Fwd: Re: [Insight-users] 3D deformable registration

Luis Ibanez luis.ibanez at kitware.com
Fri May 4 10:07:17 EDT 2007


Hi Vijay,

I would suggest you to explore the use of the BSplineDeformable
registration method.

This method fits better with the standard registration framework of ITK,
and *it will* manage without problems images of different number of
pixels as well as different pixel spacing.

You will find instructions on how to use this method in the ITK
Software Guide:


           http://www.itk.org/ItkSoftwareGuide.pdf


towards the end of the "Image Registration" chapter.



     Regards,



         Luis



--------------------------
vijay kumar wrote:
> Hi Luis,
> I know you are very busy, but could you respond to this query of mine? I 
> am waiting for your response to proceed further. Which deformable 
> registration method would you suggest for me to register two 3D images 
> of dissimilar sizes. Image1: [X1,Y1,Z1]; Image2: [X2,Y2,Z2], where X1, 
> Y1 and Z1 are not equal to X2, Y2 and Z2 respectively. If ITK doesn't 
> have a suitable method, I shall then start exploring other 
> options/programs. Otherwise, ITK is what I would like to use.
>  
> Thank you very much.
>  
> Regards,
> Vijay
> 
> */vijay kumar <desivij at yahoo.com>/* wrote:
> 
>     Date: Thu, 3 May 2007 13:08:27 -0700 (PDT)
>     From: vijay kumar <desivij at yahoo.com>
>     Subject: Re: [Insight-users] 3D deformable registration
>     To: Luis Ibanez <luis.ibanez at kitware.com>
> 
>     Hi Luis,
>     Thank you for your detailed response and suggestions. I looked into
>     DeformableRegistration11.cxx and also into the parameter file -
>     FiniteElementRegistration3.txt. From what I see, the parameter file
>     expects both input 3D images to be of the same size. In my case, the
>     images are of size (160x256x256) and (256x256x200), with voxel sizes
>     of (1.0x1.0x1.0 mm) and (1.21x1.21x1.23 mm). As you mentioned, the
>     voxel sizes are obtained from the Analyze header. How about the
>     image sizes? In the section called 'Information about the image
>     inputs' there is option for only one set of Nx, Ny and Nz. Does this
>     mean that I have to scale the images to the same size before I use
>     deformable registration? If so, that may be messy. I was under the
>     assumption that deformable registration would take care of reslicing.
>      
>     Thank you.
>      
>     Regards,
>     Vijay
> 
>     */Luis Ibanez <luis.ibanez at kitware.com>/* wrote:
> 
> 
>         Hi Vijay,
> 
>         If you are interested in a 3D version of the FEM deformable
>         registration example, you may want to look at the example:
> 
>         Insight/Examples/Registration/
>         DeformableRegistration11.cxx
> 
>         and the associated input file:
> 
> 
>         Insight/Examples/Data/
>         FiniteElementRegistration3.txt
> 
> 
>         The example: DeformableRegistration11.cxx is the 3D version
>         DeformableRegistration1.cxx
> 
> 
> 
>         The pixel dimensions (pixel spacing) are taken from the
>         meta information of the image. There is no need for you
>         to put it in the parameters file.
> 
> 
> 
>         The values of Elasticity and RhoC must be found through
>         experimentation.
> 
>         As a rule of Thumb, high elasticity implies that a very
>         strong force applied to the material will produce very
>         small deformations. Therefore, you could start running
>         experiments with large values of elasticity. If you find
>         that the deformation field has vectors with very small
>         magnitudes, then you know that the Elasticity can be
>         reduced. If you find that the deformation vectors are too
>         large (e.g. the anatomical tissue gets displaced too far)
>         then you will know that the Elasticity is too low.
> 
>         Note that you may have to play with the orders of magnitude
>         of the Elasticity.
> 
> 
>         The RhoC factor is used for computing the Mass matrix,
>         and its final effect is to multiply the forces applied
>         to the model. Therefore, large values of RhoC will result
>         on larger deformations, while small values of Rhoc will
>         result in smaller deformations.
> 
> 
> 
> 
>         Regards,
> 
> 
> 
>         Luis
> 
> 
> 
>         ------------------
>         vijay kumar wrote:
>          > Hi,
>          > I would like to register two 3D images that are of different
>         sizes ([nx1
>          > ny1 nz1] & [nx1 ny2 nz2] voxels) and different voxel
>         dimensions. I
>          > modified DeformableRegistration1.cxx to accept 3D images, but
>         I have
>          > questions about the parameters file
>          > (FiniteElementRegistrationParameters1.txt):
>          >
>          > 1. How do I specify the sizes of both images? The section -
>         Information
>          > about image inputs has the option to specify only one set of
>         sizes. How
>          > do I specify the second set?
>          >
>          > 2. How do I include voxel dimensions? Or, will the program
>         get this
>          > info from the image header?
>          >
>          > 3. How do we determine appropriate E, RhoC and Alpha values?
>         Are there
>          > some guidelines for their choice?
>          >
>          > Thanks,
>          > Vijay
>          >
>          >
>          >
>          >
>         ------------------------------------------------------------------------
>          > Ahhh...imagining that irresistible "new car" smell?
>          > Check out new cars at Yahoo! Autos.
>          >
>          >
>          >
>          >
>          >
>         ------------------------------------------------------------------------
>          >
>          > _______________________________________________
>          > Insight-users mailing list
>          > Insight-users at itk.org
>          > http://www.itk.org/mailman/listinfo/insight-users
> 
> 
>     ------------------------------------------------------------------------
>     Ahhh...imagining that irresistible "new car" smell?
>     Check out new cars at Yahoo! Autos.
>     <http://us.rd.yahoo.com/evt=48245/*http://autos.yahoo.com/new_cars.html;_ylc=X3oDMTE1YW1jcXJ2BF9TAzk3MTA3MDc2BHNlYwNtYWlsdGFncwRzbGsDbmV3LWNhcnM->
> 
> 
> 
> ------------------------------------------------------------------------
> Ahhh...imagining that irresistible "new car" smell?
> Check out new cars at Yahoo! Autos. 
> <http://us.rd.yahoo.com/evt=48245/*http://autos.yahoo.com/new_cars.html;_ylc=X3oDMTE1YW1jcXJ2BF9TAzk3MTA3MDc2BHNlYwNtYWlsdGFncwRzbGsDbmV3LWNhcnM-> 
> 


More information about the Insight-users mailing list