Department of Psychology, University of Miami, Coral Gables, FL, USA.
Department of Psychology, University of Miami, Coral Gables, FL, USA.
Neuroimage. 2018 Jan 15;165:158-169. doi: 10.1016/j.neuroimage.2017.10.007. Epub 2017 Oct 10.
Brain-behavior associations in fMRI studies are typically restricted to a single level of analysis: either a circumscribed brain region-of-interest (ROI) or a larger network of brain regions. However, this common practice may not always account for the interdependencies among ROIs of the same network or potentially unique information at the ROI-level, respectively. To account for both sources of information, we combined measurement and structural components of structural equation modeling (SEM) approaches to empirically derive networks from ROI activity, and to assess the association of both individual ROIs and their respective whole-brain activation networks with task performance using three large task-fMRI datasets and two separate brain parcellation schemes. The results for working memory and relational tasks revealed that well-known ROI-performance associations are either non-significant or reversed when accounting for the ROI's common association with its corresponding network, and that the network as a whole is instead robustly associated with task performance. The results for the arithmetic task revealed that in certain cases, an ROI can be robustly associated with task performance, even when accounting for its associated network. The SEM framework described in this study provides researchers additional flexibility in testing brain-behavior relationships, as well as a principled way to combine ROI- and network-levels of analysis.
功能磁共振成像（fMRI）研究中的大脑-行为关联通常仅限于单一的分析水平：要么是一个限定的脑区域（ROI），要么是更大的脑区域网络。然而，这种常见的做法可能并不总是考虑到同一网络的 ROI 之间的相互依存关系，或者分别考虑到 ROI 层面上的潜在独特信息。为了考虑这两个信息源，我们结合了测量和结构方程模型（SEM）方法的结构成分，从 ROI 活动中推导出网络，并使用三个大型任务 fMRI 数据集和两个独立的大脑分割方案，评估个体 ROI 及其各自的全脑激活网络与任务表现的关联。工作记忆和关系任务的结果表明，当考虑到 ROI 与其对应网络的共同关联时，众所周知的 ROI 与表现的关联要么不显著，要么相反，而整个网络则与任务表现稳健相关。算术任务的结果表明，在某些情况下，即使考虑到与 ROI 相关的网络，ROI 也可以与任务表现稳健相关。本研究中描述的 SEM 框架为研究人员测试大脑-行为关系提供了额外的灵活性，以及一种将 ROI 和网络分析水平相结合的原则方法。
