FedSemiDG: Domain Generalized Federated Semi-supervised Medical Image Segmentation

TL;DR

This paper introduces FedSemiDG, a novel federated semi-supervised learning framework for medical image segmentation that enhances domain generalization to unseen data by combining adaptive model aggregation, pseudo label refinement, and perturbation-invariant learning.

Contribution

The paper proposes FGASL, a new framework with GAA, DR, and PIA strategies, addressing domain shift and improving model generalization in federated semi-supervised medical image segmentation.

Findings

Outperforms state-of-the-art FSSL methods on four medical segmentation tasks.

Achieves robust generalization to unseen domains.

Effectively handles domain shift with novel aggregation and pseudo-label strategies.

Abstract

Medical image segmentation is challenging due to the diversity of medical images and the lack of labeled data, which motivates recent developments in federated semi-supervised learning (FSSL) to leverage a large amount of unlabeled data from multiple centers for model training without sharing raw data. However, what remains under-explored in FSSL is the domain shift problem which may cause suboptimal model aggregation and low effectivity of the utilization of unlabeled data, eventually leading to unsatisfactory performance in unseen domains. In this paper, we explore this previously ignored scenario, namely domain generalized federated semi-supervised learning (FedSemiDG), which aims to learn a model in a distributed manner from multiple domains with limited labeled data and abundant unlabeled data such that the model can generalize well to unseen domains. We present a novel framework, Federated Generalization-Aware SemiSupervised Learning (FGASL), to address the challenges in FedSemiDG by effectively tackling critical issues at both global and local levels. Globally, we introduce Generalization-Aware Aggregation (GAA), assigning adaptive weights to local models based on their generalization performance. Locally, we use a Dual-Teacher Adaptive Pseudo Label Refinement (DR) strategy to combine global and domain-specific knowledge, generating more reliable pseudo labels. Additionally, Perturbation-Invariant Alignment (PIA) enforces feature consistency under perturbations, promoting domain-invariant learning. Extensive experiments on four medical segmentation tasks (cardiac MRI, spine MRI, bladder cancer MRI and colorectal polyp) demonstrate that our method significantly outperforms state-of-the-art FSSL and domain generalization approaches, achieving robust generalization on unseen domains.