>I 'm working with PET images in order to superimpose normal patient to patients with desease.
>- Do I need to realign different PET scans from different normal patients (one scan for eash patient)? If yes, in the realignment, the smoothing (FWHM) to be used for PET images is 5 or 7? In the SPM manual it is written 7 but the default value for PET is 5.
No, the "realign" option is for within-subject, between-scans realignment (i.e. if you had a H2[15O] or fMRI time series). As you're working with FDG PET, if you only have one scan per subject, you would normalise them to a stereotaxic space directly (I think we discussed this in a previous mail - Unified Segmentation via their MRI and SEGMENT in SPM5 (or NEW SEGMENT in SPM8) preferred, otherwise NORMALIZE to an FDG PET template, ideally from the same scanner).
If you have, now or in the future, two or more FDGs per subject, you would bring them in register prior to one of those normalisations via COREGISTER:ESTIMATE ONLY (i.e. without reslicing).
>- When I'm doing the normalisation, in the writing options, I have to use "preserve amount" or "preserve concentrations"? My feeling is for the first one. And about the voxel size, I have to leave the default values (2 2 2)?
No, it should normally be "preserve concentrations" - see SPM archives for the reasoning behind it (probably under "modulation").
You don't have to leave the default voxel size of 2x2x2mm, but it works rather well...
>- it is fine 12 for the FWHM of the smmothing?
Should be ok, particularly for looking for spatially extended changes like in Alzheimer's.
Hope this helps,
All the best,