Abstract

Research on amultiscale approach to tubular structure\ detection in 3D biomedical images

Author(s): Shihai Zhu

Image detection is an important technology which has been applied extensively to geometric measurement, industrial quality inspection, three-dimensional surface detection and the like. Because most of image detection algorithms are effective for specific problems so far, there is no detection algorithm can be applied to solve all image detection problems. Detection of tubular structures in 3D medical images is an important issue for vascular medical imaging. In this paper, we propose a new adaptive medialness measure for detection of tubular structures in 3D medical images. The adaptiveness of the medialness is based on the Hessian matrix of the image, its eigenvectors and eigenvalues. First we describe the proposed measure of medialness and ridge, Second we use a simple model of cylindrical vessel with circular Gaussian cross-section to illustrate our detection solution, meanwhile we give the relationship between the size of the structure and its selected scales. From this relationship, we explain the extraction of local extrema and make a full reconstruction of the vessels network. Thirdly synthetic and real images are used to verify our study under the suggested tubular structures. At last we draw some conclusions and give some areas for future research