Do Segmentation Models Understand Vascular Structure? A Blob-Based XAI Framework

2025-04-11

AI Summary

Medical image segmentation models suffer from limited clinical trustworthiness due to their black-box nature, particularly in vascular segmentation—where it remains unclear whether models leverage global anatomical structures (e.g., connectivity, branching topology). To address this, we propose a novel interpretability framework integrating vessel-graph-guided keypoint localization with multi-scale 3D blob analysis, augmented by gradient-based attribution and a customized blob detector to systematically evaluate models’ perception of global vascular structure. Experiments on the IRCAD and Bullitt datasets reveal highly localized attributions, indicating strong reliance on local texture rather than global topological features such as vessel thickness, tubularity, or connectivity—demonstrating a lack of anatomically consistent reasoning. This work is the first to expose fundamental structural limitations of mainstream 3D vascular segmentation models and establishes a reproducible, topology-aware evaluation paradigm for trustworthy medical AI.

Technology Category

• Computer Vision: Segmentation
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Deep learning models have achieved impressive performance in medical image segmentation, yet their black-box nature limits clinical adoption. In vascular applications, trustworthy segmentation should rely on both local image cues and global anatomical structures, such as vessel connectivity or branching. However, the extent to which models leverage such global context remains unclear. We present a novel explainability pipeline for 3D vessel segmentation, combining gradient-based attribution with graph-guided point selection and a blob-based analysis of Saliency maps. Using vascular graphs extracted from ground truth, we define anatomically meaningful points of interest (POIs) and assess the contribution of input voxels via Saliency maps. These are analyzed at both global and local scales using a custom blob detector. Applied to IRCAD and Bullitt datasets, our analysis shows that model decisions are dominated by highly localized attribution blobs centered near POIs. Attribution features show little correlation with vessel-level properties such as thickness, tubularity, or connectivity -- suggesting limited use of global anatomical reasoning. Our results underline the importance of structured explainability tools and highlight the current limitations of segmentation models in capturing global vascular context.

Problem

• Assessing if segmentation models understand global vascular anatomy
• Evaluating model reliance on local vs. global vessel features
• Developing XAI tools to analyze 3D vessel segmentation decisions

Innovation

• Blob-based analysis of Saliency maps
• Graph-guided point selection method
• Combines gradient-based attribution with blob detection