A scoping review published by a team of researchers from the University of Queensland, Monash University, CSIRO, Queensland Cyber Infrastructure Foundation and the Queensland University of Technology has revealed that federated learning (FL) in healthcare, while promising enhanced patient privacy, requires comprehensive governance frameworks that are currently lacking in the field.
The research, which is part of the National Infrastructure for federated learNing in DigitAl health (NINA) project led by the Queensland Digital Health Centre at UQ, examined 39 studies to develop the first consolidated framework for the governance of federated learning in healthcare settings.
Unlike traditional machine learning that centralises patient data, federated learning allows AI models to be trained across multiple healthcare institutions without sharing sensitive patient information.
Lead author of the study, Dr Rebekah Eden from the Queensland Digital Health Centre, said that while federated learning minimises data-sharing challenges, it introduces new governance complexities that healthcare organisations are not adequately prepared for.
“As part of the study we [the research team] examined federated learning governance alongside related techniques, including machine learning and federated data networks,” Dr Eden said.
“This review highlighted that, contrary to the view that federated learning solves all data sharing problems, there are some governance challenges that need to be addressed to realise the potential of federated learning in healthcare and safeguard our health consumers and health professionals.
“Successful implementation will require improvements like greater coordination between participating institutions, enhanced training for local data stewards at each participating site, more complex contractual arrangements due to multiple participating parties and new ethical considerations for distributed AI model development.”
Significantly, the research found that only seven out of 39 studies specifically addressed federated learning governance in healthcare, with most research focusing on traditional machine learning approaches that may not adequately address federated learning's unique challenges.
Dr Eden and co-authors suggest that failure to address these governance challenges could undermine public trust and impede the translation of federated learning into clinical practice.
“Healthcare organisations implementing federated learning initiatives need structured governance approaches that balance technical capabilities with robust ethical and regulatory frameworks,” she said.
“We need to get this right because FL provides great potential to improve the sustainable and equitable delivery of healthcare by enhancing diagnosis and treatment decisions.
“This research provides the critical first step to forming an evidence base to facilitate the effective governance of federated learning in healthcare.”
The research was funded by the Medical Research Future Fund (MRFF) National Critical Research Infrastructure Scheme.
Read the full paper here - https://doi.org/10.1111/imj.70097
