Multivariate statistical analysis in fMRI

The paper briefly discussed different statistical analysis in functional magnetic resonance imaging (fMRI). Multivariate regression analysis with multiple comparisons corrections allows the determination of activated voxels that can then be grouped into regions of interest (ROIs). Principal component analysis (PCA) is useful in extracting common temporal response features of an ROI as well as differentiating the temporal response of groups of commonly responding ROI. It can also be used to examine differences in the temporal response of subgroups of subjects in the study. Structural equation modeling (SEM) is a technique that requires a priori knowledge of the connections and their direction between ROIs. It is particularly useful in identifying changes in connectivity that result from different interventions or different classes of patients.