Most of these factors and their impact on the detection of MEG RSNs have been discussed in our recent publications (de Pasquale et al., 2010 and Mantini et al., 2011). Several other features of the present methodology bear discussion. First, it may be argued that, because the MCW procedure identifies epochs of high within-network correlation, NVP-AUY922 datasheet the
presently observed MEG RSNs (Figure 1) are predisposed to replicate the RSN priors derived from fMRI. This concern applies to the diagonal blocks in Figures 2 and 3 representing within-network correlation, but not to the off-diagonal blocks indicating cross-network interactions. In fact cross-network interactions are computed on node pairs that were not used to define the MCWs. Several features of the present methodology mitigate this concern. MCW identification for one seed in one subject is just the first step of the analysis. In the second step, Z-score maps are computed by testing the mean correlation value across subjects
in each voxel against the mean and variance of connectivity across the rest of the brain and over subjects. Importantly, to minimize the influence of fMRI priors, MCWs are identified using only a subset of three nodes from each network, whereas MEG RSN identification 3-MA mw requires that all Z-score maps corresponding to four distinct node subsets ( Table S2) pass a statistical criterion. For example, in the case of the DAN, left FEF was
not input to the MCW algorithm when left PIPS was a seed ( Figure 1C), and similarly, right FEF was not input when right PIPS was a seed. Thus, the maps in Figure 1 include only voxels that are both significant Montelukast Sodium over subjects and survive a logical AND conjunction over node subsets. A related analysis addresses the logical converse. Specifically, providing the MCW procedure with arbitrary combinations of nodes (“fake RNSs”) yielded very few MCWs and no MEG RNSs comparable to the results shown in Figure 1. The “fake RSN” analyses are presented in the Supplemental Information ( Figure S2A) and confirm a similar, previous control analysis (using a different arbitrary node combination; de Pasquale et al., 2010). More broadly, even if the MCW procedure may be predisposed to replicate RSNs corresponding to fMRI priors, this does not account for the frequency specificity of both within- and across-network interactions (e.g., Figure 2). A second question concerns the influence of the external node. Because the MCW procedure contrasts within- versus external-to-network correlations, it is critical to demonstrate that the presently obtained results are independent of the locus of the external node. Accordingly, we repeated several of the present analyses using a total of three different external nodes (RSFG, EXT2, and EXT3), in each case obtaining very similar results (Supplemental Information and Figure S4).