XAI-Driven Diagnosis of Generalization Failure in State-Space Cerebrovascular Segmentation Models: A Case Study on Domain Shift Between RSNA and TopCoW Datasets

2025-12-15

AI Summary

State-space models (SSMs) for medical imaging—e.g., UMamba—exhibit severe cross-domain generalization failure, particularly in the RSNA→TopCoW intracranial vascular segmentation task. To diagnose this, we propose a two-stage framework: first, quantitatively characterizing domain shifts in Z-axis resolution and noise distribution between source and target domains; second, introducing Seg-XRes-CAM—a novel interpretability method that applies attention analysis to localize anatomical origins of SSM generalization failure. Results reveal that, on the target domain, model attention maps achieve only 0.101 IoU with ground-truth vessel masks—well below the 0.3 interpretability threshold—while attaining high overlap (IoU ≈ 0.282) with erroneous predictions. This confirms the model learns spurious correlations rather than anatomically grounded features. Our work establishes an explainable diagnostic paradigm for trustworthy medical AI, enabling root-cause analysis of SSM failures in clinical deployment.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The clinical deployment of deep learning models in medical imaging is severely hindered by domain shift. This challenge, where a high-performing model fails catastrophically on external datasets, is a critical barrier to trustworthy AI. Addressing this requires moving beyond simple performance metrics toward deeper understanding, making Explainable AI (XAI) an essential diagnostic tool in medical image analysis. We present a rigorous, two-phase approach to diagnose the generalization failure of state-of-the-art State-Space Models (SSMs), specifically UMamaba, applied to cerebrovascular segmentation. We first established a quantifiable domain gap between our Source (RSNA CTA Aneurysm) and Target (TopCoW Circle of Willis CT) datasets, noting significant differences in Z-resolution and background noise. The model's Dice score subsequently plummeted from 0.8604 (Source) to 0.2902 (Target). In the second phase, which is our core contribution, we utilized Seg-XRes-CAM to diagnose the cause of this failure. We quantified the model's focus by measuring the overlap between its attention maps and the Ground Truth segmentations, and between its attention maps and its own Prediction Mask. Our analysis proves the model failed to generalize because its attention mechanism abandoned true anatomical features in the Target domain. Quantitative metrics confirm the model's focus shifted away from the Ground Truth vessels (IoU~0.101 at 0.3 threshold) while still aligning with its own wrong predictions (IoU~0.282 at 0.3 threshold). This demonstrates the model learned spurious correlations, confirming XAI is a powerful diagnostic tool for identifying dataset bias in emerging architectures.

Problem

• Diagnosing generalization failure in cerebrovascular segmentation models due to domain shift
• Using XAI to identify why attention mechanisms fail on external datasets
• Quantifying model focus shift from anatomical features to spurious correlations

Innovation

• Using XAI to diagnose generalization failure in medical imaging models
• Quantifying domain shift between datasets to explain performance drop
• Applying attention map analysis to identify spurious model correlations