[Insight-users] Re: deformable registration : Kernel Transform references

[Luis Ibanez]

Hi Silvester,

 From the documentation of the KernelTransform class

http://www . itk . org/Insight/Doxygen/html/classitk_1_1KernelTransform . html

The ITK implementation of KernelTransforms is based on
the following papers:

     IEEE TMI paper by
     Davis, Khotanzad, Flamig, and Harms,
     Vol. 16, No. 3 June 1997.

        and

     R. Sprengel, K. Rohr, H. Stiehl.
     "Thin-Plate Spline Approximation for Image Registration".
     In 18th International Conference of the IEEE Engineering
     in Medicine and Biology Society. 1996.


Regards,


    Luis


------------------------
Silvester Czanner wrote:

> Hi Luis,
> 
> could you, please write me, where can I find more information about 
> KernelSpline transforms (ThinPlate splines). I checked the ITK 
> UserGuide. Unfortunately,  there is not too much written about it.
> 
> Thank you,
> 
> Silvester
> 
> Luis Ibanez wrote:
> 
>> Hi Silvester,
>>
>>
>> It depends on what deformable registration method you use from ITK.
>>
>>
>> A) Demons
>> B) FEM
>> C) BSpline transform
>> D) KernelSpline transform
>>
>>
>> In the case of  DEMONS and FEM, you get ast output a deformation field.
>> Using this deformation field you could identify the displacements of
>> particular points in the moving image.
>>
>> These two methods, however will not use any landmark information
>> during the registration process. In practice you can simply connect
>> an observer to the registration and at each iteration print out the
>> displacement vectors corresponding to your set of landmarks.
>>
>>
>> In the case of BSpline transforms used with the basic registration
>> framework, you can use the BSpline transform as computed at each
>> iteration of the optimization process, and again map a set of
>> landmarks in  order to evaluate their displacements.  The landmarks
>> will not be used for stering the registration process, they will
>> only be probes / observers that will allow you to see how the
>> registration is progressing.
>>
>>
>> In the case of KernelBase transforms, like the ThinPlate spline
>> and its family, you actually use landmarks as part of the registration
>> process. In this context you could associate your landmarks to
>> the source and target landmarks used by the transform.  The optimizer
>> will update the coordinates of these landmarks in order to optimize
>> the image metric.
>>
>>
>>
>> Please let us know if you have further questions,
>>
>>
>> Thanks
>>
>>
>>   Luis
>>
>>
>> --------------------------
>> Silvester Czanner wrote:
>>
>>> Hi Luis,
>>>
>>> I have two grayscale volumes, which I would like to register using 
>>> deformable registration methods from ITK.
>>> Is it possible in ITK to localize some correspondent landmarks before 
>>> deformable registration and during the registration process to track 
>>> the trajectories of those landmarks?
>>>
>>> Silvester
>>>
>>
>>
>>
>> _______________________________________________
>> Insight-users mailing list
>> Insight-users at itk . org
>> http://www . itk . org/mailman/listinfo/insight-users
>>
>