Hello-
I have also noticed that vol_homocor.m works only with 2D planes, and
therefore it does a poor job with 3D images of brains (as unrelated
solutions are computed for neighbouring slices).
One simple solution is to use SPM2 'SEGMENT' function. It computes a 3D
intensity inhomogeneity correction. With SPM2, the corrected image is
created with the prefix 'm', so if you segment the image 'filename.img' a
corrected image named 'mfilename.img' is created. With spm99, the prefix is
'corr_', so you would get 'corr_filename.img'.
I have been impressed with SPM's homogeneity correction. However, there are
two other tools that might be worth a try (though I have not used them
myself):
1.) N3, from the MNI. Well documented, requires images to be in MINC
format.
http://www.bic.mni.mcgill.ca/software/N3/
2.) FAST is designed to do the same thing as SPM's 'SEGMENT' function
(correct inhomogeneity in order to separate different tissue types):
http://www.fmrib.ox.ac.uk/fsl/fast/index.html
-chris
p.s. One mystery remains: On Kalina?s web page for vol_homocor appears to do
a good job correcting a 3D volume. I tried several 3D volumes and never
ended up with a nice result. Looking at the code, it only seems to consider
the 2D slices, so it is a mystery why the code did a good job on the 3D
volume illustrated on the web page:
http://www-psych.stanford.edu/~kalina/SPM99/Tools/vol_homocor.html
On Tue, 12 Aug 2003 10:49:24 -0400, Lijun Ding <[log in to unmask]>
wrote:
>Hi All,
>
>We use Kalina Christoff's program vol_homocor.m to
>correct our structural images. There are some
>inhomogeneity between slices of the corrected image.
>Does someone have a solution to deal with this?
>
>Thank you very much!
>
>Lijun Ding
|