Learning Site-specific Styles for Multi-institutional Unsupervised Cross-modality Domain Adaptation

TL;DR

This paper introduces a method for multi-institutional unsupervised cross-modality domain adaptation in medical imaging, using site-specific style transfer and self-training to improve segmentation performance across different data sources.

Contribution

The paper proposes a dynamic network for site-specific image translation and a self-training approach, advancing multi-institutional domain adaptation in medical imaging.

Findings

Achieved 1st place in the crossMoDA 2023 challenge.

Effective site-specific style transfer improves domain adaptation.

Self-training further reduces domain gap.

Abstract

Unsupervised cross-modality domain adaptation is a challenging task in medical image analysis, and it becomes more challenging when source and target domain data are collected from multiple institutions. In this paper, we present our solution to tackle the multi-institutional unsupervised domain adaptation for the crossMoDA 2023 challenge. First, we perform unpaired image translation to translate the source domain images to the target domain, where we design a dynamic network to generate synthetic target domain images with controllable, site-specific styles. Afterwards, we train a segmentation model using the synthetic images and further reduce the domain gap by self-training. Our solution achieved the 1st place during both the validation and testing phases of the challenge. The code repository is publicly available at https://github.com/MedICL-VU/crossmoda2023.