ECE3070- AI AND DIGITAL HEALTH

Theory course – LTPC (3003)

MODULE-2

DR. K. GOWRI,
ASSISTANT PROFESSOR- SENIOR SCALE,
PRESIDENCY UNIVERSITY,
BANGALORE
 Module 2- Applying Artificial Intelligence in health
care

Health data management, Treatment pathway

design, Transforming diagnostics, Health
assistance and administration, Patient
management, Precision medicine, Supporting
pharma: drug creation and clinical trials, FDA-
approved Algorithms in Healthcare.
 2.1 Health Data Management
• Data is everywhere. As we have more connected devices, embedded sensors,
and IoT devices, the data which created will become big, as Big Data.
• Data – Character, Text, Numbers, Pictures, Sound, or Video.
• Each may be structured or Unstructured.
• Information gathering and expertise sharing, analysing huge datasets and
stumbling upon correlations or clusters that are invisible to the human eye.
The types of tasks, the smart algorithms can carry out in medicine and
healthcare:
1. Information gathering and expertise sharing : A.I. can help medical teams
collaborate and design programs that support patient care and treatment
success.
 2.1 Health Data Management
2. Findings from data: After A.I. analyzes once invisible data, it can evaluate
findings based on the leading medical literature, and update current medical
knowledge and evidence-based guidelines, as well as patient care and
treatment pathways.
3. Helping insurance providers: Smart algorithms analysing huge patient
datasets could help insurance providers manage preventive care for large
populations.
4. Performance Improvement: In the case of pharma, A.I. can improve the
quality, the time management and reach of clinical trials.
5. Streamlining the process: Regarding administration, smart algorithms can
streamline administrative processes in the hospital, help in patient
management and support the management of challenging electronic medical
records systems.
 2.1 Health Data Management
• Data is a set of values of qualitative or quantitative variables.
• To become information, data requires interpretation.
• Information is organized or classified data, which has some meaningful value
(or values) for the receiver.
• Information is the processed data on which decisions and actions should be
based.
Healthcare is undergoing a data revolution, with two huge shifts:
1. The need to tame (to bring something wild under the control) growing cost
pressures, which is generating new incentives and reimbursement
structures, and
2. Digital health, which is democratizing healthcare by empowering
consumers through their digital devices.
 2.1 Health Data Management
• Clinical trends are changing.
• The advent of social media and immediate access to information through the
Internet and health communities mean that patients are clued up about their
health.
• Patients are generating data constantly and want to own and transfer their data
between services.
• Concurrently, medicine is striving for personalized care, led through an evidence-
based approach to decision-making.
• Patients and healthcare professionals alike are vocal in their appetite for all
available clinical data to make more effective treatment decisions based on current
evidence.
• A [Link] study in 2018 demonstrated that one in three patients would like
to share their data with their healthcare professional, but only one in five patients
can do so.
 2.1 Health Data Management
• As medicine evolves to become evidence based and personalized, data use
can be used to improve
Patient and population wellness
Patient education and engagement
Prediction of disease and care risks
Medication adherence
Disease management
Disease reversal/remission
Individualization and personalization of treatment and care
Financial, transactional, and environmental
forecasting, planning, and accuracy
Data facilitates information. Information leads to insights, and insights lead to the
making of better decisions.
 2.1 Health Data Management
Types of Data
[Link] data
[Link] data
[Link]-structured data
 2.1 Health Data Management
1. Structured data:
• Structured data typically refers to something stored in a
database—in a structure that follows a model or schema.
• Almost every organization will be familiar with this form of data
and may already be using it effectively.
• Most organizations will store at least some form of data in Excel
spreadsheets, for example.
• Within a clinical setting, Electronic Health Record (EHR)s may
also take a similar, structured form.
• Readings from embedded sensors, smartphones, smartwatches,
