A topology-preserving three-stage framework for fully-connected coronary artery extraction

TL;DR

This work tackles the problem of fully connecting coronary artery trees from CTA data, which is hindered by thin distal vessels, tortuous topology, and low contrast. It introduces CorSegRec, a topology-preserving three-stage framework comprising vascular segmentation with a centerline-aware loss, centerline reconnection via a DPC walk guided by distance, probability, and cosine similarity, and implicit neural representation (INR)–based reconstruction to synthesize missing vessels. The approach yields state-of-the-art results on ASOCA and PDSCA datasets, achieving Dice scores of $88.53\%$ and $85.07\%$, with Hausdorff distances of $1.07$ mm and $1.63$ mm, respectively, outperforming existing methods. This framework enhances vascular connectivity and topology fidelity, with potential applications to other tubular structures, albeit with higher computational cost that the authors plan to address in future work.

Abstract

Coronary artery extraction is a crucial prerequisite for computer-aided diagnosis of coronary artery disease. Accurately extracting the complete coronary tree remains challenging due to several factors, including presence of thin distal vessels, tortuous topological structures, and insufficient contrast. These issues often result in over-segmentation and under-segmentation in current segmentation methods. To address these challenges, we propose a topology-preserving three-stage framework for fully-connected coronary artery extraction. This framework includes vessel segmentation, centerline reconnection, and missing vessel reconstruction. First, we introduce a new centerline enhanced loss in the segmentation process. Second, for the broken vessel segments, we further propose a regularized walk algorithm to integrate distance, probabilities predicted by a centerline classifier, and directional cosine similarity, for reconnecting the centerlines. Third, we apply implicit neural representation and implicit modeling, to reconstruct the geometric model of the missing vessels. Experimental results show that our proposed framework outperforms existing methods, achieving Dice scores of 88.53\% and 85.07\%, with Hausdorff Distances (HD) of 1.07mm and 1.63mm on ASOCA and PDSCA datasets, respectively. Code will be available at https://github.com/YH-Qiu/CorSegRec.

Figures (6)

• Figure 1: The reconstruction of the missed vessel models for disconnected branches. (1) The yellow arrow points to the disconnected region. Due to presence of calcified plaque, the corresponding location in the ground truth shows significant luminal narrowing, and our initial segmentation also shows discontinuity at the corresponding location. (2) Two yellow arrows point to two ends of large disconnected segment on the right coronary artery. At the starting end, the vessel appears "unclear" due to presence of soft plaque, while, near the terminal end, the vessel shows mild narrowing.
• Figure 2: Overview of the proposed framework. CorSegRec consists of three stages: vascular segmentation, vascular reconnection, and vascular reconstruction.
• Figure 3: Candidate branch filtering.
• Figure 4: Coronary artery luminal contour extraction model based on Implicit neural representation (INR).
• Figure 5: The process of coronary artery reconstruction based on implicit extrusion surfaces (IES).