文章详情

摘要

心脏MRI图像的精准分割对心血管疾病诊疗至关重要，但由于心脏结构复杂、边界模糊以及组织对比度较低的问题，使得传统的U-Net网络实现准确分割仍面临挑战。文章提出一种多尺度小波变换网络MS-WTUNet，该网络以U-Net为骨架，在编码与解码各层嵌入了与注意力相结合的小波块，在频域中强化纹理与边缘信息，有效提升了模型对复杂边界的表征能力。此外，模型辅以跨层多尺度特征融合与分层深度监督损失，进一步优化了模型从局部细节到全局语义的学习过程。在公开ACDC数据集上的实验表明，MS-WTUNet能够将心肌等边界模糊结构的分割精度提升至91.70%，为心脏MRI图像的自动分割提供了一种性能优异的解决方案。

关键词

医学图像分割；U-Net；小波变换；MRI图像

参考文献

1. [1] CHEN C，QIN C，QIU H Q，et al. Deep Learning for Cardiac Image Segmentation: A Review [J/OL].arXiv:1911.03723 [eess.IV].（2019-11-09）.https://arxiv.org/abs/1911.03723.
2. [2] NEHA F，BHATI D，SHUKLA D K，et al. U-Net in Medical Image Segmentation: A Review of its Applications Across Modalities [J/OL].arXiv:2412.02242 [eess.IV].（2024-12-03）.https://arxiv.org/abs/2412.02242.
3. [3] GUO T T，ZHANG T P，LIM E，et al. A Review of Wavelet Analysis and its Applications: Challenges and Opportunities [J].IEEE Access，2022，10：58869-58903.
4. [4] XU Y，QUAN R X，XU W T，et al. Advances in Medical Image Segmentation: A Comprehensive Review of Traditional, Deep Learning and Hybrid Approaches [J].Bioengineering，2024，11（10）：1034（2024-10-16）[2025-10-17].https://doi.org/10.3390/bioengineering11101034.
5. [5] BACCOUCH W，OUESLATI S，SOLAIMAN B，et al. A Comparative Study of CNN and U-Net Performance for Automatic Segmentation of Medical Images: Application to Cardiac MRI [J].Procedia Computer Science，2023，219：1089-1096.
6. [6] RONNEBERGER O，FISCHER P，BROX T. U-Net: Convolutional Networks for Biomedical Image Segmentation [C]//18th International Conference on Medical Image Computing and Computer-assisted Intervention.Munich：Springer International Publishing，2015：234-241.
7. [7] OKTAY O，SCHLEMPER J，FOLGOC L L，et al. Attention U-Net: Learning Where to Look for the Pancreas [J/OL].arXiv:1804.03999 [cs.CV].（2018-04-11）.https://arxiv.org/abs/1804.03999.
8. [8] DIAKOGIANNIS F I，WALDNER F，CACCETTA P，et al. ResUNet-a: A Deep Learning Framework for Semantic Segmentation of Remotely Sensed Data [J/OL].arXiv:1904.00592 [cs.CV].（2019-04-01）.https://arxiv.org/abs/1904.00592.
9. [9] LI X M，CHEN H，QI X J，et al. H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation from CT Volumes [J/OL].arXiv:1709.07330 [cs.CV].（2017-09-21）.https://arxiv.org/abs/1709.07330.
10. [10] ZHOU Z W，SIDDIQUEE M M R，TAJBAKHSH N，et al. UNet++: A Nested U-Net Architecture for Medical Image Segmentation [J/OL].arXiv:1807.10165 [cs.CV].（2018-07-18）.https://arxiv.org/abs/1807.10165.
11. [11] JIN Y P，ANGELINI E，LAINE A. Wavelets in Medical Image Processing: Denoising, Segmentation, and Registration [M]//SURI J S，WILSON D L，LAXMINARAYAN S. Handbook of Biomedical Image Analysis: Segmentation Models Part A.Boston：Springer US，2005：305-358.
12. [12] PHAM T H，LI X Q，NGUYEN K-D. seUNet-Trans: A Simple Yet Effective UNet-Transformer Model for Medical Image Segmentation [J].IEEE Access，2024，12：122139-122154.
13. [13] ZHANG Z J，LI X H，MA X H，et al. E-Transunet: Enhanced Transunet for Medical Image Segmentation [C]//2024 IEEE International Symposium on Biomedical Imaging (ISBI). Athens：IEEE，2024：1-5.
14. [14] CAO H，WANG Y Y，CHEN J，et al. Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentation [J/OL].arXiv:2105.05537 [eess.IV].（2021-05-12）.https://arxiv.org/abs/2105.05537.
15. [15] XU G P，WU X G，ZHANG X，et al. LeViT-UNet: Make Faster Encoders with Transformer for Medical Image Segmentation [J/OL].arXiv:2107.08623 [cs.CV].（2021-07-19）.https://arxiv.org/abs/2107.08623.

基金项目

国家自然科学基金项目（61401383）；陕西省自然科学基础研究计划项目（2021JM-518）

作者简介

黄佳敏（2002—），女，汉族，陕西咸阳人，硕士在读，研究方向：图像处理；张小波（1979—），男，汉族，陕西咸阳人，教授，博士，研究方向：图像处理和自然语言处理。