CSC2541HS: Topics in Machine Learning: Machine Learning for Health

Overview

Instructor: Dr. Marzyeh Ghassemi
Time: Thursdays 1 PM - 3 PM
Room: GB 221
TAs: Taylor Killian, Nathan Ng, Haoran Zhang
Markus: https://markus.teach.cs.toronto.edu/csc2541-2020-01
Piazza: https://piazza.com/utoronto.ca/winter2020/csc2541

Course description

This course will give a broad overview of machine learning for health. We begin with an overview of what makes healthcare unique, and then explore machine learning methods for clinical and healthcare applications through recent papers. We discuss the recent successes of of graphical models, deep learning, time-series analysis, and transfer learning in the context of health. We also broadly cover concepts of learning, algorithmic fairness, interpretability, and causality. We emphasize the importance of collaboration between technical and non-technical researchers, and consider the implications of machine learning in healthcare governance and policy. Students will choose and complete a course project, and make project presentations at the end of the course.

This course requires a strong background in linear algebra and probability theory, or strong grades in the machine learning course. Familiarity with Python programming and software engineering is required.

CSC2541 will be capped to students who have an appropriate background this semester. If you are interested in taking the course, please come to our first lecture and fill out the course application https://forms.gle/6VL4mhGM4quSN8qd7

Grading

Course grade will be made up of:

10% Weekly Reflections
20% Problem Sets
10% Paper Presentation
60% Course project

Schedule

Week	Date	Lecture Materials	Assignments
1	Jan 9, 2020	Why is healthcare unique? [Slides] Lecture 1 [Reading] Opportunities in Machine Learning for Healthcare [Reading] Biases in electronic health record data due to processes within the healthcare system: retrospective observational study
2	Jan 16, 2020	Supervised Learning for Classification, Risk Scores and Survival [Slides] Lecture 2 [Slides] Projects Overview [Reading] Recurrent Neural Networks for Multivariate Time Series with Missing Values [Reading] Understanding vasopressor intervention and weaning: risk prediction in a public heterogeneous clinical time series database [Reading] Deep Survival Analysis	Week 2 Reflections
3	Jan 23, 2020	Clinical Time Series Modelling [Slides] Lecture 3 [Reading] Clinical Intervention Prediction and Understanding with Deep Neural Networks [Reading] Towards Understanding ECG Rhythm Classification Using Convolutional Neural Networks and Attention Mappings [Reading] Disease-Atlas: Navigating Disease Trajectories using Deep Learning [Reading] Identifying and mitigating biases in EHR laboratory tests [Reading] Inference and missing data [Reading] Recommender Systems: Missing Data and Statistical Model Estimation [Reading] Semi-supervised Biomedical Translation with Cycle Wasserstein Regression GANs	Week 3 Reflections
4	Jan 30, 2020	Causal inference with Health Data --- Guest Lecture by Dr. Shalmali Joshi (Vector Institute) [Slides] Lecture 4 [Reading] Statistics and Causal Inference [Reading] A targeted real-time early warning score (TREWScore) for septic shock. [Reading] Reliable Decision Support using Counterfactual Models	Week 4 Reflections Problem Set 1 (Feb 3 at 11:59pm)
5	Feb 6, 2020	Fairness, Ethics, and Healthcare [Slides] Lecture 5 [Reading] Learning Fair Representations [Reading] Why is My Classifier Discriminatory? [Reading] Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations	Week 5 Reflections Project proposals (Feb 6 at 5pm)
6	Feb 13, 2020	Deep Learning in Medical Imaging -- Guest Lecture by Dr. Joseph Paul Cohen (MILA) [Slides] Lecture 6 (pdf) [Slides] Lecture 6 (Google Slides) [Reading] Deep Semantic Segmentation of Natural and Medical Images: A Review [Reading] Chester: A Web Delivered Locally Computed Chest X-Ray Disease Prediction System [Reading] Count-ception: Counting by Fully Convolutional Redundant Counting [Reading] Distribution Matching Losses Can Hallucinate Features in Medical Image Translation	Week 6 Reflections
7	Feb 20, 2020	Clinical Reinforcement Learning -- Guest Lecture by Taylor Killian (UofT) [Slides] Lecture 7 [Reading] Evaluating Reinforcement Learning Algorithms in Observational Health Settings [Reading] Guidelines for Reinforcement Learing in healthcare [Reading] Tree-based batch mode reinforcement learning [Reading] Clinical data based optimal STI strategies for HIV: a reinforcement learning approach [Reading] Informing sequential clinical decision-making through reinforcement learning: an empirical study [Reading] Continuous State-Space Models for Optimal Sepsis Treatment - a Deep Reinforcement Learning Approach [Reading] A Reinforcement Learning Approach to Weaning of Mechanical Ventilation in Intensive Care Units	Week 7 Reflections
8	Feb 27, 2020	Clinical NLP and Audio -- Guest Lecture by Dr. Tristan Naumann (Microsoft Research) [Slides] Lecture 8 [Reading] Multi-Label Learning from Medical Plain Text with Convolutional Residual Models [Reading] Natural Language Processing, Electronic Health Records, Clinical Research [Reading] Mayo clinical Text Analysis and Knowledge Extraction System (cTAKES): architecture, component evaluation and applications	Week 8 Reflections Problem Set 2 (Mar 1 at 11:59pm)
9	Mar 5, 2020	Interpretability / Humans-In-The-Loop --- Guest Lecture by Dr. Rajesh Ranganath (NYU) [Slides] Lecture 9 [Reading] Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission [Reading] Rationalizing Neural Predictions [Reading] The Mythos of Model Interpretability	Week 9 Reflections
10	Mar 12, 2020	Disease Progression Modelling and Generalization/Transfer Learning -- Guest Lecture by Irene Chen (MIT) [Slides] Lecture 10 [Reading] Implications of non-stationarity on predictive modeling using EHRs [Reading] Multi-task Prediction of Disease Onsets from Longitudinal Lab Tests [Reading] Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference	Week 10 Reflections
11	Mar 19, 2020	Project Help Session [Reading] Paving the COWpath: Learning and visualizing clinical pathways from electronic health record data [Reading] PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene–disease associations [Reading] Type 2 diabetes genetic loci informed by multi-trait associations point to disease mechanisms and subtypes: A soft clustering analysis	Week 11 Reflections
12	Mar 26, 2020	Course Presentations (Online)
13	Apr 2, 2020	Course Presentations (Online)	Project Report (Apr 10 at 11:59pm)

