Paper Title
Ocular Disease Detection Using Deep Learning: A Comparative Study of CNN Architectures With Local Binary Pattern Feature Extraction

Abstract
Globally, ocular diseases are a serious health problem, with about 2.2 billion individuals contracting the diseases, which the World Health Organization estimates, with at least one billion cases preventable through early detection and intervention. Manual diagnosis of ocular pathologies using fundus images is subjective, time-consuming, and requires expertise (qualified specialists), which is an issue as it constitutes a bottleneck in the provision of healthcare, especially where resources are an issue. This research project reflects an in-depth method of automated detection of ocular disease by the use of deep learning methods, implemented on the Ocular Disease Intelligent Recognition (ODIR) dataset. This research proposes an all-inclusive strategy of automated ocular disease recognition through the utilization of the deep learning model (VGG-19), ResNet- 50, and Vision Transformer models on the ODIR (Ocular Disease Intelligent Recognition) dataset. The types of ocular diseases that we have researched are: Diabetes, Glaucoma, Cataract, Age-Related Macular Degeneration, Myopia, and Hypertensive Retinopathy. One of the new contributions in the work will be the use of Local Binary Pattern (LBP) feature extraction to perform a pre-processing stage to boost the ability to capture texture information over fundus images. This methodology entails extensive pre- processing based on LBP to generate salient texture characteristics of a collection of 7,000 fundus images, then training and testing deep learning models with and without LBP pre-processing to show the effectiveness of LBP pre-processing. The study is a quality addition to automated medical diagnosis systems and eventually the application of texture-based feature extraction to deep learning to process medical images, which could bring on board a more trustworthy and less costly ocular disease-screening system.