> Dear SPM programmers, (and especially John I guess)
>
> I'm doing VBM analyzes with SPM5b ... bias corr, segmentation,
> modulation, normalization in one shoot !!! (NeuroImage 2005, Vol26) ..
> as claimed in the paper it takes a 'bit' more time ;-)
>
> I report here a 'strange' behavior ... whereas 15 subjects were
> processed without pblms (2-3h per subjects), SPM ran 5 hours on the same
> subject without any bug report in the matlab window .. and was still
> computing some stuff when I stopped it. Is it possible that the
> algorithm iterates non-stop? maybe a bug somewhere (in the spm code or
> in my computer...) or could it take this time ?!
2-3 hours per subject seems like a very long time. What platform are you
running it on? Are you using the standard default settings? In the
lower-left window, you should get a plot of the log-likelihood, which is
updated after each iteration. What was happening to this? Was it relatively
flat on the right side? How many iterations had it done? It's possible that
I may need to examine the stopping criteria or regularisation.
Best regards,
-John
|