Development of Algorithm for Dual Stage Classification to Estimate Severity Level of Diabetic Retinopathy in Retinal Images using Soft Computing Techniques

Abstract

Diabetic Retinopathy is a general origin of sight-threatening complication to which the condition occurring in persons with diabetes and which makes progressive damage to the retina. Retina is a back of eye with the light sensitive lining. The severity of this Diabetic Retinopathy in retina has four stages. The identification of the severity level of Diabetic Retinopathy in retina is a critical task. In order to make ease this task, a severity analysis method is proposed in this paper. The proposed method contains the 5 stages - (1) Pre-processing phase (2) Segmentation Phase (3) Feature Extraction Phase (4) Classification Phase I and (5) Classification Phase II. The retinal images are subjected to pre-processing phase for removing the noises from the image and make clear visible image of retina. Then the optic disks and the blood vessels in the retina are segmented using Modified Region Growing method and morphological operations, respectively. Before analyzing the severity in retinal images, it is necessary to classify the images into normal and abnormal images using Neural Network classifier. For this classification of images, the features mean, variance, entropy and area are extracted from the segmented optic disk of retinal images and also the features mean, variance, entropy, area, diameter and number of regions are extracted from the segmented blood vessels of retinal images. From the abnormal images, the severity of Diabetic Retinopathy can be evaluated by using SVM classifier based on area and intensity level of Hard Exudates and Hemorrhages. These techniques are implemented on publicly available database such as STARE and on real datasets using MATLAB 7.12. The performance of our proposed method is analyzed by Sensitivity, Specificity and Accuracy. From the results, it is proved that our work outperforms other works by providing very much better accuracy by classifying the severity level in Diabetic Retinopathy.