Fast 3D skeleton extraction of airways and applications to virtual bronchoscopy

Abstract

In this paper, a novel automatic approach of computing 3D skeletons of airways is introduced, which is based on optimized double distance transform field and endpoints detection. The advanced double distance transform fields are constructed by the fast marching method (FMM). The endpoints are detected automatically combining 3D SUSAN corner detector and the local extreme value of distance from source (DFS). Furthermore, the minimum cost path scheme is used for searching a connected, one voxel thickness flight path of airways in virtual bronchoscopy. Experimental results show that the proposed method is centered, less sensitive to object noisiness, more accurate and more efficient. We have validated the robustness of the proposed method both quantitatively and qualitatively against clinical datasets.