> I have a couple of questions regarding the normalisation step in SPM5 using
> the Unified Segmentation procedure: - Is the procedure to (1) Co-Register
> EPI and anatomical T1 of the subject, (2) Normalise the T1 of the subject
> with a template delivered with SPM (T1.mnc?) (3) To apply the parameters
> derived from this normalisation to the EPIs. Is that correct or am I
> missing something?
If you use Unified Segmentation, then at no point does the algorithm use any
of the usual template images. The matching is done entirely with the three
tissue probability maps in the tpm directory. However, if you use this
approach, you need to manually reposition the images (Display button) so that
the AC is within about 3cm of the origin and the images are approximately
transversely oriented.
>
> - How does the Unified Segmentation approach compare to standard
> normalisation in particular in patients with mostly subcortical (but
> sometimes also cortcial) lesions (mostly due to stroke)
This paper may be of interest here:
Crinion J, Ashburner J, Leff A, Brett M, Price C, Friston K. (2007). "Spatial
normalization of lesioned brains: Performance evaluation and impact on fMRI
analyses". NeuroImage 37(3):866--875
>
> - If the normalisation results in bad results due to the presence of the
> lesion, is it possible to mask out areas which shouldn't be considered for
> normalisation?
This can help, but the above paper says more.
>
> - Finally, could one use Unified Segmentation normalisation in some
> patients and "classical" normalisation in other patients and then use them
> all in one statistical analysis? (i.e., mix both methods in one sample)
I wouldn't really advise this. If you use the Unified Segmentation approach
for everybody, then you should obtain more accurate results than the
"classical normalisation".
Today, it really seems like there is a demand for an automated (or at least
semi-automated) lesion segmentation approach.
Best regards,
-John
|