One model to unite them all: Personalized federated learning of multi-contrast MRI synthesis

Curation of large, diverse MRI datasets via multi-institutional collaborations can help improve learningof generalizable synthesis models that reliably translate source- onto target-contrast images. To facilitatecollaborations, federated learning (FL) adopts decentralized model training while mitigating privacy concernsby avoiding sharing of imaging data. However, conventional FL methods can be impaired by the inherentheterogeneity in the data distribution, with domain shifts evident within and across imaging sites. Here weintroduce the first personalized FL method for MRI Synthesis (pFLSynth) that improves reliability against dataheterogeneity via model specialization to individual sites and synthesis tasks (i.e., source-target contrasts).To do this, pFLSynth leverages an adversarial model equipped with novel personalization blocks that controlthe statistics of generated feature maps across the spatial/channel dimensions, given latent variables specificto sites and tasks. To further promote communication efficiency and site specialization, partial networkaggregation is employed over later generator stages while earlier generator stages and the discriminatorare trained locally. As such, pFLSynth enables multi-task training of multi-site synthesis models with highgeneralization performance across sites and tasks. Comprehensive experiments demonstrate the superiorperformance and reliability of pFLSynth in MRI synthesis against prior federated methods