Machine Learning for Healthcare 2018

Stanford University, August 16-18, 2018

2018 Tutorial Sessions

Thursday August 16th, 2018, Li Ka Shing Learning and Knowledge Center

  • Tutorial Session A: ML for Clinicians. LK120. You can check out the curriculum here.

    • 13:30-16:30 Michael C. Hughes, PhD, Harvard University: Applying ML to Multi-Modal Health Data

  • Tutorial Session B: Healthcare for ML Researchers. LK130

    • 13:30 Arnie Milstein, MD; Professor of Medicine & Director of the Clinical Excellence Research Center, Stanford University: How can ML be used to both lower costs and improve the quality of healthcare

    • 14:15 Nigam Shah, PhD, MBBS; Associate Professor of Medicine & Biomedical Data Science, Stanford University: A brief introduction to the US healthcare system

    • 15:00 Christopher Sharp, MD; CMIO at Stanford Healthcare: What data gets produced in a hospital (and where, when and why)

    • 15:45 Albert Chan, MD, MS; Chief Digital Patient Experience, Sutter Health: Last mile technologies and use cases to reach patients outside the clinic

2018 Main Conference Agenda

Friday August 17, 2018, Berg Hall, Li Ka Shing Learning and Knowledge Center
8:00- 8:45    Breakfast & Check-in

8:45 - 9:00    Welcoming Remarks - Nigam Shah

-----Session 1 - Chairs: Finale Doshi-Velez and Jenna Wiens-----
9:00- 9:45    Abraham Verghese, MD, MACP, Stanford University
"Machine Learning, Story and the Two Culture Proposition in Medicine"
To examine the nature of clinical judgment in the age of machine learning and the two cultures of paraclinical data measurement and the patient’s lived experience.
9:45-10:30    Cynthia Dwork, PhD, Harvard University
"Theory for Society: From Problem to Practice"
Latanya Sweeney famously asked, referring to privacy threats in data analysis, "Computer science got us into this mess; can computer science get us out?"  More broadly,  machine learning algorithms codify biases in the training data, exacerbating fairness concerns, and large, one-off, data sets, irreproducible by construction, are used and reused, in query after query and study after study, exacerbating validity concerns.  We survey ongoing efforts in the theoretical computer science community to tackle these problems.  We start with differential privacy, a definition of privacy tailored to the statistical analysis of large datasets and a phalanx of accompanying algorithmic techniques.  We note a surprising application of differential privacy: the Reusable Holdout set.  Finally, we close with a glimpse of the emerging theory of algorithmic fairness.
10:30-11:00    Coffee Break and Discussion

11:00-12:00    Contributed Spotlights- Research Paper Track

12:00-14:00    Posters A & Lunch
-----Session 2 - Chairs: Jim Fackler and Ken Jung-----
14:00-14:45    Russell Greiner, PhD, University of Alberta
"Working with Medical Colleagues to Produce Effective Predictor Systems" [Slides for Greiner's Presentation]
With today’s excitement over the many successful applications of machine learning to medical applications, an increasing number of medical researchers and clinicians are starting collaborations with machine learning researchers.  This is a great opportunity to produce useful results, for a wide variety of important tasks.  There are, however, many subtle issues in designing, and evaluating, these performance systems --  e.g., based on the differences between standard biostatistics vs supervised machine learning, between emulation vs objective optimization, on the focus on the performance task, related to selection bias (especially for prognosis), etc. This presentation provides information that I wish I had known, when I started -- first using simple examples to identify and characterize these distinctions then, where appropriate, suggesting solutions. We hope this information will raise awareness of these varied issues and approaches, which will facilitate many future effective collaborations.

14:45-15:30     Joyce Lee, MD, MPH, University of Michigan
"When perfect algorithms meet Imperfect healthcare systems"
 Of all chronic diseases, diabetes is perhaps the one with the greatest opportunity for reaping the benefits of machine learning, given the role of patient management and the data-intensive nature of the condition.  However, there are a multitude of barriers for realizing this opportunity.  I will describe the current state of diabetes from the patient, provider, and health system perspective, share an example of how a remarkable patient online community decided to solve some of these problems, and describe opportunities for open science collaboration among multidisciplinary stakeholders.
15:30-16:00    Coffee Break and Discussion

