I think 0.5mm is reasonable if you are using framewise displacement (i.e. the difference in motion parameters for successive TRs).
I did a quick check on some data I have on hand, and there was relatively few TRs at 0.5mm. The ones which were >0.5mm seems to be pretty clear motion spikes, and I suspect at a lower threshold (e.g. 0.3mm) you will end up reinterpolating a lot of data points. .
If you have not read it I recommend Johnathan Power’s paper http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3849338/
They define 0.5 as lenient and 0.2 as stringent, and I think the effect of using either (or an intermediate) threshold depends a lot on the character of your data.
I also think it is not unreasonable to try a few threshold and see how many TRs are censored, and use that as a guide, For example, if 05 removes 2% or TRs and 0.3 removes 5%, maybe 0.3 is a good choice (removing motion more aggressively but not destroying data integrity). Just in that case report that you did this.
Good luck!
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Joelle Zimmermann
Sent: February-04-16 9:47 AM
To: [log in to unmask]
Subject: [FSL] framewise displacement
Hi FSLers,
I was wondering whether 0.5mm was a reasonable value for frame-wise displacement cutoff for removal of motion outliers using interpolation?
Thanks,
Joelle