Mark Musen on iCat / Web Protege for ICD-11 and the Yosemite Project

11 Nov 2014

See also: IRC log


Mark Musen


Support for Building WHO Terminologies

mark WHO is working on ICD, which has been around since the 19th century, and gets an upgrade every decade
... 11 will be next major rev
... ICD9-CM is a laundry list of terms.
... there's a decimal which *implies* a hierarchy
... in ICD9, it's hard to know if things are equiv or even related.
... jeremy rogers's slide shows that there are a handfull of codes for bike accidents
... ICD10 tried to become more granular (587 for injuries to bikes)
... this explosion of codes has made people leary of ICD10

mark: ICD currently published as a book. WHO sees a need for change
... ICD9-CM is a laundry list of terms.

mark: [WHO slide: ICD-11 -- An Enriched Information Model]
... describes the high-level components
... these are part of the content model for ICD11
... these are represented as OWL classes
... the terms are OWL individuals
... WHO was afraid of the postcoordination model but more afraid of the "exploding bicycle" precoordination model
... WHO wants ICD to be useful in urban US and rural developing countries.
... "linearaizations" result in practice-specific terminologies, etc.
... [slide ICD is no longer just a list of terms ...]
... the goal is not to create a general KB of medicine but to capture enough for machines to do some reasoning.
... [slide ICD-11 "Linearizations"]
... morbidity for public health
... mortatily for death certificates
... [sldie Revision Org Structure]
... not a big yelling competition in geneva
... distributed contributions reviewed by peers
... reduces "last man standing" standards
... [iCAT slide]
... iCAT is Web Protege with some custom interface conventions
... e.g. comments left for developers
... on the right is the WHO content model
... as applied to the code in the hierarchy on the left
... [slide: iCAT Functionalities]
... some turf battles around what tag applies to the hierarchy
... intense discussions, multimedia, pointing to web docs
... shows how webprotege features are used for the development of a very large ontology
... [slide: weekly changes]
... ICD11 has gone from a prototype to what WHO calls a "data phase" ready for testing
... [slide: Post-coord]
... "what are the ways that asma can be post-coordinated?"
... specify with selections of terms
... [slide: logical definition tab]
... important to relate asthma with occurance "intermitten" with "intermittent asthma"
... [slide: WHO adopted the same approache for international classification of traditional medicine]
... same approach applied to ICTM
... e.g. TCM, korean, japanese traditional medicine
... WHO considers all medical care important.
... much of the world recieves these traditional meds but currently no representation
... [slide: Bioportal screenshot]
... ~350 ontologies from Allen Brain to Zebra fish
... goal: use the terms and value sets in bioportabl to unify WHO terminologies with existing terminologies
... [slide: iCat allows users...]
... we can map all of the bioportal terms related to meridian systems to ICTM
... [slide: Next up: CPT]
... AMA's "Current Procedural Terminology"
... currently distributed as a book
... working with AMA to clean up CPT
... also working with AMA and WHO to reconcile ICPT with ICHI (intl classif health interventions)
... discussed with WHO heatlh interventions meeting in barcelona
... can have all of these in OWL, instantiated by particular terminology, supported by bioportal
... [slide: family of OWL-based ontologies]
... Bioportal provides a modacum of the yosemite vision

dbooth: questions?

rafael: can icat link to freebase or dbpedia?

mark: no. iCAT has limited the webprotege interface.
... but a good idea.

Tony: structure of ICD11 IRIs in OWL

mark: there's no attempt to be compatible with IHTSDO
... structure dicated by WHO
... agreement between WHO and IHTSDO to have shared content, but not started yet

Tony: the SNOMED seems like the same structure.
... SNOMED has a standoff to the code
... when we examine the binding of a health record.

Hung: how can we access bioportal

mark: NCBO website, bioportal website
... no workshops planned at this point

annete: what programming langs?

mark: java/google web toolkit

claude: goal for linking terminologies?

<dbooth> Claude: How successful was iCat for use by medical experts who are not familiar with ontologies, RDF or OWL?

mark: desire for convergance, but money to be made in terminologies
... AMA makes more money from CPT than its membership dues
... IHTSDO has of course a licenseing model
... WHO has a need to do this for financing in the future
... desire is there, but the economics favor the status quo

<dbooth> How does this relate to CDISC/HL7 RDF-ization efforts?

mark: parallel play. desire for more coordinated activity

dbooth: what would it take to make iCAT support linkage to multiple standards?
... suppposed you want to link them after development.

mark: we do this in a less-than-ideal way. we can store them all in bioportal.
... anything that we put into bioportal is linked to every other ontology by mappings.
... most programmatic, but accuracy isn't bad
... surprisingly, we get lots of web interface and bioportal API use of the linkages.
... the mappings are complex (how created, provenance, accuracy via user feedback)
... we don't have a compact linkage but the mappings are a step in the right direction

dbooth: are the maps separate ontologies?

mark: instances of OWL classes. readable through the API
... a user can flag an incorrect mapping.
... can flag concepts or mappings.
... we also allow users to select mappings from particular provenance.
... we don't have a well-defined governance process.
... adjustment is done on an ad-hoc basis.

<dbooth> http://en.wikipedia.org/wiki/ISO/IEC_11179

mark: we don't have enough feedback to motivate automation
... there's been no effort to map to 11179. parallel universes.

Sal: ... clinical trials ... NCI ... 11179 ...

ericP: i think PhUSE has done a mapping

<dbooth> David: How successful was iCat for use by medical experts who are not familiar with ontologies, RDF or OWL?

dbooth: how successful was iCAT for non-RDF-expert medical folks
... folks managing these terminologies are esteemed medical professionals, not technocrats.
... we weren't getting funding for the UI, and ICD is complex, so we were pleased that the PAGs were able to do as well as they could
... that said, they shied away from iCAT, working in spreadsheets, but they were ultimately moved to iCAT and folks have used it.
... it was hard for iCAT to overcome the allure of Excel, but folks were more than happy to work online

rafael: pipelines...

<dbooth> Rafael: What new features or directions are already planned or in the works?

mark: in iCAT, lots of work done to customize to match WHO structure: governing body/topic advisory groups/...
... worried that iCAT is overfitted to WHO workflow

dbooth: planned new features?

<dbooth> David: What new features or directions are already planned or in the works?

mark: mainly in response to AMA CPT project
... AMA wants to share value sets, etc.
... precerticiation of procedures
... same work in WHO wrt ICTM
... we will have scalability issues

<dbooth> Claude: What new features or directions are already planned or in the works?

mark: also UI to deal with overwhelming scale

<dbooth> oops, bad paste

Mark: iCAT is open source
... IHTSDO could use Protege

<dbooth> claude: Could iCAT be more generalized for adoption at, say, IHTSDO, that is, for general terminology management? Is iCAT open source (sorry, I may have missed this)?

Mark: they are currently committed to their own tools and DL

<dbooth> Tony: What is the reperesentation of Value Sets? Are they just part of he Class hierarchy?

mark: considering UI enhancements to browse sets.
... when users go to iCAT or webprotege, users want to browse stuff in its context. hard at large scale.

<dbooth> What about marrying iCat with BioPortal? To what extent are they integrated? Can or should they be integrated?

mark: integration is such that iCAT and webprotege users can go to bioportal to grab terms from ontologies
... iCAT/webprotege comments can appear in bioportal
... shows user feedback in a nice way
... plan tighter coordination
... clicking a link in bioportal to bring up a webprotege interface
... also for folks using webprotege to use bioportal for the repo

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2014-11-11 18:21:37 $