[Insight-developers] bug report on DeformableModelSimplexMesh application

[Defeng WANG]

Hi, Luis,

The dataset I used is "sphere.zip", which can be downloaded from http://appsrv.cse.cuhk.edu.hk/~dfwang/tmp/sphere.zip (the size is about 30K). By uncompressing this file, you will see Analyze file of "sphere.hdr" and "sphere.img".

The parameters I used are listed as follows,

Internal Forces (Alpha): 0.8
External Forces (Beta): 0.8
Damping Forces (Gamma): 0.35
Range of Search: 2
Rigidity (Regularizing): 0.5
Number of Iterations: 1

The parameters involved in the gradient calculation of 
DeformableModelSimplexMesh are keep unchanged. That is, I did not make any 
change to the source code of DeformableModelSimplexMesh.

The steps to reproduce my results are listed as follows,

1. "File/Load File", select "sphere.hdr";
2. left click the center point of the upper left image, then click 
"File/Create Mesh", you will see a simplex spherical mesh in the window;
3. "File/Preprocess Image", wait about two minutes to finish the preprocess;
4. Click "Settings/Deform Mesh" about 25 times. Each time after your click, 
you will see that the simplex mesh deform. In this process, you can see the 
problem I described in my last email.

By the way, if I set the parameter of "Number of Iterations: 1" to 25, and 
click "Settings/Deform Mesh" only once. The result I got seems quite 
similiar to the initial mesh, and is not like the one got by seting "Number 
of Iterations: 1" to 1 and click "Settings/Deform Mesh" 25 times. I think 
this is another bug in this program.

If any part in the above desription is not clear enough, please let me know 
and I will try to provide more details.

Best wishes,
Defeng


----- Original Message ----- 
From: "Luis Ibanez" <luis.ibanez at kitware.com>
To: "Defeng WANG" <dfwang at cse.cuhk.edu.hk>
Cc: "Leila Baghdadi" <baghdadi at phenogenomics.ca>; 
<insight-developers at itk.org>; <insight-users at itk.org>
Sent: Friday, February 01, 2008 10:40 AM
Subject: Re: [Insight-developers] bug report on DeformableModelSimplexMesh 
application


>
> Hi Defeng,
>
> Can you please post in a public web site the image that
> you are using as input, and the full set of parameters
> that you are using for running this application ?
>
> Otherwise it is very hard for us to figure out how far
> from the edge of the image you are seeing the final
> contour stopping.
>
> Please give use a very clear sequence of step that we
> could follow to reproduce your experience here.
>
>
>
>   Thanks
>
>
>      Luis
>
>
> -------------------
> Defeng WANG wrote:
>> Hi, Leila,
>>
>> Thanks for your reply. It seems that the parameters in the UI of
>> DeformableModelSimplexMesh are not related to the gradient calculation. I
>> read the source code involved in DeformableModelSimplexMesh. As for the
>> gradient calculation, three filters are used successively as follows,
>>
>>  m_GradientMagnitude->SetInput(m_CastImage->GetOutput() );
>>  m_GradientMagnitude->SetSigma(0.5);
>>
>>  m_SigmoidImage->SetInput( m_GradientMagnitude->GetOutput());
>>  m_SigmoidImage->SetOutputMinimum(0);
>>  m_SigmoidImage->SetOutputMaximum(1);
>>  m_SigmoidImage->SetAlpha(230);
>>  m_SigmoidImage->SetBeta(1300);
>>
>>  m_GradientFilter->SetInput( m_SigmoidImage->GetOutput());
>>  m_GradientFilter->SetSigma( 0.5);
>>
>>
>> I have taken a look at the final gradent image. It seems that it is good
>> enough to describe the
>> edge or boundary of a sphere. So what do you think of this problem 
>> existing
>> in this simple example?
>>
>>
>> Look forard to your response.
>>
>> Best wishes,
>> Defeng
>>
>> ----- Original Message ----- From: "Leila Baghdadi"
>> <baghdadi at phenogenomics.ca>
>> To: "Defeng WANG" <dfwang at cse.cuhk.edu.hk>
>> Cc: <insight-developers at itk.org>; <insight-users at itk.org>
>> Sent: Thursday, January 31, 2008 10:54 PM
>> Subject: Re: [Insight-developers] bug report on
>> DeformableModelSimplexMesh application
>>
>>
>>> Hi Defeng,
>>>
>>> Yes I have used that code many a times. I am not sure if I understand
>>> what you mean by "it does not converge".
>>>
>>> My understanding of deformable models is
>>>
>>> 1. you must make sure your gradient image is created properly. Use
>>> paraview to look at the vector image
>>>
>>> 2. you must play with the parameters to get the model to converge,
>>>
>>> This is a specific type of deformable models which uses simplex mesh.
>>> I suggest you read about this algorithm first. This is developed based
>>> on the paper by Herve Delingette of INRIA france.
>>>
>>>
>>> Leila
>>>
>>> On Thu, 2008-31-01 at 22:29 +0800, Defeng WANG wrote:
>>>
>>>> Hello, ITK users and developers,
>>>>
>>>> I found that DeformableModelSimplexMesh, provided in
>>>> InsightApplications-3.4.0, does not converge. It seems that the
>>>> deformable mesh will not stop near the boundary of one object to be
>>>> segmented. I tried a very simple volume data set containing a sphere
>>>> only. The initial mesh I used is a spherical simplex mesh inside it.
>>>> However, by increasing the number of iterations, the mesh will
>>>> continue to expand until it reaches outside of the image and the
>>>> program reports error.
>>>>
>>>> Is there anybody tried this program successfully before? Or anybody
>>>> knows how to adjust the parameters properly to get right segmentation?
>>>>
>>>> Best wishes,
>>>> Defeng
>>>> _______________________________________________
>>>> Insight-developers mailing list
>>>> Insight-developers at itk.org
>>>> http://www.itk.org/mailman/listinfo/insight-developers
>>
>>
>> _______________________________________________
>> Insight-developers mailing list
>> Insight-developers at itk.org
>> http://www.itk.org/mailman/listinfo/insight-developers
>> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.itk.org/mailman/private/insight-developers/attachments/20080201/0450c4de/attachment.htm