Hi,
Your use of rmsdiff is completely different to how you should use it to assess the within-timeseries fmri motion. What you are doing instead is measuring the change between the structural scan (highres) and the functional scan, which will also take into account the change in FOV, and is pretty much useless.
Instead, you need to take the matrices output by mcflirt with the -mats option on and use these with rmsdiff, using the example_func as the refvol and either (a) the matrix for timepoint N and the identity matrix, or (b) the matrices for timepoint N and timepoint N-1. In (a) you get the absolute motion wrt the reference volume, whereas in (b) you get the relative motion from one timepoint to the next.
I don't quite understand how you are running fsl_motion_outliers, as INPUT=$REGDIR/example_func2highres.mat is not sensible in this context. You need to use the original 4D image data (prior to motion correction) or the motion corrected data and the --nomoco option. Also, the values from dvars are not a measure of mm motion, they are related to differences in the image intensities, so it is perfectly reasonable to get larger values. Normalising dvars does not turn it into a measurement of movement.
All the best,
Mark
On 20 Aug 2013, at 16:42, Allison Jack <[log in to unmask]> wrote:
> Hello FSL experts,
>
> I would like to report an RMS movement value for each of my participants, so I was attempting to use rmsdiff. However, the results I get from my calculation seem FAR larger than anything I have ever seen reported. For example, for participant 1, the MCFLIRT results in prestats indicate a relative mean displacement of .06mm & an absolute mean displacement of .05mm; by my calculations from the par file, their maximum deviation from one volume to the next is .31 mm. However, the same participant's result from rmsdiff is 54.3698. (!) I assume I have missed some very important step, but although I have read through the rmsdiff wiki and the technical report I cannot seem to figure out my error.
>
> I ran rmsdiff using the following, based on a thread I had seen in the listserv:
>
> REGDIR=$STUDYDIR/ANALYSIS/FSL/LEVELONE/$SUBJ/runA.feat/reg
> REFVOL=$REGDIR/highres.nii.gz
> INPUT=$REGDIR/example_func2highres.mat
> IDENT=/Applications/fsl/etc/flirtsch/ident.mat
>
> rmsdiff $INPUT $IDENT $REFVOL
>
> As a side note, my DVARS values (the text file produced by fsl_motion_outliers, e.g.:
>
> fsl_motion_outliers -i $INPUT -o $OUTPUT --dvars -s $DVARS)
>
> are also similarly large. I saw a technical note by Thom Nichols on standardizing DVARs and I was wondering if that is what I might need to implement here; if so, could someone provide some examples of use of his DVARS.sh script (http://www2.warwick.ac.uk/fac/sci/statistics/staff/academic-research/nichols/scripts/fsl/dvars.sh)?
>
> Any correct usage examples would be helpful here.
>
> Thanks so much!,
> Allison
|