> 1. How to normalize the individual whole brain anatomical image into
> standard space using SPM96? It will be nice to render individual
> statstics onto individual anatomy while maintaining the coordinate of
> common space.

Providing the anatomical image is in register (via the mat files) with
the image from which the normalization parameters were estimated, then it
is simply a matter of using the `Write Normalised Only' button.  First
select the `*_sn3d.mat' file in which the parameters are stored, and then
select your anatomical image.  To obtain high resolution spatially
normalized anatomical images, you can simply change the voxel sizes
via the defaults before writing your normalized images. 


> 2. For individual study, e.g., presurgical mapping on a case of brain
> tumor, a rationale will be to preserve the individual anatomy with
> SPM(Z) cogistered or rendered on the anatomical image without
> normalization. What is the best way to achieve this?

Please note: SPM IS A RESEARCH TOOL, AND NOT FOR CLINICAL USE.  It does
not have anything like FDA approval, and the authors can not take any
responsibility for anything that may go wrong when it is applied to
pre-surgical planning, or used as a diagnostic tool.

However, if you are not doing pre-surgical mapping, then I'm afraid that
there is not a simple way of doing this in SPM96.  The closest I can suggest
is to co-register the images, and estimate spatial normalization parameters
from one of the images using only a six parameter affine transformation.
You can then apply these parameters to all images.


> 3. We have recently come across the situation that we had to adjust basis
> function for different individual to obtain a good normalization result
> even if the scanned volume (slices) were pretty much the same. Could any
> one give us some information addressing this issue?

It happens with SPM96, particularly if the contrast in the template differs
from the contrast in the image that you are attempting to normalize.  There
are also a couple of other reasons, explained at:
http://www.mailbase.ac.uk/lists/spm/1998-01/0029.html


> 4. To run normalization correctly, should we choose mean image (mean*.img)
> from
> realignment as the image that the parameters "determined from"; or should we
> choose
> the coregistered mean (rmean*.img) for the task? We got different results
> using different reference images. It turns out both good and bad result in
> the normalized images for different individuals. Will appreciate some
> comments on this.

The mean image produced by the realignment is intended to be used for
estimating spatial normalization parameters from.  It is created as a `soft'
mean, whereby it is the integral of the realigned images, divided by the
number of images for which data could be sampled.  Any co-registered images
are likely to have zeros in the image where data in the original image could
not be sampled.  To SPM, the interface between the zeros and tissue look like
edges, so it will attempt to distort the images to correct them. 

Regards,
-John


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%