Dear Jon,
I have been looking for the same thing (longitudinal VBM methods), and have settled on using DARTEL via SPM8b (it is under the TASKS menu in both SPM5 and SPM8b). Here is my crude understanding: DARTEL warping is much fancier and more detailed than the usual SPM normalization (a LOT more parameters), so I think it will be great for quantifying subtle within-subject differences (like TBM... but everything is translated back into understandable units of volume, which I find preferable for aiming at non-imaging journals). For each subject, each timepoint is warped to a subject-specific template (something halfway between Time1 and Time2), and then you calculate the warp between each subject template to a group template. Therefore, the same warp is used on both timepoints to go from subject space into common space (as in Gaser's method). You can also calculate the warp between the group template and MNI space to put results in MNI space. The DARTEL toolbox has a way to compose warps to avoid multiple re-sampling of the images. I found this post enlightening:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0804&L=SPM&P=R48484
and here is the DARTEL manual:
http://www.fil.ion.ucl.ac.uk/~john/misc/dartel_guide.pdf
I am still figuring out details ... along the way, I plan to post any scripts that our group makes in order to make it happen. Also I want to mention the processing for DARTEL may be time-consuming. When I tried DARTELing two timepoints with the default smoothing, it took 17 hours per subject (how long will the group template with 100 subjects take?) -- without smoothing, it took about 2.5 hrs per subject. John Ashburner pointed out that the defaults were made for between-subject DARTEL; smoothing may not be necessary for the within-subject step since the images are already very similar.
Best wishes,
Dana
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf Of Jon Jackson [[log in to unmask]]
Sent: Thursday, November 20, 2008 6:49 AM
To: [log in to unmask]
Subject: [SPM] longitudinal VBM with SPM5 / SPM8B
Dear List,
I'm looking for the best way to conduct a longitudinal study VBM analysis
with SPM. Previously, under a discussion of 'longitudinal VBM with SPM5'
(Sep 2006)*, Christian Gaser recommended using the longitudinal
segmentation in the VBM2 toolbox. This is because (as I understood it) in
the VBM2 toolbox, the segmentation is performed using templates that are
aligned to the baseline and followup scans with the same registration
parameters whereas the SPM5 default segmentation will individually register
the baseline and followup scans.
Is this still the recommended approach or can this now be achieved with
tools in SPM5 or SPM8B?
Many thanks,
Jon
* http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind06&L=SPM&P=R543396
|