Fmriprep: Motion outliers FD dvars

[lenacanlab]

Dear all, I would like to know how motion outliers (motion_outlier00, motion_outlier01 etc specified in the _desc-confounds_regressors.tsv file) are calculated in fMRIprep. It seems like it is based on dvars but it also seems to use the FD values. It seems like the underlying threshold varies from subject to subject, is that correct? Thank you!

[jdkent]

Thank you for your patience @lenacanlab.
Motion outliers are determined by both FD and DVARS thresholds (meaning if the volume is over the FD threshold OR the DVARS threshold it will be counted as an outlier) which are defined as 0.5 and 1.5 by default, but you can change the defaults with --fd-spike-threshold and --dvars-spike-threshold parameters. To explicitly answer your second question, the threshold does not change from participant to participant.

[utooley]

Is this something that should go into the documentation @oesteban, along with the fact that it's the Power calculation of FD from MCFLIRT motion params that is reported in confounds.tsv? Related also to this post on Neurostars, I have once find myself digging into the repo to find where and how FD was calculated.

If so, I can tackle, not sure if anyone else cares about this, though.

[kfinc]

@utooley I had the same problem with that, as with other confounds measures. I'm already drafting some ideas on how to categorize confound variables in a table (categories: motion, physiological, other). Take a look here: https://docs.google.com/document/d/14WyhCvEn0G5fbLY-Rk7BQa9ISzYEB2ll3YCjyd-OjB8/edit. I'm wondering how to keep a balance between simple description and more technical details.