“…Aggregated models can be impaired by the heterogeneity in the data distribution naturally evident for multi-institutional datasets [29,32] due to different scanners, acquisition parameters etc. Previous studies on FL-based medical imaging have introduced several prominent approaches to cope with data heterogeneity in segmentation [5,22,23,27,31,38], classification [3,22,40], and reconstruction [10,11] tasks. However, influence of data heterogeneity on FL-based MRI contrast translation remains understudied.…”