SlideShare a Scribd company logo
Medical practice as a
recommender system
Xavier Amatriain
Curai
Healthcare Recsys Workshop
Como, 2017
Outline
● Can we improve medicine?
● What does this have to do with recsys?
● Medical decision support systems
● ML medical decision systems
● Curai
Medicine: what can be improved?
Medical decisions require knowledge & data
● Doctors have partial information
○ Of the patient’s history
○ Of the patient’s symptoms
○ Of medical knowledge
■ Different demographics
■ Latest research findings
■ Not easy to remember all past information
● They also might have cognitive biases
○ They decide mostly based on past experience, but this experience is very
limited
● Incentives of medical industry also play a role
Experts disagree
● Experts disagree among themselves.
○ E.g. oncology experts disagree on the value of colon screening
● Things treated as facts for years end up being wrong.
○ E.g. giving aspirin to reduce fever has been proved to be dangerous
● Psychiatric disorder diagnosis even lower agreement
(Cohen’s Kappa of 0.2 or 0.3 in most cases)
Experts disagree
● Even medical research is questionable
○ “Most research papers are false” - by Dr. Ioannidis at Stanford
Cost of medical errors
● 400k deaths a year can be attributed to medical errors as well as 4M serious
health events
○ This compares to 500k deaths from cancer or 40k from vehicle accidents
● Almost half of those events could be preventable
● 30% or $750B is wasted by the US Healthcare system every year
Growing complexity of knowledge
● Very slow pace for doctors to acquire new knowledge
● Increased amount of complexity
○ E.g. Coordinating specialists treating a patient is very complex and done
using very little technology
● No universally accepted processes and recommendations
to make decisions.
○ Those depend on each institution and practitioner
○ 50% of the recommendations made in guidelines based on expert
opinion, case studies, or standards of care, not systematic studies
No personalization!
● Clinical Practice Guides are not personalized
○ They might prove negative since they fail to take into account interactions
between different diseases (e.g. in older patients)
● Research done on “homogeneous”, healthy subjects
● It is very hard for doctors to “manually” personalize their
“recommendations”
What does this have to do with Recsys?
A word on Precision Medicine
Precision medicine
● According to the National Institutes of
Health (NIH), precision medicine is:
"an emerging approach for disease
treatment and prevention that takes
into account individual variability in
genes, environment, and lifestyle for
each person."
● Goal: predict more accurately treatment and
prevention given a particular disease & group
of people
● In contrast to "one-size-fits-all" approach
where treatment & prevention are developed
for average person
● Term is relatively new, but concept has been
around for many years.
○ E.g. blood transfusion is not given from a
randomly selected donor
Precision medicine
Recsys & Medicine
How to improve medicine?
● More/better personalization
● Make better use of (big) data
● Automate processes through
algorithms and machine learning
● Better user experience both for
patients and doctors
Does this sound familiar?
Medical Decision Support Systems
Medical Diagnosis
● Diagnosis:
○ “a mapping from a patient’s data ( normal and
abnormal history , physical examination , and
laboratory data ) to a nosology of disease states the
process of determining by examination the nature and
circumstances of a diseased condition”
○ “The knowledge of how to “work up” the patient
depends critically on the ability to evoke history,
symptoms, and physical examination findings,
concurrently with the ability to generate diagnostic
hypotheses that suggest how to further refine or
pursue the findings already elicited, or to pursue
completely different additional findings. In addition, this
must be done in a compassionate and cost-effective
manner” (R.A. Miller 1990)
Diagnosis Decision Support Systems
● DDSS
○ A computer - based algorithm that assists a clinician with one or more component steps of the
diagnostic process
○ It involves diverse activities:
■ information gathering
■ pattern recognition
■ problem solving
■ Decision-making
■ judgment under uncertainty
■ Empathy
○ Large amounts of highly organized knowledge are necessary
● Two categories
○ General-domain DDSS
○ Specialized DDSS
Medical Diagnosis as a Recsys
● Input signals
○ Implicit: Data coming from sensors, labs…
○ Explicit: What does the patient tell us
○ Others: Demographic, family history…
● Algorithms
○ Find what worked for “similar” patients in a “similar” situation
● Output
○ Ranked list with “likely” options
○ Need to also worry about other variables such as acuity, explanations....
Knowledge-based approaches to
medical decision systems
Decision/Diagnosis support systems
● They have been developed for decades
● Many early DDSS based on Bayesian
reasoning (60s-70s)
○ Bayesian networks (80s-90s)
○ Neural networks (lately)
● Most of them have been developed
“manually” with doctors looking at research
and manually encoding that into knowledge
bases.
An example: Internist-1/QMR/Vddx
● Internist-1 started in 1971, then turned into Quick Medical Reference (~1987),
and more recently into VDDx (Vanderbilt Differential Diagnosis)
● Started by Jack Myers (University of Pittsburgh, Chairman of the National
Board of Medical Examiners, President of the American College of
Physicians, and Chairman of the American Board of Internal Medicine)
considered (one of) the best clinical diagnostic experts in the US
