> Two more questions arise while making a template using the standard VBM
> protocol of Goode et al:
>
> 1.) When setting the normalization defaults, the program asks whether to
> mask the template brain with a "default" or "specified" brain mask. I
> have been selecting the default brain mask, but would prefer to use a
> study-specific mask. Similarly, when selecting the images for
> normaliization, the program asks to "select an object masking image". I
> have been selecting the very same image that I am normalizing, reasoning
> that the most specific mask for an image is itself. Is this a good
> approach to finding an appropriate
> brain mask ?
Are you spatially normalising by matching T1 to a T1 template, or are you
registering grey matter to the grey matter image from the SPM distribution?
If you are matching grey matter images, then you don't want to do any
masking/weighting.
Why do you want to use a study specific brain mask? What the brain masking
does is to weight the registration so that the cost function is a weighted
mean squares difference between the template and the image you are spatially
normalising (neglecting the effects of regularisation). The weights are
derived from the image you chose as the brain mask, which by default is the
apriori/brainmask.img. Each voxel of the weighting image should correspond
to the same voxel of the template image. This way, voxels within the brain
should contribute to the cost function, whereas those outside do not.
Object masking is different. Each voxel of the object mask, should correspond
with a voxel in the image you are estimating the spatial normalisation parameters
from, rather than the template image. Object masking is intended for spatially
normalising lesioned brains, so that regions containing abnormal voxel intensities
do not influence the final result (i.e., they are given zero weighting). If you
are not spatially normalising lesioned brains, then it should be disabled. See
http://www.psychology.nottingham.ac.uk/staff/cr1/mritut.html#Normalize for more
information.
>
> 2.) The second question relates to how to select images from the whole
> sample to use in the template. Goode et al chose 60 males and 60
> females, matched on age and handedness, and then matched these to the
> entire study group, also on age and
> handedness.
>
> I have selected 11 of 21 subjects for the template, whose age and
> handedness do not differ statistically from the total group or the
> other 10 subjects not used in the template. However, with such a
> small sample, couldn't I make a study specific template using
> all 21 subjects? Are there any methodological problems tto this
> approach?
A study specific template from the whole group would indeed be preferable
to a template made from a subset of the individuals. There is a little
bit more on this at:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind9904&L=spm&P=R6120
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0111&L=spm&P=R7347
Best regards,
-John
--
Dr John Ashburner.
Functional Imaging Lab., 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
|