower cost.
6
 Example from Real World

Ciox Health
Location: Alpharetta, Georgia

How it’s using machine learning in healthcare: Ciox Health uses

machine learning to enhance “health information management
and exchange of health information,” with the goal of
modernizing workflows, facilitating access to clinical data and
improving the accuracy and flow of health information.
7
 Medical Image Processing
https://colab.research.google.com/github/fadheladlansyah/chest-xra
y-pneumonia/blob/master/eda.ipynb

https://www.kaggle.com/sasakitetsuya/x-rays-classifiy-model-by-den
senet201

8
 Robotic Process Automation
❖ Artificial Intelligence is an umbrella term for technologies like RPA
and it also describes a computer’s ability to mimic human thinking.

❖ RPA is a rule-based software that has no intelligence and

automates repetitive tasks.

9
 Robotic Process Automation
Robotic Process Automation (RPA) is a software intended to automate
repetitive production tasks to streamline business processes. The
rule-based software evolved from screen scraping, workflow
automation, and artificial intelligence allowing the software to
aggregate data, trigger responses, and initiate new actions.

10
 Robotic Process Automation
The benefits of RPA prove it to be a solid tool that is achieving higher
quality yield at a lower cost and in less the time than traditional
methods. RPA tackles IT support process, workflow process, and back-
office work all while boosting productivity at a lower cost than hiring
a full-time employee.

11
 Robotic Process Automation
RPA is great for industries like insurance and banking that have a high-
transaction rate of repetitive tasks, but that’s all it can do. Without
any real intelligence but the instructed rules, it performs the tasks its
intended to do efficiently and effectively and nothing more.

12
 Robotic Process Automation
oday’s process automation is a great fit for businesses, offering:

User-friendliness — it’s easy to set to operate

Flexibility — able to adapt to your organization's needs
Non-disruptive — does not require significant changes to the
workplace
Scalability — can grow to accommodate increasing work
Centralization — bots are controllable from a central location
Ease of use — no programming expertise is required
Security — ironclad and foolproof protection

13
 Robotic Process Automation
Here are some of the reasons why you would use RPA and AI in a
single package:

Skyrocket the productivity of your workforce.

Save your workforce of painful, tedious, agitating, and repetitive
tasks.
Eliminate human errors from the workplace and be sure to reach
correct results.
Improve data management more effectively and save costs.
Acquire complete process transparency through a customized
dashboard and customized reporting support.
14
 Robotic Process Automation
Increase productivity: Automated applications and processes run
faster. Automation of applications and processes, plus the
automation of decision-making, forecasting and predictions from
multiple sources of structured and unstructured data in real time
empowers organizations with greater productivity and accuracy in
their planning cycles.

15
 Robotic Process Automation
Reduce costs: According to Deloitte, “Executives estimate intelligent
automation will provide an average cost reduction of 22%,” although
they also found that, “organizations currently scaling intelligent
automation say they have already achieved a 27% reduction in costs
on average from their implementations to date.”

16
 Robotic Process Automation
Improve accuracy: The use of both structured and unstructured data
and the automation of repetitive processes ensures better decision-
making, and less human intervention results in more precise results.

17
 Robotic Process Automation
Enrich the customer experience: Organizations that use technology
can better understand customers' needs, communicate more
effectively and bring higher-quality products to market. Customers, in
turn, are typically more satisfied in their buying experience.

18
 Misconceptions about intelligent automation (IA)

IA replaces a human workforce: IA actually supplements or

augments a human workforce by taking over repetitive tasks so the
human, thinking workforce is available to work on more complex or
more pressing matters.

19
 Misconceptions about intelligent automation (IA)

IA is nice to have, but not needed: IA is no-longer optional.

Automation-infused applications are prevalent in our daily lives, such
as speaking to Alexa or using a weather app.

20
 Misconceptions about intelligent automation (IA)

IA can make unbiased decisions: IA formulates decisions based on

input gathered and received, much of which is situational or provided
by individuals and organizations responsible for that input. Therefore,
the decisions made are inherently biased.

