Table 1.

Overview of scaling method used for assessing quality of preprocessing methods.

Good performance 1	Moderate performance 2	Moderately poor performance 3	Poor performance 4	Extreme Concern 5
Percent of Edges related to Motion
(0–10% edges correlated with motion)	(10–20%)	(20–30%)	(30–40%)	(> 40%)
36P+spkreg 36P+despike 36P 36P+scrub	ICA+GSR 9P aCompCor	ICA	wmMean	2P 6P 24P wmLocal tCompCor
Distance dependence of motion effects
QC-FC corr r > −0.15	r = −0.15 to −0.2	r = −0.2 to −0.25	r = −0.25 to −0.3	r < −0.3
ICA 36P+scrub	ICA+GSR wmLocal	24P 6P 36P+spkreg wmMean 36P+despike	acompcor 2P	9P tcompcor 36P

Percent of edges related to motion and distance dependence of motion effects are described in detail elsewhere [57]. 36 P = nuisance regressors included 6 motion estimates, mean white matter (WM), mean cerebral spinal fluid (CSF), and mean global signal (GS), along with the derivatives, quadratic terms and squares of these signals [171]. 36 P+despike = includes 36 regressors as described above, with despiking removal of high motion frames [194]. 36 P+spkreg = includes 36 regressors with spike regression of high motion frames [171]. 36 P+scrub = 36 parameters and motion scrubbing of high motion frames [59]. Scrubbing high motion frames were defined using framewise displacement (FD), computed as the sum of the absolute values of the derivatives of translational and rotational motion estimates. FD > .2 mm was flagged as high motion. 2 P = nuisance regression includes mean WM and mean CSF.

6 P = nuisance regression only includes 6 motion estimates from realignment. 9 P + GSR = nuisance regression includes 6 motion estimates, mean WM, mean CSF, and mean GSR [195, 196]. 24 P = nuisance regression includes 6 motion estimates, their temporal derivatives and quadratic expansion terms [197]. aCompCor = nuisance regression includes 5 principal components each from the WM and CSF, in addition to 6 motion parameters and their temporal derivatives [198]. tCompCor = nuisance regression includes 6 principal components from voxels with high variance over time [199]. wmLocal = nuisance regression includes a voxelwise localized WM regressor in addition to 6 motion parameters, and their temporal derivatives and despiking [200]. wmMean = nuisance regression includes mean WM in addition to 6 motion parameters and their temporal derivatives and despiking [200]. ICA = independent component analysis, removal of motion-related variance components from the BOLD data including mean WM and CSF regressors [201].