Ocular Disease Recognition Using Deep Learning Models

The integration of artificial intelligence and deep learning into ocular imaging has resulted in substantial advances in early disease identification and treatment. These improvements have significantly improved the accuracy and efficiency of eye disease identification by addressing the weaknesses of previous diagnostic procedures and reducing human errors. Previous research focused on particular deep-learning models for detecting individual eye disorders. This paper presents a comprehensive study on Ocular Disease Recognition (ODR) by leveraging multiple deep-learning architectures to analyze retinal images, enabling enhanced detection and classification of various ocular disorders. The transfer learning approach is used to optimize model fine-tuning, which improves classification accuracy while lowering processing needs and training duration. The experimental results demonstrate the effectiveness of the proposed pipeline, achieving a high AUC of 99.96% using MobileNet, a high accuracy of 92.46% with the ResNet50 model, and a low computational time of 8 seconds using MobileNet. These results highlight the system's ability to accurately and efficiently recognize eye diseases, providing crucial assistance for real-time and long-term disease detection systems.