Dear Kaike,
> I'm studying effects of general anesthesia on rCBF using [15O]H2O and PET.
> I've made a substraction analysis (n=8) of conditions "concentration1" -
> "awake". Results show a large (>16000 voxels) highly significant cluster.
> Problem is that the cluster is situated partly outside brain tissue; around
> hypophysis. I suspect this is due vascular effect/ artefact. Furthermore,
> the artefact-part of the cluster seems to be 'cut' or partly masked out from
> fronto-inferior side.
>
> I've used FullMonty model with
> proportional global scaling (mean to 50),
> no covariates or nuisance variables,
> proportional treshold masking (0.8 as treshold)
> no implicit mask,
> masked images explicitly with brainmask.img
> and - global mean calculation using fullmean/8 mask
>
> We figured that brainmask.img could be manually altered to exclude this
> artefact area from the analysis. Using result rendering I've visualised the
> cluster on brainmask.img. Although the shape of cluster resembles reasonably
> well that of the mask, the outlines do not match exactly. This made me
> suspicious; are there other factors/procedures contributing to restrict the
> area of analysis, which should be delt with before/rather than tamper with
> explicitly used mask?
Yes, there are other reasons, why voxels wouldn't be included into the
statistical analysis. Take e.g. a look at:
http://www.mailbase.ac.uk/lists/spm/2000-06/0196.html
In your case, the analysis threshold (which you set to proportional
threshold masking with threshold 0.8 of the estimated intracortical mean
image intensity) is likely to remove some voxels prior to the analysis.
You can check exactly, which voxels have been included by looking at the
image called 'mask.img' in your analysis directory. Also, check out the
'check reg' button, this will help you with displaying registered
images.
>
> Should I proceed with the initial idea of manually altering the
> brainmask.img with, say, MRIcro? If not, I would most appreciate if someone
> could suggest an alternative solution?
Yes, an alternative of modifying the analysis mask would be to take the
computed mask.img as a starting point.
Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-(0)20-7833-7478
FAX : -7813-1420
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|