Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[bioontology-support] [icat-users] Re: SNOMED CT concept search and representation problems

Trish Whetzel whetzel at stanford.edu
Mon Mar 29 08:33:47 PDT 2010


I think that we are all in the loop here, but just in case....  Last I  
heard Misha will review the requirements and consequences of making  
the requested change below and respond later today or tomorrow. This  
feature request is also logged at: https://bmir-gforge.stanford.edu/gf/project/bioportal_core/tracker/?action=TrackerItemEdit&tracker_item_id=1953

Trish


On Mar 19, 2010, at 10:01 AM, Trish Whetzel wrote:

> The search web service supports search from partial words,  
> difference in parameter between contains and exact match. However if  
> the search string contains two partial words "cand vul" that does  
> not work.
>
> Misha, is this something that Lucene can accommodate, returning  
> results from input of "cand vul"?  Also, there seems to be a  
> problemin that using the input term "candidias of vul" does not  
> return results even though only the last term is not complete. A  
> search using "candidias of" will return "Candidiasis of vulva" as  
> one of the results.
>
> Perhaps it would be worthwhile to talk in person with Misha about  
> how the Lucene index is set-up?
>
> Trish
>
>
> On Mar 19, 2010, at 8:43 AM, Tania Tudorache wrote:
>
>> I don't know if BioPortal plans to support partial word queries, or  
>> if you have suggestions how we can simulate them... Please see  
>> below what WHO needs.
>>
>> Unfortunately, Alan Rector has already concluded that we cannot use  
>> BioPortal for the Snomed search because of this missing feature,  
>> which is not good for any of the teams..
>>
>> Thanks!
>> Tania
>>
>>
>>
>> -------- Original Message --------
>> Subject:	[icat-users] Re: SNOMED CT concept search and  
>> representation problems
>> Date:	Fri, 19 Mar 2010 03:25:22 -0700 (PDT)
>> From:	Robert Chalmers <r.chalmers at man.ac.uk>
>> Reply-To:	icat-users at googlegroups.com
>> To:	iCat Users Group <icat-users at googlegroups.com>
>>
>> Dear Tania
>>
>> Thank you for this.   I should just like to emphasize a suggestion
>> made in my comments above.  Most of the SNOMED lookups I was
>> discussing were in the SNOMED references tab.
>>
>> As I previously wrote:
>>
>> CliniClue very rapidly finds terms by allowing partial word searches
>> in any order.   Thus the search string "ep bu ac" rapidly identifies
>> the correct term by interpreting the query as "string [includes word
>> "ep*"]
>> AND [includes word "bu*"] AND [includes word "ac*"].  I wonder how
>> easy it would
>> be to incorporate that into the iCAT SNOMED CT search. Otherwise I
>> think a lot of cross-mappings will be missed.
>>
>> Alternatively one will have to wade through hundreds of irrelevant
>> terms.
>>
>> I should like to give another example of where my suggestion would
>> have been helpful.
>>
>> B37.3 Candidiasis of vulva and vagina
>>
>> Searching for "Candidiasis of vulva and vagina" brings up a blank
>> Searching for "Candidiasis" brings up 64 results including
>> "Candidiasis of Vulva" and "Candidiasis of Vagina"
>>
>> IN CliniClue a search string "cand vul" and "cand vag" would have
>> immediately identified the two relevant entries.
>>
>> I think it would be immensely helpful to have this implemented in
>> iCAT.
>>
>> One other small point here relates to the display of tabs in the
>> content area. If you open iCAT in a normal (not maximised) window  
>> then
>> drag the right edge of the window to enlarge the window, the tab box
>> does not enlarge with it.  There is up to 8 cm of blank screen  
>> between
>> the tabs and the right arrow.   I imagine that there is a property
>> that could be changed easily to rectify this.   I unfortunately can't
>> do a screen dump here to illustrate this.
>>
>> Best wishes
>>
>> Robert
>>
>>
>>
>>
>> On Mar 18, 2:49 am, Tania Tudorache <tudora... at stanford.edu> wrote:
>> > Search on synonyms should work now.
>> >
>> > Tania
>> >
>> > On 03/16/2010 09:16 PM, Tania Tudorache wrote:
>> >
>> >
>> >
>> > > Hi Robert,
>> >
>> > > Thank you for the comments on the search functionality. It  
>> seems that
>> > > the search on synonyms does not work in the specific values  
>> sets (e.g.
>> > > SNOMED Anatomy that is linked from Body Part), and it should.  
>> We are
>> > > working with the BioPortal people to see why it doesn't.
>> >
>> > > The search is already made in the preferred term, and that  
>> seems to work.
>> >
>> > > We could have the search by SNOMED ID working. Several fields  
>> in iCAT
>> > > are linked to specific value sets in SNOMED (e.g. body part ->  
>> Snomed
>> > > Anatomy branch, Investigations -> Snomed Test Findings, etc.).  
>> So,
>> > > when you search for a body part, the search is restricted to a
>> > > specific value set in Snomed. It seems that there is a problem  
>> with
>> > > searching by Snomed ID and synonym in the specific values sets.
>> >
>> > > There are some fields for which the search is not restricted,  
>> e.g.
>> > > Severity. If you use the "Find term" link from Severity, it will
>> > > search in the entire Snomed, and in that field, the search by  
>> id and
>> > > synonym work fine.
>> >
>> > > So a very quick work around until we figure out the search in  
>> value
>> > > sets problem, is to remove the search in value sets for body  
