Dear Felice, You need to use a process known as 'back-normalization'. Others probably have more experience than me with this but in principle it is simple. I'm assuming you have a structural T1 MRI for your subject. 1) You need to make a copy of the spm single subject T1 template. You are going to back normalize this to your subject's T1 MRI. 2) You need to smooth the subject's T1 MRI with an 8x8x8 filter as spm uses smoothing in the normalization process. You should also set the origin of this image to the ac point (by hand- using display functions). 3) You now normalize the single subject T1 template to your subject's T1 MRI. Just remember to select the template when prompted for the object and your subject's smoothed MRI when prompted for the target. This will produce a *sn3d.mat file which can be used to transform any other data (such as thresholded t-maps), generated for that subject in MNI space, back to their own unique space. You need to bare in mind two things: 1) This is not a perfect solution, it is not the same as 'inverting' the *sn3d.mat file that transformed the subject's MRI into MNI space, I'm not sure that that can be done. 2) Of course, the coordinates that activations will now have in the subject's own space will be meaningless in general terms, but could be used to spatially compare within subject activity. Cheers. Alex. > Date: Wed, 16 May 2001 20:59:25 +0100 > From: Felice Sun <[log in to unmask]> > Subject: Normalization > To: [log in to unmask] > > Hi, > > I'm interested in projecting coordinates from a normalized image back > to the original image coordinates. Does anyone have any advice about this? > > Thanks, > Felice