An automated hybrid technique for detecting the stage of non-proliferative diabetic retinopathy

• Authors: Neera Singh;Atul Kumar;Ramesh Chandra Tripathi
• Venue: Proceedings of the First International Conference on Intelligent Interactive Technologies and Multimedia
• Year: 2010

Diabetes is fast becoming a scourge in the modern day society both in the developing and the developed societies. Diabetes related complications lead to a lot of morbidity and Diabetic Retinopathy is fast becoming the cause of preventable blindness. Early detection and treatment with Laser will go a long way in checking this disease. Non-proliferative Diabetic Retinopathy (NPDR) is the set of early changes that take place in the Retina. It is divided into 3 categories into mild, moderate and severe. Initial changes when the microaneurysms (MA) start appearing. Then it is followed by hemorrhages. Finally appearance of cotton wool spots and hard exudates categorize it into severe NPDR. The stage of neo vascularization (NV) when new blood vessels begin to appear (to compensate for the reduced blood supply and nutrition to the retina) finally qualifies for proliferative Diabetic Retinopathy. The idea is to extract the features of NPDR and depending on their intensity and frequency they can be graded into mild, moderate and severe. This automated grading can be matched with the specialist's perception and its accuracy can be tested. In this work, we have proposed a computer based automated hybrid technique for the detection of stages of Non-Proliferative Diabetic Retinopathy (NPDR) retinopathy stage using the color fundus images. The features are extracted from the Sample images using the image processing techniques and fed to the support vector machine (SVM). After color normalization preprocessing stage, an evidence value for every pixel is calculated by SVM. Then a mathematical morphological technique, a fuzzy c-means clustering technique, PCA, a support vector machine and a nearest neighbor classifier for further processing. The SVM classifier uses features extracted by combined 2DPCA instead of explicit image features as the input vector Combined 2DPCA is proposed and virtual SVM is applied to achieve the higher accuracy of classification. We demonstrate a Sensitivity of 97.1% for the classifier with the Specificity of 98.3%. Thus, an automated system for diagnosis of NPDR can be a useful tool for the Specialist to support in screening an detection of early Diabetic Retinopathy changes and hence timely intervention leading to reduced DR(Diabetic retinopathy) related blindness.