[Insight-users] Demons Registration for Multi modality Setup.

[Dawood Masslawi]

Hi Sibin,
You can use the ITK to read series of DICOM images, compute the 
histogram or gradient image, perform the registration and evaluate
the registration quality.
The "IO" folder in the ITK examples directory has some source code
samples for reading DICOM series, also chapter 7 of the software
guide contains more information about the samples and the methods used.
Computing the histogram can be performed using the histogram generator
in ITK. There are code samples included in the "statistics" folder
of the examples directory and chapter 10 of the software guide contains
the documentation regarding the samples. Once you compute the histogram 
you can access the histogram information using one of the following, 
instance identifiers, n-dimensional indices or iterators (I would 
recommend the iterators). Gradient filter is also supported by the ITK. 
Examples are included in the "filtering" folder and chapter 6 contains 
the documentation for this filter.
Evaluating the registration accuracy is a bit more tricky. If you want
to use mean squares you have two options, first is to evaluate the mean
square metric after the registration by setting the transformation of
a registration to identity which as I recall already has been discussed 
in the in the mailing list, the other option is to use the "GetRMSChange()" 
method supported by some registration functions which gives you the Root 
Mean Square (RMS) in the deformation filed. The first approach will be 
something like the "MeanSquareImageMetric" example in the "registration" 
folder, as for the second approach you can extract the deformation field 
after the registration and pass it to the registration function as a 
template parameter (along with the fixed and moving image) then get 
the RMS value.
Hope this helps,
Dawood Masslawi

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

Hi,
I am new to itk, I am currently doing project on Demons registration for
multi modality using preprocessing approaches such as Histogram matching and
gradient images.

Now let me tell you in detail.......

first,
As we know in CT, values are normalized(calibrated to housenfield unit) but
not in MR.
So to normalize MR image we go for preprocessing.....
Here i need help

two approaches:
Histogram Matching
Use of Gradient images

Please note my input would be reading a *set* of *DICOM* images...CT and MR
from directory not single images.


after normalize i need to do Demons registration and then need to measure
accuracy of demons...................How can i Do that using ITK

i think mean square will show me accuracy of Demons but can i do that using
ITK


Please help....
i know there is code in ITK DeformableRegistation2.cxx in which demons are
used but how to implement my problem....in that code....Please help...!!!


      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20110118/e43ba86c/attachment.htm>