Schedule

Jump to Feb. or Mar.

Date	Subject	Readings	Student Presentation	Deadlines
Jan 10, 2019	Lecture 1: Why is healthcare unique? Slides	[Required] Opportunities in Machine Learning for Healthcare [Required] Biases in electronic health record data due to processes within the healthcare system: retrospective observational study [Optional] A $3 Trillion Challenge to Computational Scientists: Transforming Healthcare Delivery
Jan 17, 2019	Lecture 2: Supervised Learning for Classification, Risk Scores and Survival Slides	[Required] Recurrent Neural Networks for Multivariate Time Series with Missing Values [Required] Understanding vasopressor intervention and weaning: risk prediction in a public heterogeneous clinical time series database [Required] Deep Survival Analysis (https://arxiv.org/abs/1608.02158 or https://www.mlforhc.org/s/21.pdf) [Optional] Reproducibility in critical care: a mortality prediction case study [Optional] Unfolding Physiological State: Mortality Modelling in Intensive Care Units [Optional] Predicting intervention onset in the ICU with switching state space models [Optional] Big data in health care: Using analytics to identify and manage high-risk and high-cost patients [Optional] Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available	Grey Kuling on Big Data In Health Care: Using Analytics To Identify And Manage High-Risk And High-Cost Patients	Reflection Questions on Required Papers: Contrast the predictive tasks (e.g., labels in the objective function) in each of the required papers; what are the benefits and drawbacks of each? Would you deploy any of these supervised models?
Jan 24, 2019	Lecture 3: Causal inference with observational data Slides	[Required] Statistics and Causal Inference [Required] A targeted real-time early warning score (TREWScore) for septic shock. [Required] Reliable Decision Support using Counterfactual Models [Optional] What-If Reasoning with Counterfactual Gaussian Processes.	David Burns on Deep Barcodes for Fast Retrieval of Histopathology Scans Adamo Young and Michael Dimmick on GRAM: Graph-based Attention Model for Healthcare Representation Learning Geoff Klein and Matt Hemsley on Deep MR to CT Synthesis using Unpaired Data Phil Boyer on What-If Reasoning with Counterfactual Gaussian Processes	Office Hours for Homework will be held on Wednesday, Jan 23 at 4-6pm in GB 405. Reflection Questions on Required Papers: Describe causal identifiability, and its impact on learning with observational data? What is the difference between a standard Gaussian Process (GP) and a causal GP (CGP) objective function? How does that impact learning?
Jan 31, 2019	Lecture 4: Fairness, Ethics, and Healthcare Slides	[Required] Learning Fair Representations [Required] Why is My Classifier Discriminatory? [Required] One of the following three readings: Inherent Trade-Offs in the Fair Determination of Risk Score Fair prediction with disparate impact: A study of bias in recidivism prediction instruments Equality of Opportunity in Supervised Learning [Optional] Machine bias and accompanying statistical analyses	Matthew MacKay and Amanjit Singh Kainth on Learning Fair Representations Duc Truong on Equality of Opportunity in Supervised Learning Punit Shah on Evaluating Reinforcement Learning Algorithms in Observational Health Settings Devin Singh on Use of GANs in Medical Imaging	Reflection Questions on Required Papers: What are ways that bias can enter into a machine learning model's predictions? Are there reasons that bias in medical data may be harder to detect or distentangle?
Feb 7, 2019	Lecture 5: Clinical Time Series Modelling Slides	[Required] Clinical Intervention Prediction and Understanding with Deep Neural Networks [Required] Towards Understanding ECG Rhythm Classification Using Convolutional Neural Networks and Attention Mappings [Required] Disease-Atlas: Navigating Disease Trajectories using Deep Learning [Required] Identifying and mitigating biases in EHR laboratory tests.	Allen Lee Kelvin Wong and Shun Da Suo on Disease-Atlas: Navigating Disease Trajectories using Deep Learning Seung Eun Yi and Chantal Shaib on Clinical Intervention Prediction and Understanding with Deep Neural Networks Pouria Mashouri on Towards Understanding ECG Rhythm Classification Using Convolutional Neural Networks and Attention Mappings	Homework 1 due at 11:59 PM on MarkUs The code for extracting the data from the MIMIC psql database is here
Feb 14, 2019	Lecture 6: Clinical Imaging (Radiology/Pathology) Slides	[Required] Deep learning assessment of tumor proliferation in breast cancer histological images [Required] CheXNet: Radiologist-level pneumonia detection on chest X-rays with deep learning [Required] Dermatologist-level classification of skin cancer with deep neural networks [Optional] Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs [Optional] Bedside computer vision — Moving artificial intelligence from driver assistance to patient safety [Optional] Systematic analysis of breast cancer morphology uncovers stromal features associated with survival	Sapir Labes on Systematic Analysis of Breast Cancer Morphology Uncovers Stromal Features Associated with Survival Hong Yue Sean Liu and Yan Li on Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs Srinivasan Sivanandan on CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning Vineeth Bhaskara Dermatologist-level classification of skin cancer with deep neural networks	Project proposals due at 5PM here. Reflection Questions on Required Papers: What are some cirtical differences between clinical imaging and images found in standard vision datasets? Are there standard pre-processing or modelling techniques that don't make sense in clinical images?
