Trust the Prior (or Not): Uncertainty-Aware Abdominal Aortic Aneurysm Segmentation

2026-06-30

AI Summary

This study addresses the challenges of intraluminal thrombus segmentation in abdominal aortic aneurysms, which include high thrombus heterogeneity, low contrast with surrounding tissues, and domain shift caused by variations in multicenter CTA protocols. The authors propose a patient-specific segmentation framework that integrates discriminative learning with anatomical priors. Specifically, they employ a local anatomical Gaussian mixture model for intensity normalization and introduce an uncertainty-gated anatomical attention module that adaptively modulates the influence of anatomical priors based on voxel-wise confidence. By explicitly decoupling anatomical priors from visual evidence and incorporating an uncertainty-aware mechanism, the method provides reliable guidance in ambiguous regions while suppressing unreliable priors, thereby enhancing both interpretability and generalization. Experiments demonstrate state-of-the-art performance on in-distribution test sets and significantly superior results on multicenter external CTA datasets compared to existing approaches.

Abstract

Robust segmentation of intraluminal thrombus is critical for risk assessment in Abdominal Aortic Aneurysm, yet it remains challenging due to heterogeneous thrombus features and low contrast with surrounding non-enhanced tissues. Domain shifts induced by different Computed Tomography Angiography (CTA) protocols further inhibit multi-center generalization of deep learning models. To address these challenges, we propose a patient-specific framework that integrates discriminative learning with anatomically informed priors. Our approach introduces two key components: (1) a patient-specific intensity normalization based on a Gaussian Mixture Model of local anatomy, and (2) an Uncertainty-Gated Anatomical Attention module that incorporates spatial priors while adaptively modulating their influence according to voxel-wise confidence. This design allows for anatomical guidance in ambiguous regions while suppressing unreliable priors. The proposed method achieves state-of-the-art performance on in-distribution test data and substantially outperforms existing alternatives in generalization to external multi-center CTA data, while remaining interpretable through an explicit separation of visual and anatomical evidence.

Problem

• abdominal aortic aneurysm
• intraluminal thrombus segmentation
• domain shift
• CT angiography
• multi-center generalization

Innovation

• Uncertainty-Gated Anatomical Attention
• Patient-Specific Intensity Normalization
• Anatomically Informed Priors
• Domain Generalization
• Abdominal Aortic Aneurysm Segmentation