Paper Title
A Neural Network-Based AI System Tailored for The Prior Identification and Hierarchical Classification of Diabetic Retinal Anomalies Through Advanced Ocular Imaging

Abstract
Diabetic Retinopathy (DR) presents itself as one of the main causes of vision loss among patients suffering from long-term diabetes and early detection of DR is crucial in its prevention to avoid irreparable damage to the retina. The conventional methods of diagnosis of DR largely depend on manual examination and clinical analysis of the fundal photographs. Despite effectiveness, these approaches are typically time consuming, require specialized training and are also subject to inter observer variation with respect to diagnostic delay and inconsistent grading. With this in mind, the purpose of this paper will be to outline a framework of diagnostic AI retina health to support the automatic diagnosis of DR symptoms in Convolutional Neural Networks (CNNs) written in a programming language known as, Pytorch, and in fundus photographs. The model is trained on a publicly available datasets and predicts the image in two classes i.e DR and Non-DR based on the features extracted directly from the images. The overall classification accuracy of the proposed system on testing data set was 94% which shows that the proposed system is reliable and robust. The algorithm showed great potential in helping ophthalmologists in terms of screening and early intervention, thus leading to good care for patients with lesser burden of disease. Keywords - Diabetic Retinopathy, Deep Learning, CNN, PyTorch, Medical Imaging, Fundus Images.