A Comprehensive Analysis of Deep Learning Model for the Detection of Diabetic Retinopathy Using Explainable Artificial Intelligence

Diabetic retinopathy (DR) is a leading cause of preventable blindness, and early detection is essential for effective treatment. This paper proposes a convolutional neural network (CNN) is proposed for accurate classification of DR severity levels, with a target accuracy of 98 %. Transfer learning models such as DenseNet121, InceptionV3, and Xception, along with machine learning methods including SVM and RNNs, were evaluated. Xception achieved 82 % accuracy, while the proposed CNN outperformed existing methods, reaching 95 % accuracy on a diverse dataset and near-perfect performance in binary classification. To support real-time diagnosis, a user- friendly platform is being developed to enable instant retinal image analysis. The integration of explainable AI techniques, particularly Gradient-weighted Class Activation Mapping (Grad-CAM), further enhances model interpretability and supports clinical decision-making. This research highlights the potential of deep learning in DR detection and provides practical insights into optimizing classification techniques for real-world medical applications.