Federated Medical Image Classification under Class and Domain Imbalance exploiting Synthetic Sample Generation

TL;DR

This paper introduces FedSSG, a federated learning framework that generates synthetic samples to address class and domain imbalance in medical image classification, improving model performance across diverse institutions.

Contribution

The novel contribution is a synthetic sample generation strategy integrated into federated learning to handle class and domain imbalance in medical imaging.

Findings

Significantly improves model accuracy on underrepresented classes.

Enhances generalization across heterogeneous imaging devices.

Maintains minimal computational overhead at clients.

Abstract

Exploiting deep learning in medical imaging faces critical challenges, including strict privacy constraints, heterogeneous imaging devices with varying acquisition properties, and class imbalance due to the uneven prevalence of pathologies. In this work, we propose FedSSG, a novel Federated Learning framework that addresses domain shifts caused by diverse imaging devices while mitigating the under-representation of rare pathologies. The key contribution is a strategy for generating synthetic samples and distributing them across clients to improve coverage of both underrepresented pathologies and imaging devices. Experimental results demonstrate that our approach significantly enhances model performance and generalization across heterogeneous institutions, with minimal computational overhead at the client side.