This is a very minor inconsistency. I was looking for the VR of :
0008 0040 Data Set Type
In dcmtk, I found VR=US, but I found VR=SS in dicom3tools(*).
Looking at the definition in
http://www.dclunie.com/dicom-status/status.html#AcrNema1985
I found out that the valid values can be:
IMAGE= 0
GRAPHICS= 2
TEXT= 3
OTHER= 100
NULL= 101
IDENTIFIER= 102
PRIVATE IMAGE= 8000
PRIVATE GRAPHICS= 8002
PRIVATE TEXT= 8003
Therefore because of the high values (eg: 0x8003) I believe this is
a bug to use SS for this VR.
Comments ?
Mathieu
(*) dicom3tools_1.00.snapshot.20051218
Since there is no well-defined mapping from ACR-NEMA VRs to DICOM VRs,
one can probably not call this a bug.
In fact, (0008,0040) is defined as VR "BI" in ACR-NEMA, which is described
as "16 bit binary (2's complement, except for pixels which are separately labeled)"
so there might be a good reason to map this to "SS" in DICOM.
However, as you point out, ACR-NEMA explicitly mentioned values > 0x8000
for this attribute (and not negative values), so US can't be that wrong as a choice.
In the end, this looks like an inconsistency in ACR-NEMA (1985) that still
manifests in the 2006 DICOM world. Hey, we're really good in preserving bad design :-)
Regards,
Marco Eichelberg
OFFIS