Differences

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

--- projects:workgroups:wg_meeting_02032016 [2016/03/07 21:29] – anu_gururaj
+++ projects:workgroups:wg_meeting_02032016 [2016/03/09 20:31] (current) – anu_gururaj
@@ Line 3: / Line 3: @@
 ==== Attendees ====
-Hua Xu, Jon Duke, George Hripcsak, Karthik Natarajan, Anupama Gururaj, Mark Khayter, Min Jiang, Alexandre Yahi, Noemie Elhadad, Juan M Banda, Olga Patterson, Lian
+Hua Xu, Jon Duke, George Hripcsak, Karthik Natarajan, Anupama Gururaj, Mark Khayter, Min Jiang, Alexandre Yahi, Noemie Elhadad, Juan M Banda, Olga Patterson, Lian Hu
 ==== Agenda ====
-      -IRB for use of clinical text
+{{:projects:workgroups:nlp_wg_meeting_02032016_final.pdf|}}
-      -Clinical text data storage and representation schema
-      -NLP tools/pipelines for ETL
+  - Minimal Model Presentation – Alex
-      -Use cases, e.g, phenotyping for cohort selection using NLP outputs
+  - Note-type mapping Presentation – Karthik
-      -Discussion
+  - Share existing ontologies from Vanderbilt (Hua) and Regenstrief (Jon)
+  - Share strategies for combining data from different searches – Jon
+  - Report on WG for commenting – Hua
+  - Wrappers for cTAKES and Metamap – Min
+  - Improvements to search engine set up using MT samples – Min
+  - Textual Data Representation – Discussion
+  - Goals of 2016
+  - Change of meeting time
 ===Minutes===
-  - General IRB document for use of clinical text and approval from all contributors, post online - Almost completed
+  - Minimal model presentation - Alex {{:projects:workgroups:ohdsi_nlp_wg_yahi.pdf|}}
-  - Collect minimum set of modifiers for all clinical entities that support use of rule to derive clinical concepts: Alex
+        - the model is based on the SHARE-N model and adapted to the current data structure. This model incorporates other semantic types and all of the modifiers are not available in cTAKES yet.
-      * cTAKES is being run on clinical notes programmatically. Alex will present the minimal model in the next meeting.
+        - the notes were processed from eMERGE cohort at Columbia with about 60,000 notes encompassing 1700 patients. The original patient number was 3200.
-  - Aggregate and share note-type metadata from various sources: Karthik
+        - In theory, a set containing the combination of minimal modifiers can be generated. Practically, can we trust the data enough to add it into OHDSI tables? - only highest confidence data (with maximum PPV) should be added to the tables.
-      * LOINC note type mapping would be a very useful resource. We should generate hierarchical representation of note-types as an ontology. Karthink will present his work to date at the next meeting.
+        - Next steps:
-      * Existing ontology for note types to be shared : Vanderbilt (Hua) and Regenstrief (Jon)
+          - Look at the note sections to determine the errors.
-  - Simple search set up for MT samples: MinPresentation
+          - Work with Sunny to generate the NLP outputs for the phenotyping data
-      * Presentation
+          - Evaluate by comparisons with structured data
-      * The interface being developed should present a summary with visualization for patients/notes.
+          - Make the system more robust
-      * We will add Boolean query options to improve the search
+          - Generate a protocol and/or annotation guidelines
-      * We will implement a Ranking algorithm
+          - Share the data as a Gold standard with manually annotated CUIs
-      * Assign fake patient ID's to the notes to generate the visualization portion.
+          - Alex's script is to be tried on different datasets and evaluated across notes from different institutions
-      * Generate a program like Circe to define the patient cohort
+          - Identify minimal set of notes to work with when recommending to the OHDSI community
-      * Next steps: How to move the data from textual searches stored in a table outside of OMOP to the OMOP?
+          - Identify sets of concepts that are not reliable - negation is a very good example of this idea.
-          * Structured searches from CDW and textual searches can be combined using existing strategies. Jon will share the slides of his presentation on combining data from different searches
+          - Continue discussion of NLP system evaluation across different sites
-          * Run NLP on the ElasticSearch to extract information
+   - The NLP-WG will meet on second Wednesday of every month
-  - Wrappers for cTAKES and Metamap
-  - Report on the WG - Hua will generate and share with the members for comments
-  - The best ways to represent textual data need to be determined
 ===Action Items===
-  - Minimal Model Presentation - Alex
+  - Note-type mapping Presentation - Karthik
-  - Note-type mapping Presentation - Karthink
   - Share existing ontologies from Vanderbilt (Hua) and Regenstrief (Jon)
   - Share strategies for combining data from different searches - Jon
@@ Line 45: / Line 48: @@
   - Improvements to search engine set up using MT samples - Min
   - Textual Data Representation - Discussion
+  - NLP system evaluation across different sites - Discussion