Identifying Missing Hierarchical Relations in SNOMED CT from Logical Definitions Based on the Lexical Features of...

Area of Science:

• Medical Informatics
• Computational Linguistics
• Ontology Engineering

Background:

• SNOMED CT is a comprehensive clinical terminology essential for healthcare data interoperability.
• Ensuring the completeness and accuracy of SNOMED CT hierarchies is crucial for reliable clinical decision support.
• Identifying missing hierarchical relations can improve the semantic richness and usability of SNOMED CT.

Purpose of the Study:

• To automatically identify potentially missing hierarchical (subClassOf) relations within SNOMED CT.
• To leverage lexical features of concept names for inferring ontological relationships.
• To propose a scalable method for SNOMED CT quality assurance and enrichment.

Main Methods:

• Logical definitions were created from concept names' lexical features, represented in OWL EL.
• The ELK reasoner was employed to infer hierarchical relations based on these definitions.
• The inferred hierarchy was compared against the original SNOMED CT hierarchy, with manual review for validation.

Main Results:

• The approach was applied to 15,833 disorder and procedure concepts.
• 559 potentially missing hierarchical relations were identified.
• A high validation rate of 78% was achieved for the identified relations.

Conclusions:

• A lexical approach can effectively identify missing hierarchical relations in SNOMED CT.
• This method offers an easy-to-implement, efficient, and scalable solution for quality assurance.
• The findings contribute to improving the completeness and accuracy of SNOMED CT.