[Insight-users] Re: Deformable registration tests

[Luis Ibanez]

Hi Tina,

Using transforms with 6 degrees of freedom during the
computation of the forces in the FEM method may not
be a good idea since the physical model is not using
the torques at the nodes anyways.

The argument for using only translations is that this
translations are converted into forces applied to the
nodes of the FEM mesh.  The physical model is then
applied and displacements at the FEM nodes are computed.

When you apply a deformable registration, it is assumed
that you already solved for an Affine transform. That
is, you are registering only the residual deformations
that were not compensated by a successful registration
using an affine transform. In this circumstances, we
are confident in that global translations, global rotations
and global shearings have already been resolved.

The application of simple translation in the nodes of
the mesh is justified on the assumption of using local
forces in order to deform the mesh.

Note that this doesn't mean that the FEM mesh cannot
induce some local rotations. In fact, if two adjacent
nodes in the mesh have displacements in opposite directions,
all the image pixels lying in between these two nodes
will have some rotation induced.



Regards,


   Luis



--------------------
Tina Wang wrote:
> Luis,
>  
> Could you tell me how to set up a transformation in FEM based non-linear 
> registration?
> I run the "DeformableRegistration1" and found that 
> "translationtransform" was used. I want to change it to a transform with 
> at least 6 DOF. Please help.
>  
> Thanks,
>  
> Tina 
> 
> ------------------------------------------------------------------------
> Do you Yahoo!?
> Protect your identity with Yahoo! Mail AddressGuard 
> <http://antispam . yahoo . com/whatsnewfree>