Using automated methods to identify connections across biomedical terminologies

2009 2009

Other formats: Order a copy

Abstract (summary)

Terminologies play important roles in the annotation, organization, sharing and retrieval of the biomedical data. Several biomedical informatics applications such as information retrieval, decision support and knowledge discovery depend on identification of mappings, translations, or links (collectively referred to as connections) between different terminologies. Consider, for example, the connection between Hb (SNOMED CT) and Hemoglobin (LOINC) for terminology mapping application, or between Hb (SNOMED CT) and Anemia (MeSH) for an information retrieval application. Identification of such cross-terminology connections for a given application or task is a challenging problem. Existing research methods to identify connections are based on either manual selection, or problem-specific algorithms that are labor intensive, difficult to maintain, or unscalable beyond their original domains.

Integrated Biomedical Terminology Resources (IBTRs) such the Unified Medical Language System and Open Biomedical Ontologies provide rich knowledge sources that can be potentially used towards connection identification. In this research, we propose a novel machine learning-based approach that uses the semantic and structural features of an IBTR to identify connections across biomedical terminologies. First, we model the semantic features in the IBTR based on the hierarchical organization of concepts and the set of associative relationships asserted between indirect, transitively related concept pairs. Second, we analyze the structural topology of the IBTR using network theoretic methods such as the scale-free property, clustering coefficient and topological overlap in the IBTRs to identify connections. Finally, the semantic and structural properties of the IBTR are combined into an integrated model for connection identification. A web-based tool, TermLink is developed based on the proposed methods to perform a cost-benefit analysis and disseminate the research to the wider community.

We evaluated the proposed methods across eight training datasets corresponding to the connection identification tasks such as terminology mapping, and information retrieval. The results indicate that the semantic properties provide high classification accuracy (80–90%) across all the training datasets. The structural analysis of IBTR revealed new classes of concepts such as informational and noisy hubs that provide strong cues towards connection identification. The integrated semantic and structural approach was found to be a more effective approach than each of the individual approaches in terms of classification accuracy and computational time. The cost-benefit analysis revealed that the TermLink tool could potentially reduce time for the manual approach to identify connections.

In conclusion, the existing knowledge in the IBTRs can be modeled using semantic and structural features to identify connections for different applications. The proposed approach provides a robust, automated and problem independent method that can be combined with manual methods to reduce the time and cost in identifying connections across biomedical terminologies.

Indexing (details)

Computer science
0715: Bioinformatics
0984: Computer science
Identifier / keyword
Applied sciences; Biological sciences; Biomedical terminology; Machine learning; Ontology mapping
Using automated methods to identify connections across biomedical terminologies
Patel, Chintan
Number of pages
Publication year
Degree date
School code
DAI-B 71/02, Dissertation Abstracts International
Place of publication
Ann Arbor
Country of publication
United States
Cimino, James
Columbia University
University location
United States -- New York
Source type
Dissertations & Theses
Document type
Dissertation/thesis number
ProQuest document ID
Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.
Document URL
Access the complete full text

You can get the full text of this document if it is part of your institution's ProQuest subscription.

Try one of the following:

  • Connect to ProQuest through your library network and search for the document from there.
  • Request the document from your library.
  • Go to the ProQuest login page and enter a ProQuest or My Research username / password.