and IoT devices are typically forms of structured data—whether
that be the provision of blood glucose readings, steps walked,
calories burned, heart rate, or blood pressure.
 2.1 Health Data Management
2. Unstructured data:
• Unstructured data does not have a predefined model or schema.
• Data that is unstructured has no identifiable structure within it, and
this presents problems for querying and information retrieval.
• E-mails, text messages, Facebook posts, Twitter tweets, and other
social media posts are good examples of unstructured data.
• The Gartner report has indicated that data volume is set to grow 800%
over the next 5 years, and 80% of this data will be in the form of
unstructured data.
• Any organization would benefit from a mechanism of unstructured
data analysis to reduce the resource costs unstructured data add to
any organization.
• Structured data is traditionally easier to analyze than unstructured
data due to unstructured data’s raw and unorganized form.
• However, analytics of unstructured, informal datasets is improving,
with the use of data science and machine learning methods such as
Natural Language Processing (NLP) assisting in the understanding and
classification of sentiment.
 2.1 Health Data Management
3. Semi-structured data:
• Semi-structured data fits the space between structured and
unstructured data.
• Semi-structured data does not necessarily conform with the
formal structures of data schemas associated with relational
databases or data tables.
• However, semi-structured data may contain tags or markers
to separate semantic elements and enforce grouping of
records and fields within the data.
• Languages such as JSON (JavaScript Object Notation) and
XML (Extensible Markup Language) are all forms of semi-
structured data.
 2.1 Health Data Management
Health Care Data:
• Healthcare stakeholders understand they are surrounded by masses of data from
patients, professionals, and transactions.
• It is key to know how to drive value and meet KPIs (key performance indicators)-
Reducing Re-admissions, waiting time and predictive analysis.
Electronic Health Records (EHR):
• Every patient has a digital health record consisting of their details,
demographics(statistics about population), medical history, allergies, clinical
results, and so forth.
• Records can be shared, with patient consent, via secure computer systems and are
available for healthcare providers from both public and private sectors.
• Each record comprises one modifiable file, which means that doctors can
implement changes over time with no danger of data replication or inconsistencies.
 2.1 Health Data Management
• The EHR is evolving into the blockchain, which seeks to decentralize and
distribute access to data.
• No longer are patients considered passive recipients of care.
• Healthcare is now obliged to engage patients in their health, healthcare
decision-making, care, and treatment. Engagement can be maintained
through digital means.
• Example : Blue Shield of California is improving patient outcomes by
developing an integrated system that connects doctors, hospitals, and health
coverage to the patient’s broader health data to deliver evidence-based,
personalized care.
• EHRs can also trigger warnings, alerts, and reminders for when a patient
should get a new lab test; or track prescriptions to see if a patient has been
following treatment instructions.
 2.1 Health Data Management
• Looking at the potential in artificial intelligence for information gathering and
expertise sharing, analysing huge datasets and stumbling upon correlations
or clusters that are invisible to the human eye, it is easy to discern what types
of tasks smart algorithms can carry out in medicine and healthcare.
1. Regarding information gathering and expertise sharing, A.I. can help medical
teams collaborate and design programs that support patient care and
treatment success.
2. After A.I. analyzes once invisible data, it can evaluate findings based on the
leading medical literature, and update current medical knowledge and
evidence-based guidelines, as well as patient care and treatment pathways.
3. Smart algorithms analysing huge patient datasets could help insurance
providers manage preventive care for large populations.
 2.1 Health Data Management
4. In the case of pharma, A.I. can improve the quality, the time management
and reach of clinical trials.
5. Regarding administration, smart algorithms can streamline administrative
processes in the hospital, help in patient management and support the
management of challenging electronic medical records systems.

