> Are there any serious disadvantages for coregistration (motion correction)
> and Talairach normalization if one only collects slices less than a whole
> brain? (We want to image the auditory cortex.) My impression is that the
> following obstacles might be an issue:
> (1) For motion correction, "out of plane" motion leads to motion in and out
> of the "cylinder" of (axial) slices, and perhaps the coreg. algorithm might
> not deal with that well. (To restate that, if you collect only a subset of
> slices, out-of-plane movement means that the actual tissue imaged varies at
> the "top" and "bottom" of the cylinder.)
(1) If the subjects have moved out of plane, then you are not comparing the
same brain regions in all subjects. I think it is still best to realign,
even though you may lose some data at the top and bottom of the field of view.
One thing that you may wish to tweek if you have a limited field of view is
the "Registration Quality" (via Defailts). This option allows much faster
realignment with little or no difference in registration accuracy for images
with a good field of view. However, if you have a slightly smaller FOV then
you may wish to adjust the settings slightly - increasing the quality.
> (2) For Talairach normalization, one might have to adjust the template with
> the whole brain by masking out the parts not imaged, etc.
I wouldn't really recommend adjusting the template - providing that there is
data (head or air) in the whole image volume and not empty planes of zeros.
SPM spatial normalisation minimises the mean squares difference between the
source image and template, but only for those voxels that are common to both.
A voxel in the template that maps outside the FOV of the source image will
be ignored in the computation. If the FOV is really small, then the spatial
normalisation will default to just an affine transformation. Also, it may be
worth repositioning the images (via Display) in order to have better
registration starting estimates.
One thing to watch out for is if you are doing multi-subject studies where
all of them have very small FOVs. Only those voxels where there is data for
all subjects are included in the analyses, which can mean that the analysed
region is very small.
Best regards,
-John
--
Dr John Ashburner.
Wellcome Department of Cognitive Neurology.
12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420
http://www.fil.ion.ucl.ac.uk/~john
mail: [log in to unmask]
|