Classification of Exudates for Diabetic Retinopathy Prediction using Machine Learning

摘要

These days diabetes is one of the most common disease that affects the working age group which eventually leads to complication in retina termed as diabetic retinopathy (DR). Initial screening and diagnosis of DR is essential to avoid partial or complete optical damage. Manual inspection by ophthalmologist is time consuming and tedious process. So in order to lower the work load and to reduce the chances of complete blindness an automated screening system is necessary. The paper proposes an algorithm that focuses on DR detection based on exudates using features such as local binary pattern (LBP) and wavelet transform approximation coefficient matrix. Images are distinguished as exudate and non exudate by using machine learning classification algorithms such as support vector machine (SVM), k-nearest neighbour (KNN), decision tree, random forest (RF) and artificial neural network (ANN).