Dear all,
I'm not sure whether I should use event design or block design. I have (for
all subjects) one session with 3 actual conditions (A B C) + Rest (R) (so 4
conditions so far). I also have some scans inbetween which are instruction
scans (patients were given instructions what they should do next).
First problem was with these instruction scan. I specified them as
condition, so I had 5 conditions presented as follows:
I A B I C I R six times.
The actual problem occurred when I created design matrix and specified
conditions as events (I used hrf with time derivatives as basis function).
For some subjects the t-test showed absolutely no activation (zero spots)
and for some only few tiny spots. I tried the same with hrf alone and the
result was same. When I specified the conditions as epochs, t-test showed a
lot of activation.
I know these conditions are not clear events (one condition lasts 10 scans/
25 seconds) but is it possible to analyze these as events? Or should I just
analyze them as epochs?
I want to do RFX analysis so thats why the t-test is important.
Another question about pre-processing:
The functional scans were not in ac-pc-line so I used Display to turn them
as close that line as possible (they are EPI scans, so one cannot know that
for sure). Then I coregistered them with high-resolution T1 scans (they were
on ac-pc-line) using T1 as target. After that I normalised the T1 scans
(determine parameters only) and used these parameters to normalize my
functional images (write normalised only).
Is this the right way to do when your functional images are not in
ac-pc-line? Can I trust the results? Are the activation spots in right
places?
Thank you in advance,
Eini Martikainen
University of Kuopio
Finland
|