Dear Florian,

There is a simpler way to change the settings of the outlier identification step without having to re-run the ART preprocessing step (simply applying different thresholds to the same framewise displacement and global signal change measures that ART has computed previously). You can do that in the Setup.Covariates (1st-level) menu by selecting the option Covariate tools. Compute new/derived first-level covariates and then selecting the option Compute scrubbing. You can then select the new thresholds that you would like to apply (e.g. [9 2] for an liberal level, [5 0.9] for an intermediate level, and [3 0.5] for a conservative level) and that will automatically re-create your 'scrubbing' first-level covariate (as well as the related QC second-level covariates, such as QC_ValidScans, QC_MeanMotion, etc.)

Hope this helps

Alfonso

Originally posted by lammersf:

Dear all,

I've run the CONN default preprocessing pipeline with conservative settings for outlier detection. However, I would like to change the settings to the intermediate level. I considered the "apply individual preprocessing step" a potential way to repeat the outlier identification step without running the whole default preprocessing pipeline all over again.

To my experience, CONN will always take the most recently generated files to proceed with, which would be  the swau*-files after completion of the whole default preprocessing pipeline. Accordingly, I now have two sets of files called art_regression_outliers_swau* and art_regression_outliers_au*, suggesting that CONN has taken two different input files from different preprocessing stages. However, the ART results look identical at first glance, except for the number of identified outliers.

So, does it matter if I take the *au- or *swau-files as an input for ART? And even more interesting, is there a way to tell CONN it should use the *au-files instead of the *swau files without going back to the project set-up and importing the *au-files?

Thanks in advance,

Florian