Fully automated estimation of fMRI guided SPECT brain networks and their functional network connectivity in schizophrenia patients vs controls: a NeuroMark ICA approach

Single photon emission computerized tomography (SPECT) scans have emerged as a useful imaging modality that has been explored in the literature for the last 40 years. To date, little work has focused on studying functional network connectivity utilizing SPECT data. In this study, we present a fully automated, spatially constrained ICA (sc-ICA) approach to evaluate functional network connectivity profiles in SPECT data using the NeuroMark pipeline. We then use this to compare group differences in brain function in patients with schizophrenia (SZ) and healthy controls (HC). We evaluate both the expression of brain networks as well as the whole brain SPECT connectome (assessed as inter-subject covariation among networks) to evaluate the neuroimaging links to schizophrenia. 76 healthy controls, and 137 schizophrenia patient SPECT images were acquired from Amen Clinic sites along with diagnostic information. Each patient participated in two SPECT brain scans, acquired during rest and while performing a sustained attention task across twelve clinical imaging sites. Preprocessed SPECT data were analyzed via sc-ICA using 53 spatial priors derived from functional MRI data. Results revealed group differences between healthy controls and patient SPECT data. Out of the 53 components in the resting state SPECT data, 15 total components were found to show differences in various brain regions after correcting for multiple comparisons. The vast majority of components showed reduced connectivity in patients. Relationships were associated with age, sex, hearing voices and having disjointed thoughts. Additional covariates such as medication status and scanner type were considered. Initial associations with clinical variables revealed relationships between the auditory, sensorimotor, and cognitive control networks, as well as broad network connectivity between the cortical and default mode network areas. For the task data, the components found in resting data were replicated, with patients showing reduced connectivity in similar regions irrespective of task performance. In summary, these results confirm and extend existing work highlighting large scale network disruptions noted in prior schizophrenia fMRI studies between the subthalamic, cortical and auditory networks.