Using Luke to search the drug index, It (RxNorm code 328608) doesn't seem to be inside the default rx_norm dictionary...
Don't recall the attributes that was used to build the index, but it could be due to the fact that it was 'RXN_OBSOLEETED 07/31/2009'? --Pei > -----Original Message----- > From: Masanz, James J. [mailto:[email protected]] > Sent: Thursday, January 02, 2014 2:52 PM > To: '[email protected]' > Subject: RE: what dictates full medication mention? > > cTAKES annotates all mentions it finds, so if a shorter one is part of a > longer > one, you are supposed to get both/all, just as you are seeing with Ibuprofen > and Ibuprofen 100MG. > > Offhand I don't know why the Drug NER cTAKES is not also creating a > org.apache.ctakes.typesystem.type.textsem.MedicationMention for > "Propranolol 40 MG". It's not creating one for me either. I'm using > DrugAggregatePlaintextUMLSProcessor.xml > And when I look at UMLS via the UTS, I see an rxnorm entry Propranolol 40 > MG [A10426756/RXNORM/SCDC/328608] for C1123553. > > Drug NER component uses a lucene index of terms from RxNorm. It's > possible that term is missing from the index. I can check on that later today > or tomorrow if no one else beats me to it. > > I would do it using the Luke tool to inspect the lucene index in ctakes- > resources- > 3.1.0\resources\org\apache\ctakes\drugner\lookup\rxnorm_index > > -- James > > > -----Original Message----- > From: [email protected] > [mailto:[email protected]] On > Behalf Of Chase Master > Sent: Thursday, January 02, 2014 1:30 PM > To: [email protected] > Subject: what dictates full medication mention? > > Hi, > > I have a quick question. When I run the cTAKES drugNER module on > "Ibuprofen 100 MG" and "Propranolol 40 MG", I get an annotation for both > Ibuprofen 100MG and Ibuprofen for the first case, but for the second case I > only get Propranolol, by itself. It does show the strength as 40MG for > Propranolol, but it doesn't create a medication mention with full covered text > as "Propranolol 40 MG". > Both of these have the same semantic type, I noticed ( [T200]). > > Thanks, > Chase
