Healthcare Industry 5.0: Pareto-Optimal IoT-Based Health Monitoring With Federated Learning
Ioanna Diamantoulaki, Sotiris A. Tegos, Pavlos S. Bouzinis, Panagiotis G. Sarigiannidis, Christos Chatzakis, Stamatios Petousis, Nicos Maglaveras, George K. Karagiannidis · IEEE Internet of Things Journal · 2025
With the recent expansion of patient data availability and storage capabilities, healthcare entities tend to store an increasing amount of medical data locally. In addition, ongoing advances in the era of healthcare industry 5.0 and Artificial Intelligence of Things can lead to more efficient use of medical data, resulting in better health monitoring at lower costs. However, due to strict privacy restrictions related to the sensitive nature of medical data, it is often only used locally and ultimately underutilized. To this end, federated learning (FL) offers a promising solution for the efficient use of medical data, facilitating the development of reliable and robust healthcare tools. This can be achieved thanks to its decentralized nature, which allows participating entities to collaborate and thus develop and train a centralized shared model without requiring data sharing. Taking this into account, we present a Pareto-front optimization framework for FL-based health monitoring that is able to mitigate false negative predictions for the required level of false positives. By applying the proposed framework to four different medical applications, it is shown that the risk of misdiagnosis is significantly reduced, providing an additional tool for medical professionals.