The Devil is in the Statistics: Mitigating and Exploiting Statistics Difference for Generalizable Semi-supervised Medical Image Segmentation

TL;DR

This paper introduces a novel semi-supervised domain generalization method for medical image segmentation that leverages domain shifts in feature statistics to improve unseen domain performance, using multiple and aggregated statistics branches.

Contribution

It proposes a new approach with multiple statistics-individual branches and a statistics-aggregated branch to handle domain shifts and improve generalization in semi-supervised medical image segmentation.

Findings

Outperforms recent state-of-the-art methods on three datasets.

Effectively mitigates domain shift impact on pseudo-label quality.

Enhances unseen domain generalization through statistical perturbations.

Abstract

Despite the recent success of domain generalization in medical image segmentation, voxel-wise annotation for all source domains remains a huge burden. Semi-supervised domain generalization has been proposed very recently to combat this challenge by leveraging limited labeled data along with abundant unlabeled data collected from multiple medical institutions, depending on precisely harnessing unlabeled data while improving generalization simultaneously. In this work, we observe that domain shifts between medical institutions cause disparate feature statistics, which significantly deteriorates pseudo-label quality due to an unexpected normalization process. Nevertheless, this phenomenon could be exploited to facilitate unseen domain generalization. Therefore, we propose 1) multiple statistics-individual branches to mitigate the impact of domain shifts for reliable pseudo-labels and 2) one statistics-aggregated branch for domain-invariant feature learning. Furthermore, to simulate unseen domains with statistics difference, we approach this from two aspects, i.e., a perturbation with histogram matching at image level and a random batch normalization selection strategy at feature level, producing diverse statistics to expand the training distribution. Evaluation results on three medical image datasets demonstrate the effectiveness of our method compared with recent SOTA methods. The code is available at https://github.com/qiumuyang/SIAB.