Medical AI · Track 2 of 10
Machine learning
Classical machine learning on tabular clinical data: framing the question, building a cohort without leakage, and training, explaining and checking risk models.
For clinicians who want to read ML papers critically and developers building their first clinical model.
By the end of this track you should be able to
- Frame a clinical question as prediction or causal inference, and know why the difference matters
- Build a cohort and train/test split without data leakage
- Train, explain and calibrate logistic-regression, tree-based and survival models
- Judge whether a model would be useful at the bedside with decision-curve analysis
Lessons
- Prediction or causation? Framing the clinical question
Beginner · Clinical · 25 min · Exercise
- Building a cohort and avoiding data leakage
Intermediate · Technical · 35 min · Exercise
- A logistic-regression risk score
Intermediate · Technical · 35 min · Exercise
- Gradient-boosted trees for readmission risk
Intermediate · Technical · 35 min · Exercise
- Survival analysis: time-to-event models
Intermediate · Technical · 35 min · Exercise
- Missing data and class imbalance
Intermediate · Technical · 30 min · Exercise
- Explaining models with SHAP
Intermediate · Technical · 30 min · Exercise
- Calibration and decision-curve analysis
Intermediate · Clinical · 30 min · Exercise
- Unsupervised learning for patient phenotypes
Intermediate · Technical · 30 min · Exercise
Next track: Deep learning & imaging
Educational content only. This material teaches methods; it is not clinical advice. Code, models and examples here are not validated medical devices and must not be used to make decisions about real patients without proper clinical validation, governance and any required regulatory approval. Examples use synthetic or openly licensed de-identified data only. Never paste identifiable patient information into public AI tools. Mentions of models, tools or vendors are for identification, not endorsement.