Causal Representation Learning with Observational Grouping for CXR Classification

TL;DR

This paper proposes a novel causal representation learning method using observational grouping in chest X-ray classification, enhancing model robustness and generalisability across demographic and imaging variations.

Contribution

It introduces an end-to-end framework that leverages grouping to learn identifiable causal representations for medical image classification.

Findings

Improved generalisability across diverse groups

Enhanced robustness to demographic and imaging variations

Effective causal representation learning demonstrated

Abstract

Identifiable causal representation learning seeks to uncover the true causal relationships underlying a data generation process. In medical imaging, this presents opportunities to improve the generalisability and robustness of task-specific latent features. This work introduces the concept of grouping observations to learn identifiable representations for disease classification in chest X-rays via an end-to-end framework. Our experiments demonstrate that these causal representations improve generalisability and robustness across multiple classification tasks when grouping is used to enforce invariance w.r.t race, sex, and imaging views.