Detecting Diabetic Retinopathy in Medical Imaging using Deep Learning
Main Article Content
Abstract
Early detection of diabetes-related complications, particularly diabetic retinopathy (DR), is essential to prevent from permanent loss of vision. Diagnosis by ophthalmologists is time-consuming and costly. Recent researches in deep learning has enabled automated analysis of medical images with increased accuracy and efficiency. This research demonstrates deep learning based approach for automated diabetic retinopathy detection using deep convolutional neural network architectures such as Inception-V3 and VGG-16. These architectures are used to extract discriminative features from retinal fundus images. The models are trained and evaluated on a publicly available dataset containing five DR severity classes. The results shows that the Inception-V3 model achieves higher classification accuracy than VGG-16 in identifying different stages of diabetic retinopathy. In addition, a lightweight web-based interface is developed to support practical deployment and user interaction. The findings suggest that deep learning–based methods can assist clinicians by providing consistent performance for diabetic retinopathy detection.