Automated Screening System for Grading of Retinal Abnormalities

Detection and grading of diabetic retinopathy in the at-risk population (diabetics) is crucial for providing timely treatment thereby preventing visual loss. On the basis of the most common systems, such as the National Health Service's (NHS) Program, the Scottish Grading Scheme (SGS), and the Early Treatment Diabetic Retinopathy System (ETDRS), we establish a novel automated diabetic retinopathy grading system. Medical criteria based on information from all these systems are used to grade the severity of diabetic retinopathy by calculating numbers and sizes of detected abnormalities throughout specified fields around the fovea (center of vision). The main purpose of this work is to develop a new automated diabetic retinopathy grading system based on medical systems, namely NHS program, SGS, ETDRS, and EyePACS protocol. The proposed system achieved overall success rate of 98.8% for a set of 50 images from Messidor Database and 98.4% when we use the set of 80 images from DIARETDB1. The results of the proposed system have been compared with the other existing systems in the literature and shows higher values of the overall success rate. These results assure that this system could be used for a computer-aided mass detection and grading of diabetic retinopathy as part of an automatic, fast, and accurate screening regime.