Automated Diabetic Retinopathy Detection using Convolutional Neural Network
Diabetic retinopathy (DR) is normally encountered in individuals who are affected by diabetes for longer time. The vision of patient will be regular and his retinal blood vessels get affected mildly. Protein and fluid might seep out from the blood vessels due to diabetes. Several factors are related with the development of DR like genetic parameters and metabolism control density. This work aims to establish a new automated DR recognition scheme, which involves phases such as “(i) Segmentation (ii) Feature Selection and (iii) Classification”. At first, the input fundus image is processed under segmentation phase, and once after this, the feature selection takes place, where 30 features are chosen over the 70 features by exploiting the adaboost algorithm. These selected features are then subjected to Convolutional Neural Network (CNN), from which the presence of disease is classified. Finally, the betterment of adopted scheme is compared and validated over other classifiers.