Domain Game: Disentangle Anatomical Feature for Single Domain Generalized Segmentation

TL;DR

This paper introduces Domain Game, a novel framework for single domain medical image segmentation that enhances feature disentanglement by leveraging geometric transformations to separate diagnostic features from domain-specific features, improving cross-site generalization.

Contribution

The paper proposes a new feature disentanglement framework using geometric transformations and a game-theoretic approach to improve single domain generalization in medical image segmentation.

Findings

11.8% performance boost in prostate segmentation

10.5% improvement in brain tumor segmentation

Effective separation of diagnostic and domain-specific features

Abstract

Single domain generalization aims to address the challenge of out-of-distribution generalization problem with only one source domain available. Feature distanglement is a classic solution to this purpose, where the extracted task-related feature is presumed to be resilient to domain shift. However, the absence of references from other domains in a single-domain scenario poses significant uncertainty in feature disentanglement (ill-posedness). In this paper, we propose a new framework, named \textit{Domain Game}, to perform better feature distangling for medical image segmentation, based on the observation that diagnostic relevant features are more sensitive to geometric transformations, whilist domain-specific features probably will remain invariant to such operations. In domain game, a set of randomly transformed images derived from a singular source image is strategically encoded into two separate feature sets to represent diagnostic features and domain-specific features, respectively, and we apply forces to pull or repel them in the feature space, accordingly. Results from cross-site test domain evaluation showcase approximately an ~11.8% performance boost in prostate segmentation and around ~10.5% in brain tumor segmentation compared to the second-best method.