现代信息科技2026，Vol.10Issue(1)：52-57,6.DOI:10.19850/j.cnki.2096-4706.2026.01.011
MS-WTUNet面向心脏MRI分割的多尺度小波变换网络
A Multi-scale Wavelet Transform Network for Cardiac MRI Segmentation of MS-WTUNet
摘要
Abstract
The accurate segmentation of cardiac MRI images is crucial for the diagnosis and treatment of cardiovascular diseases.However,the complex structure of the heart,blurred boundaries,and low tissue contrast pose significant challenges for traditional U-Net architectures in achieving precise segmentation.This paper proposes a Multi-Scale Wavelet Transform Network named MS-WTUNet.The network employs U-Net as its backbone,embeds wavelet blocks with attention into each layer of the encoder and decoder,and enhances texture and edge information in the frequency domain,thereby effectively improving the model's capability in characterizing complex boundaries.Furthermore,the model incorporates cross-layer multi-scale feature fusion and a hierarchical deep supervision loss,optimizing the learning process from local details to global semantics.Experiments on the public ACDC dataset demonstrate that MS-WTUNet can effectively improve the segmentation accuracy of myocardial and other boundary-ambiguous structures to 91.70%,providing a high-performance solution for automatic segmentation of cardiac MRI images.关键词
医学图像分割/U-Net/小波变换/MRI图像Key words
medical image segmentation/U-Net/Wavelet Transform/MRI image分类
信息技术与安全科学引用本文复制引用
黄佳敏,张小波..MS-WTUNet面向心脏MRI分割的多尺度小波变换网络[J].现代信息科技,2026,10(1):52-57,6.基金项目
国家自然科学基金项目(61401383) （61401383）
陕西省自然科学基础研究计划项目(2021JM-518) （2021JM-518）
