用于构建脑磁图网络的信号提取方法  

Representative Signal Extraction in MEG Network

作　　者：杨春兰[1] 吴文晓 吴水才[1] 任洁钏[2] YANG Chunlan;WU Wenxiao;WU Shuicai;REN Jiechuan(Intelligent Physiological Measurement and Clinical Translation,Beijng International Base for Scientific and Technological Cooperation,College of Life Science and Bio-engineering,Beijing University of Technology,Beijing 100124,China;Beijing Tiantan Hospital,Capital Medical University,Beijing 100050,China)

机构地区：[1]北京工业大学生命科学与生物工程学院,智能化生理测量与临床转化北京市国际科研合作基地,北京100124 [2]首都医科大学附属北京天坛医院,北京100050

出　　处：《北京工业大学学报》2020年第7期795-802,共8页Journal of Beijing University of Technology

基　　金：国家自然科学基金资助项目(31640035,81601126);北京市教委科技计划项目(KM201710005013)。

摘　　要：构建脑磁图功能网络时,一般选取各脑区功率最强的源信号代表其神经活动,这将造成信息损失.针对此问题,提出了2种改进方案:基于叠加平均的方法和基于聚类的方法.为了验证上述方案,选取了51例被试任务状态下的脑磁数据,对其各频段利用最大功率、叠加平均、聚类3种方法进行信号提取,然后对以其构建的脑功能网络进行k均值聚类分析,此外,对以上3种方法构建的脑网络特征进行分析比较.结果表明:叠加平均方法的准确率最高、最大功率次之,聚类方法准确率最低;脑网络特征分析结果发现基于叠加平均和最大功率方法构建的大脑网络具有较强的小世界属性,而使用聚类方法构建的大脑网络其小世界属性较弱.基于本研究初步得出结论,采用叠加平均和最大功率信号提取方法构建脑磁图脑网络具有可行性.For the construction of MEG brain network,the source signal with the strongest power in each brain region is generally selected to represent its neural activity,which causes the loss of information.In this study,two improvement schemes were proposed for this problem:research methods based on superposition averaging and cluster-based research methods.To evaluate the scheme,51 subjects were selected with MEG data under task state.Signals were extracted by using maximum power,superimposed average and clustering,respectively,from each band.The brain function network was constructed and clustered.Additionally,the characteristics of brain network constructed by the three methods were compared.Based on the results,the superposition averaging method acquired the highest accuracy,then the highest power method and clustering acquired lower accuracy.Brain network feature analysis found that those based on the method of superimposed average and maximum power have strong small world attributes,while brain networks constructed using clustering methods have weaker world attributes.From that,it is feasible to construct MEG brain network using the superimposed average and maximum power methods.

关 键 词：脑磁图(MEG) 源重建 信号提取 脑功能网络 网络特征 小世界属性 

分 类 号：R445.9[医药卫生—影像医学与核医学]