Soha Rawas, Agariadne Dwinggo Samala
The growing demand for real-time disease prediction in healthcare necessitates advanced AI frameworks capable of ensuring both computational efficiency and patient privacy. This study introduces an Edge-Assisted Federated Learning (EAFL) system designed for real-time disease prediction while preserving sensitive data using privacy-preserving AI techniques. Leveraging lightweight convolutional neural networks and federated learning, the framework decentralizes model training across simulated edge nodes, reducing reliance on centralized data storage. Differential privacy mechanisms are integrated to secure inter-node communications, enhancing data protection. The proposed framework was evaluated on publicly available datasets, including CheXpert and COVID-19 radiography, achieving an accuracy of 94.2% and 97.5%, respectively, with minimal latency (0.85 s per prediction). Comparative analysis reveals that the EAFL system matches the performance of centralized AI models while significantly reducing privacy risks. To demonstrate scalability, additional evaluation was conducted across different dataset distributions, revealing consistent performance across varying data characteristics and medical scenarios. This research offers a scalable, privacy-focused solution that bridges the gap between real-time disease prediction and secure data handling, paving the way for its adoption in modern healthcare applications. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.
Faculty of Science, Department of Mathematics and Computer Science, Beirut Arab University, Beirut, Lebanon; Faculty of Engineering, Universitas Negeri Padang, West Sumatra, Padang, Indonesia
