Hi Chase,

I have observed this similar issue in my use case, for example,
"... #1 Bottle(s)metformin PO undefined Methotrexate (ORAL) 15 MG (2.5MG TABLET 
Take 6) PO QWEEK , 6 tablets, once a week #24 Tablet(s)Metrocream 
(METRONIDAZOLE 0.75% Cream)..."

I thought the dosage number for Methotrexate should be "15", and the unit 
should be "MG". Instead, I got the number as "15", but the unit as "%".

And Sean Murphy, who developed the drug-ner module, gave me the following 
explanation.  Hope it is relevant and helpful! Hope this out-of-context 
signature problem could be fixed in the future.

Best,
Chen

___________________________

        The drugner pipeline attempts to assign the drug signature elements 
such as strength, strength unit, frequency, route, etc. in a very linear, 
left-to-right approach.   It doesn't have any ability to cross reference the 
mentions, so it sees each drug named entity as a new distinct mention.
        Therefore, in your example below there are actually four entities being 
parsed:
        1)   melformin from '1 Bottle(s)metformin PO undefined' which is 
assigned route as "Enteral_Oral" only
        2)  Methotrexate from ' Methotrexate (ORAL) 15 MG (2.5MG TABLET Take 6) 
PO QWEEK , 6 tablets, once a week #24 Tablet(s)' which is assigned              
route "Enteral_Oral",  dose "24", form "tablet", strength "15", strength unit 
"MG"
        3) Metrocream from 'Metrocream' with no attributes and
        4) METRONIDAZOLE from 'METRONIDAZOLE 0.75% Cream' with route "Topical", 
form "cream", strength "0.75" and strength unit "%"

        So the window that dictates where these terms are assigned is bordered 
by the actual drug mentions, themselves.

MedicationDosage pertains to the quantity, typically a numeric value followed 
by the form.
MedicationStrength pertains to the quantity, also typically a numeric value 
followed by a strength unit.

        So I think the situation you are seeing could be disjointed or out of 
context with the entity mention that owns the attribute.

        Thanks,
        ~Sean

-----Original Message-----
From: Chase Master [mailto:[email protected]] 
Sent: Wednesday, September 04, 2013 11:27 AM
To: [email protected]
Subject: Re: Incorrect frequency in Drug NER?

Hi,
I've been experimenting with the Drug NER module as well.  I noticed that for 
the phrase "Ibuprofen increased from 200 mg to 300 mg p.o. once a day."
the medicationFrequency has number = 1, but unit = null for the Ibuprofen with 
strength 300 mg.  Also, for the phrase "Aspirin 200-mg once-a-day", there is no 
medicationFrequency, whereas for the phrase "Aspirin 200-mg once a day" 
medicationFrequency has number = 1, unit = day.  I have noticed this behaviour 
on both v3.0 and the trunk.

Thanks


On Wed, Aug 28, 2013 at 12:55 PM, Kannan Thiagarajan <[email protected]>wrote:

> Thanks James.
>
>
> On Wed, Aug 28, 2013 at 11:51 AM, Masanz, James J. 
> <[email protected]
> >wrote:
>
> > This looks like a bug. I've created CTAKES-230
> > https://issues.apache.org/jira/browse/CTAKES-230
> >
> >
> > -----Original Message-----
> > From: [email protected] [mailto:
> > [email protected]] On Behalf 
> > Of Kannan Thiagarajan
> > Sent: Sunday, August 25, 2013 7:20 PM
> > To: [email protected]
> > Subject: Incorrect frequency in Drug NER?
> >
> > Hello,
> >
> > I just ran into an issue where I'm getting an incorrect Medication 
> > Frequency while running the Drug NER processor 
> > (DrugAggregatePlaintextUMLSProcessor.xml). I tried this via the CVD 
> > environment. I gave a couple of different texts and noticed the issue.
> >
> > 1) The text  "Apply Bacitracin to affected area on feet twice a day" 
> > - seems to work correctly and gives a Frequency number of "2" and 
> > unit as "day"
> >
> > 2) The text ""Apply Bacitracin to affected area on feet twice a day 
> > *for one week*" - seems to trip up and gives an Frequency number of 
> > "1" and
> unit
> > as "week".
> >
> >
> > I tried other variations such as
> >
> > "for 7 day"
> > "for 2 weeks"
> > "for 1 month"
> >
> > and they seem to work fine and in all cases the Duration seems to be
> fine.
> >
> >
> > Any thoughts?
> >
> > Just for the record, I'm running with a recently updated trunk 
> > version of the code.
> >
> > --
> > Best Regards
> > Kannan Thiagarajan
> >
>
>
>
> --
> Best Regards
> Kannan Thiagarajan
>

Reply via email to