Automatic prediction of Diabetic Retinopathy and Glaucoma through retinal image analysis and data mining techniques

Application of computational techniques in the field of medicine has been an area of intense research in recent years. Diabetic Retinopathy and Glaucoma are two retinal diseases that are a major cause of blindness. Regular Screening for early disease detection has been a highly labor - and resource- intensive task. Hence automatic detection of these diseases through computational techniques would be a great remedy. In this paper, a novel computational approach for automatic disease detection is proposed that utilizes retinal image analysis and data mining techniques to accurately categorize the retinal images as Normal, Diabetic Retinopathy and Glaucoma affected. Three feature relevance and sixteen classification Algorithms were analyzed and used to identify the contributing features that gave better prediction results. Our results prove that C4.5 and random tree classification techniques generate the maximum multi-class categorization training accuracy of 100% in classifying 45 images from the Gold Standard Database. Moreover the Fisher's Ratio algorithm reveals the most minimal and optimal set of predictive features on the retinal image training data.