● The process for adding a disease requires 2-4 weeks of full-time effort and
doctors reading 50 to 250 relevant publications
An example: Internist-1/QMR/Vddx
Data for Medical Decision Systems
Knowledge bases
● Knowledge base construction
○ Build model from existing medical knowledge
○ E.g. using doctors to read on medical literature and encode
“well-established facts”
● Data = Medical Research
○ Ingesting and understanding medical publications can be mostly
automated
Electronic Health Records
● EHR/EMRs include digital information
about patients encounters with doctors
or the health system
● Snomed Clinical Terms
○ Computer processable collection of medical terms
providing codes, terms, synonyms and definitions used in
clinical documentation and reporting.
○ Considered to be the most comprehensive, multilingual
clinical healthcare terminology
○ Primary purpose: encode the meanings that are used in
health information & support effective clinical recording of
data with the aim of improving patient care.
○ It provides the core general terminology for electronic
health records.
○ It includes: clinical findings, symptoms, diagnoses,
procedures, body structures, organisms and other
etiologies, substances, pharmaceuticals, devices and
specimens.
Ontologies
Ontologies
● ICD-10
○ 10th revision of the International Statistical Classification of Diseases and Related Health
Problems (ICD), a medical classification list by the World Health Organization (WHO)
○ It contains codes for diseases, signs and symptoms, abnormal findings, complaints, social
circumstances, and external causes of injury or diseases
○ The code set allows more than 14,400 different codes and permits the tracking of new
diagnoses.
Ontologies
● UMLS (1986)
○ Designed and is maintained by the US National Library of Medicine, updated quarterly, free
○ Compendium of many controlled vocabularies in the biomedical sciences.
○ Provides mapping structure among vocabularies
○ Allows to translate among the various terminology systems
○ It can be interpreted as a thesaurus and ontology of biomedical concepts.
○ It provides tools for natural language processing.
○ Intended for developers of systems in medical informatics.
Combining and aggregating data
● At the end, what you need is a way to process and
aggregate all these different sources of data either
manually (expert systems) or algorithmically (ML)
ML for Medical Decision Systems
Health graphs
NLP
● Understanding what doctors say
● Understanding what patients say
Diagnostic Systems
● Building diagnostic systems from data through ML
Curai
Who are we
● 3 months in
● Team of 8 people as of today
● 3 co-founders with experience in Recsys, product development, business
● 4 engineers (former FAIR, Google, Uber self-driving car, Stanford…)
● 1 full-time doctor (Stanford postdoc in medicine, formerly a software engineer)
● Several advisors (Mostly doctors, and researchers related to the topic)
● (Actively hiring)
What are we doing?
● Combining AI/ML and good product/UX practices to build
a medical tool for patients
● We are leveraging all of the previous techniques
● We are stealth, too soon to say too much about what we
have
● Although… we plan on having our friends & family-ready
prototype in the coming weeks
Challenges
● Algorithmic: e.g. combining expert rule-based and ML
● Data: quality, sparsity, and bias in data
● UX: trustworthiness and engagement of the system,
incentives…
● Legal
● …
It’s about time we overcome all of these.
Conclusions
Conclusions
● Medicine/healthcare is one of the areas in society that can benefit most from
technology in general and algorithmic approaches in particular
● Medicine has many similarities to recommender systems
○ Importance of data (both big and small)
○ Content and collaborative approaches
○ Importance of UI/UX (e.g. user understanding, feedback, and explanations)
○ Evaluation has to be sensitive to overall optimization problem
● Huge opportunity, great mission: looking forward to many advances that will
save millions of lives in the next few years
References
● “Probabilistic diagnosis using a reformulation of the INTERNIST-1/QMR knowledge base” . Shwe et al. 1991.
● “Computer-assisted diagnostic decision support: history, challenges, and possible paths forward” Miller. 2009.
● “Mining Biomedical Ontologies and Data Using RDF Hypergraphs” Liu et al. 2013.
● “Health Recommender Systems: Concepts, Requirements, Technical Basics & Challenges”, Wiesner & Pfeifer, 2014.
● “A ‘Green Button’ For Using Aggregate Patient Data At The Point Of Care” Longhurst et al. 2014.
● “Building the graph of medicine from millions of clinical narratives” Finlayson et al. 2014.
● “Comparison of Physician and Computer Diagnostic Accuracy” Semigran et al. 2016.
● “Identifiable Phenotyping using Constrained Non-Negative Matrix Factorization”. Joshi et al. 2016.
● “Clinical Tagging with Joint Probabilistic Models” . Halpern et al. 2016.
● “Deep Patient: An Unsupervised Representation to Predict the Future of Patients from EHR”. Miotto et al. 2016.
● “Learning a Health Knowledge Graph from Electronic Medical Records” Rotmensch et al. 2017.
● “Clustering Patients with Tensor Decomposition”. Ruffini et al. 2017.
● “Patient Similarity Using Population Statistics and Multiple Kernel Learning”. Conroy et al. 2017.
● “Diagnostic Inferencing via Clinical Concept Extraction with Deep Reinforcement Learning”. Ling et al. 2017.
● “Generating Multi-label Discrete Patient Records using Generative Adversarial Networks” Choi et al. 2017
References
Yes, we’re hiring!

