Unified Medical Language System, Medical Subject Headings, and Bibliographic Retrieval

1
Unified Medical Language System, Medical Subject
Headings, and Bibliographic Retrieval
2
UMLS Purpose

• Make it easy for health professionals and
  researchers to retrieve and integrate relevant
  information from disparate automated sources,
  e.g.
• computer-based patient records
• factual databanks
• bibliographic databases and full-text
• expert systems
• Antedated and Anticipated the Web
• ? A Federation of Ontologies?

3
UMLS Focus Conceptual Connections

• Build knowledge sources that can be used by
  intelligent programs to overcome
• disparities in language used by different users
  and in different information sources
• difficulties in identifying which of many
  information sources is relevant
• NOT an Ontology
• Neuraths Boat Approach

4
UMLS Knowledge Sources

• Multi-purpose Tools or Intellectual Middleware
  for System Developers
• Metathesaurus (100 Terminologies)
• 1.1 M concepts
• 5.4 M names (17 Languages)
• 17 M relationships
• SPECIALIST lexicon and lexical programs
• Semantic Network

5
UMLS Metathesaurus Finely Granular Concepts

• Concepts, terms, and attributes from many
  controlled vocabularies
• New inter-source relationships, definitional
  information, use information
• Scope determined by combined scope of source
  vocabularies
• Strict definition of synonymy (Identity)

6
Lessons from UMLS

• Context is Key
• Face Validity a Problem in Many Systems
• Special Meanings
• Mice in MeSH
• Zipfs Law
• Use Meaningless Identifiers
• Let Concepts Persist

7
Mapping Issues

• Few Able to Exploit Semantic Locality
• Direct Navigation Desirable
• Directionality
• From Core Clinical Vocabularies
• To Other Important Terminologies
• Complex Logical Expressions
• Combinatorial Explosion of Mappings

8
The Medical Subject Headings

• First Edition 1960
• Used in Indexing and Cataloging
• MEDLINE/PubMed 14 M citations
• Early Retrieval System - Long Incubation
• NOT an Ontology

9
Evolving Structure

• Headings in a Hierarchy
• Qualifiers
• Supplementals

10
Headings in a Hierarchy

• Main Headings / Descriptors
• Used to Indicate Topics
• The Role of Entry Vocabulary
• Hierarchies
• Arrangement of Descriptors
• Polyhierarchical
• 23,000 Descriptors

11
Qualifiers

• Used to Modify (refine) Descriptors
• Indicate Particular Aspects
• Examples
• Heart/transplantation
• Calcitriol/antagonists inhibitors
• Aspirin/poisoning
• 83 Qualifiers

12
Supplementals

• Added Daily
• Largely Chemicals (some protocols)
• 140,000 Records

13
Observations

• Real Use makes Terminologies Better
• Never Finished
• Perfection Seeking
• Terminology Must Work With System
• URU Criteria Helpful

14
Issues in Vocabulary Maintenance

• Clarity of Expression
• Representational Integrity
• Update

15
Clarity of Expression

• Look at Larger Arena of Discourse
• COX-2 Protein or cox2 protein
• Genetic Translation v. Translation
• Ventilation (AC) v. Ventilation

16
Representational Integrity

• One Meaning - One Place
• More than Non-redundancy
• No Unrecognized Synonymy
• No Unrecognized Expressions

17
Representational Integrity

• Entry Combinations
• Complex Expressions
• Coordination of Headings/Subheadings
• How to Identify and Test
• Each Citation a Meaning
• Description Logic Helpful but Inadequate

18
Citation Maintenance - An Update Model

• Repercussions for MEDLINE/PubMed
• No Retrospective Indexing
• Year End Processing
• Automatic Tasks
• Manual Tasks
• Now Published on Web

19
Bibliographic Retrieval

• Protein Policy
• MeSH Mapping File
• Gene RIFs

20
Protein Policy

• Specific Genes Not To be Added
• Proteins by Species
• Selected Species
• Linked to Protein Classes
• Example - ECMC2 protein, S cerevisiae
• v. ECMC2 protein, human

21
MeSH Mapping File

• Licensed UMLS Content
• Asserted Synonyms from Other Sources
• Adding and Removing Synonyms
• Behind the Scenes in PubMed

22
Gene RIFs

• Part of Entrez Gene
• Literature References
• Direct Links to PubMed
• Added at Time of Indexing

23
Godel, Escher, and Bach

• Completeness Impossible
• Lonergans Hypothesis
• An Escher Drawing
• Principles are Only Local
• A Well-Tempered Clavier
• Purpose Motivates Development

Apologies to Douglas Hofstadter
24
Choices in Terminology

• Want a Standard
• Address Formalization Gap - Blois
• Understandable by Humans
• Reproducible for Machines
• Appropriate for Intended Use
• Update Model

25
Thank you!