3-D Coronary Vessel Extraction Using A Novel Minimum Path Based Region Growing

This work presents a novel approach termed minimum path based region growing (MP-RG) to extract 3-D coronary vessels. Using the joint convexity information of orientated 1D vessel profiles to build the potential function, a single-point minimum path searching is applied to guide the region growing to achieve effective vessel lumen extraction. The searching priority for sparsely distributed vascular structures allows an efficient implementation of the MP-RG algorithm. Only one start point is required to be set to extract a 3-D vessel tree. Especially, an effective stopping criteria is developed for the region growing process. We test the effectiveness of the proposed MP-RG by using clinical coronary CT data, and the encouraging performance of the proposed method is validated.