Detection and Classfication of Non-proliferative Diabetic Retinopathy Using Retinal Images

Abstract

Diabetic retinopathy is one of the general eye disease observed in diabetic patients and which is the most important reason of vision loss. Cataracts and glaucoma are some other eye diseases found in diabetic patients due to complication in diabetes. Diabetic Retinopathy is shown generally because of the impairment of retinal blood vessels in the patients suffering from diabetic patients. So there is a need of recognition of Diabetic Retinopathy in early stage of diabetes is required and also to safeguard or to save the diabetic patient from early vision loss and for that it is extreme necessary to detect and extract the retinal image and retinal blood vessels from retinal image. Automated segmentation and recognition system for retinal images is proposed in this paper to determine the intensity of diabetes. Segmentation of retinal image from eye image, segmentation of blood vessels from retinal images, segmentation of microaneurysms and hardexudates are performed more accurately. Proposed technique uses image processing techniques and support vector machine for detection and classification of non-proliferative diabetic retinopathy grades like normal, mild, moderate and severe. The proposed method is tested on STARE dataset and results are compared using parameters of sensitivity, specificity, and accuracy, and found high accuracy of 96.32 which can help to detect and prevent diabetic retinopathy.