You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ctakes.apache.org by Deepal Dhariwal <de...@gmail.com> on 2013/02/06 17:16:25 UTC
Re: Information Extraction using ontology + Access Aggregate Plain
Text UMLS processor results - Concept Annotation
Hello all,
I am executing ctakes CPE from Eclipse i.e. ctakesCPEGUI.java. I want to
access and process the (UMLS) Concept Annotations and NumToken Annotations.
I went through the src but couldnt find the the exact class which I would
need to modify to implement my Ontology Lookup Annotator. I want to check
each UMLS concept present in document, then check whether there is a
corresponding instance in Heartfaid Ontology and if so assert a fact about
the patient.
Kindly let me know how to programmatically access Concept Annotation
results.
Thanks
Deepal Dhariwal
On Fri, Jan 18, 2013 at 12:50 PM, Chen, Pei
<Pe...@childrens.harvard.edu>wrote:
> Hi Deepal,
> I am assuming the mapping would be through UMLS CUI's.
> For example, for each CUI identified by the cTAKES UMLS Dictionary Lookup
> annotator.
>
> Just putting some ideas out there:
> One would need a data structure to represent the Cardio Classes (perhaps a
> hashmap of some sort) that maps CUI's to the classes/instances.
> And then also a process to load that structure from the RDF (from either
> flat file, db, etc.) format?
> One might also need a place holder in the CAS (type system) to store the
> identified classes?
>
> --Pei
>
> > -----Original Message-----
> > From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> > Sent: Wednesday, January 16, 2013 11:21 AM
> > To: ctakes-dev@incubator.apache.org
> > Subject: Re: Information Extraction using ontology
> >
> > Hello all,
> >
> > So what I am aiming is as follows:
> > 1. Extract UMLS medical terms from a given clinical narrative using
> ctakes
> > umls lookup annotator 2. Identify cardio vascular terms from above set.
> > These terms are classes / instances of heart failure ontology. I want
> to make
> > this mapping generic so that if in future I am targeting carcinogenic
> terms, I'll
> > would just need to change underlying ontology.
> > 3. Lastly I want to run a reasoner over the rdf triples obtained from
> above
> > step.
> >
> > For example: Find the timi score of patient 1. Extract all medical terms
> from
> > clinical narrative 2. Extract cardio vascular terms such as whether the
> patient
> > has congestive heart failure, whether he is an active smoker etc. using
> heart
> > failure ontology 3. Run the sparql query of form :
> > timi score = does patient have chf + age > 65 + active smoker?
> >
> > Hence I need guidance on an annotator that would map ctakes terms to
> > heart failure ontology and allow to run reasoner over it. Any suggestions
> > would be of great help.
> >
> >
> > Thanks
> > Deepal Dhariwal
> >
> > On Mon, Dec 31, 2012 at 9:00 AM, Lin, Chen
> > <Ch...@childrens.harvard.edu>wrote:
> >
> > > This mapping sounds very interesting and has potential usage for other
> > > projects. I will be curious to follow its progress. Thank you, Deepal,
> > > for digging this!
> > >
> > > Best,
> > > Chen
> > >
> > > Sent from my iPhone
> > >
> > > On Dec 30, 2012, at 6:43 AM, "Deepal Dhariwal"
> > > <de...@gmail.com>
> > > wrote:
> > >
> > > > Hello Pei,
> > > >
> > > > I am aiming for an annotator that provides mapping between UMLS
> > > > CUI's and ontology classes. If I could write a generic code that
> > > > provides mapping between cTakes results and any ontology then it
> > > > might be useful to the community.
> > > > Let me know your views on this annotator.
> > > >
> > > > Thanks
> > > > Deepal Dhariwal
> > > >
> > > > On Fri, Dec 28, 2012 at 10:37 AM, Chen, Pei
> > > > <Pe...@childrens.harvard.edu>wrote:
> > > >
> > > >> Hi Deepal,
> > > >> If I understand correctly, you were planning to have an annotator
> > > >> that iterates through all of the UMLS CUI's identified by cTAKES
> > > >> and then map them to the heartfaid ontology class/code.
> > > >> I am not aware of anything that does specifically this, but there
> > > >> is an open ticket that is trying to do something similar for Drugs
> > > >> classes (RxNorm to NDF-RT mapping) (
> > > >> https://issues.apache.org/jira/browse/CTAKES-107).
> > > >> It does seem like something that may be useful for others if you
> > > >> would like to contribute that to the community though...
