Design of robust vascular tree matching: validation on liver

Area of Science:

• Medical Imaging
• Computer Vision
• Computational Anatomy

Background:

• Accurate registration of hepatic vascular systems is crucial for longitudinal studies and treatment planning.
• Existing methods struggle with topological variations and deformations common in medical imaging.
• CT scans provide detailed anatomical information but require robust registration techniques.

Purpose of the Study:

• To develop an original and efficient tree matching algorithm for intra-patient hepatic vascular system registration.
• To address challenges posed by topological modifications and significant deformations in segmented vascular trees.
• To provide a reliable method for comparing vascular structures from CT scans acquired at different times.

Main Methods:

• Vascular systems are segmented from CT-scan images and modeled as trees.
• An iterative tree matching algorithm is proposed, starting from the root.
• A quality criterion is used to update and maintain the best match solutions.
• The algorithm handles topological changes and deformations robustly.

Main Results:

• The proposed algorithm demonstrates efficiency in matching bifurcations (nodes) and vessels (edges) between vascular trees.
• It shows robustness against segmentation failures and significant anatomical deformations.
• Validation on a large synthetic database confirms its performance across various challenging cases.

Conclusions:

• The developed tree matching algorithm offers an effective solution for hepatic vascular system registration.
• Its robustness makes it suitable for real-world clinical scenarios with image artifacts and patient variability.
• This work advances the field of medical image registration for vascular structures.