Is the SPM5 coregister&reslice program a tiny bit non-deterministic,
eg thru a randomized initialization of the parameter search or
something like that ?
I ran the identical files thru twice on a Sun/Solaris machine,
matlab v. 7 R14, spm rev 573 and got these 2 different translation
matrices (from the ps output). Although the numbers are very close,
the differences were enough to create small differences in the resliced
output image file as well. I used all the default coregister&reslice
options, just specifed input files. My inputs were Analyze format,
not nifti, but spm5 did not complain about that.
X1 = 0.993*X -0.083*Y +0.361*Z -7.544
Y1 = 0.089*X +0.994*Y -0.274*Z -8.826
Z1 = -0.020*X +0.018*Y +0.994*Z +0.358
X1 = 0.992*X -0.082*Y +0.373*Z -7.898
Y1 = 0.088*X +0.994*Y -0.274*Z -8.722
Z1 = -0.021*X +0.018*Y +0.994*Z +0.478
I had similar variability working on a Linux/opteron machine.
thanks
kate
|