Dear spm-ers,
I am determining the transformation parameters to normalise two -static-
Ecat files. The original files are dynamic studies. so they must be
preprocessed. The original files have voxel dimensions 3.375 x 3.375 x
2.425 mm^3. When I produce preprocessed static files with voxel sizes
2.425 x 2.425 x 2.425 the spm_sn3d routine doesn't find the correct
result: it should find a rotation, but instead it produces a scaling
parameter. When I produce preprocessed static files with voxel sizes
2.25 x 2.25 x 2.425 mm^3 (x and y zoom 1.5 and z no zoom), the routine
does find the correct result.
Although it is possible to normalise the images, I'm still not happy
with the results. It appears that the voxel size of the template image
influences the success of the normalisation. There are 2 reasons I can
think of why it goes wrong:
1. The preprocessing phase is not carried out correctly. This is rather
unlikely, because the only things that need to be done are: summing of
frames, centering the head, and zooming in on the head. The resulting
files were checked visually, using line profiles, and investigating the
result of spm_vol(). They appear to be correct.
2. The optimisation procedure in spm is very sensitive to changes in
grey value distributions. As the positions of the head differs between
the two images, they have different grey value distributions. Measures
like chi-square are sensitive to those changes.
A suggestion would be to normalise the grey values in the preprocessing
phase. One way to do this would be to multiply all voxel values. But in
Analyze files and Ecat files it is also possible to make use of scaling
and calibration factors. Canging those would be a lot faster.
My questions are:
Has anyone got a clue about the nature of this problem? Is it one of the
two things described above, or are there other considerations?
Do routines like spm_sample_vol make use of scaling factors, or is it
necessary to multiply every single voxel to normalise two grey value
distributions?
I hope someone has an answer to this (technical) problem
Thanks in advance,
Alle Meije Wink
--
Alle Meije Wink
kamer Y0203
PET-centrum Academisch Ziekenhuis Groningen
Hanzeplein 1
9713 GZ Groningen
tel: 050-3613439
email: [log in to unmask]
--
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|