首页 | 官方网站   微博 | 高级检索  
     

基于标签对齐的多模态一致性表型关联方法
引用本文:汪美玲,邵伟,张道强. 基于标签对齐的多模态一致性表型关联方法[J]. 软件学报, 2022, 33(12): 4545-4558
作者姓名:汪美玲  邵伟  张道强
作者单位:南京航空航天大学 计算机科学与技术学院,江苏 南京 211106;模式分析与机器智能工业和信息化部重点实验室,江苏 南京 211106
基金项目:国家自然科学基金(61876082,61902183,61861130366,61732006);国家重点研发计划(2018YFC2001600,
摘    要:近年来,随着脑影像和基因技术的发展,脑影像遗传学得到了广泛的关注.在脑影像遗传研究中,检验遗传变异(即单核苷酸多态性(single nucleotide polymorphisms,SNPs))对大脑结构或功能的影响是一项艰巨的任务.此外,提取的多模态脑表型和来自同一区域的一致性脑影像标志物为理解疾病(例如,阿尔茨海默病(Alzheimer’s disease,AD))的机理提供了更多的见解.利用多模态脑表型作为桥接风险基因位点和疾病状态的中间特征,设计通过标签对齐的多模态学习方法来识别AD中风险基因位点与疾病状态之间的一致性表型.首先,用标准的多模态方法去探索和AD相关的基因位点(即APOEe4 rs429358)与多模态脑影像之间关系;其次,为了利用标记样本之间的标签信息,在标准多模态方法的目标函数中添加了一个新的标签对齐正则化项,使得所有具有相同类别标签的多模态样本在映射空间中更靠近;最后,在公开的ADNI (Alzheimer’s disease neuroimaging initiative)数据集上的3种脑影像(即大脑的结构组织信息、脱氧葡萄糖正电子发射断层扫描和正电子发射断层扫描淀粉样蛋白成像)进行实验.实验结果表明:该方法可以在多模态脑影像上发现鲁棒的、一致性脑区域来解释AD的病因,并在3个模态上将相关系数分别提高了8%,9%,5%.

关 键 词:脑影像遗传学  多模态脑影像表型  单核苷酸多态性  标签对齐  阿尔茨海默病
收稿时间:2020-08-01
修稿时间:2021-02-03

Label-aligned Multi-modality Consistent Phenotype Association Method
WANG Mei-Ling,SHAO Wei,ZHANG Dao-Qiang. Label-aligned Multi-modality Consistent Phenotype Association Method[J]. Journal of Software, 2022, 33(12): 4545-4558
Authors:WANG Mei-Ling  SHAO Wei  ZHANG Dao-Qiang
Affiliation:College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China;MIIT Key Laboratory of Pattern Analysis and Machine Intelligence, Nanjing 211106, China
Abstract:Recently, with the rapid development of imaging and genomic techniques, the brain imaging genetics has received extensive attention. In the brain imaging genetic studies, it is a challenging task to examine the influence of genetic variants, i.e., single nucleotide polymorphisms (SNPs), on structures or functions of human brains. In addition, multimodal brain imaging phenotypes extracted from different perspectives and imaging markers from the same region consistently showing up in multimodalities gives more ways to understand the diseases mechanism, such as Alzheimer''s disease (AD). Accordingly, This work exploits multi-modal brain imaging phenotypes as intermediate traits to bridge genetic risk factors and disease status. Consistent phenotype between genetic risk factors and disease status is discovered via the designed label-aligned multi-modality regression method in AD. Specifically, standard multi-modality method is first applied to explore the relationship between the well-known AD risk SNP APOEe4 rs429358 and multimodal brain imaging phenotypes. Secondly, to utilize the label information among labeled subjects, a new label-aligned regularization is included into the standard multi-modality method. In such way, all multimodality subjects with the same class labels should be closer in the new embedding space. Finally, the experiments are conducted on three baseline brain imaging modalities, i.e., voxel-based measures extracted from structural magnetic resonance imaging, fluorodeoxyglucose positron emission tomography and F-18 florbetapir PET scans amyloid imaging, from the Alzheimer''s disease neuroimaging initiative (ADNI) database. Related experimental results validate that the proposed method can identify robust and consistent regions of interests over multi-modality imaging data to guide the disease-induced interpretation. Furthermore, the values of correlation coefficient have been increased by 8%, 9%, and 5% in comparison with the best results of the existing algorithms on three modalities.
Keywords:brain imaging genetics  multimodal brain imaging phenotypes  single nucleotide polymorphisms  label-aligned  
本文献已被 万方数据 等数据库收录!
点击此处可从《软件学报》浏览原始摘要信息
点击此处可从《软件学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号