Decoupling Wavelet Sub-bands for Single Source Domain Generalization in Fundus Image Segmentation

2026-03-30

AI Summary

This work addresses the degradation of generalization performance in fundus image segmentation caused by variations in imaging devices and clinical settings when only a single source domain with annotations is available. To tackle this challenge, the authors propose WaveSDG, a novel network that leverages wavelet subband decomposition to disentangle anatomical structures from domain-specific appearance features. A dedicated WISER module is introduced to separately process low-frequency global topology and high-frequency directional edges, thereby enhancing cross-domain generalization without requiring additional labeled target data. Experimental results demonstrate that WaveSDG consistently outperforms seven state-of-the-art methods across five unseen target domains in both optic cup and disc segmentation tasks, achieving the highest Dice scores, the lowest 95% Hausdorff distances, and reduced performance variance.

Abstract

Domain generalization in fundus imaging is challenging due to variations in acquisition conditions across devices and clinical settings. The inability to adapt to these variations causes performance degradation on unseen domains for deep learning models. Besides, obtaining annotated data across domains is often expensive and privacy constraints restricts their availability. Although single-source domain generalization (SDG) offers a realistic solution to this problem, the existing approaches frequently fail to capture anatomical topology or decouple appearance from anatomical features. This research introduces WaveSDG, a new wavelet-guided segmentation network for SDG. It decouples anatomical structure from domain-specific appearance through a wavelet sub-band decomposition. A novel Wavelet-based Invariant Structure Extraction and Refinement (WISER) module is proposed to process encoder features by leveraging distinct semantic roles of each wavelet sub-band. The module refines low-frequency components to anchor global anatomy, while selectively enhancing directional edges and suppressing noise within the high-frequency sub-bands. Extensive ablation studies validate the effectiveness of the WISER module and its decoupling strategy. Our evaluations on optic cup and optic disc segmentation across one source and five unseen target datasets show that WaveSDG consistently outperforms seven state-of-the-art methods. Notably, it achieves the best balanced Dice score and lowest 95th percentile Hausdorff distance with reduced variance, indicating improved accuracy, robustness, and cross-domain stability.

Problem

• domain generalization
• fundus image segmentation
• single-source domain generalization
• anatomical topology
• appearance decoupling

Innovation

• wavelet sub-band decomposition
• domain generalization
• anatomical-appearance decoupling
• WISER module
• fundus image segmentation