Comparative Analysis of Intelligent Systems using Support Vector Machine for the Detection of Diabetic Retinopathy

Abstract

Diabetic Retinopathy (DR) is an ailment in the retina of human, caused due to prolonged diabetes which damages the retina and can cause loss of vision. DR is an effect of diabetes mellitus. DR is detected, based on the existence of retinal abrasions such as microaneurysms, hemorrhages, exudates and cotton wool spots. Statistically, 80% of people suffering from diabetes are estimated to have suffered from DR. Considering such as huge population, the process of manual analysis and detection of these features are feasible, but at the same time it is a highly cumbersome task. Hence, in order to minimize the labor-intensive detection process, intelligent systems using Machine Learning (ML) are proposed. Based on the performance of different ML models, this paper proposes an intelligent system named Intelligent System for DR using Support Vector Machine (ISDRSVM), for early detection of DR. The proposed methodology uses image processing to enhance retinal fundus images and derive DR features for better extraction and classification. The model is trained on four different SVM kernels, namely linear kernel, polynomial kernel, sigmoid kernel and Radial Basis Function (RBF) kernel. The classification is categorized into five categories, namely 0-no DR, 1-mild Non-Proliferative DR (NPDR), 2-moderate NPDR, 3-severe NPDR and 4-proliferative DR. The model is compared with a previously proposed model. The proposed model enhances the classification accuracy by 5% in comparison with its previously proposed structure. Additionally, among all the SVM kernels, the proposed sigmoid kernel achieves a better accuracy of 69.09%.