🤖 AI Summary
This work addresses the limited robustness of 3D brain glioma segmentation models under cross-institutional, scanner, or protocol-induced image context shifts by proposing a counterfactual context auditing framework. It introduces counterfactual representation learning into 3D medical image segmentation for the first time, generating anatomically plausible perturbed samples that preserve lesion structure while systematically altering imaging context. This enables controlled evaluation of model prediction stability under distributional shifts. The framework explicitly uncovers model reliance on non-lesion contextual cues—overcoming the limitations of conventional Dice and HD95 metrics—and reveals context-sensitive failure modes on BraTS and UTSW datasets that standard evaluations miss. Moreover, it achieves strong performance both within and across domains.
📝 Abstract
Medical image segmentation models can achieve strong benchmark performance while remaining sensitive to scanner, protocol, and institutional variation. These context shifts alter image appearance without changing the underlying lesion, allowing models to exploit nuisance cues that Dice and HD95 fail to expose. We present TRACE-Seg3D, a counterfactual context auditing framework for robust 3D medical image segmentation. TRACE-Seg3D preserves lesion-relevant evidence and systematically varies imaging context to quantify prediction stability under controlled context shifts. The framework pairs each segmentation with audit evidence for context sensitivity and anatomical plausibility, enabling case-level reliability assessment beyond overlap-based evaluation. Experiments on BraTS and UTSW glioma segmentation benchmarks demonstrate competitive in-distribution and cross-domain performance. TRACE-Seg3D also exposes context-sensitive failure modes missed by conventional metrics. These results establish counterfactual context auditing as a practical route toward transparent and reliable 3D medical image segmentation under distribution shift. Our code is available at https://github.com/danleneurocom/Counterfactual-Representation-Network.