AUTOMATIC DIABETIC RETINOPATHY DETECTION USING DEEP LEARNING

Abstract

Diabetic Retinopathy (DR) is a severe complication of diabetes that can lead to vision impairment and permanent blindness if not detected at an early stage. Early diagnosis and timely treatment are essential to prevent disease progression. This study presents an automated deep learning-based system for detecting diabetic retinopathy from retinal fundus images. The proposed approach utilizes the MobileNetV2 convolutional neural network with transfer learning to efficiently extract features and perform classification. Retinal images obtained from a publicly available Kaggle dataset are preprocessed using techniques such as resizing, normalization, and Gaussian filtering to enhance image quality and improve model performance. Additionally, data augmentation methods including rotation, zooming, and horizontal flipping are applied to increase dataset diversity and reduce overfitting. The trained model is evaluated using a confusion matrix, achieving 259 true negatives, 264 true positives, 12 false positives, and 15 false negatives. The system attains an overall classification accuracy of 95.09%, demonstrating its effectiveness in identifying diabetic retinopathy.