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.
@article{arxiv.2407.07557,
title = {Real World Federated Learning with a Knowledge Distilled Transformer for Cardiac CT Imaging},
author = {Malte Tölle and Philipp Garthe and Clemens Scherer and Jan Moritz Seliger and Andreas Leha and Nina Krüger and Stefan Simm and Simon Martin and Sebastian Eble and Halvar Kelm and Moritz Bednorz and Florian André and Peter Bannas and Gerhard Diller and Norbert Frey and Stefan Groß and Anja Hennemuth and Lars Kaderali and Alexander Meyer and Eike Nagel and Stefan Orwat and Moritz Seiffert and Tim Friede and Tim Seidler and Sandy Engelhardt},
journal= {arXiv preprint arXiv:2407.07557},
year = {2025}
}