Liver vessel segmentation based on centerline constraint and intensity model

Abstract

Abstract Liver vessels provide lots of important information for liver-disease diagnosis and liver surgery. This paper presents an effective liver vessel segmentation method from abdominal computer tomography angiography (CTA) images. The proposed method applies two techniques including centerline constraint and intensity model for effective detection of liver vessels, in which the former aims at generating the position and distance restraints for the detection of thin vessels by the offset medialness filter and height ridge traversal algorithm, while the latter is mainly used to extract intensity feature for the detection of thick vessels based on Kernel Fuzzy C-Means (KFCM). And then, the centerline constraint and intensity model are integrated into graph cuts for ultimate liver vessel segmentation. The proposed method does not require any manual selection of the initial vessel regions, and is capable of dealing with complex liver vessel systems. The experimental results on clinical CTA data sets give an average accuracy, sensitivity, and specificity of 97.4%, 83.0%, and 98.1%, respectively, which show the efficiency of the proposed method on liver vessel segmentation.