Biomedical Physics with Applications to Disease
Interactive modules for data modeling, AI and machine learning
Every applet from the BPAD data-modeling chapter, rebuilt to run entirely in your browser. No R server, no installation, no waiting — change a parameter and the simulation, the model fit and the figures update immediately.
Module clusters
The chapter's applets are grouped into eight pedagogical clusters that follow the arc from study design through deployment.
Foundations & Study Design
Estimands, dependence structure, acquisition shift, leakage, and the design decisions that determine whether a model result means anything at all.
5 modules · Sections 1-3
Imaging Features & Preprocessing
From CT volumes and masks to quantitative features: physical units, texture discretization, augmentation, and the physics of image formation.
4 modules · Sections 4-6
Supervised Learning
Classification and regression on the kidney cohort: thresholds, ROC geometry, neighborhood methods, model comparison, and decision cost.
5 modules · Section 7
Time-to-Event & Longitudinal
Censoring, competing risks, landmark analysis, tumor growth kinetics, and joint models for repeated biomarker measurements.
6 modules · Section 8
Unsupervised Learning & Segmentation
Phenotyping, cluster stability, intensity and spectral segmentation, and the metrics used to judge a delineation.
6 modules · Section 9
Deep Learning & Image-Native AI
Capacity and overfitting, convolution kernels as physical operators, receptive fields, parameter budgets, and multimodal fusion.
4 modules · Section 10
Evaluation, Calibration & Uncertainty
Discrimination versus calibration, decision curves, conformal prediction intervals, and reproducibility auditing.
4 modules · Section 11
Translation, Fairness & Deployment
Subgroup fairness, counterfactual explanation, shortcut learning, differential privacy, human-AI teaming, and model documentation.
6 modules · Sections 12-13
Reproducible by construction
Every module exposes its random seed. Two people who set the same seed see identical numbers, which makes results quotable in a lecture or a problem set.
Physics-aware simulations
Slice thickness, reconstruction kernels, partial-volume averaging and noise are modeled explicitly rather than hand-waved, so the statistical consequences follow from the imaging physics.
Documentation alongside the demo
Each module carries the derivation, the learning objectives, discussion prompts and suggested classroom use next to the interactive figure.