[Dcmlib] Add Series in gdcmSampleData/ForSeriesTesting

Jean-Pierre Roux jpr at creatis.insa-lyon.fr
Mon Oct 3 18:53:10 CEST 2005


Mathieu Malaterre wrote:

> Jean-Pierre Roux wrote:
> <> 
>
[...]
<>       A given 'Serie', may be composed of :        - 'Scout Views', 
in 3 different incidences.
        - MIP images (as many incidences as images)
        - Slices to build a volume + a scout view, showing where the 
slices were done
       We can easyly realise that it's up to application to decide 
whether a given serie has to be splitted or not !
       --> gdcm problem is to give user tools to help him to decide :-(

>
> How about we bring the subject on dicom newsgroup ? Maybe somebody has 
> already written a decent heuristic that deals with a good chunk of the 
> case...


I think the solutions are application dependant :
gdcm::SerieHelper splits the images depending on their Series Instance 
UID, and we call that a 'CoherentFileList'.
Aware user is allowed to ask to ignore some images, using :
   void AddRestriction(uint16_t group, uint16_t elem, std::string const 
&value, int op);

Then, he can ask to sort a given 'CoherentFileList' on the slice 
location (default), or on what he feels like, using 
SetUserLessThanFunction()
and/or SetSortOrderToReverse().

Something is missing : allow user to tell the images he wants to remove 
from the 'CoherentFileList' before sorting it.
It could be, for instance,  'remove images that doesn't follow main 
Orientation'.
(I'll think for a name later ...)
Only user may decide whether he wants to do that :

Such a removal would be OK when there is one scout view inside a serie : 
the scout view is usefull to show the Medical Doctor *where* the slices 
were done; It's useless for the Computer guy that wants to build a volume !

Such a removal wouldn't have any meaning for a serie holding MIP images 
(each one has a different orientation, but *functionally* they belong to 
the same Series (you can use them to make a moovie!)

JPRx


>
> My 2 cents,
> Mathieu
>



More information about the Dcmlib mailing list