A Semi-automated Method for Domain-Specific Ontology Creation from Medical Guidelines

Abstract

The automated capturing and summarization of medical consultations has the potential to reduce the administrative burden in healthcare. Consultations are structured conversations that broadly follow a guideline with a systematic examination of predefined observations and symptoms to diagnose and treat well-defined medical conditions. A key component in automated conversation summarization is the matching of the knowledge graph of the consultation transcript with a medical domain ontology for the interpretation of the consultation conversation. Existing general medical ontologies such as SNOMED CT provide a taxonomic view on the terminology, but they do not capture the essence of the guidelines that define consultations. As part of our research on medical conversation summarization, this paper puts forward a semi-automated method for generating an ontological representation of a medical guideline. The method, which takes as input the well-known SNOMED CT nomenclature and a medical guideline, maps the guidelines to a so-called Medical Guideline Ontology (MGO), a machine-processable version of the guideline that can be used for interpreting the conversation during a consultation. We illustrate our approach by discussing the creation of an MGO of the medical condition of ear canal inflammation (Otitis Externa) given the corresponding guideline from a Dutch medical authority.