More Related Content

What's hot

02 Machine Learning - Introduction probability
02 Machine Learning - Introduction probability02 Machine Learning - Introduction probability
02 Machine Learning - Introduction probability
Andres Mendez-Vazquez
 
Disease Prediction And Doctor Appointment system
Disease Prediction And Doctor Appointment  systemDisease Prediction And Doctor Appointment  system
Disease Prediction And Doctor Appointment system
KOYELMAJUMDAR1
 
Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018
Grigory Sapunov
 
End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...
End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...
End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...
Universitat Politècnica de Catalunya
 
Heart Disease Prediction Using Data Mining Techniques
Heart Disease Prediction Using Data Mining TechniquesHeart Disease Prediction Using Data Mining Techniques
Heart Disease Prediction Using Data Mining Techniques
IJRES Journal
 
Chest X-ray Pneumonia Classification with Deep Learning
Chest X-ray Pneumonia Classification with Deep LearningChest X-ray Pneumonia Classification with Deep Learning
Chest X-ray Pneumonia Classification with Deep Learning
BaoTramDuong2
 
AI and Deep Learning
AI and Deep LearningAI and Deep Learning
AI and Deep Learning
Manoj Kumar
 
Federated learning in brief
Federated learning in briefFederated learning in brief
Federated learning in brief
Shashi Perera
 
Fairness in Machine Learning and AI
Fairness in Machine Learning and AIFairness in Machine Learning and AI
Fairness in Machine Learning and AI
Seth Grimes
 
Machine Learning and Applications
Machine Learning and ApplicationsMachine Learning and Applications
Machine Learning and Applications
Geeta Arora
 
Quantum neural network
Quantum neural networkQuantum neural network
Quantum neural network
surat murthy
 
deep learning applications in medical image analysis brain tumor
deep learning applications in medical image analysis brain tumordeep learning applications in medical image analysis brain tumor
deep learning applications in medical image analysis brain tumor
Venkat Projects
 
Fedarated learning
Fedarated learningFedarated learning
Fedarated learning
VaishakhKP1
 
Applications in Machine Learning
Applications in Machine LearningApplications in Machine Learning
Applications in Machine Learning
Joel Graff
 
Computer vision
Computer vision Computer vision
Computer vision
Dmitry Ryabokon
 
Explainable AI in Healthcare
Explainable AI in HealthcareExplainable AI in Healthcare
Explainable AI in Healthcare
vonaurum
 
AI and ML in Cybersecurity
AI and ML in CybersecurityAI and ML in Cybersecurity
AI and ML in Cybersecurity
Forcepoint LLC
 
Fairness and Bias in Machine Learning
Fairness and Bias in Machine LearningFairness and Bias in Machine Learning
Fairness and Bias in Machine Learning
Surya Dutta
 
Federated Learning
Federated LearningFederated Learning
Explainable AI
Explainable AIExplainable AI
Explainable AI
Wagston Staehler
 

What's hot (20)

02 Machine Learning - Introduction probability
02 Machine Learning - Introduction probability02 Machine Learning - Introduction probability
02 Machine Learning - Introduction probability
 
Disease Prediction And Doctor Appointment system
Disease Prediction And Doctor Appointment  systemDisease Prediction And Doctor Appointment  system
Disease Prediction And Doctor Appointment system
 
Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018Deep Learning: Application Landscape - March 2018
Deep Learning: Application Landscape - March 2018
 
End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...
End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...
End-to-end Speech Recognition with Recurrent Neural Networks (D3L6 Deep Learn...
 
Heart Disease Prediction Using Data Mining Techniques
Heart Disease Prediction Using Data Mining TechniquesHeart Disease Prediction Using Data Mining Techniques
Heart Disease Prediction Using Data Mining Techniques
 
Chest X-ray Pneumonia Classification with Deep Learning
Chest X-ray Pneumonia Classification with Deep LearningChest X-ray Pneumonia Classification with Deep Learning
Chest X-ray Pneumonia Classification with Deep Learning
 
AI and Deep Learning
AI and Deep LearningAI and Deep Learning
AI and Deep Learning
 
Federated learning in brief
Federated learning in briefFederated learning in brief
Federated learning in brief
 
Fairness in Machine Learning and AI
Fairness in Machine Learning and AIFairness in Machine Learning and AI
Fairness in Machine Learning and AI
 
Machine Learning and Applications
Machine Learning and ApplicationsMachine Learning and Applications
Machine Learning and Applications
 
Quantum neural network
Quantum neural networkQuantum neural network
Quantum neural network
 
deep learning applications in medical image analysis brain tumor
deep learning applications in medical image analysis brain tumordeep learning applications in medical image analysis brain tumor
deep learning applications in medical image analysis brain tumor
 
Fedarated learning
Fedarated learningFedarated learning
Fedarated learning
 
Applications in Machine Learning
Applications in Machine LearningApplications in Machine Learning
Applications in Machine Learning
 
Computer vision
Computer vision Computer vision
Computer vision
 
Explainable AI in Healthcare
Explainable AI in HealthcareExplainable AI in Healthcare
Explainable AI in Healthcare
 
AI and ML in Cybersecurity
AI and ML in CybersecurityAI and ML in Cybersecurity
AI and ML in Cybersecurity
 
Fairness and Bias in Machine Learning
Fairness and Bias in Machine LearningFairness and Bias in Machine Learning
Fairness and Bias in Machine Learning
 
Federated Learning
Federated LearningFederated Learning
Federated Learning
 
Explainable AI
Explainable AIExplainable AI
Explainable AI
 

Similar to Medical advice as a Recommender System

Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017
Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017
Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017
MLconf
 
ML to cure the world
ML to cure the worldML to cure the world
ML to cure the world
Xavier Amatriain
 
rti_innovation_brief_evidence_synthesis
rti_innovation_brief_evidence_synthesisrti_innovation_brief_evidence_synthesis
rti_innovation_brief_evidence_synthesis
Anupa Bir
 
