Dear Luca,
Maybe this http://www.sbirc.ed.ac.uk/cyril/SPM-course/Talks/2015/12_Morphometry_Edinburgh2015.pdf is a good start. In short, VBM is an analysis technique. Unfied segmentation allows to normalise images into a certain template space (e.g. MNI) based on a-piori tissue priors (e.g. the six tissue type images in SPM), segmentation and normalisation take place within the same generative model. Dartel or Shoot allow to register images together interatively onto their study-specific average (or an a-priori Dartel template), and the registration is higher-dimensional, they do not incorporate a segmentation step.
Now when working with Dartel one would segment the T1 images, and forward the output (GM / WM tissue images) into Dartel or Shoot. One could also segment PET images accordingly, but the tissue priors might just not be optimal for segmenting PET images.
However, one could also Dartel-register together raw T1 images of different subjects (or possibly intensity-corrected versions of the raw images) or PET images. Whether this makes sense in practice is another issue, e.g. the intensity distribution of different PET scans might be very different and accomplish the registration.
Hope this helps for the moment
Helmut
|