"Tag Path" Convention

[Mathieu Malaterre]

Hi there,

I cannot find anywhere in the standard any notion of "Tag Path". Has
there been any discussion on a notion of so called "DICOM Tag Path" ?

I can see that DCMTK people have been using the following (man
dcmodify):

{sequence[item-no].}*element

I am personnally going a bit more along XPath and using:

{>item-no>}*group,element

For example:
1.
0008,9092>1>0020,000d

would indicate the attribute (0020,000d) Study Instance UID from the
first item of nested dataset of attribute sequence (0008,9092)
Referenced Image Evidence Sequence.

2.
5200,9229>*>0020,0037

would indicate all the attributes (0020,0037) Image Orientation
Patient which can be found in item [1-n] of nested dataset if
attribute sequence (5200,9229) Shared Functional Groups Sequence.

Has there been any proposal ? Is there any already existing
convention ?

Thanks for comments,
-Mathieu

[David Clunie]

See the "native model" XML schema in the WG 23 work (Sup 118) at:

ftp://medical.nema.org/medical/dicom/supps/sup118_pc.zip
ftp://medical.nema.org/medical/dicom/supps/sup118_pc.pdf

The Xpath looks like:

"DicomAttribute[@tag='00180080']/Value[@number=1]"

or

"DicomAttribute[@keyword='RepetitionTime']/Value[@number=1]"

and similarly for sequences (where each item with a nested data set
is treated as a value).

David

On Jun 8, 11:03 am, Mathieu Malaterre <mathieu.malate...@gmail.com>
wrote:

[Joerg Riesmeier]

> I cannot find anywhere in the standard any notion of "Tag Path". Has
> there been any discussion on a notion of so called "DICOM Tag Path" ?

> I can see that DCMTK people have been using the following (man
> dcmodify):

> I am personnally going a bit more along XPath and using:

Btw, since you are talking about dcmodify and XPath: I just wrote a little
"Howto" that shows how to use DCMTK's XML tools and an XSLT processor to
change the patient's name:

http://support.dcmtk.org/wiki/dcmtk/howto/xmltools

Maybe, some of you find it interesting :-)

Regards,
Joerg

[bbennett]

For what its worth, I have been using the following convention in my
personal set of perl based DICOM tools (posda.com):

(3006,0039)[0](3006,0040)[0](3006,0042) - identifies the "Contour
Geometric Type" of the first item in the "Contour Sequence" in the
first item of the "ROI Contour Sequence"

I think it may be the same as the DCMTK conventions, but I came by it
independently. It seems pretty natual to me.

Just my $.02,

Bill Bennett