| We want to do the statistical analysis for one subject on both the
| normalised EPI data and the unnormalised data. Having performed identical
| steps for each (except normalisation in one), in separate directories we
| looked at the glass brain pictures.
|
| In the normalised images we saw what we expected - several discrete blobs
| of activation. However in the unnormalised data the whole brain was
| coloured grey for the same threshold. Having moved the threshold higher we
| could identify peaks corresponding to those seen in the normalised data.
| Is this because our subjects brain was much smaller than the template or
| are we missing something? We have performed these steps with another
| subject and the normalised/unnormalised data look much more comparable at
| the same threshold. We double checked the design matrix and estimation.
Bizarre. I don't think its to do with normalisation, because it is the
un-normalised images that appear to show a problem. Do the time courses
show anything strange? Is there anything unusual in a plot of the globals?
|
| The second question is more cosmetic. In spm96 the coregistration option
| provided columns of grey, white matter, csf and other. Is it correct that
| spm 99 just gives columns of grey and white matter?
This is true for the coregistration. Since the final step only involves
registering WM with WM, and GM with GM, then there was no need to create
a CSF image at this stage. With no image, there is nothing to display.
Good luck,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|