15:01:05 RRSAgent has joined #hcls 15:01:05 logging to http://www.w3.org/2017/10/10-hcls-irc 15:01:07 RRSAgent, make logs world 15:01:07 Zakim has joined #hcls 15:01:09 Meeting: Semantic Web Health Care and Life Sciences Interest Group Teleconference 15:01:09 Date: 10 October 2017 15:02:24 Chair: David Booth 15:05:14 Present: David Booth, Tony Mallia 15:10:55 Topic: Semantic MediaWiki for structured healthcare data (Tony Mallia) 15:11:07 Present+ Gopi 15:12:11 tony: This grew out of FHIR RDF work. 15:14:58 Present+ EricP 15:16:41 tony: Trying a binding simplification 15:16:54 ... Built a small EHR to test it out 15:17:47 ... Patient-Patient link (I think) is to say that two patients are the same person. 15:21:09 ... We can also find a valueset. Head of SNOMED tree is substance. 15:21:31 ... But it is also a subclass of FHIR:AllergyIntolerance. 15:21:47 ... I used Semantic MediaWiki to try this out. 15:22:14 ... I had used it for other things 15:22:44 ... Classess are called Categories in Semantic MediaWiki. 15:23:03 ... Nice thing is that the templates include SPARQL-like queries. 15:23:14 ... and they define visual presentation. 15:23:25 ... Took me less than a week to do this! 15:23:31 ... Exceptional productivity. 15:23:50 ... It implements some of the OMG MDA concepts. 15:24:35 ... Gartner measured productivity gain as 23x with tooling from a Spanish company using Semantic MediaWiki. 15:25:58 (Tony starts demo) 15:28:58 tony: Health record is generated from the query 15:31:43 ... Can even do chained properties with inverse properties, in the queries. 15:33:24 ... For valueset codes, we can even display the definition when you hover over. 15:34:42 ... You can select semantic codes. 15:35:23 ... I've shown Clinical Status class as the root, but you can tell it to have no root instead. 15:36:26 ... Mouse-over can also be used for explaining various fields. 15:37:02 ... The record is categories as a FHIR AllergyIntolerance. 15:37:12 ... The platform will list them all. 15:38:09 ... I can also add things that are like properties. 15:39:07 ... Pages are sort of like blank nodes. 15:40:33 eric: Any idea of scalability? All of snomed? 15:41:47 tony: Some public sites seem to be large. I don't see a scalability problem, because you're on top of MediaWiki. 15:42:12 ... It has pages and triples. 15:42:46 ... This triplestore is on top of mysql, but you can do other storage underneath. 15:43:05 gopi: Can you query using SPARQL? 15:43:15 tony: It is SPARQL-like, but not SPARQL. 15:45:10 dbooth: How do you see this being used? What would be the best use of it? 15:46:01 tony: I'm trying this out for publishing a paper. There's a cross-over from wiki to 4GL language, and an informatics community can use it. No programming needed. 15:46:43 ... I think we are about to have a revolution in software. Anywhere that you use spreadsheets etc for knowledge, it could be done this way. 15:46:53 ... I suspect this could be a new platform for developing EHRs. 15:47:03 ... Don't know who will pick it up. 15:47:37 ... Huge reduction in the dev cost. 15:47:50 ... There will be lots of resistance from status quo vendors. 15:48:39 Present+ Rob_Hausam 15:51:47 tony: This can also support inter-wiki communication. 15:53:02 ... To pull allergies from other providers. 15:54:27 dbooth: Plan to do more with this? 15:54:36 tony: Submit for publication. 15:55:32 ... Then look for sponsorship. 15:57:19 dbooth: You put in some FHIR things manually. If this were done seriously, then all of the FHIR structure definitions, valuesets, etc., could be loaded programmatically. 15:58:19 tony: Yes, though there would be a versioning issue when the FHIR spec changes. 15:59:33 ... Currently looking at styles to use for transformations. 16:02:00 tony: For CRUD on complex graph structures, this is very good. 16:04:26 gopi: Using an RDF triplestore? 16:04:44 tony: It puts them in mysql right now, but can be configured with a real triplestore. 16:04:59 https://www.semantic-mediawiki.org/wiki/Help:SPARQLStore/RepositoryConnector/Generic 16:07:14 dbooth: I could see it being used for data capture similar to Red Cap: https://www.project-redcap.org/ such as surveys. 16:08:36 ... Value that I see is in being able to rapidly create new data capture forms, for structured data capture, but have them properly semantically linked to existing records. 16:10:30 gopi: OpenEHR has a collaboration with Marant (in Slovenia), have a platform to create a model in OpenEHR, export to their platform, and it creates the forms. 16:10:49 ... This may be able to do something similar. 16:12:17 tony: Once you've built your ont with a language, you're stuck with that language. I suspect there is opportunity for standardizing it. 16:13:43 ADJOURNED 16:13:59 rrsagent, draft minutes 16:13:59 I have made the request to generate http://www.w3.org/2017/10/10-hcls-minutes.html dbooth 16:43:00 dbooth has joined #hcls 16:43:25 zakim, who is here? 16:43:25 Present: David, Booth, Tony, Mallia, Gopi, EricP, Rob_Hausam 16:43:27 On IRC I see dbooth, Zakim, RRSAgent, trackbot, ericP 16:51:01 i|ADJOURNED|Screenshots and slides from Tony's demo: https://lists.w3.org/Archives/Public/www-archive/2017Oct/0002.html 16:51:06 rrsagent, draft minutes 16:51:06 I have made the request to generate http://www.w3.org/2017/10/10-hcls-minutes.html dbooth 19:14:45 Zakim has left #hcls