Automated Detection of Microaneurysm using Textural Analysis

Diabetic retinopathy (DR), a common medical condition in diabetes patients, is a major cause of blindness. DR affects those who have diabetes for 20 years or more. By proper monitoring, major cases of DR could be rectified. The longer a person has diabetes, the higher his or her chances of developing diabetic retinopathy. Each year, DR accounts for 12% of all new cases of blindness. The algorithm proposed in this paper focuses on detection of DR using cross sectional and local binary pattern (LBP) features. Proposed system was tested on a database of 150 fundus images with mild to moderate signs of microaneurysms (MAs). Healthy and DR images were distinguished using three popular classifiers, among artificial neural network (ANN) shows highest accuracy of 92% and sensitivity of 100%.