21
 Examples with Code

https://github.com/coding-ai/EasyApply-Linkedin

https://github.com/kelaberetiv/TagUI

https://learn.aisingapore.org/courses/learn-rpa-with-tagui-beginners
-course/

https://github.com/tebelorg/RPA-Python

22
 Artificial Intelligence for Treatment Design

Algorithms based on artificial intelligence can aid in decision-making

and thus improve the accuracy of diagnoses. For example, several
studies demonstrate that radiologists can improve the accuracy of
cancer detection using radiological scans when assisted by artificial
intelligence. In the future, medical A.I. trained through reinforcement
learning may be able to discover treatments and cures for diseases
that human medical professionals are unable to.

23
 Artificial Intelligence for Treatment Design

IBM Watson launched its dedicated program for oncologists,

providing clinicians with evidence-based treatment options. Watson
for Oncology has an advanced ability to analyse the meaning and
context of structured and unstructured data in clinical notes and
reports that may be critical to selecting a treatment pathway.
By combining attributes from the patient’s file with clinical expertise,
external research, and data, the program identifies potential
treatment plans for a patient.

24
 Artificial Intelligence for Drug Creation

Clinical trials for pharmaceutical development can take up to a

decade and cost billions of dollars. Accelerating and cost-effectively
completing this process would have an impact on current healthcare
and how innovations reach everyday medicine. Artificial intelligence
significantly reduces the time and cost of drug discovery and
development.

25
 Artificial Intelligence for Drug Creation

Insilico Medicine, an artificial intelligence-powered pharmaceutical

startup, identified a potential new drug in only 46 days in 2019. Its
software accomplished this by sifting through terabytes of data that
would take humans years to process. Atomwise (a US-based
company) used its artificial intelligence algorithm to identify two
drugs with significant potential to reduce Ebola infectivity during the
2015 Ebola epidemic. This effort was completed in less than a day.

26
 Artificial Intelligence for Drug Creation

Another excellent example of using big data for patient management

is Berg Health, a biopharmaceutical company based in Boston that
mines data to determine why some people survive diseases and thus
improves existing treatments or develops new ones. They combine
artificial intelligence and biological data from patients to denote the
differences between healthy and disease-friendly environments,
thereby assisting in the discovery and development of drugs,
diagnostics, and healthcare applications.

27
 Artificial Intelligence for Drug Creation

Python Libraries

DeepChem (https://github.com/deepchem/deepchem )

Examples
https://github.com/deepchem/deepchem/tree/master/examples

https://colab.research.google.com/github/PatWalters/blog_posts/blo
b/main/chemfp_simsearch.ipynb#scrollTo=FzIW-F8R-v8p

https://colab.research.google.com/drive/1Y2kZH9iOePbk1TA73InOzU
_B8pvrGDJr?usp=sharing
28
 Digital Health Consultation
Digital health consultation tools have grown in popularity as a result
of the upsurge of smartphones and the growing demand for 24/7
accessible healthcare. With so many medical facilities only open
during standard business hours Monday through Friday, it can be
challenging for people with typical 9 to 5 schedules to fit in a doctor
visit.
Additionally, some individuals may be unaware when an illness is
severe enough to warrant a trip to the doctor rather than simply
sleeping it off and taking over-the-counter medications.

29
 Digital Health Consultation
In 2018, the Kaiser Family Foundation conducted a survey and
discovered that 45% of people aged 18 to 29 lacked a primary care
provider or regular doctor. As a result, the popularity and necessity of
digital health consultations grew. Digitized health is particularly
popular among younger working generations because it offers a
convenient alternative to visiting a doctor's office.

30
 Digital Health Consultation
Consider the following scenario: you are ill but do not own a car, and
the nearest clinic is more than an hour away via public
transportation. Is it really worth the time spent travelling there and
back to see a doctor for something that may turn out to be a
common cold?

31
 Digital Health Consultation
Not only does digital consulting enable remote health assistance, it
also reduces the amount of time patients and doctors spend
together, allowing doctors to see more patients.

32
 Digital Health Consultation
Because so many people prefer digital consultations to traditional in-
office doctor visits, AI is being used to identify, detect, and treat
health concerns remotely via smartphone applications.

33
 Digital Health Consultation
Buoy, an interactive symptom-checking chatbot, is a current digital
consultation tool that makes use of artificial intelligence. It is similar
to WebMD's Symptom Checker, but the primary difference is that
Buoy makes use of artificial intelligence, whereas WebMD does not.