> > > >>
> > > >> --Pei
> > > >>
> > > >>> -----Original Message-----
> > > >>> From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> > > >>> Sent: Thursday, December 27, 2012 10:28 PM
> > > >>> To: ctakes-dev@incubator.apache.org
> > > >>> Cc: ctakes-user@incubator.apache.org
> > > >>> Subject: Re: Information Extraction using ontology
> > > >>>
> > > >>> Hello Pei,
> > > >>>
> > > >>> The dictionary is an ontology i.e. a .owl file (Reference:
> > > >>> http://lis.irb.hr/heartfaid/ontology/) . I am not sure whether it
> > > >>> is
> > > >> subset of
> > > >>> UMLS but its definitely based on UMLS since they have relations
> > > >>> such as 'UMLS Synonym'
> > > >>> An example of mapping terms would be:
> > > >>> 1. UMLS lookup annotator returns 'Hypertension' as medical
> > > >>> concept.I
> > > want
> > > >>> to assert that it corresponds to Hypertension class on this
> > > >>> ontology
> > > >> whose
> > > >>> parent is Blood Pressure and relations include TreatedBy, Could
> > > >>> Include
> > > >> etc.
> > > >>>
> > > >>> Hence I want to build a annotator that provides mapping between
> > > >>> ctakes annotator results and owl classes. Does the dictionary
> > > >>> lookup component allow to plug in dictionaries in .owl or .rdf
> format?
> > > >>>
> > > >>> Thanks
> > > >>> Deepal Dhariwal
> > > >>>
> > > >>> On Thu, Dec 27, 2012 at 10:31 AM, Chen, Pei
> > > >>> <Pe...@childrens.harvard.edu>wrote:
> > > >>>
> > > >>>> Hi Deepal,
> > > >>>> Within the dictionary lookup component, you can also plugin other
> > > >>>> custom dictionaries in additional to the UMLS ones.
> > > >>>> The dictionary itself can be in different formats (MySQL, HSQLDB,
> > > >>>> Lucene, CSV text file) and could be configured via the xml files.
> > > >>>> Is this what you were referring to? Do you have an example of
> > > >>>> "mapping terms to the corresponding classes?"
> > > >>>> Just curious, Is the heart ontology a subset of UMLS?
> > > >>>>
> > > >>>> Thanks,
> > > >>>> --Pei
> > > >>>>
> > > >>>>> -----Original Message-----
> > > >>>>> From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> > > >>>>> Sent: Thursday, December 27, 2012 12:17 AM
> > > >>>>> To: ctakes-user@incubator.apache.org;
> > > >>>>> ctakes-dev@incubator.apache.org
> > > >>>>> Subject: Information Extraction using ontology
> > > >>>>>
> > > >>>>> Hello all,
> > > >>>>>
> > > >>>>> I have a heart ontology and I want to extract terms from a
> > > >>>>> clinical text
> > > >>>> that
> > > >>>>> correspond to classes in the ontology. cTAKES UMLS Lookup
> > > >>>>> Annotator returns UMLS terms from a given piece of text. Could I
> > > >>>>> modify the
> > > >>>> annotator
> > > >>>>> to extract terms from the ontology and is there is some other
> > > >>>>> component
> > > >>>> of
> > > >>>>> cTAKES that would allow me to do so. I am trying to set up a
> > > >>>>> cTakes
> > > >>>> pipeline
> > > >>>>> where I first extract medical terms from text using UMLS lookup
> > > >>>>> then
> > > >>>> filter
> > > >>>>> cardio vascular terms and identify relations between them using
> > > >>>>> an
> > > >>>> ontology.
> > > >>>>>
> > > >>>>> Thanks
> > > >>>>> Deepal Dhariwal
> > > >>>>
> > > >>
> > >
>
RE: Information Extraction using ontology + Access Aggregate Plain
Text UMLS processor results - Concept Annotation
Posted by "Chen, Pei" <Pe...@childrens.harvard.edu>.
Hi Deepal,
Take a peek at:
http://svn.apache.org/viewvc/incubator/ctakes/trunk/ctakes-dictionary-lookup/src/main/java/org/apache/ctakes/dictionary/lookup/ae/
DictionaryLookupAnnotator.java
And something like UmlsToSnomedConsumerImpl.java that creates/puts the UmlsConcept objects into the CAS...
But I presume you will most likely create a Brand new Annotator that will be added to the pipeline?
Also, I would encourage upgrading new development against either trunk or at least 3.x (instead of 2.5).
