Automated Detection of Proliferative Diabetic Retinopathy Using Brownian Motion Features

Diabetes is a chronic disease caused when the body does not produce enough insulin or the insulin produced fails to break down glucose in the blood. It is a non-communicable disease and the condition is irreversible. Treatment is vital to prevent the condition from worsening and complications. One of the complications of diabetes is diabetic retinopathy, a disease that affects the vision. There are four stages of diabetic retinopathy. In this paper, we focus on the last stage of diabetic retinopathy, which is Proliferative Diabetic Retinopathy (PDR). Fractal dimensions and Hurst coefficients are the features extracted from normal and proliferative diabetic retinopathy images. These features are then input to five classifiers namely, Support Vector Machine (SVM), Probabilistic Neural Network (PNN), Decision Tree (DT), K-Nearest Neighbour (KNN) and Fuzzy Sugeno (FS) to select the best classifier. FS classifier yielded the highest average accuracy of 94%, sensitivity of 92% and specificity of 96%.