Answers are grounded in this paper and cite their sources.
Demonstrates improved eye disease diagnosis accuracy in privacy-sensitive environments, indicating a practical approach for healthcare.
Early and correct identification of eye disease is essential in preventing vision loss and blindness. However, machine learning methods require centralized data, which raises concerns about healthcare privacy and security. To address these issues, this study presents a Federated Learning framework for eye disease diagnosis based on the ODIR-5K dataset. Federated learning enables collaborative model training across different organizations without disclosing sensitive patient data, and decentralization mitigates the risks associated with single points of failure. To enhance eye disease categorization while maintaining data privacy, this study proposes a Federated Learning (FL) and Decentralized Federated Learning (DFL) framework that employs the Swin Transformer, MobileNetV2, and EfficientNet-B0 for collaborative Training. The proposed framework ensures practical Training across clients using a peer-topeer model that updates based on FL and DFL. Accuracy, precision, recall, F1-score, and loss are used to evaluate performance, ensuring a dependable and scalable solution for practical medical applications. Our approach enables flexible, privacy-preserving model building by distributing Training across multiple nodes, thereby improving accuracy and data privacy. Furthermore, we optimize bi-level hyperparameters to fine-tune model parameters, resulting in optimal diagnostic performance. The approach combines cutting-edge Artificial Intelligence (AI) with practical healthcare applications, enabling safer, more efficient, and widely accessible detection of eye diseases.
No takes yet. Share an insight, caveat, or question.
Sami et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: