Image Segmentation and 3-D Reconstruction of Coronary Artery

Abstract

Segmentation of arterial wall boundaries from biomedical images is an important issue for many applications such as in the study of plaque characteristics, to extract mechanical properties of the arterial wall, its 3-D construction, and the measurements such as wall radius, lumen radius and lumen size. So here we present a solution to segmentation of images of coronary arteries and 3D construction. The structure introduced here is a set of connected vertices. An initial contour model which is defined here is allowed to deform according to an energy minimizing term with minimum number of iterations. The energy associated with the contours are depends on curvature of contour, and image features. Using training data in a properly built shape space, we are able to classify media and adventitia walls approximately using a large set of data sets using Support Vector Machines. The 3D construction of artery using point matching of successive frames is also explained here with less complexity. The tests of the presented algorithm on a large set of data depicts the effectiveness of this approach.