A Comprehensive Review of Federated Learning Applications in Healthcare
Keywords:
Learning Federated, machine learning, Privacy-Preserving, Healthcare, Electronic Health Records, Differential PrivacyAbstract
Federated Learning (FL) has evolved to become a ground breaking example of privacy-orientated medical artificial intelligence and addressing the crucial challenge of working with distributed medical information and preserving patient confidentiality. The presented large-scale review unites the research done in 2021-2025 in different areas of healthcare, including the Internet of Medical Things (IoMT), electronic health records (EHRs), medical imaging, wearable health monitoring, prediction of chronic diseases, and clinical research. The discussion highlights how FL helps in training models collaboratively and maintaining data locality, relying on secure aggregation and blockchain integration as well as on the family of different privacy methods. The studies demonstrate that FL is able to predict, achieve cross-site generalisation, and reduce the risk of privacy compared to traditional centralised learning. Nevertheless, it does have significant limitations, such as non-IID and heterogeneous data, communication bottlenecks, inconsistent quality of data, small sample sizes, and a high computational cost, especially in an architectural arrangement with blockchain enhancements. In addition, the lack of standard assessment structures and the fact that there are no large-scale-scale-scale commercial applications prevents broader application in clinical settings. This review summarises the current knowledge, highlights gaps in the methodology, and provides the recommendations on future research directions to create FL systems that can scale, provide security and interoperability, and provide intelligent healthcare applications of the next generation.
DOI: https://doi.org/10.24321/3051.4266.202507
References
A. Vaid et al., “Federated learning of electronic health records to improve mortality prediction in hospitalized patients with COVID-19: Machine learning approach,” JMIR Med Inform, vol. 9, no. 1, Jan. 2021, doi: 10.2196/24207.
F. Cremonesi et al., “The need for multimodal health data modeling: A practical approach for a federatedlearning healthcare platform,” J Biomed Inform, vol. 141, May 2023, doi: 10.1016/j.jbi.2023.104338.