Health Care Processes and Decision Making_Lecture 5_slides
Health Care Processes and Decision Making_Lecture 5_slidesHealth Care Processes and Decision Making_Lecture 5_slides
Health Care Processes and Decision Making_Lecture 5_slides
CMDLearning
 
Clinical decision support system
Clinical decision support systemClinical decision support system
Clinical decision support system
DrSiddharthSingh5
 
The Many Lives of Data
The Many Lives of DataThe Many Lives of Data
The Many Lives of Data
ljmcneill33
 
Health Care Processes and Decision Making_lecture 1_slides
Health Care Processes and Decision Making_lecture 1_slidesHealth Care Processes and Decision Making_lecture 1_slides
Health Care Processes and Decision Making_lecture 1_slides
CMDLearning
 
Krithara meetup may18_final (1)
Krithara meetup may18_final (1)Krithara meetup may18_final (1)
Krithara meetup may18_final (1)
Dataconomy Media
 
Drug Information.
Drug Information.Drug Information.
Drug Information.
varshawadnere
 
Precision and Participatory Medicine - MEDINFO 2015 Panel on big data
Precision and Participatory Medicine - MEDINFO 2015 Panel on big dataPrecision and Participatory Medicine - MEDINFO 2015 Panel on big data
Precision and Participatory Medicine - MEDINFO 2015 Panel on big data
Health and Biomedical Informatics Centre @ The University of Melbourne
 
Evidence-based medicine in clinical pharmacy
Evidence-based medicine in clinical pharmacyEvidence-based medicine in clinical pharmacy
Evidence-based medicine in clinical pharmacy
MohammedUsman79
 
JPND Research Strategy presentation T. Gasser
JPND Research Strategy presentation  T. GasserJPND Research Strategy presentation  T. Gasser
JPND Research Strategy presentation T. Gasser
jpndresearch
 
Evidence based Orthopedics
Evidence based OrthopedicsEvidence based Orthopedics
Evidence based Orthopedics
Abdulla Kamal
 
Towards online universal quality healthcare through AI
Towards online universal quality healthcare through AITowards online universal quality healthcare through AI
Towards online universal quality healthcare through AI
Xavier Amatriain
 
Healthcare analytics
Healthcare analytics Healthcare analytics
Healthcare analytics
Arun K
 
Structure of medical data
Structure of medical dataStructure of medical data
Structure of medical data
joseangelruiz21
 
Enabling Evidence Based Medicine
Enabling Evidence Based MedicineEnabling Evidence Based Medicine
Enabling Evidence Based Medicine
Athula Herath
 
Learning to speak medicine
Learning to speak medicineLearning to speak medicine
Learning to speak medicine
Xavier Amatriain
 
The Learning Health System: Thinking and Acting Across Scales
The Learning Health System: Thinking and Acting Across ScalesThe Learning Health System: Thinking and Acting Across Scales
The Learning Health System: Thinking and Acting Across Scales
Philip Payne
 
Health Care Processes and Decision Making_Lecture 4_slides
Health Care Processes and Decision Making_Lecture 4_slidesHealth Care Processes and Decision Making_Lecture 4_slides
Health Care Processes and Decision Making_Lecture 4_slides
CMDLearning
 

Similar to Medical advice as a Recommender System (20)

Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017
Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017
Xavier Amatriain, Cofounder & CTO, Curai at MLconf SF 2017
 
ML to cure the world
ML to cure the worldML to cure the world
ML to cure the world
 
rti_innovation_brief_evidence_synthesis
rti_innovation_brief_evidence_synthesisrti_innovation_brief_evidence_synthesis
rti_innovation_brief_evidence_synthesis
 
Health Care Processes and Decision Making_Lecture 5_slides
Health Care Processes and Decision Making_Lecture 5_slidesHealth Care Processes and Decision Making_Lecture 5_slides
Health Care Processes and Decision Making_Lecture 5_slides
 
Clinical decision support system
Clinical decision support systemClinical decision support system
Clinical decision support system
 
The Many Lives of Data
The Many Lives of DataThe Many Lives of Data
The Many Lives of Data
 
Health Care Processes and Decision Making_lecture 1_slides
Health Care Processes and Decision Making_lecture 1_slidesHealth Care Processes and Decision Making_lecture 1_slides
Health Care Processes and Decision Making_lecture 1_slides
 
Krithara meetup may18_final (1)
Krithara meetup may18_final (1)Krithara meetup may18_final (1)
Krithara meetup may18_final (1)
 
Drug Information.
Drug Information.Drug Information.
Drug Information.
 
