towards personalized medicine in the netherlands · selected watson for oncology to identify...
TRANSCRIPT
© 2016 E-HEALTH WEEK AMSTERDAM
CONFIDENTIAL
CONFIDENTIAL
CONFIDENTIAL
CONFIDENTIAL
Towards Personalized Medicine in The Netherlands Clinical Decision Support and Cognitive Computing in Oncology
© 2016 E-HEALTH WEEK AMSTERDAM
The Dutch Health Deal – CDSS in Oncology (June 8TH 2016)
Stimulating innovation between government and (private) partners
Impactful innovations improving quality of life, efficiency, outcomes
Initiated by ‘market’ entities
Government removes bottlenecks for the parties involved
© 2016 E-HEALTH WEEK AMSTERDAM
3
© 2016 E-HEALTH WEEK AMSTERDAM
Introduction speakers
Prof. Dr Gerrit Meijer Netherlands Cancer Institute Diagnostic oncologist, specialized in Translational Gastrointestinal Oncology
Dr Nicky Hekster IBM Netherlands Technical Leader Healthcare & LifeSciences Watson Ambassador
Prof. Dr Sabine Linn Netherlands Cancer Institute Medical oncologist, specialized in Breast Cancer
© 2016 E-HEALTH WEEK AMSTERDAM
Data is growing exponentially It demands new approaches in both technology and strategy
We are here
44 zettabytes
80% unstructured data
2016 2020
20% structured data
2010
Non-standardized data and numbers, free text, speech, video, images, pictures, …
Numbers, spreadsheets, standardized data models (semantic models), …
© 2016 E-HEALTH WEEK AMSTERDAM
Four industrial revolutions Medicine is here!
© 2016 E-HEALTH WEEK AMSTERDAM
Tabulating systems
1900 Programmable Era
1950 Cognitive Era
>2010
A new computer era is coming of age – cognitive computing From automating the world to understanding the world
© 2016 E-HEALTH WEEK AMSTERDAM
Definition of Cognition The mental action of acquiring knowledge and understanding through thought, experience, and our senses
Knowledge
Ability to understand
Ideation, conviction
Sensation, observation
Imagination
Store in and retrieve from memory
Problem solving capabilities
Think
Language
10
© 2016 E-HEALTH WEEK AMSTERDAM
How cognition works To become an expert
Evaluation
Observation
Decision Interpretation
© 2016 E-HEALTH WEEK AMSTERDAM
How our biran wkros?
I cdn'uolt blveiee taht I cluod aulaclty uesdnatnrd waht I was rdanieg: the
phaonmneel pweor of the hmuan mnid. Aoccdrnig to a rseearch taem at
Cmabrigde Uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are,
the olny iprmoatnt tihng is taht the frist and lsat ltteer be in the rghit pclae. The
rset can be a taotl mses and you can sitll raed it wouthit a porbelm. Tihs is
bcuseae the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a
wlohe.
© 2016 E-HEALTH WEEK AMSTERDAM
History of AI
1974- 1980: 1st AI “Winter”
1970s 1980s 1990s
1956: “Birth” of AI John McCarthy coins term artificial intelligence (AI) at Dartmouth Conference
1965: First Expert System Stanford team led by Ed Feigenbaum creates DENDRAL and MYCIN
1987- 1993: 2nd AI “Winter” 1950: Turing Test
Turing introduces way to test for intelligent behavior
1990s: AI on www AI-based extraction programs prevalent on www
1997: Deep Blue IBM Deep Blue defeats World Chess Champion
2016: Google DeepMind AlphaGo wins Go
2005: Autonomous car
Stanford-built autonomous car wins DARPA Grand Challenge
2014: Market changes IBM formation of Watson Group and Google acquisition of Nest Labs
2011: Watson IBM’s Watson competes and wins on Jeopardy!
1960s 1950s 2010… 2000s
2014: Facebook Recognize individuals DeepText
© 2016 E-HEALTH WEEK AMSTERDAM
The Grand Challenges
Chess – Deep Blue (1997)
• A finite, mathematically well-defined search space (10120)
• Limited number of moves and states on an 8 x 8 board
• Grounded in explicit, unambiguous mathematical rules
Human Language – Watson (2011)
• Ambiguous, contextual and implicit
• Grounded only in human cognition
• Seemingly infinite number of ways to express the same meaning
Go – DeepMind (2016) • A finite, mathematically well-defined but very large search space (10761)
• Limited number of positions and states on a 19 x 19 board
• Based on explicit, unambiguous logical rules
© 2016 E-HEALTH WEEK AMSTERDAM
Thomas J. Watson (1874 – 1956)
© 2016 E-HEALTH WEEK AMSTERDAM
Big Data & Analytics
Data Mining, Optimization, Text Analytics
Artificial Intelligence
Machine Learning, Natural Language Processing, Algorithms & Theory
Cognitive Experience
HCI, Speech, Translation, Machine Vision, Visualization
Cognitive Knowledge
Knowledge Representation, Ontologies, Semantics, Context
Computing Infrastructure
High Performance Computing, Distributed Systems, Programming Models & Tools
IBM Watson is based on
© 2016 E-HEALTH WEEK AMSTERDAM
Understands natural
language and human speech
Adapts and learns from
user selections and responses
Reasons, generates and evaluates hypothesis for
better outcomes
3
2
1
Watson is an example of a cognitive system Intelligence Amplification
Watson does not predict! Watson only explains from a very large set of data and helps human beings taking complex decisions.
