Vascular Bifurcation Detection by Symmetric Analysis of Eigenvalues

Abstract

Reconstruction of vasculature in medical images is critical for the early diagnosis of various vascular diseases. In this paper a common method of Hessian based filtering is used to enhance the vascular structures but a different method is proposed to extract the bifurcation (branch) points in 3D photoacoustic images. Bifurcation points play an important role for diagnosis of vessel stenosis, atherosclerosis and surgical planning. In order to clearly visualize the vascular structures, detection of branch points are needed, to enhance the effect of vessels at these points. In the proposed method, first the vascular structure is extracted by eigenvalues analysis of Hessian matrix. Based on the eigenvalues, we detect the bifurcation points by considering a small patch in the image. To detect the branch point, it was analyzed that if the vascular image is rotated from 00 to 1800 angles with a small step size thetas0, the branch points maintain their rotation symmetry. The algorithm is tested on synthetic data and also on photoacoustic images acquired from Optical Resolution Photoacoustic Microscopy (OR-PAM). The proposed method is shown to be highly effective at detecting branch points.