Artificial Intelligence - Potential Game Changer for Medical Technology Companies
1. This document is confidential and contains proprietary information, including trade secrets of CitiusTech. Neither the document nor any of the information
contained in it may be reproduced or disclosed to any unauthorized person under any circumstances without the express written permission of CitiusTech.
Artificial Intelligence: Potential Game
Changer for Medical Technology Companies
2 November, 2017 | Author: Poulami Chatterjee: Healthcare Business Analyst,
Akash Jha: Healthcare Consultant, CitiusTech
CitiusTech Thought
Leadership
2. 2
Agenda
AI: Overview of a potential game changer
AI: Healthcare applications & implications
AI: Key benefits for Medical Technology companies
References
3. 3
A set of statistical techniques for identifying
patterns and enable intelligent decision support.
Deep Learning
Definition: Deep
learning is a sector of
machine learning that
uses neural networks
to mimic the cognitive
capability of the human
brain.
Technology: IBM
Watson, Python, R and
Tensor flow etc.
Predictive Analytics
Definition: Predictive
analytics has a variety
of techniques such as
predictive modeling,
machine learning and
data mining to analyze
current and historical
facts to predict future
outcomes.
Technology: SPSS, SAP
HANA RapidMiner, etc.
Natural language
processing (NLP) is a field in
AI and Linguistics which is
concerned with the
interactions between
computers and human
(natural) languages
Artificial Intelligence
Applications and services are designed to simulate human senses and its derived intelligence without any
external input to perform cognitive functions such as learning and problem solving.
Machine Translation
Natural Language
understanding
Natural Language
Generation
Corpus Linguistics
Computer vision is the
capturing, processing and
understanding of visual
information similar to the
combination of the eye
and the brain.
Image Extraction
Derived Insights &
Reports
Image Processing &
Analytics
Image Sensors
Natural Language Processing Machine Learning Vision
AI: Driven By Cognitive Tools and Next-Gen Analytics
4. 4
Deep Learning
Predictive Analytics
Classification
& Clustering
Translation
Image
Recognition
Computer
Vision
Commonly used software
R Python Cortana IBM Watson
Tensorflow Apache CTAKES Rapid Miner SPSS
Natural
Language
Processing
Machine
Learning
Vision
Artificial
Intelligence
3 Pillars of AI
5. 5
Use of NLP to scan precedents and research
material for cases
Lawyer chatbots to diagnose a case and
automatically prepare an appeal
Use of complex AI systems and NLP based
research to make trading and investment
decisions
“Robo-advisers” offer portfolio management
and AI apps help in personal finance
AI based Self-driving cars under test from
Uber, Google and Tesla
Intelligent cars self-diagnosing technical
problems, locating gas stations etc.
Use of AI to analyze browsing, spending
patterns and push ads to users
Smart posters sensing people’s presence
and changing ads based on reactions
Analyzing customer emails for meaning and
sentiment and feeding data to CRM systems
Deep learning techniques to replace human
customer services with bots
Use of AI to prepare completely autonomous
reports on major events, i.e., sports,
elections
Analyzing consumer pattern and pushing
relevant content, i.e., Netflix, Facebook etc.
Today AI has influence over many aspects of our lives
Others uses of AI are Law enforcement, games, smartphones etc.
Marketing Law
Retail Finance
AutomobileMedia
AI: Significant Adoption Across Industries
6. 6
AI: Overview of a potential gamechanger
AI: Healthcare applications & implications
AI: Key benefits for Medical Technology companies
Agenda
7. 7
AI: Significant Opportunities in Healthcare (1/2)
Patient Engagement
Chatbots
Roles Market
Personalized advice
Diagnosis and Referrals
Medication reminders
Analyze mental health
Lifestyle suggestions
Dr. AI by HealthTap
Buoy app
Microsoft HealthBot
BabylonHealth App
Your.MD app
Population Health
Advanced dashboards
Roles Market
Imaging analytics
Risk Stratification
Survivability prediction
Early diagnosis
Gene sequence analysis
Customized care plans5.3
IBM Watson
Deep Genomics
Google Deepmind
Intel Lumiata
Sophia Genetics
CareSkore
8. 8
AI: Significant Opportunities in Healthcare (2/2)
Medicine
Diseases, drugs & trials
Roles Market
Drug efficacy prediction
Medication Management
New Drug discovery
Outbreak prediction
NLP based outcomes
IBM Watson with J&J
Microsoft Hanover
AtomWise
AiCure app
Insilico medicine
PathAI
User Experience
Operational efficiency
Roles Market
NLP Text mining
Natural voice search
Virtual assistants
Internet of Things (IoT)
Linguamatics
Dolbey Fusion Speech
Nuance Nina
Microsoft Cortana
Medical technology companies and application vendors are already developing and prototyping AI applications
across various healthcare use cases.