© 2016 E-HEALTH WEEK AMSTERDAM
R&D
Demonstration
Commercialization
Health and Lifesciences Applications
IBM Research Project
(2006 – )
Jeopardy! Grand Challenge
(Feb 2011)
Watson for
Healthcare (Aug 2011 –)
Watson Health Group (April 2015 – )
Watson for Financial
Services (Mar 2012 – )
Expansion
Internal start-up division
Watson IoT Group
(Jan 2016 – )
Internet of Things
Applications
Brief history of IBM Watson
Watson Group
(Jan 2014 – )
Cross-industry Applications
© 2016 E-HEALTH WEEK AMSTERDAM
IBM Bluemix PaaS platform
© 2016 E-HEALTH WEEK AMSTERDAM
Relationship Extraction
Questions &
Answers
Language Detection
Personality Insights
Keyword Extraction
Image Link
Extraction
Feed Detection
Visual Recognition
Concept Expansion
Concept Insights
Dialog Sentime
nt Analysis
Text to Speech
Tradeoff Analytics
Natural Language Classifier
Author Extraction
Speech to
Text
Retrieve &
Rank
Watson News
Language Translation
Entity Extraction
Tone Analyze
r
Concept Tagging
Taxonomy
Text Extraction
Message Resonance
Image Tagging
Face Detection
Answer Generation
Usage Insights
Fusion Q&A
Video Augmentation
Decision Optimization
Knowledge Graph
Risk Stratification
Policy Identification
Emotion Analysis
Decision Support
Criteria Classification
Knowledge Canvas
Easy Adaptati
on
Knowledge Studio Service
Statistical
Dialog
Q&A Qualification
Factoid Pipeline
Case Evaluation
The Watson that competed on Jeopardy! in 2011 comprised what is now a single API—Q&A—built on five underlying technologies.
Since then, Watson has grown to a family of 28 APIs.
By the end of 2016, there will be nearly 50 Watson APIs— with more added every year.
Natural Language Processing
Machine Learning
Question Analysis
Feature Engineering
Ontology
Analysis
Catalog will grow from 28 to 50 APIs (2016)
© 2016 E-HEALTH WEEK AMSTERDAM
American Cancer Society creates a Virtual Cancer Health Advisor with IBM Watson The advisor will anticipate the needs of people with different types of cancers, at different
stages of disease, and at various points in treatment.
It will become increasingly personalized as individuals engage with it, effectively getting
“smarter” each time it is used. The advisor will use ACS's cancer.org 14.000 pages of
detailed information on more than 70 cancer topics.
ACS and IBM also envision incorporating Watson’s voice recognition and natural language
processing technology, enabling users to ask questions and receive audible responses.
© 2016 E-HEALTH WEEK AMSTERDAM
Current diagnosis based on imaging alone or with limited context
Clinical Records
Knowledge
IBM Cognitive Capabilities
Imaging
Providing evidenced based options
IBM combines multiple data sources and cognitive capabilities to assist the physician
+ +
Applying cognitive tools to medical imaging will help assist medical experts
© 2016 E-HEALTH WEEK AMSTERDAM
Raw image Reference Highlighted anatomy Segmented arteries Arterial features
Learn from databases Annotated reference Before registration After registration Anomaly (stenosis)
Anomaly detection involves complex analytics
© 2016 E-HEALTH WEEK AMSTERDAM
Medical information doubles every 5 years. By 2020 it is expected to double every quarter.
80% of the healthcare professionals spends at most 5 hrs/month to keep abreast of his/her domain
80% of the information is unstructured
Only 20% of the knowledge doctors use is evidence based: 1 out of 5 diagnoses are wrong or incomplete.
