Segmentation of major temporal arcade in angiography images of retina using generalized hough transform and graph analysis

Abstract

Structural analysis of the vascular architecture of the retina facilitates the diagnosis of retinopathy because retinopathy affects the blood flow velocity. To this end, blood flow velocity in the major temporal arcade (MTA) vessel in retina may be measured to detect blood flow velocity changes. For this measurement, separation of the MTA vessel is a necessary step. We propose methods for identifying the MTA vessel. We use Gabor filters to detect retinal vessels and Hough transform to model the MTA vessel. Then, we use the MTA model to find a seed point for image segmentation. Next, we take advantage of graph theory to track temporal arcade vessels and separate the MTA vessel. We have evaluated our algorithm using 43 clinical angiography images, in which traces of MTA was drawn by an ophthalmologist. Experimental results show that the proposed methods correctly separate MTA in 93%.