Hi Irina
> I am running a multivariate analysis on distinguishing two conditions
> on a single-subjects level. For the purpose of multivariate analysis I
> do not normalize the data. For the analysis, i am interested in grey
> matter voxels only. So I need a grey matter image that would be
> coregistered with my functional images, to use that as a mask. What is
> an optimal sequence of operations to get there? I can 1) first segment
> the structural image, and then coregister (estimate and reslice) the
> grey matter image with the functional mean or 2) do
> coregister:estimate prior to segmentation, then segmentation, and then
> coregister:reslice after the segmentation. Or 3) I can first segment
> the structural image, then coregister:estimate using the corrected
> structural as a source, referenced to the functional mean, and
> introducing the same manipulation in the grey matter image by
> selecting it as "other images".
I think the easiest/best approach is (2) above; it makes sense to
start with your structural and functional in the same space. Note
that as a final step you may want to binarize your gray matter
segmentation; i.e., use imcalc to threshold, so that only voxels with
a certain probability of being GM are included. So for example, if
you only wanted voxels with at least a 50% chance of being GM, in
imcalc just select the GM segmentation and use the expression i1>.5 .
Hope this helps,
Jonathan
|