Dear SPM experts,
I am afraid this is a question that has extensively been discussed in
tharchives before. Nevertheless I try to ask the following questions
again because the problem is still not clear to me.
We have performed an event-related fMRI study comparing two groups
during performance of a motor task (TR 2.5s).
We are primarily interested in differential activation between both
groups for the main effect, i.e. during performance of the task vs. rest.
Having read about the problem with slice timing, we have built a design
matrix using a set of 3 basis functions, i.e. the hrf and its 1st and
2nd derivative for each subject.
Now we are not sure what ist the best way for performing the group
comparison and how to do it in SPM5.
1.) simply performing a two-sample t-test comparing the contrast-images
for the hrf between both groups? Do the derivatives have any influence
on the result in this analysis, or would it be better in this case to
'simply' perform slice timing?
2.) as we read about bringing all three basis functions onto the second
level we also tried this by building a "factorial design" with 2 factors
'basis function' (3 levels, dependent between levels) and 'group' (2
levels, independent between levels). This results in 6 cells [BF1,
Group1], [BF1, Group2], [BF2, Group1], [BF2, Group2] and [BF3, Group1]
and [BF3,Group3]. Here we entered the corresponding contrast images from
the first level for all basis-functions / subject
Is this design correct? does SPM know that for example the first scan in
cells 1, 3 and 5 or 2, 4, 6 'belong together'?
If yes, what are the correct contrasts to perform the group comparison?
Does it make any sense to simply perform a t-contrast e.g. 1 -1 0 0 0 0
for testing for stronger activation in Group A vs. B? In this case what
is the role of the other BF beyond adding more degrees of freedom?
Is it necessary to perform F-contrasts and what would be an appropriate
contrast to look for differential activations between both groups? I
assume this can only test for different shapes of the hrf. Or is there a
way to test for stronger activation using a F-contrast. In a second
step, what is the best way to see in which aspect the hrf-shape differs
(I assume plotting the BOLD response in the voxel resulting from the
group comparison for an within-group analysis for each group)?
I would very much appreciate any help even if this questions seems to be
trivial...
Thank you very much in advance!
Bernhard
--
-----------------------------------------------------------------------------------------
PD Dr. Bernhard Haslinger
Oberarzt
Neurologische Klinik und Poliklinik
Klinikum Rechts der Isar
TU-Muenchen
Ismaninger Strasse 22
D-81675 München
E-mail: [log in to unmask]
WWW: http://www.neurokopfzentrum.med.tum.de/neurologie/
-----------------------------------------------------------------------------------------
|