[Insight-users] Marker based registration

[Luis Ibanez]

Hi Jenny,

You could use several of the registration algorithms presented
in the ITK examples under:

            Insight/Examples/Registration/

Please read the registration chapter of the software guide

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

Chapter 8, pdf-page 241, paper page 267.
You will need to understand the concepts defined in
this chapter in order to use the components of the ITK
registration framework.

Please also read the section regarding the ResampleImageFilter.
Section 6.7.1, pdf-page 199.

You will find additional material about the ITK registration
framework in the sectino of Tutorials

      http://www . itk . org/HTML/Tutorials . htm


Look in particular at the PDF document

       "Registration Methods Overview"


-----

You can start using directly the example

          Insight/Examples/Registration/
                       ImageRegistration3.cxx

which combines a MeanSquare metric and a Translation
transform. This example should be able to register your
two consecutive slices.  If the missalignemt is really
small, you could switch to use the NormalizadeCorrelation
metric. The image registration example #3 is also available
in python as ImageRegistration3.py.

---

If the markers in the image are reliable, you could
restrict the registration to use the information from
only a small region around the marker. You could use
the "FixedImageRegion" setting for this purpose.

If you have many markers in the image, you could also
use the recently added masking capability of the image
metrics in order to use only the regions surrounding
the markers.  (How many markers do you have in a single
slice ?)


----

Pleaes let us know if you encounter any problems,


     Regards,


       Luis


-----------------------
Jenny Simpson wrote:
> I need to register a set of DICOM slices from a CT scan of bones.  The
> problem with the data is that every third slice is jittered, i.e. the data
> exhibits a pattern of offsets.  A straw full of liquid was scanned along
> with the bones, so we have a marker that we can use for registration. I'm
> new to registration though, and I was hoping someone could tell me what
> algorithm/code inside or outside of Itk would be best to use to solve this
> problem.
> 
> Thanks,
> 
> Jenny Simpson
> Scientific Computing and Imaging Institute
> University of Utah
> 
>