Accounting for data variability in multi-institutional distributed deep learning for medical imaging

Abstract

Our work is the first to identify and address challenges of sample size and label distribution variability in simulated distributed deep learning for medical imaging. Future work is needed to address other sources of real-world data variability.