Gruppe 3: NCBO Annotator

Anleitungen

Einleitung

The NCBO (National Center for Biomedical Ontology) Annotator, from Stanford University, is an automatic annotator of clinical texts. You provide the system with text, and the system tries to code it for you. The NCBO Annotator can use many different ontologies and terminologies to code text, including SNOMED and ICD.

Task 1

The NCBO can use many different terminologies to code text. Your task is to enter some text and make the system tag it using a number of different ontologies.

Click this link to open the NCBO Annotator (opens in a new window):

http://bioportal.bioontology.org/annotator

The annotator looks like the following:

NCBONew

To choose an ontology, just start typing in the Select Ontologies field. For example, to choose SNOMED just start typing:

NCBOSelectOntology

Or you can click select from list to see all the ontologies available:

NCBOList

Enter the following example text and use SNOMED CT to code it:

Joint pain. Muscle pain, even her skin. Patient was tired yet could not sleep. She had frequent migraine headaches. She had irritable bowel syndrome. She was severely depressed. She had fibromyalgia, anemia, endometriosis.

Now try with the sample text: Click the insert sample text link to insert this text into the text field.

NCBOSampletext

Perform a search of the sample text using SNOMED CT. Leave the Select UMLS Semantic Types field blank for now. The annotator should look like the screenshot below:

NCBOICD

Click Get Annotations and view the results.

Now we will search for specific semantic types. In the Select UMLS Semantic Types field, start typing Neoplastic Process.

NCBONeoplasms

Click Get Annotations again. Observe the results.

Task 2

Now we will use ICD to code some text. Delete SNOMEDCT from the Select Ontologies field and start typing ICD. Select ICD10 from the drop down list. Remove any semantic types from the Select UMLS Semantic Types field.

Use ICD to code this text again:

Joint pain. Muscle pain, even her skin. Patient was tired yet could not sleep. She had frequent migraine headaches. She had irritable bowel syndrome. She was severely depressed. She had fibromyalgia, anemia, endometriosis.

Task 3: Beurteilung

You will be graded according to your answers. A maximum of 5 points are available.

  1. Download the questions: g3-antwortbogen-en.docx
  2. Answer the questions as best you can.
  3. Send your answers by email.

Task 4: Presentation

Present to the rest of the class how the NCBO can be used to tag a medical text. Use two different ontologies for the same text and discuss the difference. Discuss why automatic coding is hard (Hint. Think for example of the following text: No evidence of muscle pain or irritable bowel syndrome. Hint: ICD will find irritable bowel syndrome (ICD code K58), but the text said no evidence of...). Show how you can visualise a search term (see Visualise a search term below.)

Visualise a search term

To visualise search terms:

  1. Use SNOMED to annotate/code some text (for example, the sample text: Melanoma is a malignant tumor of melanocytes which are found predominantly in skin but also in the bowel and the eye.).
  2. In the list of annotations, click a term such as Eye Structure.
  3. Click the Visualisation tab to see where this term lies in the SNOMED hierarchy.
Visualise

References/Resources

Sample text from The New York Times' Diagnosis column: http://www.nytimes.com/2011/11/13/magazine/diagnosis-hurt-all-over.html

Return to list of groups