ABSTRACT

Diabetic Retinopathy (DR) is an injury to the retina due to uncontrolled blood sugar levels. Early detection and diagnosis of this disorder is essential as it may lead to complete vision loss if left untreated. In this work, a system for detecting the various stages of DR is proposed. Image morphological operations and Thresholding techniques are used for segmenting the retinal features like microaneurysms, hemorrhages. Large set of features are extracted from the images. Classification is performed using Random forest classifiers into four distinct stages namely normal, Mild -NPDR, Moderate- NPDR and PDR for publicly available data set. The accuracy obtained is 91.2%.