Hi everyone,
I have some trouble getting the dual_regression function to produce the
correct output files. I am quite convinced it has something to do with my
inexperience with regard to FSL though, so maybe the solution is really simple.
In any case, the function seems to run properly, going through the "creating
common mask", "doing the dual regressions" and "sorting maps and running
randomise" phases, but there is no statistical output to speak of. Some
background:
We want to look at connectivity differences in different brain states and we
would like to have some statistics to support or disprove our hypothesis.
I ran the temporal concatenation ICA for all resting state data, recorded
during either brainstate 1(33 segments) or brainstate 2(10 segments)
together, without specifying a design matrix. This worked fine.
Then I ran the dual_regression as instructed by the dualreg download page,
using the .filelist file from the ICA to produce a list of inputs.
However, running the function seems to produce incorrect masks (completely
black images), doesn't generate the necessary intermediate files and
subsequently doesn't actually do any permutation tests. It does produce
empty drA.*** files, and the drC.*** and drD.*** files contain errors since
the proper input for these calculations is missing (the dr_stage1_subject00000
& dr_stage2_subject00000 files)
It seems, then, that there might be something wrong with the (format of)
input data, but these are normal zipped .nii output files of the MELODIC
analysis, and look fine on visual inspection.
In case it matters, the design matrix was created using the wizard: higher
level design, with group (33 1s and 10 2s) and 2 EVs (group A: 33 1s, 10 0s;
group B: 33 0s, 10 1s) as factors. The design.con contains 4 contrasts: group
A>groupB, groupB>A, mean groupA, mean group B.
Is there anything obvious I'm missing here? Are there known reasons that can
cause the common mask generation or the creation of the intermediate files to
fail?
I'd be happy to provide more details if necessary.
best regards,
Eelco
--
Eelco van Dongen
PhD Student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
Nijmegen, The Netherlands
|