Fmriprep standardised DVARS


Could somebody please describe how the standardised version of DVARS is computed in fmriprep? (sorry if I missed this information)

I also noticed that, by default, fmriprep proposes a threshold of 1.5 std_DVARS to identify motion outliers / spikes. Do you have any additional suggestion on how to set this threshold value for spike detection?

Thank you very much for your help!!

Welcome @agriffa!

The std_DVARS are essentially the non standard DVARS divided by the standard deviation of temporal difference time series. You can see the implementation of the DVARs algorithm in more depth here:

re: thresholds - fmriprep includes a few flags to tweak motion / spike thresholds (with defaults of 0.5 and 1.5 respectively). More info on the flags can be found on the docs:

@mgxd Thank you for your reply and references therein (Nichols2013). Much appreciated!