[Insight-users] Problem with Demons deformable registration

Xuan Zhao xuanaux at gmail.com
Fri May 28 17:40:42 EDT 2010 

Hi, guys

I finally found out the problem is at WarpImageFilter. I got the
correct result after I add a line: "warper->SetOutputDirection(
fixedImage->GetDirection() );". The problem is that the orientation of
the warp filter by default is ((1,0,0),(0,1,0),(0,0,1)), while the
orientation of my image is ((-1,0,0),(0,-1,0),(0,0,1)). Then the image
warped is beyond the extent of the moving image due to incorrect
orientation. I think it would be good to add SetOutputDirection to the
deformable registration examples of ITK in the future version.

Tom: The two images I am using are CT images of different human. I
think it will be helpful to scale two images to the same intensity
value using HistogramMatchingImageFilter before Demons algorithm. Am I
right?

Anyway, thank you very much for your advices.

Best regards,
Xuan

On Fri, May 28, 2010 at 10:33 AM, Tom Vercauteren
<tom.vercauteren at m4x.org> wrote:
> Hi Xuan
>
> You are using a HistogramMatchingImageFilter in your code. It would be
> worth checking that it is not this step that produces the constant
> value images. You should write down the image produced by this filter.
>
> As a side note, do you actually need the histogram matching? If your
> images are discrete labels, you should not change their intensities
> before doing the registration.
>
> Hope this helps,
> Tom
>
> On Fri, May 28, 2010 at 19:20, Xuan Zhao <xuanaux at gmail.com> wrote:
>> Hi, Luis
>>
>> Thank you so much for your detailed analysis. Here is my implementation.
>>
>>> A)  What pixel type are you using for the image types
>>>       in the template instantiation of the Demons filter ?
>>
>> The type of input image is unsigned short. And I casted the image to
>> float as the instantiation type of Demons filter. Cause I notice that
>> examples are mostly in float type.
>>
>>
>>> B)   Have you verified the content of the resampled moving
>>>       image after the rigid registration ?
>>>       It could well be that such image is outside of the
>>>       extent of the fixed image and it was resampled to
>>>       an empty value
>>
>> I did verify the resampled moving image after the rigid registration.
>> I wrote the resampled image to DICOM files and it seems correct.
>>
>>> C)  What value are you using for the Default pixel value
>>>       in the Resample Image filter ?
>>>       (after the rigid registration )
>>
>> The default pixel value in the resampled image filter is 0.
>>
>>> D)   For how many iterations is the Demons filter running ?
>>
>> I used only one iteration for the Demons filter just for simplicity.
>> Because images are first rigidly registered so there is large overlap
>> between them. So I think even one iteration should work although the
>> deformation maybe very small. I tried using more iterations (up to 10
>> because it took really long time for each iteration) and the resulting
>> image is still constant.
>>
>>> E)    Please try writing the image to a file format different
>>>        from DICOM (for example, .mha, .vtk, .nrrd, .nii) and
>>>        use a standard image application (such as Slicer)
>>>        to look at the content of the image.   That will help us
>>>        rule out DICOM (or not) as the source of the problem.
>>
>> I definitely could try this. But as I mentioned, I wrote resampled
>> image to DICOM files, they are correct. So I think the DICOM part
>> should not be the problem.
>>
>> Actually, I am thinking would the problem caused by the different
>> orientations of two images, in which one is in supine position and
>> another is in prone. But it is still confusing since the rigid
>> registration works well. Anyway, I will check details regarding this.
>>
>> Regards,
>> Xuan
>>
>> On Fri, May 28, 2010 at 7:45 AM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>>>
>>> Hi Xuanaux,
>>>
>>>
>>> A)  What pixel type are you using for the image types
>>>       in the template instantiation of the Demons filter ?
>>>
>>>
>>> B)   Have you verified the content of the resampled moving
>>>       image after the rigid registration ?
>>>       It could well be that such image is outside of the
>>>       extent of the fixed image and it was resampled to
>>>       an empty value
>>>
>>>
>>> C)  What value are you using for the Default pixel value
>>>       in the Resample Image filter ?
>>>       (after the rigid registration )
>>>
>>> D)   For how many iterations is the Demons filter running ?
>>>
>>>
>>> E)    Please try writing the image to a file format different
>>>        from DICOM (for example, .mha, .vtk, .nrrd, .nii) and
>>>        use a standard image application (such as Slicer)
>>>        to look at the content of the image.   That will help us
>>>        rule out DICOM (or not) as the source of the problem.
>>>
>>>
>>>
>>>    Please let us know,
>>>
>>>
>>>          Thanks
>>>
>>>
>>>             Luis
>>>
>>>
>>> --------------------------------------------------------------------------------
>>> On Wed, May 26, 2010 at 6:13 PM, xuanaux <xuanaux at gmail.com> wrote:
>>>>
>>>> Hi, guys
>>>>
>>>> I am trying to use Demons registration filter to register two 3-D volume
>>>> images of heart. Heart volume images were segmented from CT scans of whole
>>>> body. Rigid registration was first applied to make a global registration
>>>> of
>>>> two images. Transformed moving image and original fixed image were used as
>>>> moving and fixed images of Demons algorithm. However, when I write back
>>>> the
>>>> resulting image to the DICOM files, the resulting image has constant value
>>>> of 1024, which is apparently not correct. I did check the two inputs
>>>> images
>>>> to the Demons filter. They have the same origin, orientation and spacing.
>>>> I
>>>> have no idea what is going wrong.
>>>>
>>>> Could any one give me any suggestion of the possible problem? Thanks in
>>>> advance!
>>>>
>>>> Attached please find the code of Demons registration part.
>>>>
>>>> http://itk-insight-users.2283740.n2.nabble.com/file/n5106037/code-Demons.txt
>>>> code-Demons.txt
>>>> --
>>>> View this message in context:
>>>> http://itk-insight-users.2283740.n2.nabble.com/Problem-with-Demons-deformable-registration-tp5106037p5106037.html
>>>> Sent from the ITK Insight 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
>>>>
>>>> Kitware offers ITK Training Courses, for more information visit:
>>>> http://www.kitware.com/products/protraining.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
>>
>> Kitware offers ITK Training Courses, for more information visit:
>> http://www.kitware.com/products/protraining.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
>>
>

More information about the Insight-users mailing list