Healthcare and Lifesciences professionals are suffering from Infobesity
© 2016 E-HEALTH WEEK AMSTERDAM
© 2016 E-HEALTH WEEK AMSTERDAM
Creating a Corpus of Knowledge for Cancer Care Based on > 290 medical journals, > 200 textbooks and > 12 million pages free text
Ingestion of NCCN guidelines for breast cancer and lung cancer
• Roughly 500,000 unique combinations of breast cancer patient attributes.
• Roughly 50,000 unique combinations of lung cancer patient attributes.
Over 600,000 pieces of evidence ingested, from 42 different publications/publishers • The Breast Journal, National Comprehensive Cancer Network (Clinical Practice Guidelines, Drug and Biologics
compendium, et al.), American Journal Of Hematology, Annals Of Neurology, CA: A Cancer Journal For Clinicians, Cancer Journal, Cochrane, EBSCO, Hematological Oncology, Hepatology, International Journal Of Cancer, Journal Of Gene Medicine, Journal of Clinical Oncology, Journal of Oncology Practice, Massachusetts Medical Society Journal Watch, Massachusetts Medical Society New England Journal Of Medicine, Merck, Nephrology, UptoDate, Clinical Lung Cancer, Current Problems in Cancer, Cancer Treatment Reviews, Elsevier's Monographs in Cancer (multiple), Clinical Breast Cancer, European Journal of Cancer, Lung Cancer (the journal).
• YAGO, DBpedia, WordNet
© 2016 E-HEALTH WEEK AMSTERDAM
© 2016 E-HEALTH WEEK AMSTERDAM
Mayo Clinic
Selected Watson to analyze EMRs for Clinical Efficiency and
Effectiveness Program
Department of Veterans Affairs
Selected Watson to analyze EMRs in a demo project
Bumrungrad International Hospital
5 year agreement for Watson for Oncology
Watson for Oncology, trained by Memorial Sloan Kettering
available in clinical use in lung, breast, colon and rectal cancer
Baylor College of Medicine Published results of use with Watson Discovery Advisor – identified 7 targets for P53
activation within weeks
Watson Genomics Advisor Secured 13 Cancer and
academic medical centers for beta testing
MD Anderson Introduced proprietary
solution with Watson for clinical use for Leukemia and
Molecular Targeted Therapies
Mayo Clinic Completed testing with Clinical Trial Matching for lung, breast,
colon and rectal cancer
Manipal Hospitals Selected Watson for Oncology
to identify evidence-based treatment options among
200.000 patients/y
Manipal Hospitals Selected Watson for Oncology
to identify evidence-based treatment options among
200.000 patients per yearpatients/year.
Ongoing Training Partner
Watson Cognitive in de Gezondheidszorg
Metropolitan Health Uses Watson Engagement
Advisor to handle more that 12 million client interactions per
year/r.
© 2016 E-HEALTH WEEK AMSTERDAM
Sabine Linn
© 2016 E-HEALTH WEEK AMSTERDAM
Disclosures
Sabine Linn received institutional unrestricted research grants from:
•Amgen, AstraZeneca, Genentech, Roche, Sanofi
Sabine Linn is named inventor on a BRCAness signature patent
Sabine Linn was an advisory board member for Novartis, Pfizer, Roche, Sanofi,
AstraZeneca
Sabine Linn is a member (pro bono) of the scientific advisory boards of
Cergentis and Philips Health BV
© 2016 E-HEALTH WEEK AMSTERDAM
Where are we heading for?
© 2016 E-HEALTH WEEK AMSTERDAM
Begin with the end in mind (S. Covey)
© 2016 E-HEALTH WEEK AMSTERDAM
Why?
Too much scientific knowledge to keep up with as a clinician
Desirable to have continuous medical education during outpatient clinics
Standardize quality of care
More patients in clinical trials
Population-based datasets for research
© 2016 E-HEALTH WEEK AMSTERDAM
How? Case: test intelligence amplification system (IBM Watson)
Gap analysis – translation to Dutch situation
Assess cost-effectiveness
Arrange governance
Ethical, legal and social aspects
•E.g. Data ownership, intellectual property etc
© 2016 E-HEALTH WEEK AMSTERDAM
Demo IBM Watson Health for Oncology
36
© 2016 E-HEALTH WEEK AMSTERDAM
Added value Free text extraction (saves time)
Check for completeness of diagnostic information
More treatment options
Suggestions for eligible studies
Treatment overview for the patient
Continuous medical education during outpatient clinics
Data mining for self learning decision support system
© 2016 E-HEALTH WEEK AMSTERDAM
Risks?
Who is responsible?
Simplifying medical complexity
IT will never be able to capture all symptom combinations in models
It is a SUPPORT TOOL
Cost-effectiveness?
© 2016 E-HEALTH WEEK AMSTERDAM
Gerrit Meijer