Parametric images need to be normalized to a common template. When MR is available, the MR image can be normalized to the T1 weighted MNI atlas. However, in some cases it may be desirable to use a PET template that is in MNI space. Such a template is used for PET to PET normalization.
Creation of a template using SPM
PET to PET Normalization can be beneficial for subjects that do not have MRI. The basic idea is to create a PET template from the other subjects who do have MRI. To do this, MRI data from these other subjects is normalized to a MRI template (T1.mnc), and then this same normalization is applied to the PET data to fit the PET data to the same space as the MRI template. In normalization, the output is a normalized file (wrPET_mean.img) and the transformation matrix (wrPET_mean_sn.mat). The normalized PET files are then averaged and smoothed to make a PET template.
Normalization to a template
For PET to PET normalization, a mean PET image from a dynamic study is normalized to the PET template. At this point, the normalized PET images and MRI template are in the same space and ready for SPM analysis.
For MR to MR normalization, the subjects MR image is normalized to the T1 weighted MNI atlas. When a corresponding PET image is coregistered to the MR, then the same MR normalization can be used to normalize the PET to the MNI atlas.