摘要
针对目前深度学习方法应用于医学图像配准精度不高的问题,提出了增加低分辨率辅助特征的无监督3D卷积神经网络的脑部图像配准模型。使用无监督学习的卷积网络回归出位移场,再通过空间变换层对浮动图像进行变换,然后根据构建的损失函数优化网络参数,实现端到端的无监督学习。通过添加注意力模块,在网络对应层间的连接中加入低分辨率的辅助特征,增加结构特征的同时减少多余的背景信息。方法与无监督的U-Net和VoxelMorph在MICCAI2012多图谱数据中比较,结果表明,有更高的配准精度和更快的配准速度,且不需要专家标注信息,因此在医学图像配准上具有较好的应用潜力。
In view of the low accuracy of the current deep learning method in medical image registration,an unsupervised 3Dconvolutional neural network model for brain registration is proposed.The convolution network is used to regress the displacement field,and then the floating image is transformed through the spatial transformation layer.Then the network parameters are optimized according to the constructed loss function to realize the end-to-end unsupervised learning.By adding attention gate structure,low resolution auxiliary features are added to the connection between corresponding layers of the network to increase features and reduce background information.Compared with the unsupervised U-Net and VoxelMorph in MICCAI2012multi-graph data,The results show that the method has higher registration accuracy and faster registration speed,and does not require expert annotation information,so it has good application potential in medical image registration.
作者
薛湛琦
王远军
XUE Zhanqi;WANG Yuanjun(Institute of Medical Imaging Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《光学技术》
CAS
CSCD
北大核心
2021年第1期80-86,共7页
Optical Technique
基金
上海市自然科学基金资助项目(18ZR1426900)。
关键词
无监督学习
图像配准
辅助特征
卷积神经网络
脑部图像
unsupervised learning
image registration
auxiliary features
convolutional neural network
brain image