[Dcmlib] Unreadable dicom file from Peter Cech

Jean-Pierre Roux jpr at creatis.insa-lyon.fr
Tue Aug 23 17:50:58 CEST 2005


Mathieu Malaterre wrote:

> Jean-Pierre Roux wrote:
>
>> Hi !
>>
>> The trouble starts because the 'private' tag 2001-105f (and some more 
>> ) are 'Sequence' , but the Transfert Syntax is 'Implicit VR', and 
>> till now, gdcm has no way to guess it.
>> (The 'Shadow Dictionary' supplied by Philips is not usable :-(
>
> Ok this is PHILIPS-Intera.dic.
> JP can you have a look at it

I did before, that's why I said it's not usable...

> , and just tell me what I should generate for entry like:
>
> 2001 00xx LO 1 Private Creator Group 2001

> and
> 2001 xx01 FL 1 Chemical Shift

According to what I see in the the header, you should generate 2001 10.. 
instead of 2002 xx.., and 2005 10.. instead of  2005 xx..
and 2001 0010 instead of 2001 00xx.

But this will *not* solve the pb, since
2005|1083 SQ
2005|1402 SQ
are not in the PHILIPS-Intera.dic
(and a lot of other tags are missing, but we don't care of them)
Have a look at attached document t1.txt, it contains all the Private 
elements for this image

>
> I will add the cmake rule to generate appropriate rule (+option to 
> generate shadow dicts).
>
>> I just commited a patch to bypass the problem.
>> (vtkgdcmViewer displays the image, that's no so bad ;-)
>> Don't try to *write* the image, because the header is not straight, 
>> and the result wouldn't be kosher...
>
>
> what is the basic problem ?

The brutal patch I made (to be able to read all the public values and 
the pixels) messes up the private elements that are in Sequences (since 
it doesn't know they are).

A local, temporary, troubleless solution is to add the 2 lines, in the 
Dicts/dicomV3.dic.
If you generate the Shadow Dict, please add the lines 2005 in the 
skeleton you get from the PMS pdf document.

> I'm working for a clean solution, but that's not so easy.
>
>>
>> --> Peter,
>> if you add the following lines :
>>
>> 2001 105f SQ 1 Sequence
>> 2005 1083 SQ 1 Sequence
>> 2005 1402 SQ 1 Sequence
>
>
> ouch ! Only the first one can be found in PHILIPS-Intera.dic

As usual ...
(We noticed a lot of stuff was missing when we dealt with Marlene's images)

JP

>
>> in Dicts/dicomV3.dic,
>> your images will be gdcm-readable/gdcm-writable.
>
>
> Thanks JP, you rocks ! I could not fix the bug myself :(
>
> Mathieu
>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: t1.txt
URL: <http://www.creatis.insa-lyon.fr/pipermail/dcmlib/attachments/20050823/1ebc4ef4/attachment.txt>


More information about the Dcmlib mailing list