Fig. 2

Comparison of functional connectivity (FC) matrices estimated before denoising (top row, a,b) and after denoising (bottom row, c,d) on subjects with two complete resting-state runs (N = 116). On the left (a,c), the lower triangular matrices are the average FC derived from single-band resting-state acquisitions (N = 34), while the upper triangular matrices show the average FC derived from multiband resting-state acquisitions (N = 100). Note that some subjects (N = 18) have both SB and MB scans and therefore contribute to both upper and lower triangles. On the right (b,d), lower triangular matrices are derived from movie fMRI data (N = 57), while upper triangular matrices are derived from multiband resting-state acquisitions (N = 100). We used data in CIFTI format registered to the MNI152NLin2009cAsym space, processed through fMRIPrep and denoised with rsDenoise with the strategy described in⁴⁸ For each subject, two runs were concatenated before computing the average time series for each of 400 parcels of the Schaefer cortical parcellation⁸³. Parcels are grouped following the 7 resting-state networks defined in the Yeo parcellation⁸⁴. FC was computed as the pairwise Pearson’s correlation between parcel time series (color scale). For subjects with more than one session available, individual FC matrices are averaged across sessions before averaging them across subjects (so that each subject only contributed once).