Towards Fixing Vessel Segmentation Breakage: A Systematic Review

Area of Science:

• Medical Imaging
• Computational Anatomy
• Vascular Imaging

Background:

• Accurate arterial tree reconstruction from computed tomography angiography (CTA) is crucial for clinical applications.
• Vessel discontinuity, where thin branches appear disconnected, is a significant challenge in CTA segmentation.
• This problem arises from factors like bifurcations, noise, and imaging artifacts.

Purpose of the Study:

• To systematically review computational methods for addressing vessel discontinuity in CTA segmentation.
• To identify current gaps in research and methodologies for improving vascular segmentation accuracy.

Main Methods:

• A scoping review following PRISMA-ScR guidelines was performed.
• Searches were conducted on Google Scholar and PubMed for studies from March 2000 to March 2026.
• Eligible studies included peer-reviewed articles and conference proceedings on 3D vascular segmentation from CTA.

Main Results:

• Three generations of computational solutions were identified: geometric/geodesic methods, topology-aware deep learning, and hybrid post-processing frameworks.
• Topology-sensitive metrics (e.g., clDice, Topology Sensitivity) are preferred over standard metrics like Dice coefficient for evaluating clinical utility.
• Hybrid frameworks combining deep learning with geodesic restoration are the current state-of-the-art.

Conclusions:

• Vessel discontinuity remains a critical challenge in vascular CTA segmentation.
• Hybrid post-processing frameworks represent the most advanced solutions.
• Adopting topology-aware metrics is recommended for future studies to better assess clinical relevance.