Deep Learning–Based Ocular Disease Detection and Classification Using Transfer Learning Techniques

Abstract

Diabetic retinopathy, glaucoma, and age-related macular degeneration, which belong to the category of ocular diseases, rank among the primary causes of vision impairment. Early detection and diagnosis are essential to avoid irreversible blindness. Even so, manual retinal image analysis is very time-consuming and prone to human errors. This paper introduces an automated diagnosis and classification tool for ocular diseases based on image processing and machine learning algorithms. The proposed model relies on deep convolutional neural networks (CNNs) and transfer learning approaches using VGG16 and ResNet50 models to derive relevant features from fundus and Optical Coherence Tomography (OCT) images. Image preprocessing routine Noise removal, normalization, and resizing involving augmentation of image quality was performed. We are pre-processed to a dimension of 224 × 224 before feeding into the model. The proposed model was trained and tested on benchmark datasets of retinal images, and, It is noted to be highly accurate with good robustness when classified. Comparative study based on standard or existing AI models showcased its efficacy based on Parameters like Accuracy, Precision, Recall, and F1-score.The results indicate that such automated diagnostic tools can effectively assist ophthalmologists by providing faster, more consistent, and objective assessments. Future work aims to integrate this system into mobile or real-time screening platforms, making early eye disease detection more accessible in clinical and remote healthcare settings.