[vtk-developers] vtkDICOMImageReader broken in vtk 5.0 and up?

[Sven Prevrhal]

Hi Mathieu and others:

vtkDICOMImageReader or vtkImageViewer2 from 5.0 up seems to have a memory
problem with bigger files - I have a digitized X-ray of 10MB, and older vtk
versions are leaky, at least the python wrappers. This is a potential
showstopper for me 

I see two ways out - 
1. fix vtkDICOMImageReader and/or vtkImageViewer2 
2. find a tool that converts to 16-bit JPEG or other format (I don't know
any) and try to read that. 

Can you point me to where the problem might lie? I will try implementing a
bug fix then.

The DICOM file that causes a crash in my configuration is on
ftp://ftp.mrsc.ucsf.edu/pub/sven/SC.1.3.6.1.4.1.9590.100.1.1.581587240110609
28131648073180919160033  

The JPEG version (only 8bit!) is ftp://ftp.mrsc.ucsf.edu/pub/sven/x.jpg   


Sven Prevrhal


-----Original Message-----
From: Sven Prevrhal [mailto:sven.prevrhal at radiology.ucsf.edu] 
Sent: Wednesday, March 15, 2006 1:24 PM
To: 'Mathieu Malaterre'
Subject: RE: [vtkusers] vtkDICOMImageReader broken in vtk 5.0 and up?

Great! From where do I pull the cvs sources? 
:pserver:anonymous at public.kitware.com:/cvsroot/VTK or, as your text
suggests, somehow form ParaView?

- Sven


-----Original Message-----
From: Mathieu Malaterre [mailto:mathieu.malaterre at kitware.com]
Sent: Wednesday, March 15, 2006 12:15 PM
To: Sven Prevrhal
Subject: Re: [vtkusers] vtkDICOMImageReader broken in vtk 5.0 and up?

Sven,

	I believe the seg fault you saw is fixed in current VTK CVS.

	Please give it a try if you can. Thanks for the data to reproduce
the problem.


$ cvs ci -m"BUG: DICOM spec do not enforce value to be present for Slice 
Number" 
/cvsroot/ParaView/ParaView/VTK/Utilities/DICOMParser/DICOMAppHelper.cxx,v 
  <--  DICOMAppHelper.cxx
new revision: 1.23; previous revision: 1.22
/cvsroot/ParaView/ParaView/VTK/Utilities/DICOMParser/DICOMParser.cxx,v 
<--  DICOMParser.cxx
new revision: 1.15; previous revision: 1.14

HTH
Mathieu
Ps: You may want to contact your DICOM vendor as the DICOM has a very 
bizarre 0008,0008 Image Type value: "Vidar" ??? Whereas it should be 
something like:

DERIVED/SECONDARY

Sven Prevrhal wrote:
> Hi Mathieu,
> 
> You can grab the images from ftp://ftp.mrsc.ucsf.edu/pub/sven. The setup
> file is unrelated!
> 
> Thanks for looking at this!
> 
> sven
> 
> Sven Prevrhal, Ph.D.
> Department of Radiology
> University of California, San Francisco
> 
> 185 Berry St, Suite 350, San Francisco, CA 94107
> ph 415-353-9404, fax 415-353-3438
> 
> Faculty Assistant: Jessica.Mapes at radiology.ucsf.edu
> 
> 
> -----Original Message-----
> From: Mathieu Malaterre [mailto:mathieu.malaterre at kitware.com] 
> Sent: Friday, March 10, 2006 12:49 PM
> To: Sven Prevrhal
> Cc: vtkusers at vtk.org
> Subject: Re: [vtkusers] vtkDICOMImageReader broken in vtk 5.0 and up?
> 
> Sven Prevrhal wrote:
> 
>>I have DICOM images that vtk 4.4 will read just fine, but vtk 5.0 and up 
>>and nightly builds will crash.  (I built vtk with python 2.4). Anybody 
>>experiencing similar behavior?
> 
> 
> Can you send me a sample data to reproduce the problem ?
> 
> Thanks
> Mathieu
> 
> 
> 
> 


-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: testdicomread.py
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20060427/0e0ea60d/attachment.ksh>