Operationalizing federated learning in real-world healthcare applications – Fed-Ops

The practical use of federated learning (FL) in real-world healthcare applications has been so far limited in scope and extent, essentially by focusing in showcasing the feasibility of FL in ad-hoc analysis based on pre-defined models and datasets. We are in dire need for solutions to allow the long-term exploitation of a federated learning environment in the healthcare setting. The aim of Fed-Ops is to lead to a set of methodology and tools to operationalize the use federated learning in healthcare. To enable Continuous Integration and Development in FL, we will focus on the development of innovative Data Integration and Continual Learning approaches dedicated to the analysis of federated clinical and medical imaging datasets. Fed-Ops will contribute to the open-source framework Fed-BioMed, allowing clinical translation into the existing federated learning infrastructure connecting a group of hospitals of the Unicancer Consortium for the development of AI models for the prediction of immunotherapy response.