• Parallel to the increase of companies creating smart algorithms for medical

purposes, policymakers and regulators in healthcare also recognized the
immense potential in A.I. to optimize the processes in medicine for healing
patients in a more efficient way, as well as in pharma or in administration in
general.
 2.1 Health Data Management
• The rise in the number of U.S. Food and Drug Administration (FDA) -approved
algorithms shows this change in the attitude of regulators.
• As the amount of information they accumulate grows exponentially, the
assistance of computing solutions in medical decisions is imminent.
• A.I. will open new dimensions for doctors on the personal level as well as for
hospitals and other medical institutions on a more structural level.
• On the institutional level, the most obvious use of A.I. will be data
management.
• Collecting, storing, normalizing, tracing its lineage – it is the first step in
revolutionizing the existing healthcare systems.
• For example, the A.I. research branch of the search giant, Google, launched
its Google DeepMind Health project, which is used to mine the data of
medical records in order to provide better and faster health services.
 2.1 Health Data Management
• Data analytics can also streamline administrative processes in order to arrive
at better care.
• 97 percent of healthcare invoices in the Netherlands are digital, containing
data regarding the treatment, the doctor, and the hospital.
• These invoices could be retrieved easily.
• Dermatologists also recognized the huge potential of big data to bring lasting
change to their specialty.
• The American Academy of Dermatology introduced a clinical registry called
• DataDerm in 2016. The database was created by dermatologists and it
connects data on millions of patients from thousands of dermatologists
throughout the US. It eases the pain of reporting and allows medical
professionals to demonstrate the quality of care they provide to payers,
policy makers, and the medical community.
 2.1 Health Data Management
• At the same time, it gives every member a private analysis of his or her
practice’s data against national averages – down to the patient level. It is
great for setting standards in dermatology, measuring how each participant
performs and ensuring the average quality of care.
 2.2 Treatment pathway design
• IBM Watson launched its special program for oncologists able to provide
clinicians 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.
• Then, by combining attributes from the patient’s file with clinical expertise,
external research, and data, the program identifies potential treatment plans
for a patient.
• IBM launched another algorithm called Medical Sieve.
• It is an ambitious long-term exploratory project to build the next-generation
“cognitive assistant” with analytical, reasoning capabilities and a wide range
of clinical knowledge.
 2.2 Treatment pathway design
• Medical Sieve is qualified to assist in clinical decision making in radiology and
cardiology.
• The “cognitive health assistant” is able to analyze radiology images to spot
and detect problems faster and more reliably.
• Microsoft’s research machine-learning project, dubbed Hanover, also aims to
ingest all the papers and help predict which drugs and which combinations
are the most effective for cancer cases.
• Moreover, a series of start-ups, such as the Hungarian Turbine or Cambridge
Cancer Genomics aim to build smart algorithms to make oncology better.
• They are building artificial intelligence solutions to design personalized
treatments for any cancer type or patient faster than any traditional
healthcare service.
 2.2 Treatment pathway design
• Robots enabled with artificial intelligence are increasingly assisting
microsurgical procedures, too, to help reduce surgeon variations that could
affect patient recovery.
• Catherine Mohr, vice president of strategy at Intuitive Surgical and expert in
the field of surgical robotics, 23 believes surgery will also take to the next
level with the combination of surgical robotics and artificial intelligence.
• She envisioned a tight partnership between humans and machines, with one
making up for the weaknesses of the other.
• In fall of 2017, Maastricht University Medical Center in the Netherlands
already used an A.I.-assisted surgery robot to suture small blood vessels –
some no larger than .03 millimeters – and up to .08 millimeters across.
 2.2 Treatment pathway design
• In addition, A.I. is being used to provide analysis of a surgeon’s technical
abilities with products like Caresyntax’s qvident, a Web-based surgical risk
and quality management tool, designed to eliminate the “black box” of
surgical visibility.
• An intervention is considered a Clinical Path Way (CPW), if it meets four
criteria:
1) It is a structured multidisciplinary care plan;
2) It is used to channel the translation of guidelines or evidence into local
structures;
3) It details the steps in the course of treatment or care in a plan, pathway,
algorithm, guideline, protocol, or other “inventory of actions” (ie, the
intervention has time frames or criterion-based progression);
 2.2 Treatment pathway design
4) It aims to standardize care for a specific clinical problem, procedure, or
episode of care in a specific population.
• However, as an implementation strategy, CPW implementation research and
improvement involve extensive complexity and numerous challenges.
 2.2 Treatment pathway design

Fig.1 Pathway design and evaluation programme. The process is repeated until the
clinical pathway management design achieves the maximum public health benefit.
 2.2 Treatment pathway design

Fig.2 Generation of pathway elements and pathway design construction.

 2.2 Treatment pathway design
• Taking the clinical pathway for neonatal hypoglycaemia management with rooming-in at a
hospital as a hypothetical example,
“initial feed within 1 hour and screen plasma glucose 30 minutes after 1st feed”
• It generates critical pathway elements because it is the proven key decision-making step
in terms of whole costs and benefits for neonatal hypoglycaemia management with
rooming-in.
• For low blood glucose concentrations that may occur in the first hours after birth or
persist for up to several days, the duration of screening plasma glucose prior to each feed
and its impact on the whole costs and benefits of neonatal hypoglycaemia management
are uncertain.
• Thus, “Continue feeds q 2–3 hours and screen plasma glucose prior to each feed until 24
hours” generates controversial pathway elements.
• The other interventions are the standard operating procedures; they generate
conventional pathway elements.
• Each pathway element is constructed by population (pathway node) and intervention
(pathway node branch), and interlocking pathway elements ultimately form the pathway
prototype.