Personalized Federated Learning in Edge-Cloud Continuum for Privacy-Preserving Health Informatics: Opportunities and Challenges

The growing interest in advanced data-intensive models for healthcare applications presents several challenges and opportunities. Federated learning (FL) emerges as an attractive solution to allow decentralized nodes to collectively train shared machine learning models without the need of transmitting sensitive data to a central database. This can safeguard privacy while effectively leveraging the distributed computational resources available in the cloud-edge continuum. In health informatics, the need for robust privacy-preserving mechanisms is paramount, especially when the nodes of the FL system are associated with datasets from individual patients, as opposed to the case of databases that include many patients (such as those available from hospitals). This need becomes particularly significant when addressing diagnoses and predictive analytics in personalized medicine, precision medicine, risk stratification, and longitudinal monitoring. We explore the applications of FL frameworks in the context of cloud-edge in healthcare. We identify real-world settings to assess the benefits and challenges of personalized federated learning. These include data imbalance issues, usability, promoting replicability, improving security, minimizing environmental impact (greenness), and optimizing overall efficiency.