16:00-17:00   Contributed Spotlights - Research Paper Track

17:00              Posters B & Dinner  

Saturday  August 18, 2018
8:30-9:00    Breakfast

-----Session 3 - Chairs: Byron Wallace and Rajesh Ranganath----- 
9:00-9:45    Mohammed Saeed, MD, PhD, University of Michigan
"Unlocking Healthcare Data to Support Research in Machine Learning"
Modern healthcare systems produce massive volumes of data. Such large repositories offer rich and diverse data that can be leveraged with advanced machine learning to create knew knowledge and understanding about medicine and healthcare. However, fundamental barriers exist that have restricted the availability and utility of healthcare data to the machine learning community at large. In this talk, I will give an overview of some fundamental challenges in creating such data resources and creating fruitful partnerships with the machine learning community. A case study using the MIMIC Database will be presented, and the lessons learnt will be discussed.

 9:45-10:30   Danielle Belgrave, PhD, Imperial College London and MSR Cambridge
"From Endotype Discovery towards Personalised Healthcare"
Machine learning advances are opening new routes to more precise healthcare, from the discovery of disease subtypes for stratified interventions to the development of personalised interactions supporting self-care between clinic visits. As medicine pivots from treating diagnoses to treating mechanisms the need for more intelligent phenotyping and frequent interaction brings machine learning and AI into the spotlight. In this talk, I will present a flexible framework for endo-phenotype discovery through the application of probabilistic modelling to disambiguate diseases where there are heterogeneous phenomena. This strategy enables us to develop a more personalised approach to healthcare whereby information can be aggregated from multiple sources within a unified modelling framework. The work presented will be motivated within specific clinical contexts.

10:30-11:00   Coffee Break and Discussion

11:00-12:00   Contributed Spotlights - Clinical Abstract Track

12:00-13:30   Lunch & Clinical Track Posters

-----Session 4 - Chair: David Kale----- 
13:30-14:15     Andrew Ng, PhD Stanford University
"Artificial Intelligence: Transforming healthcare access and delivery"
Healthcare is one of the most exciting industries for applied artificial intelligence (AI), allowing for better prevention, detection, diagnosis, and treatment of disease. Increasingly, important progress is derived from collaborations between computer scientists and clinician scientists, translating machine learning performance into measurable impact in clinical practices worldwide. This talk will discuss case studies of these collaborations producing results in cardiology, radiology and palliative care.

14:15-15:30      Panel - Guidelines for the Safe & Meaningful Deployment of ML in Clinical Care

15:30-15:45      Closing Remarks

16:00-17:00    Feedback Discussion Session


2018 Accepted Research Track Papers

1. Prediction of Cardiac Arrest from Physiological Signals in the Pediatric ICU
Sana Tonekaboni (University of Toronto); Mjaye Mazwi (The Hospital For Sick Children), Peter Laussen (The Hospital For Sick Children), Danny Eytan (The Hospital For Sick Children), Robert Greer (The Hospital For Sick Children), Sebastian D. Goodfellow (The Hospital For Sick Children), Andrew Goodwin (The Hospital For Sick Children), Michael Brudno (University of Toronto), and Anna Goldenberg (University of Toronto, Vector Institute)

2. Boosted Trees for Risk Prognosis
Alexis Bellot (University of Oxford); Mihaela van der Schaar (University of California, Los Angeles)

3. Racial Disparities and Mistrust in End-of-Life Care
Willie Boag (MIT); Harini Suresh (MIT); Leo Anthony Celi (MIT); Pete Szolovits (MIT); Marzyeh Ghassemi (University of Toronto)

4. Sequential Pattern Analysis on Neurosurgical Simulation Data
Scott Buffett (National Research Council Canada); Catherine Pagiatakis (National Research Council Canada); Di Jiang (National Research Council Canada)

5. Multi-task multiple kernel learning reveals relevant frequency bands for critical areas localization in focal epilepsy
Vanessa D'Amario (Università degli Studi di Genova); Federico Tomasi (Università degli Studi di Genova); Veronica Tozzo (Universita' degli Studi di Genova); Gabriele Arnulfo (Università degli Studi di Genova); LIno Nobili (Ospedale Niguarda Ca' Granda); Annalisa Barla (Università degli Studi di Genova)

6. Contextual Bandits for Adapting Treatment in a Mouse Model of de Novo Carcinogenesis
Audrey Durand (McGill University); Charis Achilleos (University of Cyprus); Demetris Iacovides (University of Cyprus); Katerina Strati (University of Cyprus); Georgios Mitsis (McGill); Joelle Pineau (McGill)

