FedSODA: Federated Cross-assessment and Dynamic Aggregation for Histopathology Segmentation

TL;DR

FedSODA introduces a federated learning framework with synthetic-driven cross-assessment and dynamic aggregation to improve histopathology image segmentation across diverse and imbalanced datasets.

Contribution

It proposes novel cross-assessment and dynamic aggregation methods to address data heterogeneity and sample imbalance in federated histopathology segmentation.

Findings

Significantly improves segmentation accuracy across multiple datasets.

Effectively mitigates data heterogeneity and sample imbalance.

Outperforms existing federated learning approaches.

Abstract

Federated learning (FL) for histopathology image segmentation involving multiple medical sites plays a crucial role in advancing the field of accurate disease diagnosis and treatment. However, it is still a task of great challenges due to the sample imbalance across clients and large data heterogeneity from disparate organs, variable segmentation tasks, and diverse distribution. Thus, we propose a novel FL approach for histopathology nuclei and tissue segmentation, FedSODA, via synthetic-driven cross-assessment operation (SO) and dynamic stratified-layer aggregation (DA). Our SO constructs a cross-assessment strategy to connect clients and mitigate the representation bias under sample imbalance. Our DA utilizes layer-wise interaction and dynamic aggregation to diminish heterogeneity and enhance generalization. The effectiveness of our FedSODA has been evaluated on the most extensive histopathology image segmentation dataset from 7 independent datasets. The code is available at https://github.com/yuanzhang7/FedSODA.