Segmentation of Carotid Arteries in Computed Tomography Angiography Images Using Fast Marching and Graph Cut Methods

Carotid artery stenosis causes the narrowing of blood vessels and as a result, it leads to stroke. In this paper we use the Mean shift smoothing for uniformity of gray levels in order to distinguish the carotid artery and the adjacent tissues in an image. Then, centerline of the arteries is extracted by using 3D fast marching. After extraction centreline of vessels, segmentation is performed by 2D graph cut. The results are evaluated automatically by a physician. The proposed method applied on 14 Computed Tomography Angiography (CTA) data. Results show 87% average Dice measure and 2.17 averages Dmax with variance 2.11 for our methods. Evaluations showed that the proposed method requires minimal user intervention, since the relatively low resistance to the gray levels changes in vascular, low resistance to extreme changes in vessel diameter and vessel branch locations. It can be used in qualitative and quantitative evaluations, because of high accuracy of proposed method