User Tools

Site Tools


projects:workgroups:wg_meeting_01092019

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
projects:workgroups:wg_meeting_01092019 [2019/02/11 22:28]
anu_gururaj2 created
projects:workgroups:wg_meeting_01092019 [2019/02/11 23:37] (current)
anu_gururaj2 [Enterprise NLP needs]
Line 2: Line 2:
 Please find the slides from the meeting below: Please find the slides from the meeting below:
  
-{{ :​projects:​workgroups:​nlp_wg_meeting_20181114.pdf |}}+{{ :​projects:​workgroups:​nlp_wg_meeting_20190109_v2.pdf |}}
  
 Please find the recording of the meeting below: Please find the recording of the meeting below:
  
-{{ :​projects:​workgroups:​2018-11-14_14.00_ohdsi-nlp-wg.mp4 |}}+{{ :​projects:​workgroups:​2019-01-09_13.07_ohdsi-nlp-wg.mp4 |}}
  
-Action items updates:+====== ​Action items updates: ​====== 
 + 
 +===== 2018 Summary ===== 
 +  * Delivered NLP Tools (https://​github.com/​OHDSI/​NLPTools) 
 +  * Wrappers for converting existing NLP systems (MetaMap, cTAKES, CLAMP) outputs to CDM NLP tables 
 +  * Mapping of CUIs to standard terminology in CDM 
 +  * Other ongoing projects 
 +  * Mapping of Note Types to LOINC/​standard vocabulary  
 +  * Landscape Analysis of section identifier systems and proposal of a standard terminology for use  
 +  * Extension and standardization of concept-modifiers from NLP systems 
 + 
 +===== 2019 Aims ===== 
 +  * Evaluate and revise textual CDM tables by sharing practical issues and lessons learnt during ETL for processing textual data into CDM 
 +  * Develop tools (within Atlas) to facilitate uses of NLP data for cohort building/​phenotyping  
 +      - Collaborate with eMERGE consortium:  
 +          - Presentation by Yuan Luo on March 13th, 2019 
 +  * Conduct cross-site studies that use textual data 
 +  * Continue developing other NLP resources 
 + 
 +===== Ongoing Projects ===== 
 +  * Mapping of Note Types to LOINC/​standard vocabulary – Karthik Natarajan, Ruth Reeves, and Jon Duke  
 +    - Karthik hopes to start working on this again, last progress was an evaluation against rule based system with Med (naive Bayes approach) 
 +    - Notes shared with Jon, loaded report names and generated n-grams, need to be tagged per ideas for mapping, Karthik to find somebody to tag/​annotate,​ not accounting yet for linking them back to OMOP vocabulary, note-type needs to be standardized,​ the LOINC document ontology and Radlax are flattened here, both Note_type and Note_class are present, needs to be reviewed, leverage Jon's tool here 
 +  * Landscape Analysis of section identifier systems and proposal of a standard terminology for use – Hua Xu and Karthik Natarajan  
 +    - use SecTagger?, no other updates, Mitre has a tool called Zoner, need to review current Section Header tools, Stephane will share the terminology he has 
 +  * Mapping of CUIs to standard terminology – Juan Banda, need to add link to Juan's repo in the OHDSI NLP repo 
 +  * Standardization of term_modifiers and values – Hua Xu 
 + 
 +===== Presentation ===== 
 +==== Enterprise NLP needs ==== 
 +  * Fast deployment of NLP solutions through 
 +  - Task scheduling  
 +  - Resource management (distributed,​ scalable) 
 +  * Visual Analytics 
 +  - Search NLP outputs 
 +  - Extract information for research – phenotyping ​  
 +  * Enhanced security 
 +   - Multi-user 
 +   - Configurable 
 +  * Demo by Ergin Soysal
  
-  * Rules for defining term_exists - led by Stephane Meystre - COMPLETED 
-  * Mapping of CUIs to standard terminology - led by Juan Banda - COMPLETED, here's the link to the repo: [[https://​github.com/​thepanacealab/​OHDSIananke]] 
-  * Mapping of Note Types to LOINC/​standard vocabulary - Karthik Natarajan, Ruth Reeves, Jon Duke and Hua Xu will work together on this: Multiple note types sent to Jon. To be followed up. 
-  * Landscape Analysis of section identifier systems and proposal of a standard terminology for use - led by Hua Xu with help from Karthik Natarajan: SecTagger and section header ontology obtained from Vanderbilt. Need to figure out normalization. 
-  * Examples and rules for term_temporal - led by George Hripsack (Sunny): The CDM to be updated with the definition below: 
-        * If in the past, then term_temporal is PAST. And term exists has to also be positive. You never have term_exists=no with PAST. 
-        * If not in the past, then term_temporal is blank. 
-        * If it is from the same inpatient admission, that is not considered past, but still in the present. 
-        * More discussion needed on whether to merge term_exists and term_temporal into one variable, or to leave them apart. 
-  * Standardization of term_modifiers and values - led by Hua Xu: To define modifier list and normalized values for each modifier were the tasks. Initial review were from SHARPn modifiers (also used in cTAKES) and Wendy Chapman'​s group ([[https://​github.com/​Blulab-Utah/​resource_ontologies/​wiki]]). 
-        * Should certainty be merged with negation? - in instances where the negation is not about certainty (existential/​systemic),​ it would be better to keep them separate. Another point to think about is the utility of the categories (definite, probable, uncertain, probable negated, definite negated) and the ability of NLP systems to annotate the terms with them. Boolean negation is easier to annotate and handle. 
-        * Granularity of modifiers for subject - Family, Other, Patient may be sufficient for current datasets 
-        * Generic/​conditional modifier - Generic and Conditional as modifier may be acceptable. cTAKES does not support this currently. 
-        * Course - possible values (very limited set) presented. Same as cTAKES. Default way of dealing with it could be to assign the ontology and the value from it. Open ended for now. Can be modified later based on use cases/​usage. 
-        * Temporal - is pulled from cTAKES, could be improved. 
-        * Discovery technique - Name of NLPtool, in cTAKES currently, the methodology (gold-standard,​ rule-based, dictionary look-up etc.) used is the discovery technique. Utility? - needs to be discussed 
-        * Confidence score - how reliable 
-        * Disease specific modifiers - Body location, Severity (some values overlap with cTAKES) 
-        * Medication specific modifiers - need to maybe add strength? 
-        * Test specific - Change reference range to lower threshold, upper threshold? 
-        * Procedure specific modifiers - need examples before discussion, body location, also add laterality here. 
-  * term modifier was used to connect two tables in CDM that could not be otherwise obviously linked together. 
-  * include NLP in the Condition Type concept ID, or fact relationship table. 
-  * How has the NLP table information been used by the community? 
projects/workgroups/wg_meeting_01092019.1549924087.txt.gz · Last modified: 2019/02/11 22:28 by anu_gururaj2