[Insight-users] vtk polydata import / export: Image to Model registration

[Luis Ibanez]

Hi Imho,

Thanks for clarifying your final application.

If your goal is to perform image guided surgery,
I'm affraid that Demons registration will not
be able to deliver the speed you need.

You probably need the registration to be performed
in a matter of seconds.

In particular if the intra-surgery model is represented
as a geometrical model, the step of generating a volume
from the geometrical model will be a bottleneck in the
whole process.

I would suggest you to take a look at the PointSetToImage
registration methods, and/or to the SpatialObjectToImage
methods.

The first method registers an itkPointSet with an itkImage.
The ideal point set is generated as a band around the edges
of the object. There is a new filter that you can use for
generating the point set from a binary mask (segmented image).
http://www . itk . org/Insight/Doxygen/html/classitk_1_1BinaryMaskToNarrowBandPointSetFilter . html
Then you can use a PointSetToImage registration method.
http://www . itk . org/Insight/Doxygen/html/classitk_1_1PointSetToImageRegistrationMethod . html


The second methods registers an itkSpatialObject with an
itkImage. An example is available in the SoftwareGuide.pdf.
Section 7.14, Model Based Registration, pdf-page 286.


The itkSpatialObject is in fact designed for representing
models of anatomical organs. You can group spatial objects
in order to model complex shapes.



Regards,



   Luis



----------------
imho wrote:
> Hi Luis,
> 
> I don't understand, I've found the imageViewer where you said.
> I've a too old version I guess.
> 
> what I'm trying to do is a registration between a liver reconstructed 
> from slices and a "live liver", who will be digitalised (I don't know 
> how already) in the chirurgical room.
> But I've to simulate this live digitalisation, so I load a polyData 
> saved before, and deform it a bit.
> I don't know the structure that will have the digitalised one, but I can 
> have the reconstructed one as imageData (originally) and polyData.
> I already do a point-based registration and a ICP registration with 
> these two polyData (with vtk, using vtkLandmarkTransform, vtkTransform, 
> vtkTransformPolyDataFilter and vtkIterativeClosestPointTransform)
> 
> So I've first preferably continued using polyData, but I guess I can 
> change all of this to imageData?
> I'll look in this way.
> Thanks
> 
> Imho.
> 
>