Precision and Participatory Medicine - MEDINFO 2015 Panel on big data
Precision and Participatory Medicine - MEDINFO 2015 Panel on big dataPrecision and Participatory Medicine - MEDINFO 2015 Panel on big data
Precision and Participatory Medicine - MEDINFO 2015 Panel on big data
 
Evidence-based medicine in clinical pharmacy
Evidence-based medicine in clinical pharmacyEvidence-based medicine in clinical pharmacy
Evidence-based medicine in clinical pharmacy
 
JPND Research Strategy presentation T. Gasser
JPND Research Strategy presentation  T. GasserJPND Research Strategy presentation  T. Gasser
JPND Research Strategy presentation T. Gasser
 
Evidence based Orthopedics
Evidence based OrthopedicsEvidence based Orthopedics
Evidence based Orthopedics
 
Towards online universal quality healthcare through AI
Towards online universal quality healthcare through AITowards online universal quality healthcare through AI
Towards online universal quality healthcare through AI
 
Healthcare analytics
Healthcare analytics Healthcare analytics
Healthcare analytics
 
Structure of medical data
Structure of medical dataStructure of medical data
Structure of medical data
 
Enabling Evidence Based Medicine
Enabling Evidence Based MedicineEnabling Evidence Based Medicine
Enabling Evidence Based Medicine
 
Learning to speak medicine
Learning to speak medicineLearning to speak medicine
Learning to speak medicine
 
The Learning Health System: Thinking and Acting Across Scales
The Learning Health System: Thinking and Acting Across ScalesThe Learning Health System: Thinking and Acting Across Scales
The Learning Health System: Thinking and Acting Across Scales
 
Health Care Processes and Decision Making_Lecture 4_slides
Health Care Processes and Decision Making_Lecture 4_slidesHealth Care Processes and Decision Making_Lecture 4_slides
Health Care Processes and Decision Making_Lecture 4_slides
 

More from Xavier Amatriain

Data/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealthData/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealth
Xavier Amatriain
 
AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19
Xavier Amatriain
 
AI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 updateAI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 update
Xavier Amatriain
 
AI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approachAI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approach
Xavier Amatriain
 
Lessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systemsLessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systems
Xavier Amatriain
 
AI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for EveryoneAI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for Everyone
Xavier Amatriain
 
From one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategyFrom one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategy
Xavier Amatriain
 
Recommender Systems In Industry
Recommender Systems In IndustryRecommender Systems In Industry
Recommender Systems In Industry
Xavier Amatriain
 
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Xavier Amatriain
 
Past present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry PerspectivePast present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry Perspective
Xavier Amatriain
 
Staying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning WorldStaying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning World
Xavier Amatriain
 
Machine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleMachine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora Example
Xavier Amatriain
 
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systemsBIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
Xavier Amatriain
 
Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
Strata 2016 -  Lessons Learned from building real-life Machine Learning SystemsStrata 2016 -  Lessons Learned from building real-life Machine Learning Systems
Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
Xavier Amatriain
 
Past, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspectivePast, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspective
Xavier Amatriain
 
Barcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons LearnedBarcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons Learned
Xavier Amatriain
 
10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf
Xavier Amatriain
 
10 more lessons learned from building Machine Learning systems
10 more lessons learned from building Machine Learning systems10 more lessons learned from building Machine Learning systems
10 more lessons learned from building Machine Learning systems
Xavier Amatriain
 
Machine Learning to Grow the World's Knowledge
Machine Learning to Grow  the World's KnowledgeMachine Learning to Grow  the World's Knowledge
Machine Learning to Grow the World's Knowledge
Xavier Amatriain
 
MLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@QuoraMLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@Quora
Xavier Amatriain
 

More from Xavier Amatriain (20)

Data/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealthData/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealth
 
AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19
 
AI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 updateAI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 update
 
AI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approachAI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approach
 
Lessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systemsLessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systems
 
AI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for EveryoneAI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for Everyone
 
From one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategyFrom one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategy
 
Recommender Systems In Industry
Recommender Systems In IndustryRecommender Systems In Industry
Recommender Systems In Industry
 
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
 
Past present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry PerspectivePast present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry Perspective
 
Staying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning WorldStaying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning World
 
Machine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleMachine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora Example
 
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systemsBIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
 
Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
Strata 2016 -  Lessons Learned from building real-life Machine Learning SystemsStrata 2016 -  Lessons Learned from building real-life Machine Learning Systems
Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
 
Past, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspectivePast, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspective
 
Barcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons LearnedBarcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons Learned
 
10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf
 
10 more lessons learned from building Machine Learning systems
10 more lessons learned from building Machine Learning systems10 more lessons learned from building Machine Learning systems
10 more lessons learned from building Machine Learning systems
 
Machine Learning to Grow the World's Knowledge
Machine Learning to Grow  the World's KnowledgeMachine Learning to Grow  the World's Knowledge
Machine Learning to Grow the World's Knowledge
 
MLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@QuoraMLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@Quora
 

Recently uploaded

"$10 thousand per minute of downtime: architecture, queues, streaming and fin...
"$10 thousand per minute of downtime: architecture, queues, streaming and fin..."$10 thousand per minute of downtime: architecture, queues, streaming and fin...
"$10 thousand per minute of downtime: architecture, queues, streaming and fin...
Fwdays
 
Leveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and StandardsLeveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and Standards
Neo4j
 
LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...
LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...
LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...
DanBrown980551
 
"Choosing proper type of scaling", Olena Syrota
"Choosing proper type of scaling", Olena Syrota"Choosing proper type of scaling", Olena Syrota
"Choosing proper type of scaling", Olena Syrota
Fwdays
 
A Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's ArchitectureA Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's Architecture
ScyllaDB
 
Christine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptxChristine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptx
christinelarrosa
 
Introducing BoxLang : A new JVM language for productivity and modularity!
Introducing BoxLang : A new JVM language for productivity and modularity!Introducing BoxLang : A new JVM language for productivity and modularity!
Introducing BoxLang : A new JVM language for productivity and modularity!
Ortus Solutions, Corp
 
