Mark -- I ran this and it seems to work with one exception: In step 6. I
get a "could not find example_fun2highres.mat". Isn't it enough to run
"updatefeatreg . -gifs" under /reg? All updates seem to occur when run
as such. Thanks for your help on this. Best -- Julius
Julius Fridriksson Ph.D.
Assistant Professor
Communication Sciences & Disorders
University of South Carolina
Tel: (803) 777-5931
Fax: (803) 777-3081
>>> [log in to unmask] 03/10/05 09:53AM >>>
Hi,
The best way to do this analysis is still to analyse each session
individually
and then combine them later in a higher level analysis. This allows
both
the within-session and between-session variances to be modelled
better.
However, you are right in that you need to make the registrations work
specially, as the default in FEAT certainly won't work.
What I recommend is the following:
1 - pick a reference image (e.g. an anatomical scan from that subject)
2 - run FEAT on each individual session, but deselect all
registrations
except the one to standard space, but use the reference image
you've
selected previously as the filename for the "standard space" image
3 - after FEAT has run, register each session's example_func to the
reference image separately using FLIRT with cost function
weighting
to reduce the effect of the lesions.
4 - copy the FLIRT matrix from above into the reg sub-directory of the
appropriate FEAT folder, but call the matrix
example_func2highres.mat.
Also create a highres2standard.mat file which contains the
identity
transformation - that is the following 4 lines:
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
5 - in the reg subdirectory do "imln standard highres"
6 - do "cd .." (to get to the main .feat directory) and run
"updatefeatreg . -gifs" here
This will regenerate the example_func2standard.mat file and
all appropriate matrices and images along with it.
7 - Now you can run higher level analyses as normal.
Note that steps 4 and 5 are necessary in this case to trick
updatefeatreg
into dealing with just an example_func to standard space registration.
The key is that if you replace key matrices in the reg subdirectory
then
updatefeatreg will fix everything else in this subdirectory to
incorporate
them correctly, allowing higher level analysis to proceed correctly.
I hope this makes sense to you, and let me know if you have any
trouble
with this.
All the best,
Mark
Julius Fridriksson wrote:
>Hi -- I have a question regarding co-registration for stroke
patients:
>We employed three pre- and three post-treatment scanning sessions in
>several persons with large unilateral brain lesions. The plan is to
run
>a higher level analysis for each subject to get mean activation
across
>the three pre- and three post-treatment scans. My quesion is this:
>should we co-register the three functional runs using FLIRT before we
>run the actual first level analyses? We would like to mask out the
>lesion in each case for volume weighting. We are not interested in
>colapsing across the subjects, only across the three pre- and three
>post-treatment scanning sessions within each participant. Thanks very
>much in advance -- Julius
>
>Julius Fridriksson Ph.D.
>Assistant Professor
>Communication Sciences & Disorders
>University of South Carolina
>Tel: (803) 777-5931
>Fax: (803) 777-3081
>
>
|