MITRA OBJECT UTF8 ATTRIBUTES 1.0

[Mathieu Malaterre]

Hi folks,

Some DCS from AGFA provide private dictionaries for attributes under private creator 'MITRA OBJECT UTF8 ATTRIBUTES 1.0'. It turns out that those attributes are always encoded as UTF-8.

* https://www.agfa.com/he/global/en/internet/he/library/libraryopen?ID=54472549

[...]

Patient Name (UTF8 Encoded) (0033,xx02) PN 
Study Description (UTF8 Encoded) (0033,xx04) CS 
Referring Physician Name (UTF8 Encoded) (0033,xx06) PN 
Requesting Physician Name (UTF8 Encoded) (0033,xx08) PN 
Performing Physician Name (UTF8 Encoded) (0033,xx0A) PN 
Reason For Study (UTF8 Encoded) (0033,xx0C) LO 
Study Comments (UTF8 Encoded) (0033,xx0E) LO 

[...]

If we consider a typical P10/latin1 instance (ISO_IR 100), the following UTF-8 chars (for example, but there are more) will cause interoperability issues:

$ echo -n 'ÁÉÅÖÓ' | hexdump -C
00000000  c3 81 c3 89 c3 85 c3 96  c3 93                    |..........|
0000000a

VR LO does not allow use of control characters (range 0x80-0x9F). So a naive DICOM implementation converting from ISO_IR 100 to ISO_IR 192 (eg. WADO-RS app) would produce:

$ echo -n 'ÁÉÅÖÓ' | iconv -f latin1 | uconv -x 'any-name'
\N{LATIN CAPITAL LETTER A WITH TILDE}\N{<control-0081>}\N{LATIN CAPITAL LETTER A WITH TILDE}\N{<control-0089>}\N{LATIN CAPITAL LETTER A WITH TILDE}\N{<control-0085>}\N{LATIN CAPITAL LETTER A WITH TILDE}\N{<control-0096>}\N{LATIN CAPITAL LETTER A WITH TILDE}\N{<control-0093>}

As such I'd like to recommend (open source) implementers to skip addition of all 7 attributes defined above (keep the default VR=UN rules), or else prefer a VR of OB.

Thanks for your consideration

[Mathieu Malaterre]

Removed from GDCM as of today:

* https://github.com/malaterre/GDCM/commit/ad9bf3a95648c65001288b62f3d924d03190e640