[Vv] clitkDicomRTStruct2Image and VTK5/VTK6⁩

[Thomas BAUDIER]

Hi Neree,

Thank you for your feedback.
I opened your console images and there is a little difference but it's 
normal. I'm not totally agree with you when you said it's a gold 
standard ; it's just the algorithms used to convert RTStruct to binary 
image are different. In vv:
  - first, we use gdcm (https://sourceforge.net/projects/gdcm/) to read 
into the dicom tags the vertex coordinates
  - they are stored into a vtkPolyData 
(http://www.vtk.org/doc/nightly/html/classvtkPolyData.html)
  - We use vtkLinearExtrusionFilter 
(http://www.vtk.org/doc/nightly/html/classvtkLinearExtrusionFilter.html#details)
  - And vtkPolyDataToImageStencil 
(http://www.vtk.org/doc/nightly/html/classvtkPolyDataToImageStencil.html#details)
In the 3rd and 4th points, some assumptions are done. And if we compare 
with console, these assumptions are most likely different.

I compared the difference between the console images and the 
clitkDicomRTStruct2Image images. There is a perfect match for the tumor 
and the aorta (except for one pixel) and the some corners of the BDF are 
different due to the previous vtk functions.

I did my own test on my side, and you're right, the correct version is 
vtk5. I'll correct that bug and notify you when it will be done.

Thank you
Thomas


On 2017-05-16 20:51, Neree PAYAN wrote:
> Dear Thomas,
> 
> To complete my remarks, the DicomRT-Struct has been converted into a
> binary image (in nifti format) directly on clinical consols. After
> verification, we can use these results as a gold standard. Then, we
> should have the same binary image as output using the
> clitkDicomRTStruct2Image command, but it’s not the case.
> I join you the gold standard images for each ROI.
> 
> Thank you for your help,
> Best regards,
> 
> Nérée
>> Le 15 mai 2017 à 16:24, Neree PAYAN <nerea.payan at u-bourgogne.fr> a
>> écrit :
>> 
>> Hi Thomas,
>> 
>> Thank you for your prompt reply.
>> To answer your question, vtk5 seems to be the correct version.
>> 
>> For the second point, I have taken into account the interpolation
>> for the conversion between Mesh and image. But in my case, the
>> differences of shapes are very importante. I have three ROIs
>> (tumour, aorta and healthy tissue) and I have a perfect match for
>> the tumour ROI and big differences in the shapes for the two
>> others…
>> 
>> Thank you for your help,
>> Best regards,
>> 
>> Nérée
>> 
>> _______________________________________________
>> vv mailing list
>> vv at creatis.insa-lyon.fr
>> http://www.creatis.insa-lyon.fr/mailman/listinfo/vv
> _______________________________________________
> vv mailing list
> vv at creatis.insa-lyon.fr
> http://www.creatis.insa-lyon.fr/mailman/listinfo/vv