[Insight-users] Helo with bspline/mutual information/gradient descent

[J.X.J.]

Hi Luis,

thanks for your reply. I have added an observe to the code, the metric value
output seems to oscillate rather than a general increase for decrease in
value. How do you plot the metric values? Is there a way to output the
metric value of each iteration into a text or similar file and I could use
MATLAB to plot it, or is there a way to do that directly from ITK?

Also I've noticed that using Viola and Wells MI is REALLY slow when coupled
with BSpline Deformable Transform. I've tried using Mattes MI which is much
much faster in comparison in the order of 5-10 fold, why is that the case?

And finally (I'm sorry for all the questions, programming in C is not my
forte), I tried using Levenberg Marquardt Optimizer but I get this error but
I compile the code: error C2664:
'itk::ImageRegistrationMethod<TFixedImage,TMovingImage>::SetOptimizer' :
cannot convert parameter 1 from 'itk::LevenbergMarquardtOptimizer::Pointer'
to 'itk::ImageRegistrationMethod<TFixedImage,TMovingImage>::OptimizerType *'

Any help is appreciated
J.X.J. 


Luis Ibanez wrote:
> 
> Hi J.X.J.
> 
> The first thing that you want to do is to instantiate a Command/Observer
> and to print out the values of the Metric at every iteration of the
> optimizer.
> 
> Plot this values and/or send the output to the mailing list.
> 
> From the progression of the metric values it will be possible to determine
> if parameters of the optimizer need to be adjusted, or whether you may
> have to review other components of your registration structure.
> 
> You will find examples on how to use Command Observers in almost
> all the files in the directory:
> 
>                     Insight/Examples/Registration
> 
> 
>     Regards,
> 
> 
>           Luis
> 
> 
> --------------------------------------------------------------------
> On Sat, Aug 1, 2009 at 6:49 AM, J.X.J. <wat.a.phony at gmail.com> wrote:
> 
>>
>> Hi everyone,
>>
>> I'm current using itk::BSplineDeformableTransform,
>> itk::MutualInformationImageToImageMetric and
>> itk::GradientDescentOptimizer
>> to register 2D MRI images. My problem is that when running 2 phantom
>> images
>> the output image is almost exactly the same as the original moving image
>> (nothing like the fixed image).
>>
>> I have normalized and smoothed the 2 input images using
>> itk::NormalizeImageFilter and itk::DiscreteGaussianImageFilter with
>> SetVariance(2.0). For mutual information the standard deviation is 0.4.
>> Optimizer learning rate is 10.0, 200 iteration and with maximum on. This
>> set
>> up is kind of a mixture of a number of deformableregistration examples on
>> default (as in I copied and pasted all parameter settings etc directly).
>>
>> Does anyone know what the problem could be? The code runs all 200
>> iterations
>> is thats an indication, also the metric value of each iteration is around
>> 0.2 - 0.4 if that's any help.
>>
>> J.X.J.
>> --
>> View this message in context:
>> http://www.nabble.com/Helo-with-bspline-mutual-information-gradient-descent-tp24768056p24768056.html
>> Sent from the ITK - Users mailing list archive at Nabble.com.
>>
>> _____________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-users
>>
> 
> _____________________________________
> Powered by www.kitware.com
> 
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
> 
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
> 
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users
> 
> 

-- 
View this message in context: http://www.nabble.com/Helo-with-bspline-mutual-information-gradient-descent-tp24768056p24883109.html
Sent from the ITK - Users mailing list archive at Nabble.com.