From Denoising Training to Test-Time Adaptation: Enhancing Domain Generalization for Medical Image Segmentation

TL;DR

This paper introduces DeY-Net and DeTTA, novel methods that improve domain generalization in medical image segmentation by incorporating denoising training and test-time adaptation, effectively handling domain shifts and noise.

Contribution

The paper proposes DeY-Net with an auxiliary denoising decoder and DeTTA for test-time adaptation, advancing domain generalization in medical image segmentation.

Findings

Significant improvements in liver segmentation benchmarks.

Outperforms state-of-the-art domain generalization methods.

Effective adaptation to noisy and target domain data.

Abstract

In medical image segmentation, domain generalization poses a significant challenge due to domain shifts caused by variations in data acquisition devices and other factors. These shifts are particularly pronounced in the most common scenario, which involves only single-source domain data due to privacy concerns. To address this, we draw inspiration from the self-supervised learning paradigm that effectively discourages overfitting to the source domain. We propose the Denoising Y-Net (DeY-Net), a novel approach incorporating an auxiliary denoising decoder into the basic U-Net architecture. The auxiliary decoder aims to perform denoising training, augmenting the domain-invariant representation that facilitates domain generalization. Furthermore, this paradigm provides the potential to utilize unlabeled data. Building upon denoising training, we propose Denoising Test Time Adaptation (DeTTA) that further: (i) adapts the model to the target domain in a sample-wise manner, and (ii) adapts to the noise-corrupted input. Extensive experiments conducted on widely-adopted liver segmentation benchmarks demonstrate significant domain generalization improvements over our baseline and state-of-the-art results compared to other methods. Code is available at https://github.com/WenRuxue/DeTTA.