Automated Coronary Artery Tracking with a Voronoi-Based 3D Centerline Extraction Algorithm

Area of Science:

• Medical Imaging
• Computational Anatomy
• Cardiovascular Disease Research

Background:

• Coronary artery disease (CAD) is a major global health concern, necessitating advanced diagnostic tools.
• Medical imaging, particularly coronary artery computed tomography, plays a crucial role in CAD detection.
• Automated extraction of coronary artery features, such as centerlines, can enhance diagnostic accuracy and timeliness.

Purpose of the Study:

• To validate and benchmark a novel automated 3D centerline extraction method for coronary arteries.
• To assess the performance of the 3D divide and conquer Voronoi diagram method using a standardized dataset.

Main Methods:

• Utilized synthetically segmented coronary artery models from the Rotterdam Coronary Artery Algorithm Evaluation Framework (RCAAEF) training dataset.
• Employed a 3D divide and conquer Voronoi diagram approach for automated centerline tracking.
• Evaluated performance using standard accuracy metrics against ground truth centerlines.

Main Results:

• The 3D method achieved exceptional accuracy, with an average overlap accuracy (OV) of 99.97% and an average error distance (AI) of 0.13 mm.
• High accuracy was consistent across all four coronary artery sub-types, including right coronary arteries (OV 99.99%), left anterior descending arteries (OV 100%), and left circumflex arteries (OV 99.96%).
• The method demonstrated excellent performance in tracking all 32 coronary vessel branches within the dataset.

Conclusions:

• The proposed 3D divide and conquer Voronoi diagram method accurately and automatically extracts coronary artery centerlines.
• This validated method shows significant potential for improving the speed and precision of cardiovascular diagnostics.
• Further exploration of this automated technique is warranted due to its high performance and clinical relevance.