Dear Hans,
SPM12 has a helper utility spm_get_bbox for this, which is available in
the batch GUI (SPM->Utils->Get Bounding Box). Since you have the
normalised images with optimal bounding boxes available, there will be
no trial and error. (It would be harder to determine normalised
bounding box corners from non-normalised data.)
Best,
Volkmar
Am Freitag, den 14.09.2018, 09:39 +0200 schrieb Hans van der Horn:
> Dear Volkmar,
>
> Thanks for your reponse, this is helpful! If I understand correctly,
> it would be best to try and find a fixed bounding box size that will
> be suitable to include every brain of every subject in the group. Is
> there a tool to gauge this size, or is it a matter of trial and
> error? And is the resulting bounding box retrievable from the
> normalized images using spm? (does it show up in check reg for
> example?)
>
> thanks again!
>
> Best,
> Hans
>
> Op vr 14 sep. 2018 om 08:58 schreef Volkmar Glauche <volkmar.glauche@
> uniklinik-freiburg.de>:
> > Dear Hans,
> > if you want to use SPM voxelwise statistics, all images are
> > required to have the same voxel-to-world mapping and dimensions.
> > This is not the case when bounding boxes are different across
> > subjects. Instead of using per-subject bounding boxes you would
> > have been better off to determine the maximum bounding box
> > necessary to include all subjects.
> > There is one additional caveat: by default, SPM does not have a
> > notion of "missing values" and will only fit your model to voxels
> > that contain data in all of your subjects. Voxels that have a value
> > of zero in at least one of your subjects will be masked. One can
> > turn this of by changing the settings for implicit masking, but
> > then exact zeros will be treated as if they were meaningful voxel
> > values and not just an indicator of "missing values".
> > Hope this helps,
> > Volkmar
|