Automated Glaucoma Detection Using CNN and Transfer Learning Models on Retinal Fundus Images

Glaucoma is one of the leading causes of irreversible blindness worldwide, resulting from progressive damage to the optic nerve and retinal nerve fiber layer. Since the disease develops silently without prominent symptoms, timely diagnosis remains a critical challenge in ophthalmology. This paper proposes an SDG-driven intelligent glaucoma detection framework using Convolutional Neural Networks (CNN) and transfer learning architectures on retinal fundus images for sustainable vision healthcare. The developed methodology incorporates image preprocessing, feature extraction, deep classification, and comparative evaluation using VGG16, ResNet50, EfficientNet, and the proposed optimized CNN model. Mathematical formulations of convolution, activation, pooling, and softmax operations are integrated to establish a structured analytical model. Experimental validation is performed using standard performance measures including accuracy, precision, recall, F1-score, ROC curve, and AUC analysis. The proposed model achieves superior classification accuracy of 95.6% with improved generalization and reduced diagnostic complexity. In addition to technical performance, this study significantly contributes to United Nations Sustainable Development Goal 3 (Good Health and Well-Being) by enabling early automated glaucoma screening, reducing preventable blindness, and supporting inclusive healthcare access in rural and resource-constrained environments.