Abstract
In this work, abundant anticorrelated networks were successfully detected in rest-stating fMRI-BOLD data from 20 subjects. Spatial independent component analysis (sICA) method was applied at both individual and group levels. At the individual level, for each subject, 30 independent components (IC) were estimated, and each IC was transformed using Z-score mapping. The voxels with >5 and < -5 Z-score were denoted as positive signals (PS) and negative signals (NS) respectively. The correlation coefficients between the mean time series of the PS and NS voxels were computed; if the calculated coefficients were <-0.3, the PS and NS voxels were considered to form an anticorrelated PS-NS network. It was found that 36.5% of the ICs contained an anticorrelated PS-NS network. The spatiotemporal patterns of most PS-NS networks varied from subject to subject, but three networks displaying spatial patterns were comparably consistent among different subjects. For group-level analysis, no anticorrelated PS-NS networks were detected. Our results suggest that future investigations adopt a broader approach for negative BOLD signal characterization. Combined consideration of PS and NS systems help to better elucidate hemodynamic and neuronal brain behavior and further develop understanding of neural mechanisms of brain information processing.