[Insight-users] Atlas registration

[Luis Ibanez]

Hi Dingguo,


1) Yes, all image-based registration algorithms
    in ITK could be used for Atlas-Patient registration.
    Note that there are also PointSet to Image and
    SpatialObject to Image registration methods in ITK
    that will only be useful if you segment one of the
    two images.  However, with the rest of the registration
    framework you have a large enough coverage of the
    most common techniques used in medical image registration.

    Rather than a matter of preference, it is a matter of
    tradeoffs between the purpose of your registration,
    the precision you require, and how much memory and
    computational time are you willing to use for achieving
    a registration that satisfy your requirements.



2)  For Mutual Information we normalize images with the
     itk::NormalizeImageFilter,

http://www.itk.org/Insight/Doxygen/html/classitk_1_1NormalizeImageFilter.html

     for Demons and in general any metric that require the
     two images to have a similar distribution of graylevels
     you can use the HistogramMatching image filter.

http://www.itk.org/Insight/Doxygen/html/classitk_1_1HistogramMatchingImageFilter.html



3)  We have discussed color in image registration recently
     on the list. You may want to search the Archives for
     May and April 2004.

     To put it short: in most cases color do not add much
     information to the image metric, and you are usually
     ok by using one of the color channels, or by converting
     the color image to grayscale.

     Of course if you feel that color is really significant
     for your images, you can always implement an image metric
     that takes advantage of color information, and integrate
     your new metric with all the normal components of the ITK
     registration framework {optimizers, transforms,
     interpolators...}



Regards,


    Luis




---------------------
Dingguo Chen wrote:

> Hi Luis and everybody
> 
> I am currently working on atlas-patient registration and 3D 
> atlas modeling. We have Cerefy Electronic Brain Atlas library
> (TT88,TT93,SW) and Geometrical Model as our standard atlas. 
> And we have different cases from different modality (MRI,CT 
> etc) as another system input. I have several questions:
> 
> 1 In principle, all registration algorithms in ITK could be 
> used in atlas-patient registration. Are there any algorithms 
> that you guys prefer? In personally, I prefer Demon 
> Deformable registration.
> 
> 2.We need perform normalization before do the registration, 
> what kind of normalization you guys use? 
> 
> 3.In TT88, atlas is color images. How can I add color 
> information to my registration result?
> 
> I am new to this research field, if anybody could share his 
> precious experiences, I will very appreciate that!
> 
> Regards
> 
> Dingguo
> _______________________________________________
> Insight-users mailing list
> Insight-users@itk.org
> http://www.itk.org/mailman/listinfo/insight-users
>