Automatic Discovery of Disease Subgroups by Contrasting with Healthy Controls

TL;DR

This paper introduces Deep UCSL, a contrastive learning method that identifies disease-specific subgroups by contrasting patient data with healthy controls, improving interpretability and homogeneity.

Contribution

The paper develops a novel contrastive subgroup discovery framework using deep learning and EM optimization, explicitly ignoring shared variability with healthy controls.

Findings

Improves subgroup quality over previous methods.

Demonstrates effectiveness on medical imaging datasets.

Provides a publicly available codebase.

Abstract

In biomedical Subgroup Discovery, practitioners are interested in discovering interpretable and homogeneous subgroups within a group of patients. In this paper, assuming that healthy subjects (i.e., controls) share common but irrelevant factors of variation with the patients, we motivate and develop a Contrastive Subgroup Discovery method, entitled Deep UCSL. By contrasting patients with controls, Deep UCSL identifies subgroups driven solely by pathological factors, ignoring common variability shared with healthy subjects. Our framework employs a deep feature extractor to learn a discriminative representation space. Mathematically, we derive a novel loss based on the conditional joint likelihood of latent clusters and patient/control labels, optimized via an Expectation-Maximization strategy alternating between subgroup inference and feature encoder updates. A regularization term further encourages representations to capture disease-specific variability while ignoring variability shared with controls. Compared to previous related works, our approach quantitatively improves the quality of the estimated subgroups, as demonstrated on a MNIST example and four distinct real medical imaging datasets. Code and datasets are available at: https://github.com/rlouiset/deep_ucsl.