Placeholder Show Content


While multiple factors, both biological and external, impact disease, AI studies in medicine are often confined to small and non-diverse patient cohorts. Such limitation typically stems from obstacles of large-scale data sharing and data privacy issues. Federated learning (FL) has emerged as one potential solution for AI developments, enabling training across a network of hospitals without direct data sharing. Here, we present an FL platform for pediatric posterior fossa brain tumors, FL-PedBrain, and evaluate its performance on a diverse and realistic multi-center pediatric cohort. We target pediatric brain tumors given the overall scarcity of such datasets, even within tertiary care pediatric hospitals. Our platform orchestrates federated training that performs an end-to-end joint tumor classification and segmentation across 19 participating international sites. FL-PedBrain exhibits less than a 1.5% decrease in classification and a 3% reduction in segmentation performance compared to the traditional approach using training with centrally shared data. We find that federated training boosts performance compared to a model trained solely on the largest single site. For example, FL boosts segmentation performance from 20 to 30% on three external and out-of-network, hold-out sites. Finally, we explore the underlying sources of data heterogeneity, such as variations in image quality, and examine robustness of FL in real world scenarios due to data imbalances.


Type of resource Dataset, still image
Date modified July 3, 2024
Publication date March 27, 2024


Author Lee, Edward
Author Yeom, Kristen


Subject Federated Learning
Genre Data
Genre Image
Genre Data sets
Genre Dataset

Bibliographic information

Access conditions

Use and reproduction
User agrees that, where applicable, content will not be used to identify or to otherwise infringe the privacy or confidentiality rights of individuals. Content distributed via the Stanford Digital Repository may be subject to additional license and use restrictions applied by the depositor.
This work is licensed under a Creative Commons Attribution 4.0 International license (CC BY).

Preferred citation

Preferred citation
Lee, E. and Yeom, K. (2024). FLPedBrain. Stanford Digital Repository. Available at https://purl.stanford.edu/bf070wx6289. https://doi.org/10.25740/bf070wx6289.


Contact information

Also listed in

Loading usage metrics...