计算机应用 ›› 2020, Vol. 40 ›› Issue (6): 1799-1805.DOI: 10.11772/j.issn.1001-9081.2019101839

• 虚拟现实与多媒体计算 • 上一篇    下一篇

基于深度卷积特征光流的形变医学图像配准算法

张家岗1, 李达平1, 杨晓东1, 邹茂扬1,2, 吴锡1, 胡金蓉1   

  1. 1.成都信息工程大学 计算机学院,成都 610225
    2.中国科学院 成都计算机应用研究所,成都 610041
  • 收稿日期:2019-10-29 修回日期:2019-12-17 出版日期:2020-06-10 发布日期:2020-06-18
  • 通讯作者: 胡金蓉(1983—)
  • 作者简介:张家岗(1998—),男,山东济南人,主要研究方向:深度学习、图像处理、人工智能.李达平(1992—),男,四川达州人,硕士研究生,主要研究方向:医学图像分析、深度学习、人工智能.杨晓东(1994—),男,安徽淮南人,硕士研究生,主要研究方向:图像处理、人工智能.邹茂扬(1974—),女,四川成都人,副教授,博士,主要研究方向:数据挖掘、机器学习、图像处理.吴锡(1980—),男,四川成都人,教授,博士,主要研究方向:人工智能、认知计算、图像处理.胡金蓉(1983—),女,四川南充人,副教授,博士,主要研究方向:医学图像分析、机器学习、人工智能.
  • 基金资助:
    国家自然科学基金资助项目(61303126,61602390);四川省科技计划项目(2016RZ0051,2018RZ0072);教育部春晖计划项目(Z2015108)。

Deformable medical image registration algorithm based on deep convolution feature optical flow

ZHANG Jiagang1, LI Daping1, YANG Xiaodong1, ZOU Maoyang1,2, WU Xi1, HU Jinrong1   

  1. 1. School of Computer Science, Chengdu University of Information and Technology, Chengdu Sichuan 610225, China
    2. Chengdu Institute of Computer Application, Chinese Academy of Sciences, Chengdu Sichuan 610041, China
  • Received:2019-10-29 Revised:2019-12-17 Online:2020-06-10 Published:2020-06-18
  • Contact: HU Jinrong, born in 1983, Ph. D., associate professor. Her research interests include medical image analysis, machine learning, artificial intelligence.
  • About author:ZHANG Jiagang, born in 1998. His research interests include deep learning, image processing, artificial intelligence.LI Daping, born in 1992, M. S. candidate. His research interests include medical image analysis, deep learning, artificial intelligence.YANG Xiaodong, born in 1994, M. S. candidate. His research interests include image processing, artificial intelligence.ZOU Maoyang, born in 1974, Ph. D., associate professor. Her research interests include data mining, machine learning, image processing.WU Xi, born in 1980, Ph. D., professor. His research interests include artificial intelligence, cognitive computing, image processing.HU Jinrong, born in 1983, Ph. D., associate professor. Her research interests include medical image analysis, machine learning, artificial intelligence.
  • Supported by:
    National Natural Science Foundation of China (61303126, 61602390), the Sichuan Science and Technology Plan (2016RZ0051, 2018RZ0072), the Chunhui Plan of the Ministry of Education (Z2015108).

摘要: 光流法是一种基于光流场模型的重要而有效的形变配准算法。针对现有光流法所用特征质量不高使得配准结果不够准确的问题,将深度卷积神经网络特征和光流法相结合,提出了基于深度卷积特征光流(DCFOF)的形变医学图像配准算法。首先利用深度卷积神经网络稠密地提取图像中每个像素所在图像块的深度卷积特征,然后基于固定图像和浮动图像间的深度卷积特征差异求解光流场。通过提取图像的更为精确和鲁棒的深度学习特征,使求得的光流场更接近真实形变场,提升了配准精度。实验结果表明,所提算法能够更有效地解决形变医学图像配准问题,其配准精度优于Demons算法、尺度不变特征变换(SIFT) Flow算法以及医学图像专业配准软件Elastix。

关键词: 图像配准, 形变配准, 卷积神经网络, 特征提取, 光流法

Abstract: Optical flow method is an important and effective deformation registration algorithm based on optical flow field model. Aiming at the problem that the feature quality used by the existing optical flow method is not high enough to make the registration result accurate, combining the features of deep convolutional neural network and optical flow method, a deformable medical image registration algorithm based on Deep Convolution Feature Based Optical Flow (DCFOF) was proposed. Firstly, the deep convolution feature of the image block where each pixel in the image was located was densely extracted by using a deep convolutional neural network, and then the optical flow field was solved based on the deep convolution feature difference between the fixed image and the floating image. By extracting more accurate and robust deep learning features of the image, the optical flow field obtained was closer to the real deformation field, and the registration accuracy was improved. Experimental results show that the proposed algorithm can solve the problem of deformable medical image registration effectively, and has the registration accuracy better than those of Demons algorithm, Scale-Invariant Feature Transform(SIFT) Flow algorithm and professional registration software of medical images called Elastix.

Key words: image registration, deformation registration, convolutional neural network, feature extraction, optical flow method

中图分类号: