Retinal Vessel Segmentation Using Morphological Top Hat Approach On Diabetic Retinopathy Images

Abstract

In the diagnosis, screening, early detection and treatment of diseases like glaucoma, diabetic retinopathy (DR), hypertension, retinopathy of prematurity (ROP), age related macular degeneration (AMD) and arteriosclerosis retinal blood vessels play a major role. In the working age group of people Diabetic Retinopathy (DR) is very deadly one since it has a threat on sight, since it may lead to blindness. Retinal vessel segmentation is the fundamental step in detecting various pathologies. Hence it is very important for retinal vasculature segmentation for helping the clinicians for screening and treating various pathologies. A novel method is proposed in this paper for extracting the retinal blood vessels. Blood vessel enhancement and suppression of background information, smoothing operation is done on the retinal image using mathematical morphology and top hat transform is used. Later segmentation is carried out using two fold hysteresis thresholding algorithm. The proposed approach is evaluated on Diabetic Retinopathy images in HAGIS and HRF dataset. Experimental results show that our method is efficient as the average accuracy achieved is 95.12% and 94.37% with HAGIS and HRF dataset respectively.