Weekly Reflections

Each week, students will select one paper from the list above, and complete a series of reflection questions.
There will be ten reflections, each worth 1% of the final grade.
Note that the reading (and associated reflection) are intended to be completed prior to the lecture for that week.
Each weekly reflection will be due at 12 pm on Thursday (i.e. 1 hour prior to the start of lecture) on Markus.
Instructions and Questions

Paper Presentations

The in-class paper presentations are worth 10% of your class grade. Presentations can be done on your own, or in teams of 2, and should be 15 minutes. Plan to cover:

What motivated the work
What problem the paper is trying to solve
The approach used in the paper
The technical or clinical significance of the paper
The secret terrible thing that a casual reader might not notice

Ensure that you have 1-2 reflection questions to spark discussion at the end of the presentation.
Sign up for a presentation date and paper here.

Problem Sets

There will be two problem sets. Problem sets must be done individually.

Problem Set 1 - Clinical Time Series

problem set 1 (last updated: Jan 16)
Due: Feb 3 at 11:59 pm on Markus
Worth 6.7% of the final grade
Please complete the steps required to get access to MIMIC data outlined here. One of the steps will require you to fill out a Data Use Agreement (DUA) where you will be asked for:
- A reference name. Write Marzyeh Ghassemi.
- The general research area for which the data will be used: Write CS 2541 Homework

Problem Set 2 - Fairness in Healthcare

problem set 2
Due: Mar 1 at 11:59 pm on Markus
Worth 13.3% of the final grade

Final Project

Project Deadlines

Project proposals (one per group): Feb 6 at 5pm
Project presentations: Mar 19, Mar 26 and Apr 2 in class
Project report (one per group): April 10 at 11:59pm

Groups

Collaboration: We strongly prefer 3-4 registered student per group, but we do permit groups with fewer students. Doing something related to your research is fine, but your class project should be distinct and you should be able to isolate your contributions to the project from those of any collaborators outside of the class.

Proposal

15% of the project grade
Due Feb 6 at 5pm on Markus.
Should be at most 3 pages, one per group
Clearly state the following:

Problem you wish to tackle
Description of data you plan to use
Proposed approach and methods
Evaluation plan
Timeline
What each student in the group will do