FQPDR: Federated Quantum Neural Network for Privacy-preserving Early Detection of Diabetic Retinopathy

Researchers have proposed FQPDR, a novel federated learning-based quantum neural network designed for the early detection of Diabetic Retinopathy (DR), a common diabetes complication that can lead to blindness. Detecting mild DR is challenging due to the tiny, low-contrast nature of microaneurysm dots, which are the earliest signs of the disease. To address data privacy concerns in medical image processing, the system utilizes federated learning, allowing collaborative model training by sharing only parameters rather than sensitive patient data. The study implemented lightweight models with limited samples and few learnable parameters using the E-ophtha and Retina MNIST datasets. Cross-evaluation on Kaggle dataset images demonstrated the robustness and efficiency of the FQPDR system. The results indicate inspiring performance compared to existing non-federated and standard federated learning methods, highlighting the potential of combining quantum computing concepts with privacy-preserving techniques for accurate, early-stage medical diagnosis without compromising patient confidentiality.