Dear Experts
I have an unusual analysis pipeline in which an alignment issue arises:
I have 4D fmri datafiles for a group of subjects. I align the subject's
individual 4D fmri files to a standard MNI space, but keep the original
scanner space resolution (64 x 64 x 36). I am able to do this using
flirt with applyxfm using the example_func2standard.mat from the
two-step registration for each subject, which I manually adjust to
obtain the correct center of origin for the image (this is documented
elsewhere). I also use a reference image using fslcreatehd.
I then run my homebrew stats in which I compare across subjects. This
gives me a map of t values at the group level.
The problem is how to transform this t value map into the 'real' higher
resolution standard space (91 x 109 x 91). This problem arises because
this t value map does not belong to any specific subject, but
corresponds to the aggregate of all subjects.
My current method is:
1. Edit the header of the tvalue map using fsledithd and put the correct
voxel dimenions (4mm), put the intent_code = '1001', and the qform_code
= '1'.
2. Next I run flirt with -ref standard.nii.gz and with applyxfm using a
example_func2standard.mat file from one of the subjects.
This works quite well, but I am wondering what the experts have to say
about this method. Specifically, I am wondering about the fsledithd's
that I make, and whether there is a more overall recommended method to
transform group level t-value maps to standard space. What about the
possibility of creating an average example_func2standard.mat across all
subjects? Is this possible?
I know obviously that the recommended method of doing the analyses
within subject, and then transforming the t value maps of each subject
to standard space is more robust, but I would like to know the optimal
solution for my specific given problem.
Thanks for any help
Niels
--
Niels Janssen
Cognitive Neuroscience and Psycholinguistics Laboratory
University of La Laguna
Tenerife, Spain
http://www.neurocog.ull.es/en
|
