A multivariate source-based morphometry (SBM) method for processing fractional anisotropy (FA) data is presented. SBM utilizes independent component analysis (ICA) and decomposes an FA image into spatial maps and loading coefficients. The loading coefficients represent the relative degree each component contributes to a given subject's FA map. We hypothesized that SBM analysis on a large dataset of age- and gender-matched patients with schizophrenia (n=65, ages 18–60 years) and healthy controls (n=102, ages 18–60 years) would show a similar, specific pattern of frontal and temporal group differences as a recent voxel-based morphometry meta-analysis. Two approaches using (a) the loading coefficients obtained from the ICA analysis and, alternatively, (b) the weighted mean FA values obtained from the ICA-defined clusters were compared for group analysis. Six of the 10 selected components had significant group differences with the loading coefficients. Each component was composed of several white matter tracts distributed throughout the brain. Nine of the 10 nonartifactual components had significant group differences with the weighted mean FA values. The weighted mean FA values for each ICA spatial map generally had larger effects sizes relative to the loading coefficients. These networks were consistent with regions identified in previous voxel-based studies of schizophrenia. SBM identified several components that covered disjoint brain regions and multiple white matter tracts that would not have been possible with previous voxel-based univariate techniques. Overall, these results suggest the importance of utilizing multivariate approaches in morphometric studies in schizophrenia.
- diffusion tensor imaging (DTI)
- independent component analysis (ICA)
ASJC Scopus subject areas