Glaucoma Detection Using Deep Learning on Retinal Fundus Images with Explainable AI

This paper presents a deep learning-based Glaucoma Detector for automated analysis of retinal fundus images. The proposed system classifies retinal images into normal and glaucoma categories using a convolutional neural network with transfer learning. Image preprocessing is applied to standardize the input, while model evaluation is performed using classification metrics and graphical analysis. A major feature of the proposed system is the integration of Grad-CAM-based explainable artificial intelligence. The generated heatmaps provide visual information about the retinal regions contributing to the model's prediction, improving transparency of the automated classification process. The trained model is integrated into a Flask-based web application that provides user authentication, image upload, prediction, confidence scores, class probabilities, and visualization. The system also supports optional optic-disc and optic-cup segmentation and cup-to-disc ratio calculation. The developed framework demonstrates the practical application of deep learning and explainable AI in retinal image analysis. It can serve as a preliminary screening and decision-support system, while clinical diagnosis and treatment decisions should remain under the supervision of qualified eye-care professionals.

 [1]  Song et al., “Lightweight deep learning framework for optic disc and optic cup segmentation for glaucoma detection,” as cited in the project literature survey. [2]  Chiang et al., “Deep learning-based glaucoma detection in highly myopic populations,” as cited in the project literature survey. [3]  Govindan et al., “Deep learning framework for early glaucoma detection using retinal fundus images,” as cited in the project literature survey. [4]  Septiarini et al., “Automated optic disc segmentation using SSD MobileNetV2 and U-Net,” as cited in the project literature survey. [5]  Saha et al., “Fast and fully automated glaucoma detection using color fundus photographs,” as cited in the project literature survey. [6]  Kavitha et al., “Glaucoma detection using traditional machine learning techniques,” as cited in the project literature survey. [7]  Pascal et al., “Multi-task deep learning framework for glaucoma detection using REFUGE,” as cited in the project literature survey. [8]  Ajitha et al., “Glaucoma identification using deep learning with SoftMax and SVM classifiers,” as cited in the project literature survey. [9]   Sandoval-Cuéllar et al., “ROI-based glaucoma classification using a convolutional neural network,” as cited in the project literature survey. [10]  Akter et al., “Deep learning methods for glaucoma detection: A comprehensive review,” as cited in the project literature survey.