Layer-wise Noise Guided Selective Wavelet Reconstruction for Robust Medical Image Segmentation

2025/11/20 by Lu, Yuting, Wang, Ziliang, Xu, Weixin +4
Computer Science · #Advanced Image Processing Techniques #Adversarial Robustness in Machine Learning #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Graphics (cs.GR)

paper · doi:10.48550/arxiv.2511.16162

openalex publication_date 2025/11/20 · openalex created_date 2025/11/23 · openalex updated_date 2026/07/28

Abstract

Clinical deployment requires segmentation models to stay stable under distribution shifts and perturbations. The mainstream solution is adversarial training (AT) to improve robustness; however, AT often brings a clean--robustness trade-off and high training/tuning cost, which limits scalability and maintainability in medical imaging. We propose Layer-wise Noise-Guided Selective Wavelet Reconstruction (LNG-SWR). During training, we inject small, zero-mean noise at multiple layers to learn a frequency-bias prior that steers representations away from noise-sensitive directions. We then apply prior-guided selective wavelet reconstruction on the input/feature branch to achieve frequency adaptation: suppress noise-sensitive bands, enhance directional structures and shape cues, and stabilize boundary responses while maintaining spectral consistency. The framework is backbone-agnostic and adds low additional inference overhead. It can serve as a plug-in enhancement to AT and also improves robustness without AT. On CT and ultrasound datasets, under a unified protocol with PGD-L_∞/L₂ and SSAH, LNG-SWR delivers consistent gains on clean Dice/IoU and significantly reduces the performance drop under strong attacks; combining LNG-SWR with AT yields additive gains. When combined with adversarial training, robustness improves further without sacrificing clean accuracy, indicating an engineering-friendly and scalable path to robust segmentation. These results indicate that LNG-SWR provides a simple, effective, and engineering-friendly path to robust medical image segmentation in both adversarial and standard training regimes.

Citations

• CARE: Anti-entanglement Ultrasound Image Segmentation via Channel-Aware Region Extrication
• U-KAN Makes Strong Backbone for Medical Image Segmentation and Generation
• VM-UNet: Vision Mamba UNet for Medical Image Segmentation
• CMUNeXt: An Efficient Medical Image Segmentation Network based on Large Kernel and Skip Fusion
• Frequency Domain Model Augmentation for Adversarial Attack
• Frequency-driven Imperceptible Adversarial Attack on Semantic Similarity
• UNeXt: MLP-based Rapid Medical Image Segmentation Network
• Holistic Adversarial Robustness of Deep Learning Models
• Transformers in Medical Imaging: A Survey
• Mixed Transformer U-Net For Medical Image Segmentation
• A Frequency Perspective of Adversarial Robustness
• TransAttUnet: Multi-level Attention-guided U-Net with Transformer for Medical Image Segmentation
• DS-TransUNet:Dual Swin Transformer U-Net for Medical Image Segmentation
• Image Segmentation Using Deep Learning: A Survey
• A Fourier Perspective on Model Robustness in Computer Vision
• Towards Deep Learning Models Resistant to Adversarial Attacks
• Rethinking Atrous Convolution for Semantic Image Segmentation
• Adversarial Machine Learning at Scale
• Towards Evaluating the Robustness of Neural Networks
• U-Net: Convolutional Networks for Biomedical Image Segmentation
• Explaining and Harnessing Adversarial Examples
• Fully Convolutional Networks for Semantic Segmentation
• Intriguing properties of neural networks
• The Lung Image Database Consortium (LIDC) and Image Database Resource Initiative (IDRI): A Completed Reference Database of Lung Nodules on CT Scans

Related

• Adaptive Adversarial Training to Improve Adversarial Robustness of DNNs for Medical Image Segmentation and Detection
• CRISP: Constrained Refinement via Iterative Squeezing Process for Robust Medical Image Segmentation under Domain Shift
• MambaEviScrib: Mamba and Evidence-Guided Consistency Enhance CNN Robustness for Scribble-Based Weakly Supervised Ultrasound Image Segmentation
• Robustification of Segmentation Models Against Adversarial Perturbations\n In Medical Imaging
• FaRMamba: Frequency-based learning and Reconstruction aided Mamba for Medical Segmentation