What is an RPA CoE? Session 2 – CoE Roles
What is an RPA CoE?  Session 2 – CoE RolesWhat is an RPA CoE?  Session 2 – CoE Roles
What is an RPA CoE? Session 2 – CoE Roles
DianaGray10
 
Session 1 - Intro to Robotic Process Automation.pdf
Session 1 - Intro to Robotic Process Automation.pdfSession 1 - Intro to Robotic Process Automation.pdf
Session 1 - Intro to Robotic Process Automation.pdf
UiPathCommunity
 
From Natural Language to Structured Solr Queries using LLMs
From Natural Language to Structured Solr Queries using LLMsFrom Natural Language to Structured Solr Queries using LLMs
From Natural Language to Structured Solr Queries using LLMs
Sease
 
Demystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through StorytellingDemystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through Storytelling
Enterprise Knowledge
 
"What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w..."What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w...
Fwdays
 
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectors
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectorsConnector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectors
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectors
DianaGray10
 
Christine's Supplier Sourcing Presentaion.pptx
Christine's Supplier Sourcing Presentaion.pptxChristine's Supplier Sourcing Presentaion.pptx
Christine's Supplier Sourcing Presentaion.pptx
christinelarrosa
 
GraphRAG for LifeSciences Hands-On with the Clinical Knowledge Graph
GraphRAG for LifeSciences Hands-On with the Clinical Knowledge GraphGraphRAG for LifeSciences Hands-On with the Clinical Knowledge Graph
GraphRAG for LifeSciences Hands-On with the Clinical Knowledge Graph
Neo4j
 
Northern Engraving | Nameplate Manufacturing Process - 2024
Northern Engraving | Nameplate Manufacturing Process - 2024Northern Engraving | Nameplate Manufacturing Process - 2024
Northern Engraving | Nameplate Manufacturing Process - 2024
Northern Engraving
 
AWS Certified Solutions Architect Associate (SAA-C03)
AWS Certified Solutions Architect Associate (SAA-C03)AWS Certified Solutions Architect Associate (SAA-C03)
AWS Certified Solutions Architect Associate (SAA-C03)
HarpalGohil4
 
Biomedical Knowledge Graphs for Data Scientists and Bioinformaticians
Biomedical Knowledge Graphs for Data Scientists and BioinformaticiansBiomedical Knowledge Graphs for Data Scientists and Bioinformaticians
Biomedical Knowledge Graphs for Data Scientists and Bioinformaticians
Neo4j
 
GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...
GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...
GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...
GlobalLogic Ukraine
 
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
AlexanderRichford
 

Recently uploaded (20)

"$10 thousand per minute of downtime: architecture, queues, streaming and fin...
"$10 thousand per minute of downtime: architecture, queues, streaming and fin..."$10 thousand per minute of downtime: architecture, queues, streaming and fin...
"$10 thousand per minute of downtime: architecture, queues, streaming and fin...
 
Leveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and StandardsLeveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and Standards
 
LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...
LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...
LF Energy Webinar: Carbon Data Specifications: Mechanisms to Improve Data Acc...
 
"Choosing proper type of scaling", Olena Syrota
"Choosing proper type of scaling", Olena Syrota"Choosing proper type of scaling", Olena Syrota
"Choosing proper type of scaling", Olena Syrota
 
A Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's ArchitectureA Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's Architecture
 
Christine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptxChristine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptx
 
Introducing BoxLang : A new JVM language for productivity and modularity!
Introducing BoxLang : A new JVM language for productivity and modularity!Introducing BoxLang : A new JVM language for productivity and modularity!
Introducing BoxLang : A new JVM language for productivity and modularity!
 
What is an RPA CoE? Session 2 – CoE Roles
What is an RPA CoE?  Session 2 – CoE RolesWhat is an RPA CoE?  Session 2 – CoE Roles
What is an RPA CoE? Session 2 – CoE Roles
 
Session 1 - Intro to Robotic Process Automation.pdf
Session 1 - Intro to Robotic Process Automation.pdfSession 1 - Intro to Robotic Process Automation.pdf
Session 1 - Intro to Robotic Process Automation.pdf
 
From Natural Language to Structured Solr Queries using LLMs
From Natural Language to Structured Solr Queries using LLMsFrom Natural Language to Structured Solr Queries using LLMs
From Natural Language to Structured Solr Queries using LLMs
 
Demystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through StorytellingDemystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through Storytelling
 
"What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w..."What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w...
 
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectors
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectorsConnector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectors
Connector Corner: Seamlessly power UiPath Apps, GenAI with prebuilt connectors
 
Christine's Supplier Sourcing Presentaion.pptx
Christine's Supplier Sourcing Presentaion.pptxChristine's Supplier Sourcing Presentaion.pptx
Christine's Supplier Sourcing Presentaion.pptx
 
GraphRAG for LifeSciences Hands-On with the Clinical Knowledge Graph
GraphRAG for LifeSciences Hands-On with the Clinical Knowledge GraphGraphRAG for LifeSciences Hands-On with the Clinical Knowledge Graph
GraphRAG for LifeSciences Hands-On with the Clinical Knowledge Graph
 
Northern Engraving | Nameplate Manufacturing Process - 2024
Northern Engraving | Nameplate Manufacturing Process - 2024Northern Engraving | Nameplate Manufacturing Process - 2024
Northern Engraving | Nameplate Manufacturing Process - 2024
 
AWS Certified Solutions Architect Associate (SAA-C03)
AWS Certified Solutions Architect Associate (SAA-C03)AWS Certified Solutions Architect Associate (SAA-C03)
AWS Certified Solutions Architect Associate (SAA-C03)
 
Biomedical Knowledge Graphs for Data Scientists and Bioinformaticians
Biomedical Knowledge Graphs for Data Scientists and BioinformaticiansBiomedical Knowledge Graphs for Data Scientists and Bioinformaticians
Biomedical Knowledge Graphs for Data Scientists and Bioinformaticians
 
GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...
GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...
GlobalLogic Java Community Webinar #18 “How to Improve Web Application Perfor...
 