7. Predicting Smoking Events with a Time-Varying Semi-Parametric Hawkes Process Model
Matthew Engelhard (Duke University); Hongteng Xu (Duke University); Lawrence Carin (Duke University); Jason A Oliver (Duke University); Matt Hallyburton (Duke University); F Joseph McClernon (Duke University)

8. Modeling "Presentness" of Electronic Health Record Data to Improve Patient State Estimation
Jacob Fauber (University of California, Riverside); Christian R Shelton (University of California, Riverside)

9. Learning to Summarize Electronic Health Records Using Cross-Modality Correspondences
Jen J Gong (MIT); John V Guttag (MIT)

10. Towards Understanding ECG Rhythm Classification Using Convolutional Neural Networks and Attention Mappings
Sebastian D Goodfellow (The Hospital For Sick Children); Andrew Goodwin (The Hospital For Sick Children); Robert Greer (The Hospital For Sick Children); Peter C Laussen (The Hospital For Sick Children); Mjaye Mazwi (The Hospital For Sick Children); Danny Eytan (The Hospital For Sick Children);

11. 3D Point Cloud-Based Visual Prediction of ICU Mobility Care Activities
Bingbin Liu* (Stanford University); Michelle Guo* (Stanford University); Edward Chou (Stanford University); Rishab Mehra (Stanford University); Serena Yeung (Stanford University); N. Lance Downing (Stanford University); Francesca Salipur (Stanford University); Jeffrey Jopling (Stanford University); Brandi Campbell; Kayla Deru; William Beninati (Intermountain Healthcare); Arnold Milstein (Stanford Univesity); Li Fei-Fei (Stanford University)
*equal contribution

12. Reproducible Survival Prediction with SEER Cancer Data
Stefan Hegselmann (University of Münster); Leonard Greulich (University of Münster); Julian Varghese (University of Münster); Martin Dugas (University of Münster)

13. Bayesian Trees for Automated Cytometry Data Analysis
Disi Ji (UC Irvine); Eric Nalisnick (UC Irvine); Yu Qian (J. Craig Venter Institute); Richard Scheuermann (J. Craig Venter Institute); Padhraic Smyth (UC Irvine)

14. Disease-Atlas: Navigating Disease Trajectories using Deep Learning
Bryan Lim (University of Oxford); Mihaela van der Schaar (University of Oxford)

15. Automated Treatment Planning in Radiation Therapy using Generative Adversarial Networks
Rafid Mahmood (University of Toronto); Aaron Babier (University of Toronto); Andrea McNiven (Princess Margaret Cancer Centre); Adam Diamant (York University); Timothy Chan (University of Toronto)

16. Representational Learning Approaches for ECG Dynamics to Detect False Arrhythmia Alarms
Eric Lehman (Northeastern University); Rahul Krishnan (MIT); Xiaopeng Zhao (University of Tennessee); Roger Mark (MIT); Liwei H Lehman (MIT)

17. Deep Spine: Automated Lumbar Vertebral Segmentation, Disc-Level Designation, and Spinal Stenosis Grading Using Deep Learning
Jen-Tang Lu (MGH-BWH); Stefano Pedemonte (Massachusetts General Hospital); Bernardo Bizzo (MGH-BWH); Sean Doyle (MGH-BWH); Katherine Andriole (MGH-BWH); Mark Michalski (Partners Healthcare); R. Gilberto Gonzalez (MGH); Stuart Pomerantz (MGH-BWH)

18. Integrating Hypertension Phenotype and Genotype with Hybrid Non-negative Matrix Factorization
Yuan Luo (Northwestern University)

19. Computer Vision-based Descriptive Analytics of Seniors' Daily Activities for Long-term Health Monitoring
Zelun Luo (Stanford University); Jun-Ting Hsieh (Stanford University); Niranjan Balachandar (Stanford University); Serena Yeung (Stanford University); Guido Pusiol (Stanford university); Li-Jia Li (Google); N Lance Downing (Stanford); Arnold Milstein (Stanford Univesity); Li Fei-Fei (Stanford University)

20. Integrating Machine Learning and Optimization Methods for Imaging of Patients with Prostate Cancer
Selin Merdan (University of Michigan); Khurshid R Ghani (University of Michigan); Brian  Denton (University of Michigan)

