Classification of Normal and Abnormal Retinal Images by Using Feature-Based Machine Learning Approach

Abstract

The human eye is one of the most beautiful and important sense organs of human body as it allows visual perception by reacting to light and pressure. Human eyes are capable of differentiating approximately 10 million colors. It contains more than 2 million tissues and cells. Along with these entire specialties, human eyes are the most delicate and sensitive organ. If not taken proper care, it may be infected with various diseases like glaucoma, myopia, hyper-myopia, diabetic retinopathy, age-related macular disease. Therefore, early-stage detection of these diseases could help in curing them completely and prevent from complete blindness. In this paper, we propose an approach to classify the normal (healthy) and abnormal (disease-infected) retinal images by using retinal image feature-based machine learning classification approach. The performance of proposed approach by using SVM classifier is 77.3%, which is found better with respect to the other classifiers like k-NN, linear discriminant, quadratic discriminant and decision tree classifiers.