[Insight-users] Deformation Inquiry

[Luis Ibanez]

Hi Michael,

There are several techniques in ITK for performing
deformable registration.


Given that you are dealing with same-modality images
you probably can use any of these different techniques.



They are:


A) - DemonsDeformable registration
      with two implementations:

Demons
http://www.itk.org/Insight/Doxygen/html/classitk_1_1DemonsRegistrationFilter.html
SymmeticForces Demons
http://www.itk.org/Insight/Doxygen/html/classitk_1_1SymmetricForcesDemonsRegistrationFilter.html


B) - BSplineDeformable registration

      This is based on using the BSPlineDeformableTransform
      in the frame of the generic ITK Registration Framework.


C) - FEM - Based deformable registration


D) - Level Sets based registration
http://www.itk.org/Insight/Doxygen/html/classitk_1_1LevelSetMotionRegistrationFilter.html


E) - You also have some other options, such as
      Segmenting the bone using a deformable model
      in the first (free bone) image, and then use
      this initial model in order to segment the second
      image.

      The displacement of mesh point between the Meshes
      resulting from both segmentations will give you
      an idea of the bone deformation. Unfortunately this
      will only give you information about the bone surface.

      For this approach you may want to look at the application

           InsightApplications/
                    DeformableModelSimplexMesh



For the methods {A,B,C,D} you will find code examples in
the directory

           Insight/Examples/Registration





I would assume that the ammount of deformation that you
are inducing in the bone is small and that the structures
are being displaced just by a couple of pixels between the
free bone and the bone under stress...      is that right ?

If this is the case, you probably want to try Demons first.
This technique will give you a deformation field in all
the extent of your image, represented in the form of a
vector field.



You probably will be tempted to use FEM, which is indeed
a good option. Just keep in mind that the physical model
in FEM is totally disconnected from the reality of the bone.
The physical model of an elastic body is set up only as
a methaphor that allows to generate a deformation field
under a regularization constraint.


BSplines is also an interesting option but it may require
a long computation time, specially if you decide to use
a fine grid for the BSpline.


So... in order to make your choice... I would suggest you
to first evaluate how big you anticipate the deformations
to be with respect to the pixel size of your image. In
other words, how many pixels away will the bone structures
move when they are subject to stress.


How big is yor image ( in number of pixels on each dimension ?).



Please let us know if you have further question,


    Thanks



       Luis



-----------------------
Michael Hardisty wrote:

> Hello,
>          I am attempting to build an application that does a deformable 
> registration.  The registration is to be performed on two scans of the 
> same bone,  in the first scan the bone is just scanned, in the second a 
> force is applied to the bone and compressed a small amount.  I am 
> interested in obtaining the deformation field (strain field) from this 
> analysis.  I have noticed that there are several methods of approaching 
> the problem within ITK.  One, which I have implemented is to do a 
> deformable registration (I am using an affine transformation), another 
> that I have recently been reviewing in the Software guide and online in 
> the deformable registration presentation is the use of the FEM framework 
> within ITK.  I am wondering about what the advantages or disadvantages 
> of these two approaches.  Which approach would you suggest that I apply 
> to my problem?  I am assuming that the ImageMetricLoad is some way of 
> deforming an FEM as a result of some sort of image metric that measures 
> similarity of two images.  I would like the chance to integrate 
> different forces other than just Image similarity.  I am also attracted 
> by the ability to control the mesh.
> 
> Is the description of the FEM framework that I have described above 
> correct?  Would the FEM framework be appropriate for my application and 
> what should I do to begin?
> 
> Is there more documentation of how this framework works outside of the 
> software guide and the Doxygen generated documentation?
> 
> Is there another technique that I am ignoring that would be more 
> appropriate?
> 
> ---
> M.A.Sc Student
> University of Toronto
> Orthopaedic Biomechanics Laboratory
> Sunnybrook & Women's College Health Sciences Centre
> 
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 
>