You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ctakes.apache.org by Bruce Tietjen <br...@perfectsearchcorp.com> on 2014/07/03 19:32:49 UTC

Converting UMLS codes to SNOMED codes using the YTEX pipeline

I've been playing around with 3.1 and 3.2 rc1 using the YTEX pipeline.

I noticed that the output generated by the YTEX pipeline outputs the UMLS
codes. Is there an easy way to convert the UMLS codes to SNOMED codes?

I had expected that the LookupDesc_SNOMED.xml might handle that, but then
noticed that the LookupConsumer identified there specifically uses the UMLS
codingscheme.

Or is there a reason why the ytex pipeline should not produce SNOMED codes?

Thanks,

Bruce Tietjen



 [image: IMAT Solutions] <http://imatsolutions.com>
 Bruce Tietjen
Senior Software Engineer
[image: Mobile:] 801.634.1547
bruce.tietjen@imatsolutions.com

Re: Converting UMLS codes to SNOMED codes using the YTEX pipeline

Posted by vijay garla <vn...@gmail.com>.
caveat: this may have changed with the new dictionary lookup algorithm, but
fwiw this is base don the old dictionary lookup

The simple answer to 'why no snomed codes by default with ytex?' this is a
cleaner data model, and a faster dictionary lookup.  If you want to map
cuis to snomed codes, simply join with the UMLS MRCONSO table.

cTAKES supports different 'Hit Consumers'; ytex is by default configured to
use a 'plain vanilla' consumer that adds just the CUIs to the named entity.
 Other Dictionary Lookup configurations use a Snomed Hit consumer.  What
this does is for a given CUI, it looks up the the related snomed codes, and
adds the snomed codes and cuis to the named entity.

Regarding speed - the Snomed consumer runs additional database queries.
 For each cui that matches a given named entity, the Snomed consumer is
running an additional database query - obviously much slower.

Regarding data cleanliness - the Snomed hit consumer produces a
'denormalized' data model - there is a many-to-many relationship between
UMLS CUIs and SNOMED Codes - each CUI can have multiple SNOMED Codes
associated with it.  As mentioned above, if you want the SNOMED codes for a
cui, just join the ytex table with the UMLS MRCONSO table, filter by
SAB=SNOMEDCT (sorry, don't know what the SAB abbreviation is off the top of
my head).

Note that this is the default config - you can mix and match dictionary
lookup components in your pipeline as you wish.

HTH,

VJ


On Thu, Jul 3, 2014 at 7:32 PM, Bruce Tietjen <
bruce.tietjen@perfectsearchcorp.com> wrote:

> I've been playing around with 3.1 and 3.2 rc1 using the YTEX pipeline.
>
> I noticed that the output generated by the YTEX pipeline outputs the UMLS
> codes. Is there an easy way to convert the UMLS codes to SNOMED codes?
>
> I had expected that the LookupDesc_SNOMED.xml might handle that, but then
> noticed that the LookupConsumer identified there specifically uses the UMLS
> codingscheme.
>
> Or is there a reason why the ytex pipeline should not produce SNOMED codes?
>
> Thanks,
>
> Bruce Tietjen
>
>
>
>  [image: IMAT Solutions] <http://imatsolutions.com>
>  Bruce Tietjen
> Senior Software Engineer
> [image: Mobile:] 801.634.1547
> bruce.tietjen@imatsolutions.com
>