[Insight-users] Help needed with 2D/3D registration

[Luis Ibanez]

Hi Paul,

When you got a "run time" error, did you verified if
the application was actually throwing an exception ?

It is common to find that the Metrics throw exceptions
when the images have overlapping regions that are too
small.

You may want to make sure that you put the StartRegistration()
call inside a try/catch block.

Regarding the Metrics and the effect of using "degenerate"
images (3D images with a single slice), you have to avoid
the metrics that compute gradients on their images.  Many
metrics in ITK does this because in that way you can use
the Jacobian of the transform and the image gradient in
order to compute the Metric derivatives in a faster way.

Another way to get around this problem is to use optimizers
that do not require the Metric to provide derivatives. For
example the OnePlusOne evolutionary optimizer, or by modifying
any of the Metrics for computing derivatives just by using
finite differences.


Please let us know if you have further questions,



    Thanks,


      Luis



------------------------

Paul de Bruin wrote:

> Hello List,
> 
> I have been playing with 2D/3D registration (matching digitally
> reconstructed radiographs from CT to X-ray images). However, I am
> having trouble getting registration to work with most metrics.
> 
> I used the ItensityBansed2D3DRegistration application from the
> InsightApplications as an example to build my own code (which I left
> mostly intact except for ripping out the ImagePyramid). The only
> metric that I have successfully been able to use is the
> MeanReciprocalSquareDifferenceImageToImageMetric. Registration using
> all other metrics results in either a runtime error or a single step
> of the registration process after which optimization ends.
> 
> After looking at the code of some of the metrics, I suspect that the
> problems may be a result of using "flat" 3D images (i.e. 3D images
> with size 1 in one of the dimensions).
> 
> If anyone is using ITK to register DRRs and X-rays (successfully) then
> I would very much like to come into contact with you.
> 
> Thanks,
> Paul
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 
>