Dear Ben,
Ben Bowles wrote:
> Hi everyone,<br>
> <br>
> Inherent in spatial normalization in SPM is the realigment of native
> images to account for subject motion by estimating the six-parameters
> used to minimize the difference between each scan and the reference
> scan. <br>
> <br>
> At this stage, SPM also provides the ability to write the 6 parameters
> in a file, which can then be entered later into the model specification
> level (fMRI-->design), in the design matrix. <br>
> <br>
> Would anyone recommend doing this? Or, is the realignment at the
> normalization stage sufficient?
The first part (realignment aka registration) is really needed, entering
the realignment parameters as regressors in your analysis is optional.
The latter removes 'activation' which is *linearly* correlated with with
your movement parameters. There is really no theoretical basis for this
assumption / simplification. I think some people do it by default to
reduce the number of motion-related false positives, others think it is
nonsense. It is conservative in the sense that it may remove
task-related activation along the way, especially if you have
task-correlated motion. In practice, depending on your design/task, this
approach may be beneficial or detrimental. I suggest you try and see how
things work out with your data. E.g., I use a design in which I separate
the largest movements from my task of interest (tasting a liquid), by
having subjects swallow after tasting. In my case, including the rp's as
regressors seems to work out fairly well, I suppose because most
movement-related variance does not coincide with my task. In other
cases, however, this relatively crude approach may remove your precious
activations altogether.
You may find more info by searching the SPM archives at
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?S1=SPM
My 2cts, regards,
Paul
> It seems to me that there is little
> sense in entering in the 6 parameters at two separate stages, if it is
> the same information that is being applied both times. I have heard it
> argued that this a more conservative way however. <br>
> <br>
> Opinions appreciated! <br>
> <br>
> Thanks<br>
> <br>
> Ben<br>
--
Paul A.M. Smeets, PhD
Image Sciences Institute
University Medical Center Utrecht
Heidelberglaan 100, Rm Q0S.459
3584 CX Utrecht, The Netherlands
Tel: ** As of 1 October 2007: +31 88 75 56682
Fax: +31 30 - 251 3399
|