Caitlin,
I found that protocol difficult to grasp, too...
> If we understand correctly, in creating gray and white matter
> templates, this means two separate templates, one for gray matter
> and one for white matter.
That's correct. These (if you have enough subjects to create your own)
will be somewhat alike to the gray.img and white.img in the
apriori-directory of SPM99, the idea being to base gray matter
normalization on gray matter only.
> During the first segmentation to remove non-brain voxels, are
> the resulting images whole brain, or does this step create two
> separate images, one gray matter and one white, per subject?
Well, if you segment the images in native-space, you get 3 sets, gray,
white, and csf ( ..._seg/2/3). In order to remove non-brain voxels, you
need to select the _seg1.img and _seg2.img to create an individual
brain-mask that defines where there is brain tissue (i.e., gray or white
matter) and where not. This brain_...image can then be used to remove
non-brain pixels from your segmented images, both gray and white matter.
John posted an optimized formula for doing this so that pixels sum up to
100% some while ago, using imcalc.
> Because of sub-optimal intial segmentation, Good et al, state "Thus,
> we re-applied the optimal normalization parameters to the original
> structural images." Is our understanding correct that the
> function of the first segmentation and normalization is only to
> arrive at normalization parameters? And then the process begins
> again, using the original (whole brain) images, but applying the
> parameters derived from the first two steps to the normalization?
Exactly: you segment, clean, normalize only to get "optimal"
normalization parameters which you then apply to yout original image.
> If this is correct, it seems that the first normalization procedure
> occurs with separate gray and white matter templates, but the second
> to whole brain images.
Yes, but you need to normalize your brain based on the optimal gray
matter-parameters in order to get optimally normalized and segmented
gray matter iamges, and based on white matter in order to get optimally
normalized and segmented white matter. I understand that there is some
reservation on white-matter VBM, though, on the grounds of statistics
beyond me.
> How does the transfer of parameters work?
Click Normalize, write normalized only; choose your "optimal" parameters
and the original image.
Good luck!
Marko
--
===============================================
Marko Wilke, MD
Research Fellow, Imaging Research Center
Children's Hospital Research Foundation
3333 Burnet Avenue - ML 5031
Cincinnati, OH 45229
Phone +1 (513) 636-0142 FAX +1 (513) 636-3754
[log in to unmask] http://www.irc.chmcc.org
|