With regards to the motion question I think you should almost certainly
include these regressors in all subjects.
I offer a practical reason for doing so: if you do not, a reviewer likely
will shoot you down for it. They will assert, not incorrectly, that
different subjects have been treated differently, and this may have
unpredictable effects on your results.
If you are moving to a second level analysis, the loss of degrees of freedom
in the first level shouldn't matter much.
However, as a caveat, motion regressors do a poor job of controlling for
motion. The motion parameters are an estimate of overall movement within a
TR< and would be valid if people made sudden specific motions only in
between TRs. This isn't the case, and the motion parameters often poorly
estimate actual motion.
And if you look at voxel time courses, motion artifacts often peak after the
TR in which the spike appears in the parameters.
There are scrubbing options you can choose. You can censor and/or
reinterpolate TRs with motion, plus an additional TR or 2. AFNI's despike
option can also help, as it removes outlier from the voxel time series
(independent from the motion parameters). Also, there are approaches such as
Daimien Faire's and Ted Satterthwaite's approaches which include more
extensive motion paramters.
Good luck,
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
www.neuranalysis.com
[log in to unmask]
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Matthew Webster
Sent: March-01-16 11:00 AM
To: [log in to unmask]
Subject: Re: [FSL] Motion Outliers + Physiological Noise Model
Hi Ella,
Thank you for your upload, unfortunately it isn't clear what
is failing from the logs. Can you tar and upload the full FEAT directory to:
https://oxfile.ox.ac.uk/oxfile/work/extBox?id=700839B35190E48BF
Kind Regards
Matthew
> Hi all,
>
> I'm running a set of first level Feat analyses on task-related BOLD
activation in 20 participants, in which some of the participants have moved
more than a voxel-worth. I have therefore included the motion outliers
confound file in the analyses.
>
> My first question is: should motion outliers be run all participants in
the study or is it okay to only include this confound file for only those
who have moved excessively?
>
> As this confound file reduces the dof in the GLM, I am keen not to do this
unnecessarily, but my instinct was to be consistent across all participants.
Can you advise?
>
> I also ask as we also have physiological noise modelling for this dataset.
This also generates a confound file. We have both motion outliers and Pnm
confounds for 18/20 participants, but the first level analyses have failed
for 14/18 of them. It runs the Stats section, then an error occurs at Post
Stats so none are computed. Here is the error message (taken from the Stats
section into the Post-Stats section in case helpful):
> Log directory is: stats
> paradigm.getDesignMatrix().Nrows()=1598
> paradigm.getDesignMatrix().Ncols()=131
> sizeTS=1598
> numTS=117655
> Calculating residuals...
>
> Post-stats
>
> child process exited abnormally
> while executing
> "if { [ catch {
>
> for { set argindex 1 } { $argindex < $argc } { incr argindex 1 } {
> switch -- [ lindex $argv $argindex ] {
>
> -I {
> incr arginde..."
> (file "/usr/local/fsl/bin/feat" line 309) Error encountered while
> running in main feat script, halting.
> child process exited abnormally
>
> My second question is: does anyone know a reason why this error is
occurring or how I can investigate where in the script the error is
encountered? I assumed it was just the impact of including both sets of
confounds, but as 3 have so far run without issue, this can't be the reason.
>
> I would be extremely grateful for your help, With best wishes Ella
> Hinton
|