3-D Vascular Tree Segmentation Using Level-Set Deformable Model

This paper describes a novel 3-D level-set deformable model-based approach for segmentation of medical computed tomography (CT) images of human brain vascular tree. The method employs a 3-D edge detection method to establish the initial contours. Afterwards a velocity field is created using the gradient vector flow algorithm. The deformable model is then initialized and solved using a level-set method. Experimental validation of the method has been conducted on CT images of real patients. Comments on performance and possible improvements are discussed.