EUVIMEDAtrasti EUVIMED
Crossref

WaveAtt ‐ TransUNet : Spatial–Frequency Wavelet‐Transformer Network for Joint Retinal Layer and Biomarker Segmentation in OCT

Shujun Men, Lei Zhang, Jiamin Wang, Yanke Li, Li Huo

International Journal of Imaging Systems and Technology · 2026-11 · 36 (6)

DOI: 10.1002/ima.70484 ↗

At a glance

Journal article in International Journal of Imaging Systems and Technology, published online on 2026-11-15. Written by Shujun Men and 4 co-authors. 42 references. Funded by: Beijing Science and Technology Planning Project, Natural Science Foundation of Beijing Municipality, State Key Laboratory of Information Photonics and Optical Communications.

SantraukaBibliografiniai duomenysCiteRelated studies

Abstract (original language)

Accurate joint segmentation of retinal layers and pathological biomarkers in optical coherence tomography (OCT) images remains a challenging task due to inherent speckle noise and frequency‐dependent boundary ambiguity in coherent imaging. Existing deep learning‐based segmentation methods primarily rely on spatial domain feature learning, lacking explicit modeling of the frequency domain characteristics of OCT signals. To address this limitation, we propose WaveAtt‐TransUNet, a space‐frequency‐aware wavelet‐transformer network for joint segmentation of retinal layers and biomarkers in OCT images. The proposed framework combines multi‐scale wavelet decomposition with Transformer‐based global context modeling to enhance boundary‐sensitive feature representations. Specifically, a Multi‐Scale Wavelet Attention Convolution (MS‐WAC) module is introduced at skip connections to selectively emphasize useful frequency components, while a Dual‐Mode Downsampling (DMD) module is introduced to preserve high‐frequency structural details as resolution decreases. Furthermore, the interactive prompting mechanism is introduced to improve adaptability to small‐scale or low‐contrast biomarkers. Extensive experiments on the AROI and OIMHS datasets demonstrate that the proposed method significantly outperforms current representative segmentation models based on convolutional neural networks and Transformers on multiple evaluation metrics.

Funding

  • Beijing Science and Technology Planning Project
  • Natural Science Foundation of Beijing Municipality
  • State Key Laboratory of Information Photonics and Optical Communications

Cite

APA

Men, S., Zhang, L., Wang, J., Li, Y., & Huo, L. (2026). WaveAtt ‐ TransUNet : Spatial–Frequency Wavelet‐Transformer Network for Joint Retinal Layer and Biomarker Segmentation in OCT. International Journal of Imaging Systems and Technology, 36(6). https://doi.org/10.1002/ima.70484

Vancouver

Men S, Zhang L, Wang J, Li Y, Huo L. WaveAtt ‐ TransUNet : Spatial–Frequency Wavelet‐Transformer Network for Joint Retinal Layer and Biomarker Segmentation in OCT. International Journal of Imaging Systems and Technology. 2026;36(6). doi:10.1002/ima.70484

BibTeX
@article{Men2026,
  author = {Men, Shujun and Zhang, Lei and Wang, Jiamin and Li, Yanke and Huo, Li},
  title = {WaveAtt ‐ TransUNet : Spatial–Frequency Wavelet‐Transformer Network for Joint Retinal Layer and Biomarker Segmentation in OCT},
  journal = {International Journal of Imaging Systems and Technology},
  year = {2026},
  volume = {36},
  number = {6},
  doi = {10.1002/ima.70484}
}
Start for free →