3-D centerline extraction for virtual cardiovascular endoscopy

Abstract

This paper discusses centerline extraction algorithms for virtual cardiovascular endoscopy system. A fully automatic extraction method is proposed which is significantly enhanced from current distance mapping method. The new approach consists of four steps: preprocessing; construction of improved Distance From Boundary field; centerline extraction; mean value smoothing of centerline. Based on the proposed automatic approach, an interactive extraction method is presented in which a centerline connects specified points selected by cardiologists. This interactive approach makes the viewport closer to pathological changes in the aorta and human heart, so cardiologists can observe focuses more clearly. The two algorithms have been applied to the virtual cardiovascular endoscopy system, and the experiments prove that our methods work well in the process of auxiliary diagnosis.