Research Information System University of Greifswald

Tolle M*, Garthe P, Scherer C, Seliger J, Leha A, Kruger N, Simm S¹, Martin S, Eble S, Kelm H, Bednorz M, Andre F, Bannas P, Diller G, Frey N, Gross S², Hennemuth A, Kaderali L¹, Meyer A, Nagel E, Orwat S, Seiffert M, Friede T, Seidler T, Engelhardt S

1 - Institut für Bioinformatik

2 - Zentrum für Innere Medizin / Klinik für Innere Medizin B

Federated learning is a renowned technique for utilizing decentralized data while preserving privacy. However, real-world applications often face challenges like partially labeled datasets, where only a few locations have certain expert annotations, leaving large portions of unlabeled data unused. Leveraging these could enhance transformer architectures' ability in regimes with small and diversely annotated sets. We conduct the largest federated cardiac CT analysis to date (n = 8, 104) in a real-world setting across eight hospitals. Our two-step semi-supervised strategy distills knowledge from task-specific CNNs into a transformer. First, CNNs predict on unlabeled data per label type and then the transformer learns from these predictions with label-specific heads. This improves predictive accuracy and enables simultaneous learning of all partial labels across the federation, and outperforms UNet-based models in generalizability on downstream tasks. Code and model weights are made openly available for leveraging future cardiac CT analysis.

Tolle, Malte Garthe, Philipp Scherer, Clemens Seliger, Jan Moritz Leha, Andreas Kruger, Nina Simm, Stefan Martin, Simon Eble, Sebastian Kelm, Halvar Bednorz, Moritz Andre, Florian Bannas, Peter Diller, Gerhard Frey, Norbert Gross, Stefan Hennemuth, Anja Kaderali, Lars Meyer, Alexander Nagel, Eike Orwat, Stefan Seiffert, Moritz Friede, Tim Seidler, Tim Engelhardt, Sandy eng England NPJ Digit Med. 2025 Feb 6;8(1):88. doi: 10.1038/s41746-025-01434-3.