基于深度学习的医学图像配准研究进展

Advances in medical image registration based on deep learning

图像配准是图像引导手术、图像融合、器官图谱生成、肿瘤和骨骼生长监测等临床任务应用的关键技术,也是一个极具挑战性的问题。近年来,深度学习技术对医学图像处理方法的研究产生重要的影响,在医学图像配准领域发展迅速。对使用深度学习技术实现医学图像配准的研究进行综述,首先按照深度学习模型将医学图像配准方法分为3类,包括监督、弱监督和无监督医学图像配准;然后分别介绍国内外研究进展,并总结这些研究方法的优缺点;在此基础上,阐述常用的深度学习配准框架以及评价标准,并总结常用的开源医学影像数据集;最后对深度学习技术在医学配准图像领域中存在的问题进行分析,展望未来发展的方向。

Image registration is a key technology for image-guided surgery,image fusion,organ map generation,tumor and bone growth monitoring and other clinical applications.It is also a very challenging problem.In recent years,deep learning technology has exerted an important influence on the research of medical image processing methods,and has developed rapidly in the field of medical image registration.Research on medical image registration using deep learning technology was reviewed.Firstly,according to the deep learning model,medical image registration methods were divided into three categories,including supervised,weakly supervised and unsupervised medical image registration.Then the research progress at home and abroad was introduced,and the advantages and disadvantages of these research methods were summarized.On this basis,the commonly used deep learning registration framework and evaluation criteria were described,and the commonly used open source medical image data sets were summarized.Finally,the existing problems of deep learning technology in the field of medical registration image were analyzed,and the future development direction was forecasted.