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
 

Medical advice as a Recommender System

  • 1. Medical practice as a recommender system Xavier Amatriain Curai Healthcare Recsys Workshop Como, 2017
  • 2. Outline ● Can we improve medicine? ● What does this have to do with recsys? ● Medical decision support systems ● ML medical decision systems ● Curai
  • 3. Medicine: what can be improved?
  • 4. Medical decisions require knowledge & data ● Doctors have partial information ○ Of the patient’s history ○ Of the patient’s symptoms ○ Of medical knowledge ■ Different demographics ■ Latest research findings ■ Not easy to remember all past information ● They also might have cognitive biases ○ They decide mostly based on past experience, but this experience is very limited ● Incentives of medical industry also play a role
  • 5. Experts disagree ● Experts disagree among themselves. ○ E.g. oncology experts disagree on the value of colon screening ● Things treated as facts for years end up being wrong. ○ E.g. giving aspirin to reduce fever has been proved to be dangerous ● Psychiatric disorder diagnosis even lower agreement (Cohen’s Kappa of 0.2 or 0.3 in most cases)
  • 6. Experts disagree ● Even medical research is questionable ○ “Most research papers are false” - by Dr. Ioannidis at Stanford
  • 7. Cost of medical errors ● 400k deaths a year can be attributed to medical errors as well as 4M serious health events ○ This compares to 500k deaths from cancer or 40k from vehicle accidents ● Almost half of those events could be preventable ● 30% or $750B is wasted by the US Healthcare system every year
  • 8. Growing complexity of knowledge ● Very slow pace for doctors to acquire new knowledge ● Increased amount of complexity ○ E.g. Coordinating specialists treating a patient is very complex and done using very little technology ● No universally accepted processes and recommendations to make decisions. ○ Those depend on each institution and practitioner ○ 50% of the recommendations made in guidelines based on expert opinion, case studies, or standards of care, not systematic studies
  • 9. No personalization! ● Clinical Practice Guides are not personalized ○ They might prove negative since they fail to take into account interactions between different diseases (e.g. in older patients) ● Research done on “homogeneous”, healthy subjects ● It is very hard for doctors to “manually” personalize their “recommendations”
  • 10. What does this have to do with Recsys?
  • 11. A word on Precision Medicine
  • 12. Precision medicine ● According to the National Institutes of Health (NIH), precision medicine is: "an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person."
  • 13. ● Goal: predict more accurately treatment and prevention given a particular disease & group of people ● In contrast to "one-size-fits-all" approach where treatment & prevention are developed for average person ● Term is relatively new, but concept has been around for many years. ○ E.g. blood transfusion is not given from a randomly selected donor Precision medicine
  • 15. How to improve medicine? ● More/better personalization ● Make better use of (big) data ● Automate processes through algorithms and machine learning ● Better user experience both for patients and doctors Does this sound familiar?
  • 17. Medical Diagnosis ● Diagnosis: ○ “a mapping from a patient’s data ( normal and abnormal history , physical examination , and laboratory data ) to a nosology of disease states the process of determining by examination the nature and circumstances of a diseased condition” ○ “The knowledge of how to “work up” the patient depends critically on the ability to evoke history, symptoms, and physical examination findings, concurrently with the ability to generate diagnostic hypotheses that suggest how to further refine or pursue the findings already elicited, or to pursue completely different additional findings. In addition, this must be done in a compassionate and cost-effective manner” (R.A. Miller 1990)
  • 18. Diagnosis Decision Support Systems ● DDSS ○ A computer - based algorithm that assists a clinician with one or more component steps of the diagnostic process ○ It involves diverse activities: ■ information gathering ■ pattern recognition ■ problem solving ■ Decision-making ■ judgment under uncertainty ■ Empathy ○ Large amounts of highly organized knowledge are necessary ● Two categories ○ General-domain DDSS ○ Specialized DDSS
  • 19. Medical Diagnosis as a Recsys ● Input signals ○ Implicit: Data coming from sensors, labs… ○ Explicit: What does the patient tell us ○ Others: Demographic, family history… ● Algorithms ○ Find what worked for “similar” patients in a “similar” situation ● Output ○ Ranked list with “likely” options ○ Need to also worry about other variables such as acuity, explanations....
  • 21. Decision/Diagnosis support systems ● They have been developed for decades ● Many early DDSS based on Bayesian reasoning (60s-70s) ○ Bayesian networks (80s-90s) ○ Neural networks (lately) ● Most of them have been developed “manually” with doctors looking at research and manually encoding that into knowledge bases.
  • 22. An example: Internist-1/QMR/Vddx ● Internist-1 started in 1971, then turned into Quick Medical Reference (~1987), and more recently into VDDx (Vanderbilt Differential Diagnosis) ● Started by Jack Myers (University of Pittsburgh, Chairman of the National Board of Medical Examiners, President of the American College of Physicians, and Chairman of the American Board of Internal Medicine) considered (one of) the best clinical diagnostic experts in the US ● The process for adding a disease requires 2-4 weeks of full-time effort and doctors reading 50 to 250 relevant publications
