6.7930: Machine Learning for Healthcare
We learned about machine learning applications in healthcare. Our project was on early diagnosis of Alzheimer’s disease, with a focus on explainability.
From the course catalog: Introduces students to machine learning in healthcare, including the nature of clinical data and the use of machine learning for risk stratification, disease progression modeling, precision medicine, diagnosis, subtype discovery, and improving clinical workflows. Topics include causality, interpretability, algorithmic fairness, time-series analysis, graphical models, deep learning and transfer learning. Guest lectures by clinicians from the Boston area, and projects with real clinical data, emphasize subtleties of working with clinical data and translating machine learning into clinical practice.
Cheatsheet
Table of Contents
- Lecture 1: History/Impact of ML in Healthcare, What makes healthcare different?
- Reading 1: AI in Health and Medicine
- Lecture 2: Goals of Healthcare, Extended Example, Cycles of Care
- Reading 2: Machine Learning in Medicine
- Recitation 1: Review, MIMIC-IV, Sci-kit learn tutorial
-
CITI Course: Belmont Report, Human Subjects Research, Institutional Review Boards (IRB), Records-Based Research, Genetic Research, Additional Protections, Health Insurance Portability and Accountability Act (HIPAA), Conflicts of Interest (COI), GPT Responsible Use
- Lecture 3: Sources of & Exploring Clinical Data, Goals of Clinical Data Science, Challenges
- Reading 3: Electronic Health Records (EHR) Safari - Data is Contextual
- Lecture 4: Risk Stratification Intro, Type 2 Diabetes Detection, Supervised Learning, Evaluation Metrics/Interpretability/Generalization
- Reading 4: Sepsis ML Early Warning System (EWS) adoption factors
-
Recitation 2: Bayes’ Theorem, Naïve Bayes, Evaluation Metrics
- Lecture 5: Review, ML for Risk Stratification, Physiological Time-Series
- Reading 5: Transformer network for ECG processing & arrhythmia classification
- Recitation 3: ICU Software, Project Proposals
-
Pset 1: Data Usage Agreement (DUA), Differential Diagnosis, ICD Codes, Patient Notes, Length of Stay Prediction
- Lecture 6: NLP History, LLMs, NLP in Healthcare
- Reading 6: Expert-Level Medical Question Answering with LLMs
- Lecture 7: Generative/Extractive LLM Applications, Limitations
- Reading 7: RAG for Clinical Trial Screening
-
Recitation 4: PyTorch Tutorial, NLP Review/Implementation
- Lecture 8: Project Team, Challenges of AI, Designing Human-AI Systems
- Lecture 9: Survival Analysis Intro, Censored Data (Kaplan-Meier), Covariates (Cox), Evaluation and Extensions
- Reading 9: Deep Cox Mixtures for Survival Regression, Survival Analysis Surveys
-
Recitation 5: Missing Data, Survival Analysis & Implementation
- Lecture 10: Causal Inference Intro, Potential Outcomes Framework, Covariate Adjustment
- Reading 10: Using big data to emulate an RCT, Distillation for data efficient treatment effect estimation
- Meeting 1: Initial Alzheimer classification project details.
- Lecture 11: Covariate Adjustment (Linear/Nonlinear), Matching, Propensity Scores
- Reading 11: ML on a claims dataset for Urinary Tract Infections, LLM causal reasoning
-
Recitation 6: Causal Inference Review
- Lecture 12: Learning Causal Policies, Applications, Dynamic Treatment Regimes, Economics-based Methods for Unobserved Confounders
- Reading 12: Physical activity/survival recommendations using the parametric g-formula, Reinforcement learning guidelines in healthcare
- Meeting 2: More project details.
- Lecture 13: Motivating/Formalizing Dataset Shift, Testing for/Mitigating Dataset Shift
-
Reading 13: Using ML to detect temporal shift in health insurance claims
- Lecture 14: Image Interpretation, Multimodal Info/Modeling
- Reading 14: Transfer learning for vision using natural language supervision
- Lecture 15: Computational Pathology (slides were never posted :/)
- Reading 15: Generative AI copilot for human pathology, foundation models, multimodal cancer prediction
-
Recitation 7: Pneumothorax, Image and Multimodal Training, Saliency Maps
- Lecture 16: Interpretability, Global/Local Interpretability Methods
- Lecture 17: FDA Regulation, HIPAA Lag, Practical Legal Guidance
- Meeting 3: Project check-in, details about the part that I’m doing
- Lecture 18: Humans and AI, How people think about/interact with AI
- Lecture 19: Coalition for Health AI Panel
- Recitation 8: Interpretability, Linear/Non-linear Models
-
Project: Work on our reasoning model ruleset generation idea
- Lecture 20: Privacy, Confidentiality, Security, Differential Privacy, Federated Learning, Privacy Concerns
- Lecture 21: Validation and Acceptance Criteria, Regulation
- Lecture 22: Disease Subtyping, Disease Progression Modeling
-
Meeting 4: Project progress updates
- Lecture 23: Bias in ML, Data Collection, Outcome Definition, Algorithm Development, Post Deployment
- Lecture 24: ML and Deep Learning in Antibiotic Discovery
- Lecture 25: Causal Inference and Genetics
- Recitation 9: Final Exam Review
- Final Cheatsheet