Automatic Detection and Classification of Microaneurysms for Early Detection of Diabetic Retinopathy in Color Fundus Images

Diabetic Retinopathy (DR) is one of the main causes of visual disorder in patients affected with diabetes. Prior diagnosis is needed to reduce the visual impairment, so that damage to eye can be minimized. In the DR, Microaneurysm (MA) is the earliest medical sign which appears as tiny individual retinal patterns. So, powerful computer aided diagnose techniques for MA detection are needed. In this paper, a new approach for the automatic detection of MAs in eye fundus images is proposed. Eleven features based on shape and intensity characteristics are extracted from MA candidates and true MAs are classified from false candidates using KNN, SVM and NB classifiers. This proposed approach is evaluated on a publicly available dataset (E-ophtha). The performance of this method is measured by using sensitivity, specificity, and accuracy metrics. The experimental outcome demonstrated that the proposed method is efficient to diagnose clinically.