Coronary centerline extraction based on ostium detection and model-guided directional minimal path

Extracting centerlines of coronary arteries is a challenging but important task in clinical applications of cardiac CTA. In this paper, we propose a new method to fully automatically extract the coronary centerlines. This method is based on automatic detection of ostia and the model-guided direction minimal path (DMP) framework. Both the reference positions of the ostia and the direction information are provided from a pre-constructed model. This model is first registered to the unseen CTA. Based on the registered initial positions, the two ostia of the coronary arteries are then automatically detected from the unseen CTA, using a learning-based algorithm. The detected ostia are then used as the starting points in the DMP, where the directional information of the coronary model is incorporated, to guide the path tracing procedure. This directional tracing improves the robustness of the coronary extraction. Finally, with an improved end point selection algorithm, the proposed method can extract the longest coronary centerline from a branch. We validated the proposed method by extracting the three main coronary branches.