A similarity measuring service for SNOMED-CT: Structural analysis of concepts in ontology

SNOMED-CT is one of the most comprehensive medical and clinical ontology that has been adopted by several countries. Developed using a tractable dialect of the Description Logic ELH, the development and maintenance of this ontology has benefited from the standard logical reasoning service of subsumption, i.e. checking whether one concept is more general/specific than another concept. Subsumption has proved inevitable in the process of ontology design, but real-world application of such a large-scale ontology deems more demanding. A nonstandard reasoning of measuring similarity between concepts appears to be useful, for instance, to recommend related drugs or treatment cases. This paper reports on the development of the previously introduced "similarity measure" for the Description Logic ELH, describes a Web service application for checking similarity among various medical terms in SNOMED-CT, and presents some preliminary results on a usability study of the similarity measure.