Learning Semantic Directions for Feature Augmentation in Domain-Generalized Medical Segmentation

TL;DR

This paper introduces a domain generalization framework for medical image segmentation that uses implicit feature perturbations and adaptive constraints to improve robustness across unseen clinical domains, addressing domain shift challenges.

Contribution

The proposed method uniquely employs semantic direction selection and covariance-based intensity sampling for feature augmentation in medical segmentation.

Findings

Outperforms existing domain generalization methods on public benchmarks.

Achieves consistent segmentation accuracy across diverse clinical domains.

Enhances robustness to domain-specific variations in medical images.

Abstract

Medical image segmentation plays a crucial role in clinical workflows, but domain shift often leads to performance degradation when models are applied to unseen clinical domains. This challenge arises due to variations in imaging conditions, scanner types, and acquisition protocols, limiting the practical deployment of segmentation models. Unlike natural images, medical images typically exhibit consistent anatomical structures across patients, with domain-specific variations mainly caused by imaging conditions. This unique characteristic makes medical image segmentation particularly challenging. To address this challenge, we propose a domain generalization framework tailored for medical image segmentation. Our approach improves robustness to domain-specific variations by introducing implicit feature perturbations guided by domain statistics. Specifically, we employ a learnable semantic direction selector and a covariance-based semantic intensity sampler to modulate domain-variant features while preserving task-relevant anatomical consistency. Furthermore, we design an adaptive consistency constraint that is selectively applied only when feature adjustment leads to degraded segmentation performance. This constraint encourages the adjusted features to align with the original predictions, thereby stabilizing feature selection and improving the reliability of the segmentation. Extensive experiments on two public multi-center benchmarks show that our framework consistently outperforms existing domain generalization approaches, achieving robust and generalizable segmentation performance across diverse clinical domains.