9. 9
AI in Healthcare: Industry Adoption (1/2)
Google DeepMind – NHS
Deep learning algorithms interpret visual information in the form of de-personalized scans (head and neck scans
at University College London Hospitals NHS Foundation Trust, eye scans at Moorfields Eye Hospital NHS
Foundation Trust) to identify potential issues.
IBM – Cleveland Clinic Lerner College of Medicine research
IBM WatsonPaths parses the medical records for facts and test results, then knits them together into competing
theories that might explain the patient's symptoms and communicate with Physicians in a natural way.
IBM – Memorial Sloan Kettering Cancer Center
IBM Watson ingests and analyzes tens of thousands of the renowned cancer center patient records and clinical
research providing treatment options with degrees of confidence for each, along with the supporting evidence.
Deep Genomics
Leverage deep learning algorithms to decode the meaning of the Genome trying to predict the effects of a
particular mutation based on the analysis of other mutations. Developed a database for prediction on how 300
million genetic variations could affect a genetic code.
Enlitic
Interprets a medical image and classify malignant tumors, patient risk and provides decision support. Also does
regressive retrospective analysis to judge clinical performance. Detection rate is 50% more accurate and 10,000
times faster than a Radiologist.
Turbine
Models cell biology on the molecular level to identify the best drug for a specific tumor, complex biomarkers and
design combination therapies by performing millions of simulated experiments guided by an AI identifying
biomarkers signaling sensitivity to treatment.
Research ProductPrimary Care
ResearchOphthalmology
Research ProductOncology
ResearchGenomics
ProductOncology
ProductPharmaceutical
10. 10
AI in Healthcare: Industry Adoption (2/2)
Your.MD – NHS
AI-powered chatbot asks users about their symptoms and provides easy-to-understand information about their
medical conditions. Uses machine learning, natural language processing and generation. App’s diagnosis has
been approved by UK NHS.
Babylon Health
Physician trained AI chatbot provides diagnosis of symptoms, continuously analyzes data and cross-references
with other patients, sets up video consultations with selected Providers and manages prescriptions and
histories.
Stanford University
In-house AI algorithm, trained with over 130,000 images of moles, rashes, and lesions, diagnoses skin cancer
rivalling professional doctors.
AiCure
AI algorithms set up medication reminders, offers facial recognition and visual confirmation of medication
ingestion, adapts to patient lifestyle patterns and set up automated interventions in case of deviant behavior.
Arterys – GE Healthcare
AI driven cloud-based medical imaging platform uses MR images to draw up the contours of the heart’s four
chambers, measuring how much blood they move with each contraction, usually hand-drawn by Cardiologists.
GE plans inclusion in its MRI machines soon.
23andMe – Rthm
23andMe have genetic diagnostic tools to help individuals understand their genetic makeup while Rthm allows
users to leverage the insights produced from their genetic test to implement changes to their everyday routine
ProductPrimary Care
Primary care
ResearchOncology
Medication
ProductCardiology
ProductGenomics
Product
Product
Research
11. 11
Key AI Focus Areas for Medical Technology Companies
Data Mining and Regression Care Plans
Reduce Redundancy
Population Health
Patient Coordination
Decision Support
Pharmacy OptimizationGenomics/Precision Medicine
AI can predict unseen & intelligent outcomes
with attached risk using historical data
Use cases may include ESRD prediction rates
and co-morbidity risk & survival
AI uses guided learning to identify repetitive,
redundant processes like utilization of scanners
Streamline processes, i.e., auto-booking provider
appointments in making a diagnosis
AI based web crawlers can find latest research
and treatment for Oncology and Cardiology
Analyze long-term effectiveness of treatment
and have alternate care workflows
AI enables Genome sequence mining, identifying
potential incompatibilities in medication
Helps save provider time by making educated
predictions on risks and prescriptions
AI uses wearables & smartphone to analyze
behavioral patterns, risk and medication adherence
Create intelligent tailored care plans keeping in
mind payor, risks and past incompatibilities
AI chat bots converse with patients in English and
diagnose basic illnesses using databases
Intelligent notifications to patients on change in
schedule, i.e., medication adherence
AI can identify usage patters for EHR like frequently
used tabs, i.e., open demographics on log-in
Use past analysis to suggest ideal next-steps for a
particular patient, i.e., follow-up on BMI etc.
