Combining expert knowledge and knowledge automatically acquired from electronic data sources for continued ontology...

Area of Science:

• Biomedical Informatics
• Ontology Engineering
• Medical Knowledge Representation

Background:

• Evaluating biomedical ontologies often requires extensive domain expert input for gold standard creation, posing a significant bottleneck.
• This study addresses the challenge of knowledge acquisition in ontology evaluation.

Purpose of the Study:

• To present a novel semi-automated method for evaluating the concept coverage and accuracy of biomedical ontologies.
• To minimize reliance on expensive domain expertise for gold standard generation by integrating diverse knowledge sources.

Main Methods:

• Developed a bacterial clinical infectious diseases ontology (BCIDO) for clinical decision support.
• Employed a semi-automated method integrating clinical practice guidelines, electronic health records, and expert scenarios to create a knowledge compendium.
• Used the compendium to evaluate BCIDO's accuracy and coverage.

Main Results:

• BCIDO contains 593 concepts and 2345 relationships.
• The semi-automated method generated a knowledge compendium with 637 concepts and 1554 relationships.
• BCIDO achieved 79% concept and 89% relationship coverage of the compendium, with variations across categories (e.g., 92% for antibiotics, 72% for bacteria).

Conclusions:

• The semi-automated method is cost-effective for creating a knowledge compendium with minimal expert input.
• This approach facilitates the continued development and evaluation of biomedical ontologies, enhancing their accuracy and coverage.