Dear Ignazio,

>many thanks for your help.
>I have much appreciated it.
>Here some my further naif questions.

>1. take representative subjects (controls of similar age, or a mix of patients and controls);
--It would very difficult for me to have 10/12 volunteers of similar age. It is fine from 30 to 50/55? 

I just meant "similar to the patients" :-) - between ~20 and ~55 nothing much is happening, but after that you get quite a few changes so if all your patients are say 70-80 having a template made of 15 year olds might not be such a good idea...

>3. "coregister only" each subject's FDG PET to their MRI
--I suppose "Coregister only" means "coregister:estimate", and reference image=MRI and source image=PET;

Correct - in SPM5/8, make sure you do these manipulations in a "fresh" directory as the .hdr will get changed to take account of the coregistration

>4. "SEGMENT" (SPM5, SPM8) their MRI - this will create a *_seg_sn.mat file
--I have SPM5; Data=PET. Can I delete all the output files: gray,white and bias-field corrected (I suppose I don't need them)?

No, data would be MRI. (Actually it _might_ work with FDG PET, come to think of it, but that would be highly experimental!). The idea is to do this to the MRIs, get the reasonable SPM5 Unified Segmentation (Ashburner and Friston 2005 - that's what you access through the SEGMENT button), and apply this to the PET.

However, you are correct that for the purpose of building your FDG template you won't need gray, white and bias-field corrected images. If you wanted you could specify while doing SEGMENT that you don't even want them written - but it might be worth having a quick look at them by way of quality check.

>5. Use the *_seg_sn.mat file to write the PET into MNI/ICBM152 space -> wPET ("normalise write"; you may want to adjust the bounding box so the resulting images have an MNI/ICBM152 matrix; this can be very useful if you switch between software packages)
--Which bounding box do you suggest? The default one are: -78 -112 -50;78 76 85.

Fine if you'll be staying in SPM for the rest of your analysis - unless you're interested in inferior cerebellum which gets cut off with this BB. To get to the "classical" 91x109x91 MNI/CIBM152 matrix you'd need to change the numbers.

>6. Create a soft mean of all wPETs (see archives)
--I found a post by John Ashburner
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind04&L=SPM&P=R115230

Correct. An excellent post. In fact I forgot about the intensities mentioned in it; it's unlikely to be a problem (unless you use fixed dosing and have subjects of radically varying body sizes/compositions or unless they haven't been fasted systematically).

All the best
Alexander


Many thanks
Regards
Ignazio