Ocular Disease Detection Using Advanced Neural Network Based Classification Algorithms

Abstract

One of the most challenging tasks for ophthalmologists is early screening and diagnosis of ocular diseases from fundus images. However, manual diagnosis of ocular diseases is difficult, time-consuming and it can be prone to errors. That is why a computer-aided automated ocular disease detection system is required for the early detection of various ocular diseases using fundus images. Due to the enhanced image classification capabilities of deep learning algorithms, such a system can finally be realized. In this study, we present four deep learning-based models for targeted ocular tumor detection. For this study, we trained the cutting-edge image classification algorithms such as Resnet-34, EfficientNet, MobileNetV2, and VGG-16 on the ODIR dataset consisting of 5000 fundus images that belong to 8 different classes. Each of these classes represents a different ocular disease. The VGG-16 model achieved an accuracy of 97.23%; the Resnet-34 model reached an accuracy of 90.85%; the MobileNetV2 model provided an accuracy of 94.32%, and the EfficientNet classification model achieved an accuracy of 93.82%. All of these models will be instrumental in building a real-time ocular disease diagnosis system.