Deep Learning Approaches for Retinal Disease Diagnosis: Insights from Fundus and OCT Analysis

School of Computer Science Engineering and Technology, Bennett University, Plot Nos 8-11, TechZone 2, Greater Noida, 201310, Uttar Pradesh, India.

^* Corresponding author: madhuri.gupta@bennett.edu.in

Published online: 18 June 2025

Abstract

The study investigates the application of deep learning to diagnose eye diseases like glaucoma and diabetic retinopathy based on imaging with fundus images, 2D OCT, and 3D OCT imaging. We investigate state-of-the-art architectures, including Vision Transformers (ViT), Convolutional Neural Networks (CNN), and hybrid frameworks on various datasets, including OCT2017, EyePACS, ACRIMA, and others. Model performance comparison — metrics used are accuracy, precision, recall, specificity, and F1-score. Testing accuracies reach as high as 98% (Vision Transformers → more generalizable), while EfficientNet-based models accomplish even higher accuracy (up to 99.6%) in the multilabel classification of 2D OCTIM series. Advanced frameworks like ViT-large + GRU and AlterNet-K demonstrate promising results in glaucoma detection using 3D OCT imaging. The findings underscore the transformative role of deep learning in ophthalmology and highlight the need for standardized datasets and interpretability in future research.