WDFFU-Mamba: A Wavelet-guided Dual-attention Feature Fusion Mamba for Breast Tumor Segmentation in Ultrasound Images

TL;DR

This paper tackles the challenge of breast tumor segmentation in ultrasound images amid speckle noise and blurred boundaries. It introduces WDFFU-Mamba, a Mamba-based U-Net-like network enhanced with a Wavelet denoising High-Frequency guided Feature (WHF) module and a Dual Attention Feature Fusion (DAFF) module to improve boundary delineation and multi-level feature fusion. Across two public BUS datasets, the method achieves superior Dice scores and HD95, with ablations validating the effectiveness of WHF and DAFF. The model demonstrates strong generalization and computational efficiency, highlighting its potential for real-world clinical deployment in breast tumor ultrasound analysis.

Abstract

Breast ultrasound (BUS) image segmentation plays a vital role in assisting clinical diagnosis and early tumor screening. However, challenges such as speckle noise, imaging artifacts, irregular lesion morphology, and blurred boundaries severely hinder accurate segmentation. To address these challenges, this work aims to design a robust and efficient model capable of automatically segmenting breast tumors in BUS images.We propose a novel segmentation network named WDFFU-Mamba, which integrates wavelet-guided enhancement and dual-attention feature fusion within a U-shaped Mamba architecture. A Wavelet-denoised High-Frequency-guided Feature (WHF) module is employed to enhance low-level representations through noise-suppressed high-frequency cues. A Dual Attention Feature Fusion (DAFF) module is also introduced to effectively merge skip-connected and semantic features, improving contextual consistency.Extensive experiments on two public BUS datasets demonstrate that WDFFU-Mamba achieves superior segmentation accuracy, significantly outperforming existing methods in terms of Dice coefficient and 95th percentile Hausdorff Distance (HD95).The combination of wavelet-domain enhancement and attention-based fusion greatly improves both the accuracy and robustness of BUS image segmentation, while maintaining computational efficiency.The proposed WDFFU-Mamba model not only delivers strong segmentation performance but also exhibits desirable generalization ability across datasets, making it a promising solution for real-world clinical applications in breast tumor ultrasound analysis.

Figures (6)

• Figure 1: (a) Overall architecture of the WDFFU-Mamba network, which consists of a Mamba-based backbone, a WHF module for capturing edge details in feature representations, and a DAFF module that fuses deep and shallow features through dual attention mechanisms; (b) Structure of the VSS Block.
• Figure 2: (a) Wavelet denoising process in the WHF module, including wavelet transform, Gaussian filtering, and inverse wavelet transform; (b) High-Frequency Guided Feature Extraction (HGFE) in the WHF module.
• Figure 3: Structure of the DAFF module. Two different inputs are processed through spatial and channel attention modules, respectively, and the resulting attention weights are multiplied with the fused features.
• Figure 4: Visualization results of the ablation study. The labels on the left indicate the dataset (including BUSI and BUS) and the tumor type (benign or malignant), while the labels below indicate different combinations of network components. Green contours represent the ground truth, and red contours represent the predicted results.
• Figure 5: Attention heatmaps of the network. The blue regions indicate the areas of primary focus by the network.