--Pei
> -----Original Message-----
> From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> Sent: Wednesday, February 06, 2013 11:17 AM
> To: ctakes-dev@incubator.apache.org
> Subject: Re: Information Extraction using ontology + Access Aggregate Plain
> Text UMLS processor results - Concept Annotation
>
> Hello all,
>
> I am executing ctakes CPE from Eclipse i.e. ctakesCPEGUI.java. I want to
> access and process the (UMLS) Concept Annotations and NumToken
> Annotations.
> I went through the src but couldnt find the the exact class which I would
> need to modify to implement my Ontology Lookup Annotator. I want to
> check each UMLS concept present in document, then check whether there is
> a corresponding instance in Heartfaid Ontology and if so assert a fact about
> the patient.
> Kindly let me know how to programmatically access Concept Annotation
> results.
>
> Thanks
> Deepal Dhariwal
>
>
> On Fri, Jan 18, 2013 at 12:50 PM, Chen, Pei
> <Pe...@childrens.harvard.edu>wrote:
>
> > Hi Deepal,
> > I am assuming the mapping would be through UMLS CUI's.
> > For example, for each CUI identified by the cTAKES UMLS Dictionary
> > Lookup annotator.
> >
> > Just putting some ideas out there:
> > One would need a data structure to represent the Cardio Classes
> > (perhaps a hashmap of some sort) that maps CUI's to the classes/instances.
> > And then also a process to load that structure from the RDF (from
> > either flat file, db, etc.) format?
> > One might also need a place holder in the CAS (type system) to store
> > the identified classes?
> >
> > --Pei
> >
> > > -----Original Message-----
> > > From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> > > Sent: Wednesday, January 16, 2013 11:21 AM
> > > To: ctakes-dev@incubator.apache.org
> > > Subject: Re: Information Extraction using ontology
> > >
> > > Hello all,
> > >
> > > So what I am aiming is as follows:
> > > 1. Extract UMLS medical terms from a given clinical narrative using
> > ctakes
> > > umls lookup annotator 2. Identify cardio vascular terms from above set.
> > > These terms are classes / instances of heart failure ontology. I
> > > want
> > to make
> > > this mapping generic so that if in future I am targeting
> > > carcinogenic
> > terms, I'll
> > > would just need to change underlying ontology.
> > > 3. Lastly I want to run a reasoner over the rdf triples obtained
> > > from
> > above
> > > step.
> > >
> > > For example: Find the timi score of patient 1. Extract all medical
> > > terms
> > from
> > > clinical narrative 2. Extract cardio vascular terms such as whether
> > > the
> > patient
> > > has congestive heart failure, whether he is an active smoker etc.
> > > using
> > heart
> > > failure ontology 3. Run the sparql query of form :
> > > timi score = does patient have chf + age > 65 + active smoker?
> > >
> > > Hence I need guidance on an annotator that would map ctakes terms to
> > > heart failure ontology and allow to run reasoner over it. Any
> > > suggestions would be of great help.
> > >
> > >
> > > Thanks
> > > Deepal Dhariwal
> > >
> > > On Mon, Dec 31, 2012 at 9:00 AM, Lin, Chen
> > > <Ch...@childrens.harvard.edu>wrote:
> > >
> > > > This mapping sounds very interesting and has potential usage for
> > > > other projects. I will be curious to follow its progress. Thank
> > > > you, Deepal, for digging this!
> > > >
> > > > Best,
> > > > Chen
> > > >
> > > > Sent from my iPhone
> > > >
> > > > On Dec 30, 2012, at 6:43 AM, "Deepal Dhariwal"
> > > > <de...@gmail.com>
> > > > wrote:
> > > >
> > > > > Hello Pei,
> > > > >
> > > > > I am aiming for an annotator that provides mapping between UMLS
> > > > > CUI's and ontology classes. If I could write a generic code that
> > > > > provides mapping between cTakes results and any ontology then it
> > > > > might be useful to the community.
> > > > > Let me know your views on this annotator.
> > > > >
> > > > > Thanks
> > > > > Deepal Dhariwal
> > > > >
> > > > > On Fri, Dec 28, 2012 at 10:37 AM, Chen, Pei
> > > > > <Pe...@childrens.harvard.edu>wrote:
> > > > >
> > > > >> Hi Deepal,
> > > > >> If I understand correctly, you were planning to have an
> > > > >> annotator that iterates through all of the UMLS CUI's
> > > > >> identified by cTAKES and then map them to the heartfaid ontology
> class/code.