Feb 21, 2019	Lecture 7: Clinical NLP and Audio Slides	[Required] Multi-Label Learning from Medical Plain Text with Convolutional Residual Models [Required] Natural Language Processing, Electronic Health Records, Clinical Research [Required] Mayo clinical Text Analysis and Knowledge Extraction System (cTAKES): architecture, component evaluation and applications [Optional] Deep Probabilistic Logic: A Unifying Framework for Indirect Supervision	Sumeet Ranka and Vaibhav Saxena on Mayo clinical Text Analysis and Knowledge Extraction System (cTAKES): architecture, component evaluation and applications Yingying Fu and Eric Wan on Multi-Label Learning from Medical Plain Text with Convolutional Residual Models Jienan Yao and Matthew Wong on Natural Language Processing, Electronic Health Records, and Clinical Research	Reflection Questions on Required Papers: How could external knowledge could be added to clinical NLP models and for what tasks might that make sense? How could clinical NLP benefit from using structured EHR data?
Feb 28, 2019	Lecture 8: Clinical Reinforcement Learning Slides	[Required] Informing sequential clinical decision-making through reinforcement learning: an empirical study [Required] A Reinforcement Learning Approach to Weaning of Mechanical Ventilation in Intensive Care Units [Required] Meaningless comparisons lead to false optimism in medical machine learning [Required] Evaluating Reinforcement Learning Algorithms in Observational Health Settings [Optional] Behaviour policy estimation in off-policy policy evaluation: Calibration matters [Optional] The stratified micro-randomized trial design: sample size considerations for testing nested causal effects of time-varying treatments [Optional] Scalable and accurate deep learning with electronic health records	Sindhu C M Gowda on A reduced dimension fMRI shared response model Daniel Dastoor and Joanna Pineda on Scalable and accurate deep learning with electronic health records Brenna Li on Meaningless comparisons lead to false optimism in medical machine learning Sneha Desai on A Reinforcement Learning Approach to Weaning of Mechanical Ventilation in Intensive Care Units	Reflection Questions on Required Papers: What are particular problems with off-policy evaluation in a clinical RL setting? What problems can arise with the action and state spaces chosen in the papers?
Mar 7, 2019	Lecture 9: Missingness and Representations Slides	[Required] Inference and missing data [Required] Recommender Systems: Missing Data and Statistical Model Estimation [Required] Semi-supervised Biomedical Translation with Cycle Wasserstein Regression GANs	Jacob Kelly and Jeevaa Velayutham on Semi-supervised Biomedical Translation with Cycle Wasserstein Regression GANs Pulkit Mathur and Zhen Gou on Recommender Systems: Missing Data and Statistical Model Estimation Yuyang Liu and Chris Meaney on Recurrent Neural Networks for Multivariate Time Series with Missing Values Angeline Yasodhara and Marta Skreta on Why is my Classifier Discriminatory?	Reflection Questions on Required Papers: Briefly discuss the techniques used in these papers to address the missingness/sparsity in electronic medical records. How does the current patient state affect the types of data that are collected in the future? Discuss if/how the models capture and propogate the patients' previous states for predicting clinical events.
Mar 14, 2019	Lecture 10: Generalization and transfer learning Slides	[Required] Implications of non-stationarity on predictive modeling using EHRs [Required] Multi-task Prediction of Disease Onsets from Longitudinal Lab Tests	Alex Lu and Amy Lu on Implications of non-stationarity on predictive modeling using EHRs Fizza Ahmad Sheikh and Daniel Hidru on Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs Sean Segal and Sergio Casas on Multi-task Prediction of Disease Onsets from Longitudinal Lab Tests Zhen Yang and Chenzi Qie on Improving Prediction of Suicide and Accidental Death After Discharge From General Hospitals With Natural Language Processing	Reflection Questions on Required Papers: Does multi-task learning inherently create more generalizable models? Are there conditions for which non-stationarity could be ignored in model training and deployment?
Mar 21, 2019	Lecture 11: Interpretability / Humans-In-The-Loop / Policies and Politics	[Required] Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission [Required] Rationalizing Neural Predictions. [Required] The Mythos of Model Interpretability. [Optional] Differential Privacy: A Survey of Results [Optional] Learning Cost-Effective and Interpretable Treatment Regimes. [Optional] Why policymakers should care about “big data” in healthcare	Pouria Mashouri on Towards Understanding ECG Rhythm Classification Using Convolutional Neural Networks and Attention Mappings Siva Manivasagam and Min Bai on Rationalizing Neural Predictions Shagun Gupta and Chun-Hao Chang on Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission Angad Kalra on Why policymakers should care about “big data” in healthcare	Reflection Questions on Required Papers: What are possible ways that a model could be verified as interpretable? Should machine learning models deployed in healthcare settings be held to a higher standard than other application areas? What concerns should policy makers have about deploying models to minimize hospital costs?
Mar 28, 2019	Course Presentations
April 4, 2019	Course Presentations
April 11, 2019	Projects Due			Project report due 11:59PM

Machine Learning for Health

Schedule