Hi Britt,
it is very likely that your MPRAGE and SPGR images will differ in terms of intensity distribution,
signal-to-noise and contrast-to-noise ratio. You can check this issue by creating a single
template for all subjects and comparing the subjects between the MPRAGE and SPGR sequence. If
no differences can be found you might combine both sequences (use a single template) and
simply calculate the correlation to your behavioral measures (simple or multiple regression). In
case of differences you can try to use a statistical model with 2 groups (MPRAGE/SPGR) and 1 or
more behavioral measure. The correlation for each group can be combined using a conjunction
approach (e.g. conjunction of contrasts 1 0 and 0 1), see http://dx.doi.org/10.1016/j.neuroimage.2004.10.021. Maybe it's sufficient to simply test for the
average correlation (contrast: 1 1).
An alternative and maybe more elegant way is to use SPM5 and the DARTEL toolbox by John
Ashburner. The use of the high-resolution diffeomorphic registration approach might help in
combining data from different sequences or even scanners (see http://dx.doi.org/10.1016/j.neuroimage.2007.09.066). However, you should still define each
sequence as separate group in the statistical model, but you don't need to use a conjunction and
can simply test for the average correlation (contrast 1 1).
Best,
Christian
-
____________________________________________________________________________
Christian Gaser, Ph.D.
Assistant Professor of Computational Neuroscience
Department of Psychiatry
Friedrich-Schiller-University of Jena
Philosophenweg 3, D-07743 Jena, Germany
Tel: ++49-3641-935805 Fax: ++49-3641-935280
e-mail: [log in to unmask]
http://dbm.neuro.uni-jena.de
On Mon, 14 Jan 2008 22:26:42 +0000, Britt Carlson Emerton <[log in to unmask]> wrote:
>Hello. I am working on a VBM project with some data from a heterogenous psychiatric sample
>that was given to me as a part of the master's portion of my graduate training and I've run into a
>methodological issue with which I need some assistance.
>
>The full data set is made up of 41 subjects all with behavioral measures of aggression and high
>resolution anatomical MR images. 18 subjects have MPRAGE and 23 have SPGR. In my efforts to
>learn VBM I started with the MPRAGE subjects and conducted an optimized VBM analysis using
Dr.
>Gaser's toolbox, but now I am interested in trying to analyze the whole sample together. Of
>course my mentors and I realize that having two acquisition sequences is not ideal but I think
that
>there may be some interesting results and wish to try understand the neural correlates of
>aggression in our full sample if we can minimize this error. I am a bit stuck regarding the best
>way to preprocess the data from the two acquisition sequences in SPM2. Specifically, I am
unsure
>whether it would be best to segment and normalize all images (regardless of scan type) to the
>same template or to separate templates.
>
>It seems to us that in order to maximize the probability of proper tissue segmentation that it
>would be reasonable to create two separate customized templates for each patient group based
>on scan type and to segment and normalize each subject to the appropriate templates rather
than
>to use a single template for all images. Then, once each subject's data are segmented and
>normalized, they can all be smoothed and statistically analyzed together, including scan type as
a
>nuisance variable.
>
>Does this sound like a valid approach? Is there a better way to deal with this issue (other than
just
>switching to SPM5 or giving up)?
>
>In my searching for an answer in the SPM archives I've been reading a lot about SPM5
>segmentation and standard templates being superior to customized templates and negating a
>need for the VBM2 "optimized" protocol. The only reason I'm hesitant to switch to SPM5 is
because
>I am not sure if I can use MARINA to specify custom ROIs in SPM5 - Does anyone know whether
or
>not this is possible?
>
>Any help will be greatly appreciated.
>
>Sincerely,
>
>Britt Carlson Emerton
>Graduate Student in Clinical Psychology and Neuropsychology
>Department of Psychology
>Suffolk University, Boston, MA
>===========================================================
=============
|