Automated Diabetic Retinopathy Detection and Multi-Stage Classification Using Convolutional Neural Networks

This paper presents an automated classification and identification of diabetic retinopathy (DR) based on CNNs. The model identifies four types of DR: no DR, moderate DR, severe non-proliferative DR, and proliferative DR. The model was trained and tested after preprocessing procedures applied to it in order to improve image quality and remove noise from the public dataset, famous for this purpose. Segmentation of the categorization process into two phases reduces complexity and increases accuracy. The proposed model achieved a class accuracy of 89.7<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> and achieved a sensitivity level of 92.1%, greater than any other methodologies reported in the literature so far. The approach provides clinicians with a more accurate method to detect DR that can lead to improved patient outcomes in the long term.