计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (2): 264-272.doi: 10.11896/jsjkx.250300137

• 计算机图形学&多媒体 • 上一篇    下一篇

聚焦边界和多尺度特征融合的脑卒中病灶分割

刘晨红1, 李凤莲1, 阳佳2, 王夙喆1, 陈桂军1   

  1. 1 太原理工大学电子信息工程学院 太原 030024
    2 中国运载火箭技术研究院研究发展中心 北京 100076
  • 收稿日期:2025-03-25 修回日期:2025-05-22 发布日期:2026-02-10
  • 通讯作者: 李凤莲(lifenglian@tyut.edu.cn)
  • 作者简介:(liuchenhong0402@link.tyut.edu.cn)
  • 基金资助:
    国家自然科学基金(62171307);山西省自然科学基金(202103021224113)

Boundary-focused Multi-scale Feature Fusion Network for Stroke Lesion Segmentation

LIU Chenhong1, LI Fenglian1, YANG Jia2, WANG Suzhe1, CHEN Guijun1   

  1. 1 College of Electronic Information Engineering,Taiyuan University of Technology,Taiyuan 030024,China
    2 R&D Department,China Academy of Launch Vehicle Technology,Beijing 100076,China
  • Received:2025-03-25 Revised:2025-05-22 Online:2026-02-10
  • About author:LIU Chenhong,born in 2000,postgra-duate,is a member of CCF(No.Z5771G).Her main research interest is medical image segmentation.
    LI Fenglian,born in 1972,Ph.D,professor,Ph.D supervisor,is a member of CCF(No.65512S).Her main research interest is medical signal processing and analysis.
  • Supported by:
    National Natural Science Foundation of China(62171307) and Natural Science Foundation of Shanxi Province,China(202103021224113).

PDF (PC)

82

摘要： 借助计算机辅助诊断技术定位脑卒中发病区域,有助于提升临床医生的诊断与治疗效率。目前,医学图像里脑卒中病变与健康组织边界常不清晰,而现有多数基于深度学习的分割方法在识别小尺寸病变及处理模糊边界方面存在不足。为此,提出了一种创新的边界感知多尺度特征集成网络(Boundary-Aware Multi-Scale Feature Integration Network,BAMFNet),用于更准确地进行脑卒中病灶分割。BAMFNet中设计了多尺度特征提取模块,该模块利用卷积神经网络和Transformer的混合架构来捕获局部和全局多尺度特征,通过内卷积机制有效减少冗余信息。此外,提出了边界增强和融合模块,其在特征融合过程中有效增强了边界区域的特征。融合模块集成了多层次的信息交互机制,增强了边界特征表示,实现了深层特征和浅层特征的有效结合。在ATLAS v1.2,ATLAS v2.0和ISLES 2022卒中数据集上的实验证明,BAMFNet的骰子相似系数分别达到了62.93%,61.79%和86.66%,优于对比方法。

关键词: 深度学习, 病灶分割, 多尺度特征融合, 边界增强, Transformer

Abstract: Computer-aided diagnosis helps clinicians locate stroke-affected brain regions,improving diagnostic and therapeutic efficiency.Currently,the boundaries between stroke lesions and healthy tissues in medical images are often unclear,and most exis-ting deep learning-based segmentation methods lack effectiveness in identifying small-sized lesions and handling blurred boundaries.To address this,the boundary-aware multi-scale feature integration network(BAMFNet) is proposed for more accurate stroke lesion segmentation.In BAMFNet,the multi-scale feature extraction module combines convolutional neural networks and Transformers to capture local and global features at multiple scales and uses involution to reduce information redundancy.The boundary enhancement and fusion module strengthens boundary-region features during fusion and integrates a multi-level information interaction mechanism.This enhances the boundary feature representation and combines deep and shallow features effectively.Experiments on the ATLAS v1.2,ATLAS v2.0 and ISLES 2022 stroke datasets show BAMFNet achieves Dice similarity coefficients of 62.93%,61.79%,and 86.66% respectively,outperforming other methods.

Key words: Deep learning, Lesion segmentation, Multi-scale feature fusion, Boundary enhancement, Transformer

中图分类号: 

