Functional Connectivity Analyses

Functional connectivity maps were derived from statistical analysis of resting state functional magnetic resonance imaging (fMRI) data collected from 497 healthy adult participants (293 female) as part of the Human Connectome Project (HCP). Maps were derived from the HCP minimally processed data included in the WU-Minn Q6 500 + MEG2 release. The HCP fMRI image preprocessing steps included image distortion correction, motion correction, within-subject registration to a structural volume, and nonlinear registration to the MNI152 standard brain template space. The CONN toolbox was used to further preprocess the data, including artifact rejection/scrubbing, detrending, aCompCor correction for removal of physiological noise and residual movement artifacts, and band-pass filtering (.01Hz to 0.10Hz), as well as for functional connectivity analyses (Whitfield-Gabrieli & Nieto-Castanon; 2012). FreeSurfer was used to segment a T1 structural image of each subject’s brain and to generate a representation of the cortical surface. The pre-processed blood oxygen level dependent (BOLD) responses from functional volume cortical voxels were then mapped to the corresponding vertex of the cortical surface. The cortical speech network was broken into a set of anatomically defined seed regions, indicated by white patches in the figures that follow. Individual functional connectivity maps for each seed region were determined by calculating the Pearson’s correlation between the mean BOLD time series in the seed region to that of all cortical vertices. Group-level statistical tests of connectivity strength were done by transforming the resulting surface correlation maps to an approximate normal distribution using the Fisher Z transform, then performing vertex-wise two-tailed t-tests to determine whether the correlation at each voxel differed significantly from 0. The t statistical map for each seed was first thresholded at the vertex level, then an additional correction was done to ensure that the family-wise probability of any cluster of supra-threshold vertices being a false positive was less than 5%. In the figures below, areas shown in red survived the t = 5 voxel-level threshold and areas in yellow survived the more stringent t = 15 voxel-level threshold.