Bayesian maximal paths for coronary artery segmentation from 3D CT angiograms.

Area of Science:

• Medical Imaging
• Computational Biology
• Radiology

Background:

• Accurate delineation of coronary arteries from 3D CT angiograms (cardiac CTA) is crucial for diagnosing cardiovascular diseases.
• Existing methods may struggle with image quality variations and anatomical anomalies.

Purpose of the Study:

• To develop and evaluate a novel recursive Bayesian model for robust coronary artery delineation.
• To optimize the model using discrete minimal path techniques for efficient computation.

Main Methods:

• A recursive Bayesian model was developed for joint optimization of vascular centerline and radius in 4D space+scale.
• Discrete minimal path techniques were employed for efficient propagation of model realizations.
• A multiscale, oriented gradient flux-based feature was used for local data likelihood, incorporating prior knowledge on radius variations.

Main Results:

• The proposed method achieved an excellent average overlap of 97.5% with expert-delineated ground truth.
• Quantitative evaluation on a large cardiac CTA database demonstrated high robustness to anomalies and poor image quality.
• Comparison with existing techniques using the Rotterdam Coronary Artery Algorithm Evaluation Framework showed superior performance.

Conclusions:

• The recursive Bayesian model with discrete minimal path optimization offers a highly accurate and robust solution for coronary artery delineation in cardiac CTA.
• The method shows significant potential for clinical application due to its efficiency, versatility, and performance in challenging imaging scenarios.