[Insight-users] about multi-modality 3D registration

[Luis Ibanez]

Hi Yixun,

You are right,

If you have images from two different subjects, then you will need
scaling in addition to rotation and translation. The scaling will
probably be anisotropic scaling.


Your choice of using first Affine, and then BSplines is a good one.


You will find useful that the BSpline transform actually accepts
a pre-transform, so it will be very easy for you to use the Affine
transform resulting from the first registration as the initial
pre-transform to be used by the second stage of registration where
you use the BSpline transform.


You will find examples on how to use both transforms in the
directory:


          Insight/Examples/Registration




Regards,


    Luis



-------------------
Yixun Liu wrote:

> Hi,
> 
> Thank you! 
> If I have a 3D Brain MRI and a Virtual Human 3D Brain dataset and the
> 
MRI comes from a patient, who is different from the virtual human. How
to register them? I think I should first rigidly register and then
non-rigidly rigister.In my opinion, the VersorRigid3D may be failed
because it cannot scale and the two brains are different in size. I want
to use a transform, which can translate, rotate, scale at the same time.
Should I use the Affine firstly and then BSpline?
> 
> Thank you!
> 
> 
> Best regards,
> 
> Yixun Liu
> 
> 
> 
> 
> 
> ----- Original Message ----- 
> From: "Luis Ibanez" <luis.ibanez at kitware.com>
> To: "Yixun Liu" <yxliu at fudan.edu.cn>
> Cc: <insight-users at itk.org>
> Sent: Tuesday, December 21, 2004 5:43 AM
> Subject: Re: [Insight-users] about multi-modality 3D registration
> 
> 
> 
>>Hi Yixun
>>
>>1) for MRI to CT, you may want to start with the VersorRigid3DTransform.
>>http://www.itk.org/Insight/Doxygen/html/classitk_1_1VersorRigid3DTransform.html
>>
>>   if after registering with this transform, you still find that
>>   there are corrections that need to be made, then you can refine
>>   with the AffineTransform.  At that point you want to inintialize
>>   the AffineTransform with the Matrix produced from your VersorRigid3D
>>   registration
>>
>>   Note that you only need Affine refinement if you find one of the
>>   two remaining missregistration:
>>
>>         A   Scaling
>>         B   Shearing
>>
>>
>>   If what you see is that there is still some rotation or translation
>>   to correct, then what you need to do is to further refine the
>>   parameters of the Rigid registration.
>>
>>
>>
>>2) For Deformable registration you may want to try BSplines first.
>>   you will have the advantage that you can introduce the Transform
>>   resulting from your registration in (1) as the pre-transform to
>>   be applied before your BSpline transform. That will give you
>>   some continuity in your work flow.
>>
>>   Note that in both cases you will have a lot of parameters
>>   to fine tune, and in both cases you will have to deal with
>>   long computational times.
>>
>>
>>
>>  Regards,
>>
>>
>>     Luis
>>
>>
>>--------------
>>Yixun Liu wrote:
>>
>>
>>>Hi,
>>> 
>>>I have a 3D MRI and a 3D CT dataset. I want to register them. Their 
>>>differences are  the rotation, translation and scaling. The metric is 
>>>definitely  the MI, but which transform will I take? The Affine or the 
>>>Versor?  If one of the dataset deforms, I also need to non-rigidly 
>>>register them. Bspline or FEM?
>>> 
>>> 
>>>Thank you in advance!
>>> 
>>>Regards,
>>> 
>>>Yixun Liu
>>>
>>>
>>>------------------------------------------------------------------------
>>>
>>>_______________________________________________
>>>Insight-users mailing list
>>>Insight-users at itk.org
>>>http://www.itk.org/mailman/listinfo/insight-users
>>
>>
>>
>>
>>
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 
>