CIV-DG: Conditional Instrumental Variables for Domain Generalization in Medical Imaging
This addresses the issue of robust medical AI by resolving structural confounding in domain generalization, though it is incremental as it builds on existing causal methods.
The paper tackled the problem of cross-site generalizability in medical AI being compromised by selection bias, proposing CIV-DG, a causal framework using Conditional Instrumental Variables to disentangle pathological semantics from scanner artifacts, which significantly outperformed leading baselines on benchmarks like Camelyon17 and Chest X-Ray datasets.
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.