  • 24. Data for Medical Decision Systems
  • 25. Knowledge bases ● Knowledge base construction ○ Build model from existing medical knowledge ○ E.g. using doctors to read on medical literature and encode “well-established facts” ● Data = Medical Research ○ Ingesting and understanding medical publications can be mostly automated
  • 26. Electronic Health Records ● EHR/EMRs include digital information about patients encounters with doctors or the health system
  • 27. ● Snomed Clinical Terms ○ Computer processable collection of medical terms providing codes, terms, synonyms and definitions used in clinical documentation and reporting. ○ Considered to be the most comprehensive, multilingual clinical healthcare terminology ○ Primary purpose: encode the meanings that are used in health information & support effective clinical recording of data with the aim of improving patient care. ○ It provides the core general terminology for electronic health records. ○ It includes: clinical findings, symptoms, diagnoses, procedures, body structures, organisms and other etiologies, substances, pharmaceuticals, devices and specimens. Ontologies
  • 28. Ontologies ● ICD-10 ○ 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD), a medical classification list by the World Health Organization (WHO) ○ It contains codes for diseases, signs and symptoms, abnormal findings, complaints, social circumstances, and external causes of injury or diseases ○ The code set allows more than 14,400 different codes and permits the tracking of new diagnoses.
  • 29. Ontologies ● UMLS (1986) ○ Designed and is maintained by the US National Library of Medicine, updated quarterly, free ○ Compendium of many controlled vocabularies in the biomedical sciences. ○ Provides mapping structure among vocabularies ○ Allows to translate among the various terminology systems ○ It can be interpreted as a thesaurus and ontology of biomedical concepts. ○ It provides tools for natural language processing. ○ Intended for developers of systems in medical informatics.
  • 30. Combining and aggregating data ● At the end, what you need is a way to process and aggregate all these different sources of data either manually (expert systems) or algorithmically (ML)
  • 31. ML for Medical Decision Systems
  • 33. NLP ● Understanding what doctors say ● Understanding what patients say
  • 34. Diagnostic Systems ● Building diagnostic systems from data through ML
  • 35. Curai
  • 36. Who are we ● 3 months in ● Team of 8 people as of today ● 3 co-founders with experience in Recsys, product development, business ● 4 engineers (former FAIR, Google, Uber self-driving car, Stanford…) ● 1 full-time doctor (Stanford postdoc in medicine, formerly a software engineer) ● Several advisors (Mostly doctors, and researchers related to the topic) ● (Actively hiring)
  • 37. What are we doing? ● Combining AI/ML and good product/UX practices to build a medical tool for patients ● We are leveraging all of the previous techniques ● We are stealth, too soon to say too much about what we have ● Although… we plan on having our friends & family-ready prototype in the coming weeks
  • 38. Challenges ● Algorithmic: e.g. combining expert rule-based and ML ● Data: quality, sparsity, and bias in data ● UX: trustworthiness and engagement of the system, incentives… ● Legal ● … It’s about time we overcome all of these.
  • 40. Conclusions ● Medicine/healthcare is one of the areas in society that can benefit most from technology in general and algorithmic approaches in particular ● Medicine has many similarities to recommender systems ○ Importance of data (both big and small) ○ Content and collaborative approaches ○ Importance of UI/UX (e.g. user understanding, feedback, and explanations) ○ Evaluation has to be sensitive to overall optimization problem ● Huge opportunity, great mission: looking forward to many advances that will save millions of lives in the next few years
  • 42. ● “Probabilistic diagnosis using a reformulation of the INTERNIST-1/QMR knowledge base” . Shwe et al. 1991. ● “Computer-assisted diagnostic decision support: history, challenges, and possible paths forward” Miller. 2009. ● “Mining Biomedical Ontologies and Data Using RDF Hypergraphs” Liu et al. 2013. ● “Health Recommender Systems: Concepts, Requirements, Technical Basics & Challenges”, Wiesner & Pfeifer, 2014. ● “A ‘Green Button’ For Using Aggregate Patient Data At The Point Of Care” Longhurst et al. 2014. ● “Building the graph of medicine from millions of clinical narratives” Finlayson et al. 2014. ● “Comparison of Physician and Computer Diagnostic Accuracy” Semigran et al. 2016. ● “Identifiable Phenotyping using Constrained Non-Negative Matrix Factorization”. Joshi et al. 2016. ● “Clinical Tagging with Joint Probabilistic Models” . Halpern et al. 2016. ● “Deep Patient: An Unsupervised Representation to Predict the Future of Patients from EHR”. Miotto et al. 2016. ● “Learning a Health Knowledge Graph from Electronic Medical Records” Rotmensch et al. 2017. ● “Clustering Patients with Tensor Decomposition”. Ruffini et al. 2017. ● “Patient Similarity Using Population Statistics and Multiple Kernel Learning”. Conroy et al. 2017. ● “Diagnostic Inferencing via Clinical Concept Extraction with Deep Reinforcement Learning”. Ling et al. 2017. ● “Generating Multi-label Discrete Patient Records using Generative Adversarial Networks” Choi et al. 2017 References