Organizations: Research Center for Federated Learning in Medicine, Indiana University School of Medicine, Indianapolis, IN, USA. · Division of Computational Pathology, Department of Pathology and Laboratory Medicine, Indiana University School of Medicine, Indianapolis, IN, USA. · Indiana University Melvin and Bren Simon Comprehensive Cancer Center, Indianapolis, IN, USA. · Sage Bionetworks, Seattle, WA, USA. · MLCommons, Dover, DE, USA. · NVIDIA, Santa Clara, CA, USA. · Departments of Radiology and Imaging Sciences; Neurological Surgery; Biostatistics and Health Data Science, Indiana University School of Medicine, Indianapolis, IN, USA. · Department of Computer Science, Luddy School of Informatics, Computing, and Engineering, Indiana University, Indianapolis, IN, USA
International challenges have become the standard for developing and benchmarking biomedical image analysis methods using centralized multi-institutional data. Federated learning (FL) describes a paradigm-shift on multi-institutional collaborations, where models are trained across institutions while their data are always retained locally. The Federated Tumor Segmentation (FeTS) challenge, first held in 2021, was the first challenge dedicated to FL in medical imaging and editions in 2022 and 2024 pushed the frontiers of real-world method validation. FeTS++ 2027 returns after a two-year gap with a substantially larger and more diverse cohort, and with the two "pluses" that name the challenge: an official classification task alongside segmentation, and a move of the infrastructure to NVIDIA FLARE (NVFlare). Built on the clinically acquired, multi-institutional multi-parametric MRI data of the BraTS 2025 Lighthouse challenge, FeTS++ spans gliomas and meningiomas across patient cohorts from institutions in North America, Europe, Asia, and Africa, and focuses on ranking submitted methodological approaches according to their model utility and an efficiency-aware score. Task 1 asks participants to find the best strategy for aggregating knowledge from per-institution segmentation models, while Task 2 asks them to distinguish glioma from meningioma under the same federated protocol, in which the entity label is heavily confounded with the site of origin.
Figures & tables
Figure 1: One federated round in the FeTS++ 2027 NVFlare simulator. The server broadcasts the global model, each client receives it with flare.receive() , trains it locally with the participant’s local_train() , and returns the updated model with flare.send() . The server then aggregates the accepted updates into the next global model, and the cycle repeats while flare.is_running() . Blue elements are participant-editable, namely the local training, the aggregation, and the serializable FLModel payload shown at the bottom, which is the only information that crosses the link. Schematic drawn from the challenge starter code, namely the client runner and the participant aggregator. Each simulated site represents a logical client associated with an institutional data partition. The figure illustrates the challenge’s federated execution protocol within the simulator.
Figure 2: Training cases per site in the FeTS++ 2027 cohort. Bars are ordered by decreasing size and the sites are renumbered for presentation, so the indices do not correspond to the identifiers in the released metadata. The dashed line marks 50 cases, below which 19 of the 26 sites fall, and per-site counts span 4 to 542 cases.
Figure 3: Entity composition per site in the FeTS++ 2027 training cohort. Each bar is one site, split into glioma and meningioma cases, in the same renumbered order as Figure 2 . 16 sites are glioma-only and two are meningioma-only, as marked by the arrows, and the percentages above the 4 largest mixed sites give their glioma share. Sites are renumbered for presentation, and the indices do not correspond to the identifiers in the released metadata.
Department of Electrical and Computer Engineering, The University of Hong Kong, Hong Kong, China · School of Communication Engineering, Hangzhou Dianzi University, Hang Zhou, China · Department of Radiation Oncology, Stanford University, California, USA +1
National Institute for Data Science in Health and Medicine, Xiamen University, Xiamen, 361005, Fujian, China · Jarvis Research Center, Tencent YouTu Lab, Shenzhen, 518075, Guangdong, China · Department of Computer Science at School of Informatics, Xiamen University, Xiamen, 361005, Fujian, China +1