[Insight-users] Surface Registration & Evaluation of Registration

[Luis Ibanez]

Hi Marc,

Thanks for your detailed and clear questions.

Here are some comments:

1) Yes, you can generate Meshes, but first you will need to segment
    the brain (or the structures of the brain that you are interested
    in). For an account of the segmentation methods available in ITK
    you may want to look at the ITK Software Guide

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

    in particular to the "Segmentation" Chapter.


2) Yes, you can conver VTK polydata, or unstructured grids into
    ITK Meshes, and vice versa.

    You will find examples of this in the directories

            InsightApplications/
                    Auxiliary/vtk

    and

            InsightApplications/
                    DeformableModelSimplexMesh




3) You may want to try the example code in

             Insight/Examples/Patented/
                  IterativeClosestPoint1.cxx
                  IterativeClosestPoint2.cxx
                  IterativeClosestPoint3.cxx


For the generic Segmentation problem, you will find
many methods intended for brain segmentation.

In particular statistical classification, and LevelSets.

You will find validation studies of brain segmentation
in the directory:


          InsightApplications/IBSRValidation/
                                   IBSRClassification/
                                   AtlasSegmentation/
                                   FEMAtlasSegmentation/


4) Doing progressive registration is the way to go.
    As you pointed out, you want to start with simple
    Transforms such as Rigid, and then move into
    Affine, and finally into deformable ones like
    BSplines.


5) You may want interesting to consider the registration
    of a Mesh against an Image. This combination gives
    you a good tradeof between memory allocation and
    comptational speed.  You may want to use the Mesh
    as the Fixed object and the Image as the Moving object.

    The reason is that The Mesh would have less nodes than
    pixels in the image, and that quering the image for
    a pixel value is more efficient that searching in the
    Mesh for a closest point.



If you get something working, you are *strongly* encouraged
to post it in the Insight Journal:


  http://www.insightsoftwareconsortium.org/InsightJournal/




    Regards,


       Luis




------------------
Marc Ruiz wrote:
> Hi!
>  
> My purpose is to register two brain surfaces. Related with this I have 
> some questions:
>  
> 1) Is it possible to obtain the mesh of  the surface of a brain with 
> itk, using for example a MRI as input image?
>  
> 2) Can I take a surface, or better, the mesh of a surface from VTK and 
> convert or use it to ITK?
>  
> 3) Has ITK a kind of atlas for my goal or do you know one that I could use?
>  
> 4) Which classes could be interesting to develope this application?
>  
> Any other idea or problem to comment will be grateful.
>  
> On the other hand, I do not know how to quantify the accuracy or the 
> error when I realize a registration with different steps, it is, first a 
> rigid registration to align the brains and later a deformable 
> registration ( I use ImageRegistration8 and DeformableRegistration7).
>  
> Thanks for your help!
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users