Diabetic Retinopathy Detection Using VGG-16 Deep Learning Architecture

Abstract

In the human body, the light-sensitive tissue is the retina, which is located at the back of the eye. In recent years, the impact of diabetes on the retina has gradually increased with a disease called diabetic retinopathy (DR), an ocular disorder. The long-lasting high level of sugar proves that diabetic patients may damage the retina's small blood vessels; this results in illness and blindness if it is not detected in the early stages. Reducing the risk factor of vision loss is possible with prompt detection and therapy of DR. Recent developments in healthcare systems adopting machine learning (ML) and deep learning (DL) models have gained a lot of popularity for image processing and analysis, early detection, and predictions with available data in a variety of applications. The performance of DL models on unbalanced data results is less accurate because most datasets related to DR are unbalanced to train the deep learning model. To overcome this problem, proposed the Pre-Synthetic minority oversampling technique (Pre-SMOTE) approach to converting unbalanced data into balanced data. And we used the visual geometry group -16 (VGG-16) model to evaluate the proposed approach. The experimental results demonstrate that the suggested model performs better in terms of accuracy (79.99%) when compared to state-of-the-art methods.