ResNet-18 for Automated Eye Disease Diagnosis: A Neural Network Solution

Eye diseases including diabetic retinopathy, cataracts, and glaucoma account globally for the main causes of visual disability and blindness. Early diagnosis of these diseases will help to prevent major vision loss and guarantee suitable therapy. In order to meet the demand for good health and well-being, we propose in this study an automated ResNet-18 deep learning model-based eye disease categorization system. Extensively sourced from IDRiD, Ocular Recognition, HRF, and other databases, the dataset consists of over 1000 retinal images per class spanning Normal, Diabetic Retinopathy, Cataract, and Glaucoma. The method trains the ResNet-18 model on these images so that it may accurately classify them into the suitable categories, therefore enabling infrastructure and innovation in the healthcare industry. Our approach confirmed the efficiency of the model in discriminating among many eye diseases with a remarkable accuracy of 91%. This outstanding accuracy highlights how well deep learning models could assist ophthalmologists with early diagnosis and patient outcome enhancement, thereby supporting sustainable cities and communities and global alliances for health enhancements.