Evaluating federated learning approaches for mammography under breast density heterogeneity.

Abstract

Purpose: Breast density is a key factor that influences mammography interpretation and is a major source of heterogeneity in multicenter datasets. Such heterogeneity poses challenges for collaborative machine learning across institutions, particularly in federated learning (FL). We aim to evaluate the impact of breast density-induced heterogeneity on FL for mammography image classification and to assess the robustness of common FL algorithms in realistic clinical settings.

Approach: We conducted experiments on mammography datasets under two scenarios: (1) a strongly heterogeneous setting where each participating site contributed exclusively low- or high-density cases, based on the BI-RADS density score, and (2) a population-based setting simulating breast density distributions observed in White and Asian populations. For the strongly heterogeneous setting, we evaluated two configurations: one with two clients, where the cases were grouped as BI-RADS A-B and C-D, and one with four clients, where each site contained cases of a single BI-RADS density. We compared three FL methods (i.e., FedAvg, FedProx, and SCAFFOLD) against centralized training, local-only training, and naïve aggregation approaches, including model ensembling and weight averaging.

Results: Across both scenarios, FL consistently achieved performance comparable to centralized training, whereas local models and naïve aggregation approaches underperformed in the presence of strong heterogeneity. Notably, FedAvg achieved accuracy on par with or exceeding centralized training, demonstrating resilience to breast density-induced data imbalance without requiring specialized heterogeneity mitigation algorithms.

Conclusions: These findings show that FL can effectively address breast density-related heterogeneity, supporting its feasibility for real-world mammography workflows. The demonstrated robustness of FedAvg underscores the potential for broad clinical deployment of FL, enabling collaborative model development while maintaining data privacy.