Large language models (LLMs) integrated with Retrieval-Augmented Generation (RAG) can enhance clinical decision support and triage. However, semantic retrieval often fails to capture the structured relationships of medical knowledge, especially in complex scenarios. This study explores the use of ontology-based knowledge graphs to improve retrieval, combining SNOMED CT and rheumatology guidelines within a graph database. Retrieval was evaluated on clinical queries using three configurations: embeddings-only, knowledge graph-only, and a hybrid approach. Results indicate that while embeddings suffice for simple cases, ontology-based knowledge graphs are crucial for complex reasoning.
Ontology-Enriched Guidelines Retrieval for Complex Rheumatological Cases
Buonocore, Tommaso Mario;Marino, Sara;Albi, Giuseppe;Sakellariou, Garifallia;Dagliati, Arianna;Parimbelli, Enea;Sacchi, Lucia
2026-01-01
Abstract
Large language models (LLMs) integrated with Retrieval-Augmented Generation (RAG) can enhance clinical decision support and triage. However, semantic retrieval often fails to capture the structured relationships of medical knowledge, especially in complex scenarios. This study explores the use of ontology-based knowledge graphs to improve retrieval, combining SNOMED CT and rheumatology guidelines within a graph database. Retrieval was evaluated on clinical queries using three configurations: embeddings-only, knowledge graph-only, and a hybrid approach. Results indicate that while embeddings suffice for simple cases, ontology-based knowledge graphs are crucial for complex reasoning.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
