CoDG-Net: Structure-Guided Style Diffusion and Collaborative Learning to Mitigate Catastrophic Forgetting in Medical Image Domain Generalization

2026-09-01

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence

AI Summary

This study addresses the problem of catastrophic forgetting in source domains caused by the pursuit of cross-domain robustness in domain generalization for medical image segmentation. To mitigate this issue, we propose a Structure-Guided Style Diffusion and Collaborative Learning Network. Methodologically, we introduce a novel style diffusion augmentation strategy constrained by frequency-domain structural priors to enhance data diversity. Architecturally, we design a dual-branch collaborative interaction mechanism based on learning bias, which facilitates adaptive knowledge transfer to alleviate source-domain forgetting. Extensive experiments demonstrate that the proposed method outperforms existing state-of-the-art approaches on both single-source and multi-source benchmarks, significantly improving target-domain segmentation performance while effectively suppressing source-domain forgetting.

Abstract

Domain Generalization (DG) for medical image segmentation is both highly challenging and critically important. However, existing medical DG methods largely overlook the issue of Catastrophic Forgetting (CF): Models often sacrifice their ability to retain source-domain knowledge while pursuing cross-domain robustness. This can directly threaten diagnostic safety in already-deployed clinical scenarios. To address this, we investigate data augmentation strategies and catastrophic forgetting for medical image DG segmentation. First, we propose a structure-guided style diffusion augmentation method. Constrained by anatomical structure consistency in the frequency domain, this method performs cross-domain diffusion on the amplitude spectrum, generating samples with more diverse and broader style coverage to better support domain generalization. Then, we design a collaborative learning network with a dual-branch interactive architecture (CoDG-Net), together with a novel learning bias-guided strategy that adaptively regulates knowledge transfer at both the layer level and the task level, thereby effectively mitigating catastrophic forgetting on the source domain. Experiments and ablation studies on single-source and multi-source medical DG benchmark datasets demonstrate that CoDG-Net not only outperforms existing state-of-the-art methods in target-domain segmentation performance, but also achieves a lower forgetting rate on the source-domain data. The code is available at: https://github.com/wangprocess/CoDG-Net.

Problem

• Domain Generalization
• Medical Image Segmentation
• Catastrophic Forgetting
• Data Augmentation

Innovation

• Domain Generalization
• Catastrophic Forgetting
• Style Diffusion
• Collaborative Learning
• Medical Image Segmentation