CIV-DG: Conditional Instrumental Variables for Domain Generalization in Medical Imaging

TL;DR

This paper introduces CIV-DG, a causal framework using Conditional Instrumental Variables to improve domain generalization in medical imaging by disentangling pathology from scanner artifacts, addressing selection bias.

Contribution

CIV-DG is the first method to incorporate conditional instrumental variables with deep learning for robust domain generalization in medical AI.

Findings

CIV-DG outperforms existing baselines on Camelyon17 and Chest X-Ray datasets.

The approach effectively disentangles pathological signals from scanner artifacts.

Experimental results validate the causal framework's robustness in real-world scenarios.

Abstract

Cross-site generalizability in medical AI is fundamentally compromised by selection bias, a structural mechanism where patient demographics (e.g., age, severity) non-randomly dictate hospital assignment. Conventional Domain Generalization (DG) paradigms, which predominantly target image-level distribution shifts, fail to address the resulting spurious correlations between site-specific variations and diagnostic labels. To surmount this identifiability barrier, we propose CIV-DG, a causal framework that leverages Conditional Instrumental Variables to disentangle pathological semantics from scanner-induced artifacts. By relaxing the strict random assignment assumption of standard IV methods, CIV-DG accommodates complex clinical scenarios where hospital selection is endogenously driven by patient demographics. We instantiate this theory via a Deep Generalized Method of Moments (DeepGMM) architecture, employing a conditional critic to minimize moment violations and enforce instrument-error orthogonality within demographic strata. Extensive experiments on the Camelyon17 benchmark and large-scale Chest X-Ray datasets demonstrate that CIV-DG significantly outperforms leading baselines, validating the efficacy of conditional causal mechanisms in resolving structural confounding for robust medical AI.