Semantic Web Health Care and Life Sciences Interest Group Teleconference

03 Jan 2017

See also: IRC log


David_Booth, EricP, Michael_van_der_Zel, Harold_Solbrig
David Booth


Use Cases

David: Need 1 or more use cases to drive the rewrite of the Linked Data Module pages.

eric: At SWAT4LS I combined info model with clinical model and searched for terms more broadly than how they were coded, using subsumption.
... It didn't require much of FHIR ont. More about combining FHIR RDF with rich terminology model.
... Use cases for a FHIR ont beyond that will be mostly around workflow.

dbooth: Written up?

eric: yes.

harold: We had to make one significant change, which we need to make in the FHIR RDF spec.
... Want to change fhir:concept ID to a type.

dbooth: Other use cases?

michael: I've been working on technical stuff of converting FHIR to RDF and back.
... Also working on ISO 11179 stuff to do mapping between models and maybe even a query using SNOMED CT.
... Want to be able to figure out what is already represented in FHIR.

eric: I haven't figured out how to exploit the ont in the info model.

harold: I was originally going to use FHIR RDF to find iso-semantic models, or models whose info content was a subset of another model.
... But decided not to do that because it was too geeky for SWAT4LS folks.
... it would give you the ability to deal with models of different granularity, such as pre vs post coordinated.

eric: Or when FHIR has more specific resource for something, but a general purpose Observation could have been used.

harold: Linda and Grahame are working on the ability to transform local DBs to FHIR and back.
... Like the Yosemite model. But you need to align the terminology for it to work.
... probably need to get more closely connected to Linda Byrd and Grahame's work, re mapping FHIR into SNOMED using a combination of the mapping tool and templates.
... Instead of dealing with RDF they're doing it with the idiom that Grahame's mapping tool presents. What is the relationship between Grahame's mapping tool and RDF?
... Claude Nanjo and Cognitive group have been implementing Grahame's mapping spec. Might want to touch base with him at some time and see how it relates.

eric: If we consider a hierarchy of Observation, AdverseReaction, Intolerance, you are supposed to use the most specific one, but being able to recognize it using Harold's work would be good.

harold: At the moment Grahame and Linda are using the SNOMED template language. Either FHIR will have to come closer to SNOMED, or the SNOMED language would have to broaden to cover other terminologies.
... The templating language has a set of slots and fillers that align with different description models in SNOMED. THey're focusing on Observable right now.
... it has a primary object of observation, finding site, thing being measured, etc. -- about 10-12 facets.
... Individual SNOMED concepts are instances of that generic model. It's job is to fill in th eslots, so that you can put in the diagnosis, laterality, body site, etc.
... So that you can take a SNOMED concept or a logic description and get something that you can put into a spreadsheet.
... Template language on the SNOMED site includes a sort of valueset def -- a list of values that can go into the slots, so there is a bit of validation creeping in.
... Linda was intially trying to support all the nuances of FHIR, but decided we needed a two step process.
... First step is from FHIR-specific to template grammar,.
... Template grammar itself is very SNOMED specific. Nobody looking at using Disease Ontology, for example.
... Need to look at relationship between this mapping language and ShExMap

<ericP> http://rawgit.com/shexSpec/shex.js/backtick/doc/shex-simple.html

eric: Once we deal with SNOMED stuff you might be interested in this, that I did.
... Click on protein record, then delete green box contents and do control-enter to see schema.
... If we had an Obs in SNOMED, turned it into an RDF graph that has relationship that come from the SNOMED OWL, then we could use those in this back-tick language and validate that graph.

harold: Other use cases we've touched on: not all the world is FHIR.
... We looked at connecting clinical data with environmental data and other stuff in the LOD cloud, but we were disappointed with lack of LOD data.
... Might want to look at that again.

michael: Also need to look at logical models. FHIR now also has logical models -- differnt way to present the data.
... Can express other data as FHIR logical data and then use it.

dbooth: Make up your own structure definitions?

michael: yes.

dbooth: Using FHIR as a schema language.

<mvdzel> https://fhirblog.com/2016/10/17/logical-models-in-fhir/

eric: The logical models have several expressions (JSON, XML, RDF), and RDF is closest to capturing the logical model directly.

dbooth: Okay to use Eric's example to rewrite the Linked Data module pages?

harold: Another good example would be connecting with the LOD cloud, like in the Yosemite vision.

<mvdzel> http://build.fhir.org/structuredefinition.html#logical

dbooth: Good! I need a working example for it.

eric: FDA data from 3 ROund Stones is now available.
... Could get toxin data from that.
... Need the location of the patient or encounter.
... Find the time and releases of toxins, and locations of clinics (from dbpedia?), and the clinics appeared in the FHIR encounters, then you could look at the incidence of a respiratory disorder 3 weeks after chemical release versus other times.
... These studies happen all the time anyway. This would make the task easier in RDF.

<scribe> ACTION: Eric to ask Daniella Meeker to see if she has ideas for chem release use case [recorded in http://www.w3.org/2017/01/03-hcls-minutes.html#action01]

<trackbot> Created ACTION-73 - Ask daniella meeker to see if she has ideas for chem release use case [on Eric Prud'hommeaux - due 2017-01-10].

San Antonio meetings

dbooth: What should we try to do there?

harold: I could demo the SWAT4LS example.
... Classifying an instance to decide if a diagnosis is for cancer, using subsumption.
... For example, to enlist patients for studies.

dbooth: Sounds good.

<scribe> ACTION: dbooth to find out the day and time of f2f meeting [recorded in http://www.w3.org/2017/01/03-hcls-minutes.html#action02]

<trackbot> Error finding 'dbooth'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.


<ericP> https://www.w3.org/2016/Talks/1205-swat4ls-egp/reasoning

<ericP> (dbooth)


Summary of Action Items

[NEW] ACTION: dbooth to find out the day and time of f2f meeting [recorded in http://www.w3.org/2017/01/03-hcls-minutes.html#action02]
[NEW] ACTION: Eric to ask Daniella Meeker to see if she has ideas for chem release use case [recorded in http://www.w3.org/2017/01/03-hcls-minutes.html#action01]

Summary of Resolutions

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.148 (CVS log)
$Date: 2017/01/03 17:09:57 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.148  of Date: 2016/10/11 12:55:14  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Present: David_Booth EricP Michael_van_der_Zel Harold_Solbrig
Found Date: 03 Jan 2017
Guessing minutes URL: http://www.w3.org/2017/01/03-hcls-minutes.html
People with action items: dbooth eric

[End of scribe.perl diagnostic output]