A study of spatial variation in fMRI brain networks via independent vector analysis: Application to schizophrenia

Spatial variability in intrinsic brain networks has not been well studied in fMRI. Independent vector analysis (IVA), is a blind source separation approach that can be used for segregating fMRI data into temporally coherent, maximally spatially independent networks enabling comparison among subjects similar to group independent component analysis (GICA). Using simulated and small sample real data, it has been shown that spatial independence in IVA is achieved while jointly maximizing the dependence across subjects. This study was motivated by the fact that IVA has not yet been applied to a large sample size or to analyze multi-group data for spatial differences. We introduce several new ways to quantify differences in variability of IVA-derived connectivity networks between schizophrenia patients (SZ = 82) from healthy controls (HC = 89) in a large (N=171) data set. Results show that IVA identified significant group differences in the auditory cortex, the basal ganglia, the sensorimotor network and medial visual cortex. Variance maps of the spatial networks showed that there is greater variability in the patients primarily in sensory networks whereas the default mode network showed more variability in the controls. In summary, IVA enables the study of spatial variation in intrinsic brain networks, an area that has not been in focus.