[Insight-users] deformable registration in binary image

[Luis Ibanez]

Hi Marc,


I'm not sure I understand your description.

I would assume that if you initialize the transform
properly then the images would be close to each other...



In any case....


You seem to be going in the right direction, since you are first
solving the Rigid registration, then the Affine registration and
finally moving into the Deformable registration.

Please not that if you are using the BSplineDeformableTransform
http://www.itk.org/Insight/Doxygen/html/d7/d3b/classitk_1_1BSplineDeformableTransform.html

this Transform let you define a "Bulk" Transform. If you
provide this bulk transform, the BsplineDeformable transform
will use it as a pre-composed transform.

         (see the SetBulkTransform() method...)


What it is usually done, is to take the Affine transform
resulting from your initial registration, and plug it in
as the bulk transform for the deformable registration.

In this way, you can initialize the deformable transform
as an identity transform (e.g. no deformations initially),
and have the bulk transform represent the compensation for
the general misalignment of the two images.



   Please give it a try,
   and let us know if you find any problems.



     Thanks



       Luis



----------------
Marc Ruiz wrote:
> Hello!
>  
> The 2 images are first rigid and then affine registered. If they are 
> initialized they are far one from the other.
>  
> So... I have to launch the registration without doing these two steps?? 
> But then, I read deformable registration is not capable of doing big 
> translations or rotations.
>  
> Any suggest??
> 
>  
> On 1/16/06, *Luis Ibanez* <luis.ibanez at kitware.com 
> <mailto:luis.ibanez at kitware.com>> wrote:
> 
> 
>     Hi Roger,
> 
> 
>     Your modifications to the code seem to be reasonable.
> 
>     The main difference between the examples
> 
> 
>          DeformableRegistration7
> 
>     and
> 
>          DeformableRegistration8
> 
> 
>     is that the first one is intended for images of the
>     same modality, while the second is intended for images
>     of different modality.
> 
>     Given that you replaced the Metric with the MatchCardinality
>     metric, then there is no major difference left between the
>     two examples.
> 
> 
>     I'm not sure about the problem you find with the
>     infinity norm of the projected gradient,.. but if
>     you are using the LBFGSB optimizer, it is known
>     that it has some issues if you start it on a local
>     minimum (maximum), in other words, if you start the
>     optimization problem in a place where the derivative
>     of the metric is null, then the optimizer can not
>     complete the first iteration.
> 
> 
>     Please give us more details on how you are initializing
>     the Transform, and how similar the Fixed and Moving
>     images are.
> 
> 
>        Thanks
> 
> 
> 
>           Luis
> 
> 
> 
> 
> 
> 
>     --------------------
>     Roger Alvaredo wrote:
>      > Hi!
>      >
>      > I tested  deformableRegistration7.cxx or
>     deformableRegistration8.cxx in
>      > MRI's and they work correctly. Now I would like to re adapt def8 for
>      > working with binary images.
>      >
>      > I change the metric to MatcthCardinality and the interpolator to
>      > NearestNeighbor.
>      >
>      > In the case of the metric I add the
>     method:  metric->MeasureMatchesOff()
>      > because the optimizer minimizes the metric, and I remove the lines of
>      > MattesMutualInformation as the number of histograms and so on. It
>     is ok??
>      >
>      > So, I would like to know...:
>      >
>      > 1. Advantatges and disvantatges of each one (def7 and def8).
>      > 2. I have a problem because the change of metric...:
>      > The initial values are: GetValue()= 0.0709776
>      > and GetInfinityNormOfProjectedGradient()=0
>      > why is 0 the norm? it stops before doing an iteration for this...
>      >
>      > Any advice?
>      >
>      > Thanks!
>      >
>      >
>      >
>      >
>      >
>     ------------------------------------------------------------------------
>      >
>      > _______________________________________________
>      > Insight-users mailing list
>      > Insight-users at itk.org <mailto:Insight-users at itk.org>
>      > http://www.itk.org/mailman/listinfo/insight-users
>     <http://www.itk.org/mailman/listinfo/insight-users>
> 
>     _______________________________________________
>     Insight-users mailing list
>     Insight-users at itk.org <mailto:Insight-users at itk.org>
>     http://www.itk.org/mailman/listinfo/insight-users
> 
> 
> 
> 
> -- 
> 
> MaRC//