>> part,
>> > > investigations, etc. and instead search in the entire Snomed.
>> >
>> > > Tania
>> >
>> > > On 03/13/2010 03:08 AM, Robert Chalmers wrote:
>> > >> I have been doing some cross-mapping to SNOMED CT (mainly in  
>> section
>> > >> LB of Chapter 12).  I realise that it is very easy to miss  
>> SNOMED CT
>> > >> concepts in the way that the search is set up and have had to  
>> resort
>> > >> to searching in CliniClue.
>> >
>> > >> Take the example of epidermolysis bullosa acquisita (LB44) or  
>> EBA.
>> > >> This is present "as is" in SNOMED CT as a synonym to "acquired
>> > >> epidermolysis bullosa" of which it is essentially a lexical  
>> variant.
>> > >> A search of "epidermolysis bullosa acquisita"  does produce a  
>> range of
>> > >> entries in SNOMED CT, all of which are children of EBA, but  
>> does not
>> > >> find not the parent term.
>> >
>> > >> Drug-induced epidermolysis bullosa acquisita
>> > >> Epidermolysis bullosa acquisita, Brunsting-Perry type
>> > >> Epidermolysis bullosa acquisita, bullous pemphigoid-like
>> > >> Epidermolysis bullosa acquisita, cicatricial pemphigoid-like
>> > >> Epidermolysis bullosa acquisita, classical acral type
>> > >> Epidermolysis bullosa acquisita, oral mucosal involvement
>> >
>> > >> In order to find EBA you need to know that SNOMED CT's preferred
>> > >> descriptor is "acquired epidermolysis bullosa"
>> >
>> > >> If I had just put in "epidermolysis" then I would have found the
>> > >> correct term amongst 47 others.   With other elements which  
>> are more
>> > >> commonly used, such as "erythema" or "dermatitis", I should  
>> have had
>> > >> to wade through 246 and 328 terms respectively to see if the  
>> specific
>> > >> term I was looking for was present.
>> >
>> > >> CliniClue very rapidly finds terms by allowing partial word  
>> searches
>> > >> in any order.   Thus the search string "ep bu ac" rapidly  
>> identifies
>> > >> the correct term by interpreting the query as "string  
>> [includes "ep*"]
>> > >> AND [includes"bu*"] AND [includes "ac*"].  I wonder how easy  
>> it would
>> > >> be to incorporate that into the iCAT SNOMED CT search.  
>> Otherwise I
>> > >> think a lot of cross-mappings will be missed.
>> >
>> > >> The same will obviously apply to all other elements of the  
>> content
>> > >> model where SNOMED CT is it will be even more important to  
>> ensure that
>> > >> the most appropriate term is found.  In these cases it is not  
>> just a
>> > >> matter of finding the same disease descriptor as is being used  
>> for
>> > >> ICD.   The editor will not be likely to know exactly how the  
>> concept
>> > >> he or she is looking for is represented in SNOMED CT.  Thus  
>> under Body
>> > >> Part a search for "toe web" would bring a blank but "toe"  
>> alone would
>> > >> bring up 563 entries from SNOMED ANATOMY including "toe of  
>> hoof".
>> > >> "Toe web" is in SNOMED CT as a synonym to "structure of  
>> interdigital
>> > >> space of foot", a further synonym of which is "web space of  
>> toe".  Had
>> > >> I searched for "web", I should have found 32 concepts but  
>> again not
>> > >> the one I was looking for.   The CliniClue search of "toe web"  
>> would
>> > >> have identified both "toe web" and "web space of toe".   I  
>> would then
>> > >> have decided that the latter was a preferable descriptor and  
>> opted for
>> > >> that.
>> >
>> > >> I think that my experience so far brings up a number of issues.
>> >
>> > >> 1.  A search of SNOMED CT synonyms as well as preferred terms is
>> > >> essential.
>> > >> 2.  Ideally a search query should replicate that provided in  
>> CliniClue
>> > >> 3.  Should iCAT permit synonym descriptors to be displayed  
>> (e.g. "web
>> > >> space of toe") rather automatically select the SNOMED CT  
>> preferred
>> > >> term (in this instance "structure of interdigital space of  
>> foot")?
>> > >> For clarity I think it should be possible to override use of the
>> > >> preferred descriptor if necessary.
>> > >> 4.  If none of these is going to be possible then should it be
>> > >> possible to search by SNOMED CT Concept ID or Description ID  
>> copied
>> > >> and pasted from CliniClue or other SNOMED browser.  (It would  
>> perhaps
>> > >> be possible for an integer to be recognised in the search box as
>> > >> either a concept ID or a description ID rather than be treated  
>> as a
>> > >> text string).
>> > >> 5.  As 80% of concepts from our principal source terminology are
>> > >> already cross-mapped to SNOMED CT and as the majority of  
>> concepts
>> > >> which will be entered into the ICD-11 foundation layer will  
>> also be
>> > >> represented in that source terminology, it would be very  
>> useful to
>> > >> have feature 4 available.
>> >
>> > >> Thanks to all who have persevered and read to the end of this  
>> message.
>> >
>> > >> Robert
>>
>> To unsubscribe from this group, send email to icat-users 
>> +unsubscribegooglegroups.com or reply to this email with the words  
>> "REMOVE ME" as the subject.
>
> _______________________________________________
> bioontology-support mailing list
> bioontology-support at lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/bioontology-support

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.stanford.edu/pipermail/bioontology-support/attachments/20100329/892bc3c7/attachment.html>


More information about the bioontology-support mailing list