Adaptive Aggregation Weights for Federated Segmentation of Pancreas MRI

TL;DR

The paper addresses domain shift across institutions in federated learning for pancreas MRI segmentation and proposes adaptive aggregation weights that dynamically adjust each client's influence based on local validation loss gaps. By updating aggregation weights with the loss-gap metric $G_i = Q_i - P_i$ and a decaying step size $s = 0.1\left(1 - t/T\right)$, the method improves generalization over standard FedAvg while preserving privacy. Experimental results across seven centers show notable gains in Dice and Jaccard on T1-weighted MRI and reductions in HD95 and ASSD on T2-weighted MRI, with performance approaching a no-FL upper bound. The approach enhances robustness of FL-based pancreas segmentation in heterogeneous clinical data, enabling privacy-preserving, cross-site deployment that better supports clinical decision-making.

Abstract

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 generalizing across domains due to variations in imaging protocols and patient 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 incorporates 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).

Figures (1)

• Figure 1: Adaptive aggregation weights for federated pancreas MRI segmentation. Unlike the traditional FedAvg algorithm, where aggregation weights remain fixed, our method updates weights dynamically based on the validation loss gap before and after aggregation, ensuring that clients contributing positively to model performance are given more weight in subsequent rounds.