Vessel Extraction by Graph Cut Method based on Centerline Estimation

Vessel extraction is of great importance in the diagnosis and surgery planning of vascular related disease. The commonly used Hessian matrix method has the false positive effect on sharp boundaries in non-vascular region. The graph cut method characterized by fast and accurate segmentation in natural images is susceptible to initialization and priors. In this paper, we conducted vessel extraction by graph cut based on centerline estimation, which takes the advantage of Hessian matrix and graph cut segmentation. The centerline points act as a role of initialization and priors for the graph cut method. The experiment on simulated vessel data demonstrated the performance of the proposed method in vessel extraction.