《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (12): 3891-3899.DOI: 10.11772/j.issn.1001-9081.2021101737

• 多媒体计算与计算机仿真 • 上一篇    

面向医学图像分割的多注意力融合网络

李鸿, 邹俊颖(), 谭茜成, 李贵洋   

  1. 四川师范大学 计算机科学学院,成都 610066
  • 收稿日期:2021-10-09 修回日期:2022-01-04 接受日期:2022-01-24 发布日期:2022-04-08 出版日期:2022-12-10
  • 通讯作者: 邹俊颖
  • 作者简介:李鸿(1995—),男,四川眉山人,硕士研究生,主要研究方向:机器学习、医学图像分割
    谭茜成(1996—),男,湖南怀化人,硕士研究生,主要研究方向:数据挖掘、迁移学习
    李贵洋(1975—),男,四川宜宾人,副教授,博士,主要研究方向:大数据存储编码、机器学习。
  • 基金资助:
    国家自然科学基金资助项目(11905153)

Multi-attention fusion network for medical image segmentation

Hong LI, Junying ZOU(), Xicheng TAN, Guiyang LI   

  1. School of Computer Science,Sichuan Normal University,Chengdu Sichuan 610066,China
  • Received:2021-10-09 Revised:2022-01-04 Accepted:2022-01-24 Online:2022-04-08 Published:2022-12-10
  • Contact: Junying ZOU
  • About author:LI Hong, born in 1995, M. S. candidate. His research interests include machine learning, medical image segmentation.
    TAN Xicheng, born in 1996, M. S. candidate. His research interests include data mining, transfer learning.
    LI Guiyang, born in 1975, Ph. D., associate professor. His research interests include big data storage encoding, machine learning.
  • Supported by:
    National Natural Science Foundation of China(11905153)

摘要:

在深度医学图像分割领域中,TransUNet是当前先进的分割模型之一。但其编码器未考虑相邻分块之间的局部联系,在解码器上采样过程中缺乏通道间信息的交互。针对以上问题,提出一种多注意力融合网络(MFUNet)模型。首先,在编码器部分引入特征融合模块(FFM)来增强模型对Transformer中相邻分块间的局部联系并且保持图片本身的空间位置关系;其次,在解码器部分引入双通道注意力(DCA)模块来融合多级特征的通道信息,以增强模型对通道间关键信息的敏感度;最后,通过结合交叉熵损失和Dice损失来加强模型对分割结果的约束。在Synapse和ACDC公共数据集上进行实验,可以看出,MFUNet的Dice相似系数(DSC)分别达到了81.06%和90.91%;在Synapse数据集上的Hausdorff距离(HD)与基线模型TransUNet相比减小了11.5%;在ACDC数据集中右心室和心肌两部分的分割精度与基线模型TransUNet相比分别提升了1.43个百分点和3.48个百分点。实验结果表明,MFUNet在医学图像的内部填充和边缘预测方面均能实现更好的分割效果,有助于提升医生在临床实践中的诊断效率。

关键词: 医学图像分割, 多器官, 自注意力, 通道注意力, 深度学习

Abstract:

In the field of deep medical image segmentation, TransUNet (merit both Transformers and U-Net) is one of the current advanced segmentation models. However, the local connection between adjacent blocks in its encoder is not considered, and the inter-channel information is not interactive during the upsampling process of the decoder. To address the above problems, a Multi-attention FUsion Network (MFUNet) model was proposed. Firstly, a Feature Fusion Module (FFM) was introduced in encoder part to enhance the local connections between adjacent blocks in the Transformer and maintain the spatial location relationships of the images themselves. Then, a Double Channel Attention (DCA) module was introduced in the decoder part to fuse the channel information of multi-level features, which enhanced the sensitivity of the model to the key information between channels. Finally, the model's constraints on the segmentation results was strengthened by combining cross-entropy loss and Dice loss. By conducting experiments on Synapse and ACDC public datasets, it can be seen that MFUNet achieves Dice Similarity Coefficient (DSC) of 81.06% and 90.91%, respectively. Compared with the baseline model TransUNet, MFUNet achieved an 11.5% reduction in Hausdorff Distance (HD) on the Synapse dataset, and improved segmentation accuracy by 1.43 and 3.48 percentage points on the ACDC dataset for both the right ventricular and myocardial components, respectively. The experimental results show that MFUNet can achieve better segmentation results in both internal filling and edge prediction of medical images, which can help improve the diagnostic efficiency of doctors in clinical practice.

Key words: medical image segmentation, multiple organs, Self-Attention (SA), channel attention, deep learning

中图分类号: