Dear Urs,
On Tue, Nov 22, 2011 at 3:03 PM, Urs Bachofner <[log in to unmask]> wrote:
> Dear SPM Experts,
>
> I need help building a statistical analysis in SPM8. I know exactly what I
> want to do, but there seem to be a lot of different ways of doing it.
>
> Here's the plot: Of 40 subjects of different age groups (Age 8 to 18) I have
> activation images (.nii) of an oddball task, generated by 3D source
> reconstruction of EEG files. They all did the same oddball task in the
> evening and in the morning, after sleep.
>
> In earlier studies my research group found that the GMFP in such a task
> decreases over night, so now I wanted to test if the involved areas also
> decrease over night.
>
> My first idea was to use imCalc to subtract the morning-images from the
> evening-images, but I encountered the problem, that the areas involved in
> solving the task differ in differerent age groups, so of course those
> differential images also show different voxels involved.
> So even though I find a greater neuronal response in the evening in all
> subjects, it makes no sense to t-test the differential images, since the
> remaining activation is in different voxels.
>
> The next idea was to define ROIs of the same size in all differential images
> (evening minus morning) and average the activation therein. Does this make
> sense? What functions of the fMRI part of SPM8 is able to do this?
>
It's not completely clear to me whether you are talking about very
close areas that just don't overlap precisely or about completely
different areas. In the former case you could use group inversion to
try to make sure that the sources are the same or even use fMRI priors
option (with a mask for your areas of interest). You can also increase
the smoothing parameter (available in the batch tool 'M/EEG source
reconstruction results'). You could also average over an ROI covering
the voxels where you see an effect in all the subjects as long as you
could justify the choice of that ROI in advance and not based on the
data (which should be OK in your case as there is a lot of literature
about the sources of MMN). I think the tools to work with ROIs are not
in the main SPM but in toolboxes like MarsBar but I'm not familiar
with these toolboxes myself.
If you are talking about completely different and non-adjacent areas I
don't see any physiological sense in putting them in the same test. I
would then just do separate tests for the different age groups which
do share the same sources and report the results of these tests
separately (with Bonferroni correction across age groups).
> Apart from this, as I am primarily interested in activation of frontal
> regions, is it necessary to cut away the rest of the activation, for exmaple
> by using MarsBaR? Or is there a tool in SPM8 wherein I can set all areas I'm
> not interested in to zero?
You don't need to set them to zero, but there is an option to use
explicit mask when you set up a statistical test or you can do small
volume correction with a mask image to define your volume.
>
> Another idea was to simply use a two-paired dependant t-test using both
> morning and evening images, but I think this won't get me any information
> about the decrease of activation that happened over night.
>
Paired t-test is equivalent to single-sample t-test on difference images.
>
> Also, I'd be interested if the decrease of activation (or of involved
> clusters, respectively) changes with age. How should this be done? Do I have
> to use the differential images or can I work with the images of evening and
> morning?
I think you should correlate the differential images with age but
it'll only make sense if you are looking at the same area/ROI across
subjects.
Best,
Vladimir
>
>
> Any hints, solutions and clues are greatly appreciated.
>
> Thank you,
>
> Urs Bachofner
>
>
> --
> NEU: FreePhone - 0ct/min Handyspartarif mit Geld-zurück-Garantie!
> Jetzt informieren: http://www.gmx.net/de/go/freephone
|