北京生物医学工程

基于耦合配准网络的MR脑图像标签迁移算法

Label transfer algorithm for MR brain image based on coupled registration network

作者：崔鹏程李恩慧李振宇张童禹张唯唯

单位：中国医学科学院基础医学研究所，北京协和医学院基础学院（北京 100005）<br />通信作者：张唯唯。E-mail：weiwei.zhang@ibms.pumc.edu.cn

分类号：R318. 04

出版年·卷·期（页码）：2023·42·1（1-8）

摘要：

目的提出一种单图谱标签迁移算法并命名为Multi-Angle，以期在队列分析中快速有效提取与神经退行性疾病相关的MR脑影像标记物和解剖结构。方法首先对初始图谱图像施加旋转变换，获得旋转图谱图像组；其次为主配准网络送入合并后的初始图谱图像与个体图像，预测形变场及候选标签；再次为副配准网络送入合并后的旋转图谱图像与个体图像，结合主网络相关特征预测候选标签；最后通过投票法融合多个候选标签获得个体图像标签。结果在Mindboggle101和HCP数据集的实验结果显示， Multi-Angle算法在两个测试集上重要解剖结构Dice相似性系数均值分别为76%和82%，Precision均值为74.0%和77.8%，ASD均值为0.83 mm和0.69 mm，均优于目前主流算法Voxelmorph和Ants-SyN。结论本文提出的Multi-Angle模型可以快速有效实现脑神经图谱标签迁移并提高评价指标准确度，对神经退行性疾病分析所需的影像特征提取具有潜在的临床应用价值。

Objective In order to quickly and efficiently extract imaging markers and anatomical structures of magnetic resonance brain image related to the neurodegenerative disease in cohort analysis, this paper proposes a single-atlas label transfer algorithm named Multi-Angle. Methods Firstly, small rotation transformations are applied to the initial atlas image to obtain a rotated atlas image sets; secondly, individual image combined with the initial atlas image is sent to the primary registration network to predict the deformation field and the candidate label; then, individual image combined with the rotated atlas image sets is sent to the secondary registration network to predict the candidate labels; finally, individual image label is obtained by fusing the multiple candidate labels with voting method. Results The experimental results on the Mindboggle101 and human brain connection group project HCP dataset show that the average values of Dice similarity coefficients of important anatomical structures on the two test sets of the Multi-Angle algorithm proposed in this paper reach 76% and 82%, the average values of Precision reach 74.0% and 77.8%, the average values of Average Surface Distance reach 0.83mm and 0.69mm respectively, which outperform the current mainstream algorithms Voxelmorph and Ants-SyN. Conclusion The proposed Multi-Angle model can effectively implement the label transferring and enhance the evaluation performance; thus, has the potential value of clinical application for image feature extraction required for neurodegenerative disease analysis.

参考文献：

[1] Xu J, Zhang Y, Qiu C, et al. Global and regional economic costs of dementia: a systematic review[J]. The Lancet, 2017, 390: S47.
[2] Mckhann GM, Knopman DS, Chertkow H, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease[J]. Alzheimer's & Dementia, 2011, 7(3): 270-279.
[3] Du AT, Schuff N, Kramer JH, et al. Higher atrophy rate of entorhinal cortex than hippocampus in AD[J]. Neurology, 2004, 62(3): 422-427.
[4] Angelina S, Suresh LP, Veni S. Image segmentation based on genetic algorithm for region growth and region merging[C]// International Conference on Computing. IEEE, 2012, E95-D(8): 970-974.
[5] Zhao M, Lin HY, Yang CH, et al. Automatic threshold level set model applied on MRI image segmentation of brain tno[J]. Applied Mathematics & Information Sciences, 2015, 9(4): 1-10.
[6] 麦永锋, 孙启昌, 贾鹏飞, 等. 基于DRR及相似性测度的2D-3D医学图像配准算法[J]. 北京生物医学工程,2021, 40(3): 263-272．
Mai YF, Sun QC, Jia PF, et al. 2D/3D medical image regulation algorithm based on DRR and the similarity measure[J]. Beijing Biomedical Engineering, 2021, 40(3): 263-272.
[7] Avants BB, Tustison N, Song G. Advanced normalization tools (ANTS)[J]. ?Insight Journal, 2008, 2008：1–35.
[8] 孙若凡, 张唯唯. 基于3D U-Net实现人体耳软骨MRI图像的解剖结构分割[J]. 中国生物医学工程学报, 2021, 40(5): 531-539.?
Sun RF, Zhang WW. Anatomical structure segmentation of human auricular cartilage MRI Images based on 3D U-Net[J]. Chinese Journal of Biomedical Engineering, 2021, 40(5): 531-539.
[9] 杨斌斌, 刘霖雯, 张唯唯. 基于DenseMedic网络的脑皮层下结构的语义分割[J]. 中国生物医学工程学报, 2020, 39(6): 652-666.
Yang BB, Liu LW, Zhang WW. Semantic segmentation of subcortical brain structures based on DenseMedic network[J]. Chinese Journal of Biomedical Engineering, 2020, 39(6): 652-666.
[10] Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation[C]// Medical Image Computing and Computer-Assisted Intervention, 2015, 234-241.?

[11]Jaderberg M, Simonyan K, Zisserman A, et al. Spatial Transformer Networks[J]. MIT Press, 2015, 2: 2017-2025.?

[12]Balakrishnan G, Zhao A, Sabuncu MR, et al. VoxelMorph: a learning framework for deformable medical image registration[J]. IEEE Transactions on Medical Imaging, 2019, 38(8): 1788-1800.
[13] Dalca AV, Balakrishnan G, Guttag J, et al. Unsupervised learning for fast probabilistic diffeomorphic registration[C]// Medical Image Computing and Computer-Assisted Intervention, 2018, 729-738.
[14] Dubost F, Bruijne MD, Nardin M, et al. Multi-atlas image registration of clinical data with automated quality assessment using ventricle segmentation[J]. Medical Image Analysis, 2020, 63: 101698.
[15] Arno K, Jason T. 101 Labeled brain images and a consistent human cortical labeling protocol[J]. Frontiers in Neuroscience, 2012, 6: 171.

[16] Fischl B. FreeSurfer[J]. NeuroImage, 2012, 62(2): 774-781.
[17] Van Essen DC, Ugurbil K, Auerbach E, et al. The human connectome project: a data acquisition perspective[J]. NeuroImage, 2012, 62(4): 2222-2231.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录