A higher-order tensor vessel tractography for segmentation of vascular structures

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computer Vision

Background:

• Vessel segmentation is crucial for analyzing vascular structures.
• Branching points in vessels pose challenges for existing segmentation algorithms.
• 3D tensor models with antipodal symmetry are limited in handling asymmetric junctions.

Purpose of the Study:

• To present a new vascular structure segmentation method based on a cylindrical flux-based higher-order tensor (HOT).
• To overcome limitations of 3D HOT models in representing asymmetric vessel junctions.
• To improve the accuracy of vessel segmentation in complex vascular networks.

Main Methods:

• Developed a higher-order tensor (HOT) model embedded in 4D to capture asymmetric junction scenarios.
• Utilized orientation vectors on the unit 3-sphere for 4D tensor construction.
• Employed a seed-based vessel segmentation algorithm using principal directions from the 4D HOT for tractography.

Main Results:

• The 4D HOT model effectively represents asymmetric junctions, unlike 3D models.
• The proposed seed-based segmentation algorithm demonstrated quantitative validation.
• Successful segmentation of synthetic tubular structures and real vascular data (MRA, CTA).

Conclusions:

• The 4D HOT-based method offers a robust solution for vascular structure segmentation, especially at complex bifurcations.
• This technique enhances the analysis of cerebral vasculature and coronary arteries.
• The approach shows promise for improved medical image analysis in angiography.