ML4H Hub: A Structured Guide to Machine Learning for Healthcare

Inspired by Sik-Ho Tsang and Technion’s 2024 ML4H course, this hub bridges technical AI with clinical reality across EHR, imaging, causality, ethics and more.

0. Clinical Foundations & Evaluation Metrics

Learn how healthcare data are represented and medical ML models are evaluated. These foundations help build useful and trustworthy systems.

0.1 Clinical data systems: [FHIR], [DICOM] 0.2 Medical Statistics: [Foundations], [Diagnostics], [Risk], [Survival], [Causality], [Validation], [Reliability]

1. Overview & Healthcare Trends

A systems view of medical AI: promise, clinical validation, workflow change, patient-facing tools, and safe deployment.

2019: [DL in healthcare], [High-performance medicine]

2. EHR Analytics & Structured Records

Structured health records can reveal early risk signals before diagnosis. EHR analytics helps detect disease earlier using routine clinical data.

2023: [Pancreatic EarlyDetect]

3. Clinical NLP & Medical LLMs

Medical LLMs help process clinical language for question answering, summarization, documentation, and decision support.

2025: [Med-PaLM 2]

4. Audio Processing & Acoustic Biomarkers

Acoustic biomarkers turn sound signals into clinical information, supporting non-invasive monitoring, screening, and diagnosis.

2024: [HeartSound]

5. Temporal & Multi-Dimensional Signals

Temporal signals reveal physiological patterns over time. Models use these sequences to support clinical prediction.

2019: [ECGRhythm]

6. Medical Computer Vision (Imaging)

Medical computer vision analyzes imaging data for classification, detection, and segmentation in clinical workflows.

2015: [U-Net]

7. Survival Analysis & Event Forecasting

Survival models estimate time-to-event risk, helping forecast clinical outcomes for prognosis and treatment planning.

2018: [DeepSurv]

8. Causal Reasoning in Healthcare

Causal reasoning separates correlation from effect, helping make healthcare AI more reliable and clinically meaningful.

2020: [Image Causality]

9. Explainable AI (XAI) & Interpretability

XAI helps explain model decisions, but medical explanations must be reliable enough to support real clinical trust.

[2021]: [False Hope]

10. Multimodal Learning & Data Fusion

Multimodal learning combines clinical data sources to create richer, more context-aware medical predictions.

2022: [HAIM]

11. Human-AI Collaboration (HCI)

Human-AI collaboration studies clinicians and AI working together. It aims for better decisions through oversight and complementary strengths.

2024: [Med Agents] 2025: [1+1>2?]

12. Implementation & Real-World Impact

Clinical AI becomes useful when it fits real workflows, supports clinicians, and improves patient outcomes.

2022: [TREWS]

13. Ethics, Equity & Fairness

Fairness in medical AI means identifying bias, testing across populations, and reducing unequal clinical impact.

2019: [Race Bias]