计算机应用 ›› 2021, Vol. 41 ›› Issue (3): 891-897.DOI: 10.11772/j.issn.1001-9081.2020060783

所属专题: 前沿与综合应用

• 前沿与综合应用 • 上一篇    下一篇

基于特征融合和动态多尺度空洞卷积的超声甲状腺分割网络

胡屹杉1,2,3, 秦品乐1,2,3, 曾建潮1,2,3, 柴锐1,2,3, 王丽芳1,2,3   

  1. 1. 山西省医学影像与数据分析工程研究中心(中北大学), 太原 030051;
    2. 中北大学 大数据学院, 太原 030051;
    3. 山西省医学影像人工智能工程技术研究中心(中北大学), 太原 030051
  • 收稿日期:2020-06-09 修回日期:2020-08-09 出版日期:2021-03-10 发布日期:2020-12-22
  • 通讯作者: 曾建潮
  • 作者简介:胡屹杉(1996-),男,山西运城人,硕士研究生,主要研究方向:机器学习、计算机视觉、医学影像分析;秦品乐(1978-),男,山西长治人,教授,博士,CCF会员,主要研究方向:医学影像分析、大数据、机器视觉;曾建潮(1963-),男,山西太原人,教授,博士生导师,博士,CCF会员,主要研究方向:复杂系统的维护决策、健康管理;柴锐(1985-),男,山西运城人,讲师,博士,主要研究方向:医学影像处理;王丽芳(1977-),女,山西长治人,副教授,博士,主要研究方向:机器视觉、大数据、医学图像处理。
  • 基金资助:
    山西省研究生教育创新项目(2020SY381)。

Ultrasound thyroid segmentation network based on feature fusion and dynamic multi-scale dilated convolution

HU Yishan1,2,3, QIN Pinle1,2,3, ZENG Jianchao1,2,3, CHAI Rui1,2,3, WANG Lifang1,2,3   

  1. 1. Shanxi Medical Imaging and Data Analysis Engineering Research Center(North University of China), Taiyuan Shanxi 030051, China;
    2. School of Data Science and Technology, North University of China, Taiyuan Shanxi 030051, China;
    3. Shanxi Medical Imaging Artificial Intelligence Engineering Technology Research Center(North University of China), Taiyuan Shanxi 030051, China
  • Received:2020-06-09 Revised:2020-08-09 Online:2021-03-10 Published:2020-12-22
  • Supported by:
    This work is partially supported by the Innovation Program of Graduate Student Education of Shanxi Province (2020SY381).

摘要: 针对甲状腺超声影像中甲状腺组织大小和形态的多样性以及周边组织的复杂性,提出了一种基于特征融合和动态多尺度空洞卷积的超声甲状腺分割网络。首先,利用不同膨胀率的空洞卷积和动态滤波器来融合不同感受野下的全局语义特征与不同范围的上下文详情的语义特征,从而提升网络对多尺度目标的适应性与准确度;然后,在特征降维时采用混合上采样方式,以增强高维语义特征的空间信息和低维空间特征的上下文信息;最后,采用空间注意力机制来优化图像的低维特征,并采用高低维特征融合的方式使高低维特征信息在保留重要特征的同时摒弃冗余信息以及使网络对于图像前背景的区分能力得到增强。实验结果表明,所提方法在甲状腺超声影像公开数据集上达到了0.963±0.026的准确率、0.84±0.03的召回率和0.79±0.03的dice系数。可见所提方法能较好地解决组织形态差异性大以及周边组织复杂的问题。

关键词: 图像分割, 注意力机制, 空洞卷积, 超声影像, 特征融合

Abstract: Concerning the the size and morphological diversity of thyroid tissue and the complexity of surrounding tissue in thyroid ultrasound images, an ultrasound thyroid segmentation network based on feature fusion and dynamic multi-scale dilated convolution was proposed. Firstly, the dilated convolutions with different dilation rates and dynamic filters were used to fuse the global semantic information of different receptive domains and the semantic information in the context details with different ranges, so as to improve the adaptability and accuracy of the network to multi-scale targets. Then, the hybrid upsampling method was used to enhance the spatial information of high-dimensional semantic features and the context information of low-dimensional spatial features during feature dimensionality reduction. Finally, the spatial attention mechanism was introduced to optimize the low-dimensional features of the image, and the method of fusing high- and low-dimensional features was applied to retain the useful features of high- and low-dimensional feature information with the elimination of the redundant information and improve the network's ability to distinguish the background and foreground of the image. Experimental results show that the proposed method has an accuracy rate of 0.963±0.026, a recall rate of 0.84±0.03 and a dice coefficient of 0.79±0.03 in the public dataset of thyroid ultrasound images. It can be seen that the proposed method can solve the problems of large difference of tissue morphology and complex surrounding tissues.

Key words: image segmentation, attention mechanism, dilated convolution, ultrasound image, feature fusion

中图分类号: