J Shanghai Jiaotong Univ Sci ›› 2023, Vol. 28 ›› Issue (4): 536-.doi: 10.1007/s12204-022-2525-3

• • 上一篇    


陈培芝1,2, 郭逸凡1, 王大寒1,2, 陈金铃1,3,4   

  1. (1. 厦门理工学院 计算机与信息工程学院,福建厦门 361024;2.福建省模式识别与图像理解重点实验室,福建厦门361024;3. 长春理工大学 信息工程学院,长春 130600;4. 朝阳科技大学 计算机科学与信息工程系,台湾台中 41349)
  • 收稿日期:2021-01-27 接受日期:2021-09-17 出版日期:2023-07-28 发布日期:2023-07-31

Dlung: Unsupervised Few-Shot Diffeomorphic Respiratory Motion Modeling

CHEN Peizhi1,2* (陈培芝), GUO Yifan1 (郭逸凡),WANG Dahan1,2 (王大寒), CHEN Chinling1,3,4* (陈金铃)   

  1. (1. College of Computer and Information Engineering, Xiamen University of Technology, Xiamen 361024, Fujian, China; 2. Fujian Key Laboratory of Pattern Recognition and Image Understanding, Xiamen 361024, Fujian, China; 3. School of Information Engineering, Changchun Sci-Tech University, Changchun 130600, China; 4. Department of Computer Science and Information Engineering, Chaoyang University of Technology, Taichung 41349, Taiwan, China)
  • Received:2021-01-27 Accepted:2021-09-17 Online:2023-07-28 Published:2023-07-31

摘要: 肺部图像配准在肺部分析应用中具有重要作用,比如呼吸运动建模。基于无监督学习的图像配准方法可以在不需要监督的情况下计算变形,因此受到了广泛的关注。但是,需要注意的是,它们有两个缺点:它们不能处理数据不足的问题,也不能保证微分同胚(保留拓扑结构)属性,特别是当肺部扫描中存在较大形变时。本文提出了一种基于无监督少样本学习的微分同胚肺部图像配准方法,称为Dlung。我们采用微调解决数据不足的问题,并采用缩放-平方层来实现微分同胚配准。在实验中,我们进行了空间时间(4D)图像上的配准,并与基准方法进行了全面比较。Dlung取得了最高的微分同胚准确率,它使用有限的数据构建了准确和快速的呼吸运动模型。这项研究扩展了我们对呼吸运动建模的认识。

关键词: 无监督少样本学习,呼吸运动建模,微分同胚配准

Abstract: Lung image registration plays an important role in lung analysis applications, such as respiratory motion modeling. Unsupervised learning-based image registration methods that can compute the deformation without the requirement of supervision attract much attention. However, it is noteworthy that they have two drawbacks: they do not handle the problem of limited data and do not guarantee diffeomorphic (topologypreserving) properties, especially when large deformation exists in lung scans. In this paper, we present an unsupervised few-shot learning-based diffeomorphic lung image registration, namely Dlung. We employ fine-tuning techniques to solve the problem of limited data and apply the scaling and squaring method to accomplish the diffeomorphic registration. Furthermore, atlas-based registration on spatio-temporal (4D) images is performed and thoroughly compared with baseline methods. Dlung achieves the highest accuracy with diffeomorphic properties. It constructs accurate and fast respiratory motion models with limited data. This research extends our knowledge of respiratory motion modeling.

Key words: unsupervised few-shot learning, respiratory motion modeling, diffeomorphic registration