Streamline Pharmacy operations by reordering
stock of most used medications
AI can predict new drug behavior based on
regression data from its chemical constituents
1
2
4
3
8
7
6
5
12. 12
Agenda
AI: Overview of a potential gamechanger
AI: Healthcare applications & implications
AI: Key benefits for Medical Technology companies
13. 13
Leveraging AI in Medical Technology: Quick Wins
Making Search More
Efficient & Accurate
Providers belonging to varied specialties may be using multiple EHRs from
different vendors
Natural Language (NLP) search service integrated into the EHR can reduce time
spent in searching historical records
It also helps Providers do a more focused searched as compared to manual
filters
Improving Care-
Coordination
Providers often go through a repetitive process of collecting and interpreting
information at multiple points in the care co-ordination process, i.e.,
appointment, reception, diagnosis and treatment
AI chat bots integrated into Patient Portals can mostly diagnose the illness and
book appointments even before the patient reaches the clinic
Enhancing Patient
Engagement
Improved turnaround time through use of chatbots and automatic
appointments reduce frustration
Patient behavioral analytics and focused notifications, i.e., sleep time, eating
habits etc. improve wellness and builds trust
Intelligent notifications and alerts to providers help long-term patient health
and improves adherence
14. 14
Leveraging AI in Medical Technology: Long-Term Impact
Significant Gains in
Population Health
AI initiatives such as chat bots help reduce probability of miscommunication and
time to treatment
Predictive analytics on patient behavioral patterns and vitals helps customize
better care plans
Advanced risk scoring helps prioritize patients for priority care
AI mining of health records and scanning of new research content helps find new
treatment avenues
Reduction in
Cost of Care
Long-term population health improvements result in less average treatment
costs
Predictive analytics, i.e., tumor analysis, imaging nodule, risk scorings help
identify critical illnesses at an early stage
Regression based AI systems can identify under and over utilization of medical
resources, i.e., scanners etc.
15. 15
AI in Medical Technology: Critical Success Factors
Big Data Challenge: Unsupervised learning requires training samples from a huge volume
of data to be successful. Healthcare data in itself in very size intensive
Defining Powerful Use Cases: Use cases will vary significantly, i.e., Patient Engagement
may be a use case for a EHR vendor but not for a PACS vendor
Building a Cognitive Data Ecosystem: For AI to succeed, total healthcare data needs to
behave as a single entity and completely accessible to the AI subsystems
Customer IT Readiness: Requires investments in Data Scientists, Predictive analytics tools
and big-data technologies or partnerships with 3rd party entities
Managing Variability: Advanced AI chatbots may sometimes have unpredictable behavior
(MS Tay bot) and needs to be supervised
Supervised Learning: Most AI applications today are supervised learning requiring
unbiased data and fast processing for optimal result
Human Challenges: Collection of data for AI implementations will raise security and ethical
challenges
16. 16
AI in Medical Technology: Critical Success Factors (1/2)
Big Data Challenge Unsupervised learning requires training samples from a huge volume of data
to be successful. Healthcare data in itself is very size intensive.
Defining Powerful
Use Cases
Use cases will vary significantly, i.e., Patient Engagement may be a use
case for a EHR vendor but not for a PACS vendor.
Building a Cognitive
Data Ecosystem
For AI to succeed, total healthcare data needs to behave as a single entity and
completely accessible to the AI subsystems.
Customer IT
Readiness
Requires investments in Data Scientists, Predictive analytics tools and big-
data technologies or partnerships with 3rd party entities.
Managing Variability Advanced AI chatbots may sometimes have unpredictable behavior (MS Tay
bot) and needs to be supervised.
17. 17
AI in Medical Technology: Critical Success Factors (2/2)
Supervised Learning Most AI applications today are supervised learning requiring unbiased data
and fast processing for optimal result.
Human Challenges Collection of data for AI implementations will raise security and ethical
challenges.
18. 18
AI in Medical Technology: Key Takeaways
AI can have huge implications in the Healthcare market as it has been having in other
spheres
Most of the innovation in AI is happening under startups with few big players, i.e., IBM,
Google
Most popular use cases are Risk Prediction, Medical Image Analytics, Patient Engagement
and Medication Adherence
We recommend Medical Technology companies to first identify very strong use cases, i.e.,
for a Pop Health vendor, Risk Prediction using AI can have huge long-term benefits
Companies will also need to define their strategy for skillsets, i.e., either develop in-house
or outsources key skills such as Data Science, Big-Data, NLP etc.
Companies should take a holistic view of the future and not be bogged down by short-
term issues in terms of cost, technology and training