This is how unknown values are handled whenever possible. The NaNs are in
regions outside the FOV of the original images, and only present in image
formats that can represent them.
A modification to lines 225 and 239 of spm_reslice.m, changing NaN to 0 should
change the behaviour to putting zeros in these regions rather than NaNs,
although I haven't tested this.
Best regards,
-John
> Using SPM peprocessing tools we noticed that after the realignment of
> functional to first volume (and after slice timing procedure) so many
> voxels take NaN values.
> We have read on list that this behavior could be caused by machine
> dependent bugs. We work on a Dual Xeon Custer.
>
> Is it possible to avoid this problem, without remove them by hand?
|