Hi,
On 5 Jul 2010, at 20:57, Darren Schreiber wrote:
> I'm trying to extract the time series from the CSF to use as
> nuisance covariate in a subsequent FEAT analysis. I used BET to
> extract the brain from a subject, FAST to segment out the CSF, and
> then fslmaths to threshold only the voxels that were above 80%
> likelihood (following Biswal PNAS 2010). I tried to use fslmeants
> with the filtered_func_data from a prior FEAT analysis to extract
> the time series and got the following error:
>
> fslmeants -i S02Q1h.feat/filtered_func_data.nii.gz -o CSF_S02Q1.txt
> -m S02brain_BET_5_pve_0_thr.nii.gz
> ERROR: Mask and Input volumes have different (x,y,z) size
>
> Which leads me to a few questions:
>
> 1) What space is filtered_func_data in after feat is typically
> run? I had assumed it was in the space of the subject's brain,
> rather than the standard space.
It is in the subject's *functional* space, which is different from
their *structural space*.
Typically the functional images have a resolution of something like
3mm while the
structural images will be 1mm or better.
>
> 2) What tool in FSL will allow me to see the dimensions, etc of the
> images?
fslsize shows just the size information, but fslhd shows all info from
the header.
>
> 3) How do I get the mask (which is made from the subject's brain and
> I believe in that subject's space) into the space of the
> filtered_func_data?
You need to use flirt together with the highres2example_func.mat
transformation file in
order to transform your mask. For example:
flirt -in S02brain_BET_5_pve_0_thr -ref example_func -init
highres2example_func.mat -applyxfm -out S02brain_BET_5_pve_0_thr_func
fslmaths S02brain_BET_5_pve_0_thr_func -thr 0.9 -bin
S02brain_BET_5_pve_0_thr_func
you need the fslmaths line to re-binarise the mask after
transformation, and I've used a conservative
threshold to exclude the partial volume edges (post-interpolation).
Once other thing I would recommend for your analysis is to avoid using
the filtered_func_data as it will have been temporally and spatially
filtered,
which is not really what you want in this case. For example, the
spatial filter
will mean that you are including non-CSF timecourses in your average.
However, you also want to generally have motion correction run (although
it may not make too much difference). So the best thing would be to run
a separate FEAT GUI where you select just pre-stats and turn off
everything
except the motion correction, and then in the feat directory that this
run produces
you will have a filtered_func_data file with motion correction but
without
temporal or spatial filtering. That is the best data to extract your
CSF signal
from.
All the best,
Mark
> Thanks in advance!
>
> Darren
>
>
> *******************************************************************************
> Darren Schreiber, J.D., Ph.D.
> Assistant Professor of Political Science, U.C. San Diego
> Assistant Adjunct Professor of Law, University of San Diego
> Political Science, SSB 367
> 9500 Gilman Drive
> La Jolla, CA 92093-0521
> dmschreiber (at) ucsd (dot) edu
> *******************************************************************************
>
|