For a longitudinal analysis, our group did a within-subject template [your #1 below] and then made a template for the whole sample (#3, sort of) out of the templates made in step 1 for each subject. It is not preferred to do #2 because then the registration will be biased by the first timepoint. If there are changes over time (between time1 and time2), and change differently between subject groups, results could be attributed to differences in registration instead of actual differences.
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Alexis Hervais-Adelman
Sent: Thursday, September 02, 2010 5:10 AM
To: [log in to unmask]
Subject: Re: [SPM] VBM comparison of two groups using DARTEL
Hi Michael, (or anyone else so-inclined)
I'm also currently finding this issue confusing, and I would be grateful if you could explain (briefly) the rationale for creating a template between groups? I can understand why one would warp all the images to the same template, but why the intermediate step of warping them all to what is essentially a mean template?
This also takes me back to a question I asked a few days ago about what the appropriate DARTEL approach would be for a between-groups longitudinal design. Would you have any suggestions to help me choose between:
1) Make a template per participant, collapsing over time
2) Make a "template" from the first time-point for each participant
3) Make a template from all the data at once.
The approach in the VBM8 toolbox appears to be one of beginning by producing a warp for one image per participant to the standard template.
Thanks in advance for any help,
Alexis
|