Automatic full conversion of clinical terms into SNOMED CT concepts

Area of Science:

• Medical Informatics
• Natural Language Processing
• Clinical Terminology

Background:

• SNOMED CT is a comprehensive clinical ontology crucial for automated reasoning.
• Existing methods lack the ability to convert clinical terms into novel, post-coordinated SNOMED CT concepts.
• Automated reasoning over clinical text requires effective clinical term-to-concept mapping.

Purpose of the Study:

• To present a novel, complete method for converting clinical terms into SNOMED CT concepts.
• To enable the creation of post-coordinated concepts not currently present in SNOMED CT.
• To facilitate automated reasoning over clinical text by improving concept mapping.

Main Methods:

• The method identifies defining relations within clinical terms to map them to SNOMED CT concepts.
• It learns solely from existing SNOMED CT term-concept pairs, requiring no manual annotations.
• The approach handles both existing and novel (post-coordinated) clinical concepts.

Main Results:

• The study proposes the first complete conversion method for clinical terms to SNOMED CT concepts.
• The method successfully generates post-coordinated concepts beyond the existing SNOMED CT hierarchy.
• Evaluation on large-scale and small-scale datasets demonstrates the method's efficacy.

Conclusions:

• This novel method significantly advances automated reasoning capabilities in clinical text analysis.
• It provides a scalable solution for mapping diverse clinical terminology to SNOMED CT.
• The approach unlocks the full potential of SNOMED CT for clinical informatics applications.