Rigid-body registration does what it says in its name. It uses a rigid-body
model, with no shears or zooms. It is suboptimal for inter-subject
registration, but it may do the job OK (as an ad hoc procedure). There
would be no need to do any reslicing. The registration itself will change
the position and orientation information in the image headers. As long as it
gets the origin in roughly the right place and the orientation about right,
then it is probably fine.
However, I am reluctant suggest it as the definative solution as I think this
is one that would probably need a bit of empirical exploration (but not for
submission as a methodology paper, as it would not be considered very
interesting). If it works, then fine. If not, then there is always the
manual re-oriontation option - which should be fairly trivial to do. If all
your data are aquired with roughly the same orientation and positioning in
the scanner, and you have converted your files with a NIfTI-compatible piece
of software, then the same translations and rotations should be roughly
applicable to all subjects. Simple.
I am actually working on (among other things) a more robust and general
purpose initial affine registration for the segmentation in SPM, which I hope
will make life easier.
Best regards,
-John
On Tuesday 30 September 2008 10:05, Cannon, Dara wrote:
> Apologies for the re-post but we haven't heard a peep on this one and
> we'd really like any insight anyone has into the issue, thank you.
>
>
>
> When reorienting images for VBM there appear to be two approaches in use
> that may be roughly equivalent and we're attempting to better understand
> the decision making process in choosing between them. We'd love
> confirmation of our interpretation of the basis for differences between
> the approach using rigid-body co-registration versus manually
> reorienting by translations and rotations alone (using display). In
> particular in the case where one is registering to a template (T1 MNI)
> rather than intra-subject registration in order to achieve optimal AC_PC
> alignment for segmentation.
>
>
>
> Our understanding is that the manual / display approach is optimal based
> on the fact that rigid-body co-registration is designed for
> intra-subject registration and therefore would involve potential
> additional (shears/zooming?) changes to the source image after
> transformation which would not be involved when using display/reorient.
> What kind of additional changes over and above manual reorientation does
> the rigid-body co-registration impart to the subjects MR properties
> (intensity, variance?). How might these changes affect subsequent
> segmentation/VBM analyses and to what extent if known?
>
>
>
> Thanks so much for your time,
>
> Dara
>
>
>
>
>
> _____________________________________
>
> Dara M. Cannon, Ph.D.
>
>
>
> Lecturer in Neuroimaging
>
> Assistant Director, Clinical Neuroimaging Laboratory
>
> Department of Psychiatry
>
> National University of Ireland, Galway
>
>
>
> Phone: 353 91 495692
>
> Mobile: 353 85 7325345
>
> E-mail: [log in to unmask] <mailto:[log in to unmask]>
>
> Fax: 353 91 494563
>
> Web:
> http://www.nuigalway.ie/psychiatry/research/neuroimaging_lab/index.html
|