Publication

Abstract : Diabetic retinopathy (DR) is a leading cause of blindness, and early detection is critical for preventing vision loss. This paper proposes a secure and highly accurate framework for DR detection in IoT-enabled healthcare environments. Fundus images are first enhanced and segmented to highlight disease regions, after which a Hamiltonian quantum graph multi-relational generative adversarial network is employed for robust classification of DR severity. To further improve performance, the model’s hyperparameters are automatically tuned using a bio-inspired optimization strategy, and patient data are protected with quantum-resistant encryption methods. Experimental evaluation on benchmark datasets (IDRiD and DIARETDB1) demonstrates that the proposed framework achieves up to 99.9% accuracy and 99.8% precision, significantly outperforming existing methods. These findings demonstrate the suggested method's potential as a scalable and private solution for early DR recognition in an actual healthcare system.