Hi everybody,

would you please allow me three quick questions on the scrubbing procedure as implemented in DPARSF (I have been using version V2.3, 130615, for this particular work)?

In the GUI, I have ticked the "head motion scrubbing regressors" box under "nuisance covariates regression", with the default settings (Power et al., FD threshold .5, time points before: 1 and after: 2, method: use each bad time point as a regressor).

a) Is this suffienct to correct for motion artifacts in the sense of the concerns expressed by Power et al. (2012) and Yan et al. (2013)?

b) my final images (after subsequent normalization and spatial smoothing) still contain as many volumes as in the begining. Is it correct that no volumes have been removed in any subject during scrubbing?

c) What ist the purpose of the "scrubbing" box in the analysis section of the GUI?

The more I think of it, the more concerned I get that I have only estimated some parameters for scrubbing without applying them?

All help and comments are greatly appreciated.





Hi Sebastian,
1. Usually, using Friston 24 parameter regression + group-level mean FD covariate should be enough, especially you have used scrubbing regressors here.

2. The number of time points will not change in scrubbing regression, as those just modeled the bad time points in regression model. However, those time points will be given 0 values. Again, the scrubbing is applied even the number of time points remains the same.

3. The "scrubbing" check box is similar to the scrubbing procedure in Power et al., 2012, by deleting the bad time points, which is less optimal. If you want to use this strategy, you can skip "scrubbing regression" and check this "scrubbing" box.



Dear Chao-Gan,

thank you so much! I appreciate your help.

All best,