Enhancing eye disease classification through synergistic deep learning approaches

Abstract

The number of people living with blindness is about 43 million people and 295 million people are living with moderate-to-severe visual impairment. The leading causes of most blindness are macular degeneration, diabetic retinopathy, and glaucoma. Moreover, the early stages of most eye diseases are asymptomatic. As a result, determining the cause becomes very difficult, and if left untreated, there can be irreversible damage to vision. This paper discusses a hybrid structure that combined ResNet50 and VGG19 to successfully classify and predict various eye diseases accurately. In addition, we used transfer learning and multi-class classification, which gave us an accuracy of 94.7%, whereas previous approaches with traditional CNN only gave an accuracy of less than 85%. This study has the potential to significantly contribute to the timely identification and precise categorization of ocular disorders, hence leading to advancements in patient treatment, increased overall well-being, and a more promising outlook for individuals affected by visual disabilities. Moreover, it indicates the possibility of wider utilization of sophisticated deep learning methods in the field of medical image analysis.