[Ctk-developers] CTK Dashboard: A DICOM data collection for testing ?

[Mathieu Malaterre]

Dear all,

On Fri, Apr 23, 2010 at 2:54 PM, OFFIS DICOM Team <dicom at offis.de> wrote:
> Dear all,
>
> Am 22.04.2010 17:21, schrieb Marco Nolden:
>>
>> Am 09.04.2010 16:45, schrieb Luis Ibanez:
>>>
>>> [...] 2) We could host a set of realistic DICOM series in the MIDAS
>>> server, and setup CMake to download them into testing machines as part
>>>  of the superbuild. It will be important to include in that collection,
>>>  a set of "bad" datasets that exercise error conditions in the code.
>>>
>>> That is, we want to include DICOM images that are not compliant, and we
>>> want to verify that the code rejects them accordingly.
>>>
>>>
>> That's a good idea. I've seen there is already a CTK repository on MIDAS.
>> How can I download this from CMake?
>>
>>> Does DMTK has a data-testing collection that we could use ? and if so,
>>>  it is redistributable ?
>
> Unfortunately we don't have a systematic collection of DICOM data which
> would fit the requirements of CTK which lies especially on working on a set
> of inter-related images for fancy image processing :-)
>
> Of course we have tons of "normal" images but most of them should not be
> publicly redistributable. Also we have an internal, artificially produced
> test collection of images which pushes to the limits of the DICOM image
> model in order to test our 2D image rendering pipeline (with stuff inside
> like 65535 bits assigned per color channel). We call that the horror cabinet
> -- I doubt it's worth sharing for CTK but I will discuss wheather this is
> possible before coming to the hackfest.
>
> Few more complex examples (say: inter-related images) can be found at the
> NEMA and are referenced from David Clunies website http://www.dclunie.com.
> Scroll down to "images". Besides this some other ressources are listed on
> our wiki: http://support.dcmtk.org/wiki/dicom/images
>
> For pure data parsing tests interesting mostly for DICOM toolkit
> providers:-) there is a collection of images (correct and incorrect DICOM)
> from Mathieu Malaterre, the main author of the GDCM DICOM toolkit. The links
> should be also in the Wiki.
>
> The best "nice" DICOM collections I know are indeed the published by the
> Osirix people.

Most of the data in gdcmData serve a very special purpose: aggressive
testing of GDCM. This is similar to Michael's remark, I doubt this is
of any interest to CTK's folks (maybe the j2k compressed ones ?)

I would recommend dataset from nema's official website. You'll find
there D. Clunie compression dataset, or instances from the new
enhanced image storage class. For large series I use dataset from
Osirix public dataset.

Bill Lorensen's dataset can be of interest for validating the
orientation widget.

HTH
-- 
Mathieu