Convolutional Neural Network for Diabetic Retinopathy Detection

The world's projected blind population will reach 40 million by 2025. A modern fundus-based algorithm that approves the classification of retinal tissue needs to be improved in the early stages of healthy and diabetic retinopathy. In this experiment, we have introduced a convolution neural network approach to detecting diabetic retinopathy. We used the publicly accessible Apatos Blindness Detection database to train a convolution neural network, where the image is processed at an early stage, primarily involving image resizing, pixel rescaling, and label encoder. After that, an image is given to the convolution neural network model, to decide whether the patient has diabetic retinopathy or not. About 3789 color retinal images are used in experiments to train the proposed model and about 948 images are collected to test its efficiency in classification. Accuracy of 96.15\%, Sensitivity 79\%, Precision 89\%, and F1-Score 84.1\% and Area Under Score 0.82 is achieved using the Convolution Neural Network-based method.