Meeting minutes
ITS meeting
DICOM
Erich: Met w Detlef. INt in working together.
DBooth: What end product?
erich: Their working on an RDF-driven imaging system.
… INt in having an official RDF spec for DICOM. They have their own cut of it, as basis. They seemed open to that idea.
eric: Other stakeholders who should be included?
erich: IDK. They'ave posted an OWL ont. SHACL defined.
… Normative files for DICOM are XML. Have to parse them to get tables out.
eric: The first FHIR RDF was XSLT that converted FHIR XML to Turtle.
erich: Not sure how much to invest, because DICOM puts out a new version every quarter.
… There's an open source python to parse the XML, but it's out of date.
DBooth: Is the XML really the authoritative source?
… Or do they gen it from something else?
erich: It's the authoritative.
eric: W First FHIR RDF, Josh Mandel did something interesting.
erich: They only maintain RDF URI for certain instnaces of data, not in general.
eric: Josh did XSLT magic, parsing structure defs, and the XSLT walked through that and instance data to emit stuff. So the XSLT was scaffolding for generated schema.
… That meant when you changed the spec you didn't update the XSLT, you just regenerated from the FHIR spec.
eric: https://
erich: Also they have value representations, defs for their datatypes. They have a few numeric types that are representated as strings. Would want ints stored as ints, dates stored as dates, etc.
eric: FHIR has some places that require microparsing, like dates. For FHIR RDF we decided we wanted fully defined types.
erich: W 20B triples you notice the performance quickly.
… I started needing a SPARQL rule set to parse and convert things.
DBooth: Does the XML have well defined patterns, or idiosyncratic?
erich: Fairly well defined.
… Int in anyone else int in this.
eric: Is the python still effective?
erich: Might be worth contacting the innocolitics group. https://
eric: In FHIR, representations are defined as representations of an abstract syntax.
erich: Everything is tables of data. CHatGPT converted some of their defs to SHACL
eric: The browser covers everything in the spec? Erich: Yes.
erich: In my implementation, I prefix the hexadecimal as IDs.
erich: Here's their repo: innolitics/
DBooth: If they're already unhappy w XML as the authoritative, might they consider changing to a more readable abstract format? If so, what?
Remaining HAPI updates
eric: Hope to work w Claude in Liden, 2 weeks from now.
ADJOURNED