Evaluation and comparison of retinal blood vessels extraction using edge detectors in diabetic retinopathy

Abstract

Diabetic retinopathy (DR) is a vascular disorder that affects the microvasculature of the retina. Early detection of DR helps the ophthalmologists in diagnosis and treatment to save the vision of the patient. The change in appearance of retinal blood vessel structures can be detected by edge detectors. In this paper different edge detectors have been evaluated to find the abnormalities in diabetic retinal blood vessels and the results are compared with a healthy retinal image. The peak signal-to-noise ratio (PSNR), mean square error values are computed for quantitative evaluation and the modulation-transfer function has been performed to represent the results graphically for edge detectors. The evaluated result shows that the Canny operator gives better output with high PSNR values in detecting tiny blood vessels of retina compared to other edge detectors like Robert, Prewitt and Sobel operator. The average value of the sensitivity and specificity are 15.6% and 85.4%.