Explainable Framework for Ontology-Based Similarity: A Use Case on SNOMED CT

Area of Science:

• Biomedical Informatics
• Artificial Intelligence in Healthcare
• Clinical Data Management

Background:

• Semantic interoperability in healthcare relies on efficient concept manipulation in ontologies.
• Current ontology similarity functions lack actionable justification, hindering clinical trust and workflow integration.

Purpose of the Study:

• To develop an explainable similarity framework for healthcare ontologies.
• To enhance the trust and utility of similarity computations in clinical workflows.

Main Methods:

• Introduced a novel framework providing similarity scores with named pivots and readable concept-to-pivot paths.
• Combined intrinsic topological specificity and path-length proximity for pivot contribution.
• Utilized SNOMED CT for qualitative case analysis.

Main Results:

• Achieved an internal AUC of 0.998, comparable to classical measures.
• Demonstrated specific pivots for clinically close concept pairs and broader pivots for distant pairs.
• Explainability metrics showed shorter paths, deeper pivots, and higher pivot impact for closely related concepts.

Conclusions:

• The proposed framework offers explainable similarity scores and justifications for clinical data.
• Enhanced explainability improves the trustworthiness and applicability of ontology similarity in healthcare.
• This approach supports semantic interoperability by providing actionable insights into concept relationships.