A modular workflow architecture for coronary centerline extraction in computed tomography angiography data

Abstract

Efficient and reliable extraction of coronary artery centerline from computed tomography angiography data is a prerequisite for a variety of medical imaging applications. Many authors have combined minimum-cost path algorithms and vesselness measures to extract coronary centerlines. We propose a modular decomposition of this extraction process, in order to facilitate the implementation and comparison of different minimum-cost path strategies allowing users (radiologists and developers) to focus on subsequent image analysis tasks. Evaluation results show a good overlap (>84%), and small distances with regard to reference centerlines (on average, not larger than the voxel size) in multi-vendor datasets, for two combinations of algorithms that follow this framework. Therefore, it can serve as a starting point to subsequent image analysis stages that require coronary centerlines.