For each participant, correlation map was calculated respectively for each seed ROI by a voxel-wise multiple-regression. Regressors included the template time course extracted by averaging time courses of all the voxels in the seed ROI under consideration, as well as the average time course in white matter and the average time course in cerebrospinal fluid (as nuisance signals). The masks of white matter were determined from each participant’s high-resolution structural image using FAST segmentation program of fMRIb software library (FSL) (www.fmrib.ox.ac.uk). The resulting white matter segmentations were then thresholded to ensure 80% tissue type probability. The cerebrospinal fluid mask was manually drawn according to the anatomic boundaries of the high-resolution three-dimensional structural images of each participant. These nuisance signals were used to account for fluctuations unlikely to be relevant to neuronal activities (Birn et al., 2006; Di Martino et al., 2008; Fox et al., 2005). The resultant t-score maps of the seed ROIs were then converted to z-score maps hereafter referred to as “correlation maps”.