21. ConvSCCS: Convolutional Self-Controlled Case Series Model for Lagged Adverse Event Detection
Maryan Morel (Ecole Polytechnique); Emmanuel Bacry (Ecole polytechnique); Stephane Gaiffas (Universite Paris Diderot); Agathe Guilloux (Universite d'Evry Val d'Essone / Ecole Polytechnique)

22. Deep Survival Analysis: Nonparametrics and Missingness
Xenia Miscouridou (University of Oxford); Adler Perotte (Columbia); Noémie Elhadad (Columbia); Rajesh Ranganath (NYU)

23. Learning to Exploit Invariances in Clinical Time-Series Data using Sequence Transformer Networks
Jeeheh Oh (University of Michigan); Jiaxuan Wang (University of Michigan); Jenna Wiens (University of Michigan)

24. Learning from the experts: From diagnostic expert systems to machine learning diagnosis models
Murali Ravuri (Curai); Anitha Kannan (Curai); Geoff Tso (Curai); Xavier Amatriain (Curai)

25. Effective Use of Bidirectional Language Modeling for Medical Named Entity Recognition
Devendra S Sachan (Petuum Inc.); Pengtao Xie (Petuum Inc.); Mrinmaya Sachan (CMU); Eric Xing (Petuum Inc.)

26. A Domain Guided CNN Architecture for Predicting Age from Structural Brain Images
Pascal Sturmfels (University of Michigan); Saige Rutherford (University of Michigan); Mike Angstadt (University of Michigan); Mark Peterson (University of Michigan); Chandra Sripada (University of Michigan); Jenna Wiens (University of Michigan)

27. Phenotyping Endometriosis through Mixed Membership Models of Self-Tracking Data
Iñigo Urteaga (Columbia University); Mollie McKillop (Columbia University); Sharon Lipsky-Gorman (Columbia University); Noemie Elhadad (Columbia University)

28. Preference Learning in Assistive Robotics: Observational Repeated Inverse Reinforcement Learning
Bryce Woodworth (UC, San Diego); Francesco Ferrari (UC, San Diego); Teofilo Zosa (UC, San Diego); Laurel Riek (UC, San Diego)

29. Reinforcement Learning with Action-Derived Rewards for Chemotherapy and Clinical Trial Dosing Regimen Selection
Gregory Yauney (MIT); Pratik Shah (MIT)

30. Multi-Label Learning from Medical Plain Text with Convolutional Residual Models
Xinyuan Zhang (Duke University); Ricardo Henao (Duke University); Zhe Gan (Duke University); Yitong Li (Duke University); Lawrence Carin (Duke University)

31. Chronic Disease Prediction Using Medical Notes
Jingshu Liu (NYU); Zachariah Zhang (NYU); Narges Razavian (NYU)

2018 Accepted Clinical Abstracts

Pilot implementation of a patient-centered e-health tool in gynecological cancer care
Saima Ahmed (McGill University); Carmen G. Loiselle (McGill University)

Finding a needle in a haystack: Machine Learning, Visual Analytics and Rare Diseases
Sharon Hensley Alford (IBM Watson Health); Piyush Madan (IBM Research); Italo Buleje (IBM Research); Fang F. Lu (IBM Research); Yanyan Han (IBM Research); Shilpa Mahatma (IBM Research)

Multimodality Machine Learning for Breast Cancer Detection: Synergistic Performance with Upstream Data Fusion of Digital Breast Tomosynthesis and Ultrasound
Emily Ambinder (Johns Hopkins); Paul H Yi (Johns Hopkins); David Porter (Johns Hopkins); William Walton (Johns Hopkins); Ferdinand Hui (Johns Hopkins); Susan Harvey (Johns Hopkins)

Clustering Anesthesiology Case Data for Future Reinforcement Learning Decision Analysis
Michael Burns (University of Michigan); Anik Sinha (University of Michigan); Yuwei Bao (University of Michigan); Sean Meyer (University of Michigan); John Vandervest (University of Michigan); Sachin Kheterpal (University of Michigan)

PYTHIA: Automated Surgical Outcomes Data Pipeline and Prediction Engine
Kristin M Corey (Duke); Sehj Kashyap (Duke); Elizabeth Lorenzi (Duke); Krista Whalen (Duke); Mark Sendak (Duke); Suresh Balu (Duke); Shelley McDonald (Duke); Mitchell  Heflin (Duke); Madhav Swaminathan (Duke); Katherine Heller (Duke); Sandhya Lagoo-Deenadayalan (Duke)

