Kim Jieun, Zhu Wei, Chang Linda, Bentler Peter M, Ernst Thomas
Department of Applied Mathematics and Statistics, State University of New York at Stony Brook, Stony Brook, New York, USA.
Hum Brain Mapp. 2007 Feb;28(2):85-93. doi: 10.1002/hbm.20259.
The ultimate goal of brain connectivity studies is to propose, test, modify, and compare certain directional brain pathways. Path analysis or structural equation modeling (SEM) is an ideal statistical method for such studies. In this work, we propose a two-stage unified SEM plus GLM (General Linear Model) approach for the analysis of multisubject, multivariate functional magnetic resonance imaging (fMRI) time series data with subject-level covariates. In Stage 1, we analyze the fMRI multivariate time series for each subject individually via a unified SEM model by combining longitudinal pathways represented by a multivariate autoregressive (MAR) model, and contemporaneous pathways represented by a conventional SEM. In Stage 2, the resulting subject-level path coefficients are merged with subject-level covariates such as gender, age, IQ, etc., to examine the impact of these covariates on effective connectivity via a GLM. Our approach is exemplified via the analysis of an fMRI visual attention experiment. Furthermore, the significant path network from the unified SEM analysis is compared to that from a conventional SEM analysis without incorporating the longitudinal information as well as that from a Dynamic Causal Modeling (DCM) approach.
脑连接性研究的最终目标是提出、测试、修改并比较特定的脑内定向通路。路径分析或结构方程建模（SEM）是此类研究的理想统计方法。在本研究中，我们提出了一种两阶段统一的SEM加GLM（一般线性模型）方法，用于分析带有个体水平协变量的多主体、多变量功能磁共振成像（fMRI）时间序列数据。在第一阶段，我们通过一个统一的SEM模型，分别分析每个主体的fMRI多变量时间序列，该模型结合了由多变量自回归（MAR）模型表示的纵向通路和由传统SEM表示的同期通路。在第二阶段，将得到的个体水平路径系数与个体水平协变量（如性别、年龄、智商等）合并，通过GLM来检验这些协变量对有效连接性的影响。我们的方法通过对一个fMRI视觉注意力实验的分析进行了例证。此外，将统一SEM分析得到的显著路径网络与未纳入纵向信息的传统SEM分析以及动态因果建模（DCM）方法得到的路径网络进行了比较。