34
Digital Health Consultation

35
The Opportunities Machine Learning
for Healthcare

36
 Clinical Decision Support Systems
(CDSS)
Clinical decision support tools aid in the analysis of large amounts of data
in order to diagnose a disease, determine the next stage of treatment,
identify any potential complications, and ultimately improve patient care
efficiency. CDSS is a powerful tool that enables physicians to perform their
duties more efficiently and effectively, while also decreasing the likelihood
of making an incorrect diagnosis or prescribing ineffective treatment.

This application of machine learning in medicine (healthcare) has existed

for some time but has gained traction in recent years. It is due to the
increased acceptance of electronic health records (EHRs) and the
digitization of various data points, including medical images.

37
 Smart Recordkeeping
Keeping all patient records up to date on a regular basis presents a
challenge, as data entry is a tedious task. It is, however, critical for
effective decision-making and improved patient care.

One application of machine learning in healthcare is the use of optical

character recognition (OCR) technology to read physician's
handwriting, which speeds up and simplifies data entry. Other
machine learning tools can then be used to analyse this data in order
to improve decision-making and patient care.

38
 Machine Learning in Medical Imaging
For the longest period of time, medical images such as X-rays were
analogue. This has limited the use of technology for identifying
anomalies, grouping cases, and conducting overall disease research.
Fortunately, the process's digitization has increased the opportunities
for performing these types of data analysis, including through the use
of machine learning. And, according to a recent meta-analysis,
machine learning algorithms perform as well as (and in some cases,
better than) human specialists, with a sensitivity of 87.0 percent and
a specificity of 92.5 percent for deep learning algorithms versus 86.4
percent and 90.5 percent for human physicians.

39
 Machine Learning in Medical Imaging
One of the most well-known successful applications of machine
learning in healthcare is Microsoft's InnerEye project. Its initial focus
was on three-dimensional radiological images, where machine
learning tools were developed to distinguish healthy cells from
tumours.

40
 Personalized Medicine
What makes medicine such a complicated and resource-intensive
field is the fact that each case is unique. Individuals frequently have a
sleuth of conditions that require concurrent treatment. As a result,
complex decisions must be made in order to develop an effective
treatment plan that takes drug interactions into account and
minimises potential side effects.

How can machine learning be used to solve this problem in

healthcare? IBM has figured this out with their Watson Oncology
system, which utilises the patient's medical history to generate
multiple treatment options.
41
 Behavior Adjustments
In healthcare, prevention is just as critical as disease treatment. One
of the most critical components of preventive medicine is behaviour
modification in order to break unhealthy habits and establish a
healthy lifestyle.

One of the advantages of machine learning in healthcare is that it can

be used to alert us to something we are unaware of. That is precisely
what Somatix accomplishes. This machine learning-based application
tracks the patient's daily activities and identifies unconscious habits
and routines, allowing the patient to concentrate on breaking them.

42
 Predictive Approach to Treatment
When it comes to the most dangerous diseases, detecting them early
significantly increases the likelihood of successful treatment. This also
assists in identifying the possibility of a patient's condition
deteriorating prior to its occurrence.

One example of machine learning's value in healthcare is its ability to

accurately predict some of the most dangerous diseases in at-risk
patients. This includes detecting diabetes symptoms (using a Naive
Bayes algorithm), liver and kidney diseases, and oncology.

43
 Data Collection
One of the most critical responsibilities of a physician is to properly
collect a patient's history. This is frequently difficult, as the patient is
not a specialist and is unaware of which data is pertinent to disclose.

Healthcare professionals can use machine learning to determine the

most pertinent questions to ask a patient based on a variety of
indicators. This will aid in the collection of pertinent data while also
providing a prediction of the most likely conditions.

44
 Elderly and Low-Mobility Groups Care
Machine learning and medicine can assist low-mobility groups
(including the elderly and people who use wheelchairs) in improving
their daily lives through smart reminders and scheduling assistance,
predicting and avoiding potential injuries through the identification of
common obstacles and determining the optimal paths, and obtaining
assistance as needed.

While these solutions are effective, they are insufficiently

widespread. Healthcare companies, on the other hand, are already
taking steps to make them widely available. For instance, in Japan,
there is a plan to have an AI perform 75% of elderly care.
45
 Robotic Surgery
Surgical procedures require extreme precision, adaptability to
changing circumstances, and an ability to maintain a consistent
approach over an extended period of time. While skilled surgeons
possess all of these characteristics, one of the potential applications
of machine learning in healthcare is for robots to perform these tasks.