Automatic Title and Abstract Screening in Healthcare Systematic Reviews
Alexander Lee (Institute for Health Metrics and Evaluation); Ashkan Afshin (Institute for Health Metrics and Evaluation); Joseph Salama (Institute for Health Metrics and Evaluation )

Leveraging Deep Learning and Rapid Response Team Nurses to Improve Sepsis Management
Anthony L Lin (Duke); Joseph Futoma (Duke); Armando Bedoya (Duke; Nathan Brajer (Duke); Mark Sendak (Duke); Faraz Yashar (Duke); Marshall Nichols (Duke); Michael Gao (Duke); Meredith Clement (Duke); Katherine Heller (Duke); Cara O'Brien (Duke)

Deep Learning-Based Identification Of Traditional Hip, Knee, and Shoulder Arthroplasty and Application to Alternative Arthroplasty Designs
Tae Kyung Kim (Johns Hopkins); Paul H Yi (Johns Hopkins); Jinchi Wei (Johns Hopkins); Ji Won Shin (Johns Hopkins); Tae Soo Kim (Johns Hopkins); Gregory D. Hager (Johns Hopkins); Haris Sair (Johns Hopkins); Jan Fritz (Johns Hopkins)

Deep Learning Preprocessing in Process-Mining Methods for Analyzing Hospital Care Pathways
Marie-Helene Metzger (AP-HP); Francois Boue (AP-HP)

External Application of Deep Convolutional Neural Networks Trained on Radiographs to Cross-Sectional Imaging in the Spine: Applications for Semantic Labeling
Tae Kyung Kim (Johns Hopkins RAIL); Paul H Yi (Johns Hopkins Radiology Artificial Intelligence Lab (RAIL)); Jinchi Wei (Radiology Artificial Intelligence Lab); Ji Won Shin (Johns Hopkins RAIL); Tae Soo Kim (Johns Hopkins University); Gregory D. Hager (The Johns Hopkins University); Jan Fritz (Johns Hopkins RAIL); Haris Sair (Johns Hopkins RAIL)

Sparsely Sampling Vital Sign Data Limits the Accuracy of Patient State Estimation
Anusha Jegatheeswaran (Sick Kids), Danny Eytan, Azadeh Assadi, Sebastian D. Goodfellow, Andrew J. Goodwin, Robert W. Greer, Mjaye L. Mazwi, Peter C. Laussen

Redesigning Clinical Care using Risk Adjusted Target Based Approach
Rika Ohkuma (Stanford University); Purnima Krishna (Stanford University); Alicia Wilson (Stanford University); Amy Lu (Stanford University)

Patient-Level Prediction in a Learning Healthcare System: A Demonstration of OHDSI
Jenna Reps (Janssen Research and Development); Peter Rijnbeek (Erasmus University Medical Center); Patrick Ryan (Janssen Research and Development)

Bayesian Network Informed Antimicrobial Stewardship in the Safe-ICU Initiative
Tavpritesh Sethi (All India Institute of Medical Sciences New Delhi; Indraprastha Institute of Information Technology Delhi; Stanford University); Shubham Maheshwari (Indraprastha Institute of Information Technology); Anant Mittal (Indraprastha Institute of Information Technology); Rakesh Lodha (All India Institute of Medical Sciences New Delhi)

Generating Actionable Insights: Machine Learning for Causal Inference with Individual-Level Patient Generated Data
Shannon Wongvibulsin (Johns Hopkins)*; Eric J. Daza (Stanford Medicine)

Improving clinical interpretation of extreme gradient boosted ensemble tree models of cardiac data with high proportions of missingness
Anthony Li (Duke NUS Medical School)*; Anders Olof Sahlen (National Heart Center)

Detecting Insertion, Substitution, and Deletion Errors in Radiology Reports Using Sequence-to-Sequence Models
John Zech (Icahn School of Medicine at Mount Sinai); Jessica Forde (Project Jupyter); Joseph Titano (Icahn School of Medicine at Mount Sinai); Deepak Kaji (Icahn School of Medicine at Mount Sinai); Anthony Costa (Icahn School of Medicine at Mount Sinai); Eric Karl Oermann (Icahn School of Medicine at Mount Sinai)