Topology adaptive vessel network skeleton extraction with novel medialness measuring function

Area of Science:

• Medical Imaging
• Computer Vision
• Computational Anatomy

Background:

• Vessel tree skeleton extraction is crucial for vascular structure segmentation.
• Conventional methods face challenges with adjacent interferences and topological adaptability.

Purpose of the Study:

• To propose a robust, topology-adaptive framework for vessel tree skeleton extraction.
• To overcome limitations of existing methods in handling interferences and adaptability.

Main Methods:

• Utilized Gaussian affinity voting (GAV) for adaptive scale-growing and robust medialness pattern extraction (GAV-MMF).
• Employed a level-set graph with GAV-MMF for interactive initial curve skeleton generation.
• Integrated GAV-MMF with stretching open active contours (SOAC) for curve refinement.
• Implemented topological checks and network reconfiguration for accurate skeleton topology.

Main Results:

• Demonstrated continuity and scalability on synthetic and clinical multi-scale vessel images.
• Achieved acceptable topological adaptability in vessel tree skeleton extraction.
• Validated the robustness against adjacent interferences.

Conclusions:

• The proposed framework offers a robust and topologically adaptive solution for vessel tree skeleton extraction.
• The method enhances accuracy and continuity in vascular structure analysis.
• It shows promise for applications in medical imaging and computational anatomy.