Currently, robotic surgery can be used effectively to assist human

surgeons. Specifically, machine learning can be used to improve
surgical modelling and planning, as well as to assess the surgeon's
abilities and simplify surgical tasks such as suturing.

46
 Drug Discovery and Production

ML algorithms can model an active component that would work on

another similar disease based on previously acquired data on active
components in drugs and their effects on the organism.

This approach could be used to develop a customised medication for

patients who have a unique combination of illnesses or special
requirements. In the future, this machine learning technique might
be combined with nanotechnology to improve drug delivery.

47
 Clinical Research
Clinical research and trials are time-consuming and costly
endeavours. This is for a good reason – new drugs and medical
procedures should be demonstrated to be safe before being widely
used. However, there are instances when the solution must be
released immediately – as is the case with the COVID-19 vaccines.

Fortunately, there is a way to accelerate the process using machine

learning algorithms. It can be used to select the optimal sample size
for the trial, collect additional data points, analyse ongoing data from
trial participants, and minimise data-related errors.

48
 Infectious Disease Outbreak Prediction
The COVID-19 pandemic demonstrated how unprepared we were for
a pandemic of this magnitude. It's worth noting that experts in the
field have been warning the government for years about the
possibility of such an event.

Now, we have machine learning-based tools that can assist in

detecting the early signs of an epidemic. Satellite data, news reports,
and social media reports, as well as video sources, are analysed to
determine whether the disease has the potential to spread out of
control.

49
 Machine Learning in Retail
Sales & CRM
Customer Recommendation
E-Commerce
Logistic and Delivery
Payment Services
Class Project

50
Machine Learning in Retail

51
 Machine Learning in Retail
Instead of applying assumptions and a complex set of rules in an
unwieldy Excel workbook, Machine Learning models learn patterns
from the data to generate predictions. This way, if the underlying
trends change, the model can be retrained, and learn these changes.
Also, statistical learning algorithm can discover patterns missed by
business analysts.

52
 Machine Learning in Retail
Using Machine Learning instead of an Excel model makes the
forecasting process much more traceable. The prediction algorithm
can be run on a cloud Machine Learning environment or a Virtual
Machine and write predictions directly to a database. These
predictions could then be distributed to business users through
interactive dashboards. This way, there is a single source of truth for
the entire company. A centrally run and managed algorithm is more
secure, more adaptable and more efficient.

53
 Machine Learning in Retail
What is sales forecasting?
A sales forecast is an estimation of future sales. This estimation can
be based on past values, economic indicators, seasonality, weather
forecasts.

54
 Machine Learning in Retail
There are two main types of sales forecasting:
1. Rule-based forecasting
2. Machine Learning forecasting

55
 Machine Learning in Retail
Rule-based forecasting: In rule-based forecasting, predictions are
generated using a set of manually developed rules and assumptions
based on past data and known trends. As an example, if business
analysts notice that sales have been consistently growing by 5% each
year, they could apply the following forecasting rule:
Sales Tomorrow = Sales Last Year * 1.05

56
 Machine Learning in Retail
Machine Learning forecasting: Machine learning algorithms would
learn the rules that would have to be manually designed in rule-
based forecasting. This is done through a process called supervised
learning. Supervised learning is the task of learning the relationship
between outputs (sales) and inputs (past sales, economic indicator,
holiday calendar etc.) Machine Learning algorithms find these
relationships by minimising prediction error – i.e. finding the
relationships and parameters that maximise prediction accuracy.

57
 Machine Learning in Retail
How to use Machine Learning to predict sales?
In the world of Machine Learning, sales forecasting is a time-series
regression problem. A regression is any task concerned with the
estimation of a continuous quantity (i.e. sales). Time-series
regressions are a particular case of regression, with an additional
time dimension.

There are two main types of time series regressions models:

1. Auto-regressive models
2. Multivariate models

58
 Machine Learning in Retail
1. Auto-regressive models: These models predict future sales solely
based on past sales values. These models include ARIMA, SARIMAX,
and Exponential Smoothing. They generate predictions by finding
trends and seasonality patterns.

2. Multivariate models: Multivariate models are based on a variety of

inputs, including past sales, holiday calendars, or even economic
indicators. These models include Linear Regressions, Neural
Networks, Decision Tree-based methods and Support Vector
Machines.

59
Machine Learning in Retail

60
Machine Learning in Retail

61
Machine Learning in Retail

62