Automated Diabetic Retinopathy Detection Using VGG19 Deep Learning Architecture

Diabetic retinopathy (DR) is a leading cause of vision impairment and blindness among individuals with diabetes, making early detection and accurate classification essential for timely intervention. This research presents a deep learning-based approach utilizing the VGG19 architecture to develop an automated system for diabetic retinopathy classification using retinal fundus images. The methodology leverages transfer learning, freezing the pretrained layers of VGG19 while integrating custom dense and dropout layers to optimize performance for the binary classification task. A dataset comprising 500 retinal images, evenly split between positive and negative DR cases, was employed, incorporating preprocessing techniques such as data augmentation and normalization. The model achieved a training accuracy of approximately 98.45%, validation accuracy of 95.66%, and test accuracy of 95.79%, demonstrating strong generalization capabilities. Performance evaluation using precision, recall, and F1-score yielded balanced metrics of 0.94 for both classes, supported by a confusion matrix that confirmed robust predictive accuracy. These findings highlight the effectiveness of deep learning in enhancing diagnostic workflows for diabetic retinopathy, contributing valuable advancements to medical image analysis. The research underscores the potential impact of integrating automated diagnostic systems into clinical practice, improving early detection rates, reducing ophthalmologist workload, and expanding access to quality eye care in diverse healthcare settings.