  • TP391
[1]LIU M B,HE X Y,YANG X H,et al.Summary of “China Cardiovascular Health and Disease Report 2023”:Prevalence of Cardiovascular Diseases and Status of Interventional Diagnosis and Treatment[J].Chinese Journal of Interventional Cardiology,2024,32(10):541-550.
[2]FEIGIN V,BRAININ M,NOPRVING B,et al.World StrokeOrganization(WSO):Global Stroke Fact Sheet 2022[J].International Journal of Stroke:Official Journal of the International Stroke Society,2022,17(1):18-29.
[3]RONNEBERGER O,FISCHER P,BROX T.U-Net:Convolu-tional Networks for Biomedical Image Segmentation[C]//Medical Image Computing and Computer-Assisted Intervention.2015:234-241.
[4]ZHOU Z,SIDDIQUEE M,TAJBAKHAS N,et al.UNet++:Redesigning Skip Connections to Exploit Multiscale Features in Image Segmentation[J].IEEE Transactions on Medical Imaging,2020,39(6):1856-1867.
[5]OKTAY O,SCHLEMPE J,FOLGOC L,et al.Attention U-Net:Learning Where to Look for the Pancreas[C]//Medical Imaging with Deep Learning.2022.
[6]LEI T,WANG R,ZHANG Y,et al.DefED-Net:Deformable Encoder-Decoder Network for Liver and Liver Tumor Segmentation[J].IEEE Transactions on Radiation and Plasma Medical Sciences,2022,6(1):68-78.
[7]ZHOU Y,HUANG W,DONG P,et al.D-UNet:A Dimension-Fusion U Shape Network for Chronic Stroke Lesion Segmentation[J].IEEE/ACM Transactions on Computational Biology and Bioinformatics,2021,18(3):940-950.
[8]LONG J,SHELHAMER E,DARRELL T.Fully Convolutional Networks for Semantic Segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2015,39(4):640-651.
[9]WANG L B,WANG S M.Fundus vascular image segmentation algorithm based on attention mechanism[J].Computer Science,2019,51(S2):359-364.
[10]LI X W,XU W X,CHEN Y,et al.Joint image registration for automatic segmentation of stroke lesions[J].Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition),2024,36(4):729-737.
[11]SHU X,LI X,ZHANG X,et al.MRAU-net:Multi-scale residual attention U-shaped network for medical image segmentation[J].Computers and Electrical Engineering,2024,118:109479.
[12]VASWANI A,SHAZEER N,PARMARR N,et al.Attention is All you Need[C]//Advances in Neural Information Processing Systems.2017.
[13]CHEN J,LU Y,YU Q,et al.TransUNet:Transformers Make Strong Encoders for Medical Image Segmentation[J].arXiv:2102.04306,2021.
[14]ZHOU H Y,GUO J,ZHANG Y,et al.nnFormer:VolumetricMedical Image Segmentation via a 3D Transformer[J].IEEE Transactions on Image Processing,2023,32:4036-4045.
[15]WANG R,CHEN S,JI C,et al.Boundary-aware context neural network for medical image segmentation[J].Medical Image Analysis,2022,78:102395.
[16]LI C,MAO Y,GUO Y,et al.Multi-Dimensional Cascaded Net with Uncertain Probability Reduction for Abdominal Multi-Organ Segmentation in CT Sequences[J].Computer Methods and Programs in Biomedicine,2022,221:106887.
[17]XIE X,PAN X,SHAO F,et al.MCI-Net:Multi-scale context integrated network for liver CT image segmentation[J].Compu-ters and Electrical Engineering,2022,101:108085.
[18]LIN Y,ZHANG D,FANG X,et al.Rethinking Boundary Detection in Deep Learning Models for Medical Image Segmentation[C]//Information Processing in Medical Imaging.Cham:Sprin-ger,2023:730-742.
[19]CHEN Y,FAN H,XU B,et al.Drop an Octave:Reducing Spatial Redundancy in Convolutional Neural Networks With Octave Convolution[C]//IEEE/CVF International Conference on Computer Vision.2019:3434-3443.
[20]ZHOU D,KANG B,JIN X,et al.DeepViT:Towards Deeper Vision Transformer[J].arXiv:2103.11886,2021.
[21]LIEW S,LO B,DONNELLY M,et al.A large,curated,open-source stroke neuroimaging dataset to improve lesion segmentation algorithms[J].Scientific Data,2022,9(1):320.
[22]HERNANDEZ M,DE L,HANNING U,et al.ISLES 2022:A multi-center magnetic resonance imaging stroke lesion segmentation dataset[J].Scientific Data,2022,9(1):762.
[23]QI K,YANG H,LI C,et al.X-Net:Brain Stroke Lesion Segmentation Based on Depthwise Separable Convolution and Long-Range Dependencies[C]//Medical Image Computing and Computer Assisted Intervention.2019:247-255.
[24]YANG H,HUANG W,QI K,et al.CLCI-Net:Cross-Level Fusion and Context Inference Networks for Lesion Segmentation of Chronic Stroke[C]//Medical Image Computing and Computer Assisted Intervention.2019:266-274.
[25]ZHANG Y,LIU H,HU Q.TransFuse:Fusing Transformersand CNNs for Medical Image Segmentation[C]//Medical Image Computing and Computer Assisted Intervention.2021:14-24.
[26]WU Z,ZHANG X,LI F,et al.TransRender:a transformer-based boundary rendering segmentation network for stroke lesions[J].Frontiers in Neuroscience,2023,17:1259677.
[27]NI Z,BIAN G,ZHOU X,et al.RAUNet:Residual Attention U-Net for Semantic Segmentation of Cataract Surgical Instruments[C]//Neural Information Processing.2019:139-149.
[28]CHEN B,LIU Y,ZHANG Z,et al.TransAttUnet:Multi-Level Attention-Guided U-Net With Transformer for Medical Image Segmentation[J].IEEE Transactions on Emerging Topics in Computational Intelligence,2024,8(1):55-68.
[29]PETIT O,THOME N,RAMBOUR C,et al.U-Net Transfor-mer:Self and Cross Attention for Medical Image Segmentation[C]//Machine Learning in Medical Imaging.2021:267-276.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发