Evaluating Deep Learning and Traditional Approaches in the Automated Classification of Retinal Diseases

Abstract

Retinal diseases such as diabetic retinopathy, glaucoma, and age-related macular degeneration are leading causes of preventable blindness worldwide. Early and accurate diagnosis is essential, but manual interpretation of retinal images remains time-consuming and prone to variability. This study evaluates deep learning models, DenseNet121 and DenseNet201, against traditional machine learning models like Support Vector Machine (SVM) and Random Forest (RF) for the automated classification of retinal diseases using fundus images. The dataset consists of over 21,000 images across ten ocular diseases. DenseNet201 achieved the highest validation accuracy (85.77%), outpacing traditional models (SVM: 65.04%, RF: 64.87%). This research highlights the potential of AI in ophthalmology, demonstrating its ability to automate disease detection and improve diagnostic accuracy.