Adaptive Aggregation Weights for Federated Segmentation of Pancreas MRI
Federated learning (FL) enables collaborative model training across institutions without sharing sensitive data, making it an attractive solution for medical imaging tasks. However, traditional FL methods, such as Federated Averaging (FedAvg), face difficulties in general-izing across domains due to variations in imaging protocols and pa-tient demographics across institutions. This challenge is particularly evident in pancreas MRI segmentation, where anatomical variability and imaging artifacts significantly impact performance. In this paper, we conduct a comprehensive evaluation of FL algorithms for pancreas MRI segmentation and introduce a novel approach that in-corporates adaptive aggregation weights. By dynamically adjusting the contribution of each client during model aggregation, our method accounts for domain-specific differences and improves generalization across heterogeneous datasets. Experimental results demonstrate that our approach enhances segmentation accuracy and reduces the impact of domain shift compared to conventional FL methods while maintaining privacy-preserving capabilities. Significant performance improvements are observed across multiple hospitals (centers).
Authors
- Hongyi Pan (ORCID: https://orcid.org/0000-0001-9421-3936)
- Rajesh N. Keswani (ORCID: https://orcid.org/0000-0002-9196-2138)
- Michael B. Wallace (ORCID: https://orcid.org/0000-0002-6446-5785)
- Ulaş Bağcı (ORCID: https://orcid.org/0000-0001-7379-6829)
- Görkem Durak (ORCID: https://orcid.org/0000-0002-1608-1955)
- Ziyue Xu (ORCID: https://orcid.org/0000-0002-5728-6869)
- Pallavi Tiwari (ORCID: https://orcid.org/0000-0001-9477-4856)
- Marco J. Bruno (ORCID: https://orcid.org/0000-0001-9181-5499)
- Alpay Medetalibeyoğlu (ORCID: https://orcid.org/0000-0002-5829-9186)
- Candice W. Bolan (ORCID: https://orcid.org/0000-0002-9847-3599)
- Tamas A. Gonda (ORCID: https://orcid.org/0000-0002-0093-3738)
- Yury Velichko (ORCID: https://orcid.org/0000-0002-2287-5727)
- Elif Keleş (ORCID: https://orcid.org/0000-0001-8103-797X)
- Ivo G. Schoots (ORCID: https://orcid.org/0000-0002-9804-0603)
- Concetto Spampinato
- Yavuz Taktak
- Zheyuan Zhang (ORCID: https://orcid.org/0000-0002-4146-0164)
- Halil Ertugrul Aktas
- Michael G. GORRins
Institutions
- Northwestern University (US)
- University of Wisconsin–Madison (US)
- Johns Hopkins University (US)
- Jacksonville College (US)
- Erasmus MC (NL)
- University of Catania (IT)
- Nvidia (United States) (US)
- Mayo Clinic in Florida (US)
- New York University (US)
- Istanbul University (TR)
Publication Details
- Published
- 2025-04-14
- DOI
- https://doi.org/10.1109/isbi60581.2025.10981148
- Citations
- 1
- Primary Topic
- Brain Tumor Detection and Classification
- Type
- article
- Field-Weighted Citation Impact
- 0.84
Funders
- National Institutes of Health
- National Heart, Lung, and Blood Institute
- National Cancer Institute
- National Institute of Diabetes and Digestive and Kidney Diseases