> > > > >> I am not aware of anything that does specifically this, but
> > > > >> there is an open ticket that is trying to do something similar
> > > > >> for Drugs classes (RxNorm to NDF-RT mapping) (
> > > > >> https://issues.apache.org/jira/browse/CTAKES-107).
> > > > >> It does seem like something that may be useful for others if
> > > > >> you would like to contribute that to the community though...
> > > > >>
> > > > >> --Pei
> > > > >>
> > > > >>> -----Original Message-----
> > > > >>> From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> > > > >>> Sent: Thursday, December 27, 2012 10:28 PM
> > > > >>> To: ctakes-dev@incubator.apache.org
> > > > >>> Cc: ctakes-user@incubator.apache.org
> > > > >>> Subject: Re: Information Extraction using ontology
> > > > >>>
> > > > >>> Hello Pei,
> > > > >>>
> > > > >>> The dictionary is an ontology i.e. a .owl file (Reference:
> > > > >>> http://lis.irb.hr/heartfaid/ontology/) . I am not sure whether
> > > > >>> it is
> > > > >> subset of
> > > > >>> UMLS but its definitely based on UMLS since they have
> > > > >>> relations such as 'UMLS Synonym'
> > > > >>> An example of mapping terms would be:
> > > > >>> 1. UMLS lookup annotator returns 'Hypertension' as medical
> > > > >>> concept.I
> > > > want
> > > > >>> to assert that it corresponds to Hypertension class on this
> > > > >>> ontology
> > > > >> whose
> > > > >>> parent is Blood Pressure and relations include TreatedBy,
> > > > >>> Could Include
> > > > >> etc.
> > > > >>>
> > > > >>> Hence I want to build a annotator that provides mapping
> > > > >>> between ctakes annotator results and owl classes. Does the
> > > > >>> dictionary lookup component allow to plug in dictionaries in
> > > > >>> .owl or .rdf
> > format?
> > > > >>>
> > > > >>> Thanks
> > > > >>> Deepal Dhariwal
> > > > >>>
> > > > >>> On Thu, Dec 27, 2012 at 10:31 AM, Chen, Pei
> > > > >>> <Pe...@childrens.harvard.edu>wrote:
> > > > >>>
> > > > >>>> Hi Deepal,
> > > > >>>> Within the dictionary lookup component, you can also plugin
> > > > >>>> other custom dictionaries in additional to the UMLS ones.
> > > > >>>> The dictionary itself can be in different formats (MySQL,
> > > > >>>> HSQLDB, Lucene, CSV text file) and could be configured via the
> xml files.
> > > > >>>> Is this what you were referring to? Do you have an example
> > > > >>>> of "mapping terms to the corresponding classes?"
> > > > >>>> Just curious, Is the heart ontology a subset of UMLS?
> > > > >>>>
> > > > >>>> Thanks,
> > > > >>>> --Pei
> > > > >>>>
> > > > >>>>> -----Original Message-----
> > > > >>>>> From: Deepal Dhariwal [mailto:deepaldhariwal@gmail.com]
> > > > >>>>> Sent: Thursday, December 27, 2012 12:17 AM
> > > > >>>>> To: ctakes-user@incubator.apache.org;
> > > > >>>>> ctakes-dev@incubator.apache.org
> > > > >>>>> Subject: Information Extraction using ontology
> > > > >>>>>
> > > > >>>>> Hello all,
> > > > >>>>>
> > > > >>>>> I have a heart ontology and I want to extract terms from a
> > > > >>>>> clinical text
> > > > >>>> that
> > > > >>>>> correspond to classes in the ontology. cTAKES UMLS Lookup
> > > > >>>>> Annotator returns UMLS terms from a given piece of text.
> > > > >>>>> Could I modify the
> > > > >>>> annotator
> > > > >>>>> to extract terms from the ontology and is there is some
> > > > >>>>> other component
> > > > >>>> of
> > > > >>>>> cTAKES that would allow me to do so. I am trying to set up a
> > > > >>>>> cTakes
> > > > >>>> pipeline
> > > > >>>>> where I first extract medical terms from text using UMLS
> > > > >>>>> lookup then
> > > > >>>> filter
> > > > >>>>> cardio vascular terms and identify relations between them
> > > > >>>>> using an
> > > > >>>> ontology.
> > > > >>>>>
> > > > >>>>> Thanks
> > > > >>>>> Deepal Dhariwal
> > > > >>>>
> > > > >>
> > > >
> >