Abstract
Segmenting coronary arteries in Computed Tomography Angiography (CTA) remains challenging due to frequent distal disconnections caused by low contrast, calcifications, and cardiac motion artifacts. These topological errors can compromise downstream clinical applications such as stenosis assessment and flow simulation. We propose GeodVessel,a topology-aware geodesic framework for refining vascular connectivity. Vessel reconnection is formulated as a shortest-path optimization under an interpretable cost integrating anisotropic distance, intensity consistency, and vessel probability. The probability term, derived from a Gaussian Mixture Model or a segmentation model’s softmax output, acts as a soft anatomical prior that suppresses leakage while preserving plausible coronary trajectories. As a model-agnostic post-processing step, GeodVessel requires no retraining and can be applied to segmentation outputs or ground-truth annotations. Evaluated on coronary (ASOCA) and peripheral CTA datasets, it consistently improves topology-aware metrics such as centerline Dice (clDice) and connected component ratio while maintaining overlap accuracy, demonstrating that geometric optimization guided by probabilistic priors provides a robust solution for topology refinement in vascular imaging.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/STACOM2026_023.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/profile?id=~sebastien_goffart1
BibTex
@InProceedings{GofSéb_GeodVessel_MICCAISAT2026,
author = { Goffart, Sébastien AND Föllmer, Bernhard AND Bakhouche, Amel AND Lareyre, Fabien AND Raffort, Juliette AND Delingette, Hervé},
title = { { GeodVessel: A Geodesic Cost-Map Framework for Refining Coronary Vessel Connectivity } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17267},
month = {pending},
page = {pending}
}