Reproduce “Disentangled representations: towards interpretation of sex determination from hip bone” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kaifeng Zou, Sylvain Faisan, Fabrice Heitz, Marie Epain, Pierre Croisille, Laurent Fanton, Sébastien Valette
A Disentangled Variational Auto-Encoder (DVAE) system determines biological sex from 3D hip bone morphology with 99.59% accuracy. The system generates counterfactual examples that visually reveal sex-specific anatomical features, providing interpretable insights into classification decisions.
Key insights
- Disentangled latent representations can provide semantically meaningful interpretations by generating counterfactual examples that visually demonstrate class-specific morphological changes.
- Unlike traditional saliency maps, this approach reveals what fundamental anatomical differences contribute to classification decisions in a human-understandable manner.
- Explicitly factoring latent space into intrinsic and conditional (e.g., sex-specific) components enables robust interpretation and decision support, even for complex 3D anatomical data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.