Use of fMRI and Structural Equation Modeling for Studying Interconnected Brain Areas within a Hypothetical Network

Functional neuroimaging first allowed researchers to describe the functional segregation of regionally activated areas during a variety of experimental tasks. More recently, functional integration studies have described how these functionally specialized areas (i.e. areas whose activity is temporally modified) interact within a highly distributed neural network. When applied to the field of functional magnetic resonance imaging (fMRI), structural equation modeling (SEM) uses theoretical and/or empirical hypotheses to estimate the effects (path coefficients) of an experimental task within a putative network. Structural equation modeling represents a linear technique for multivariate analysis of fMRI data and has been developed to simultaneously examine ratios of multiple causality in an experimental design; the method attempts to explain a covariance structure within an anatomical (constrained) model. This method, when combined with the concept of effective connectivity, can provide information on the strength and direction of the functional interactions which take place between identified nodes of a putative network. After having provided a brief reminder of the principle of the blood oxygen level-dependent (BOLD) contrast effect, the physiological bases of brain activity and the concepts of functional integration and effective connectivity, we specify the various steps in the SEM analysis and the use of fMRI data to explore putative networks of interconnected active areas.