Back to 2025 Abstracts
Development of a Novel Agentic Artificial Intelligence Clinical Decision Support System for Craniofacial Surgery: A Proof-of-Concept Study
Berk B. Ozmen
*, Graham S. Schwarz
Department of Plastic Surgery, Cleveland Clinic, Cleveland, OH
Background: Clinical decision support systems have shown promise in various specialties, but their application in craniofacial surgery has been limited by the complexity of decision-making in this field. In this study we aim to develop the first agentic artificial intelligence clinical decision support system designed specifically for craniofacial surgery.
Methods: We developed an artificial intelligence agentic system based on the Enhanced Strategy and Cypher-driven Analysis and Reasoning using Graph of Thoughts (ESCARGOT) framework, integrating large language models with dynamic reasoning capabilities and specialized knowledge graphs. The system was trained on 8,561 full-text open access craniofacial surgery manuscripts published between January 1, 2000, and December 31, 2024. We evaluated the system's performance in providing treatment recommendations, procedure details, and similar case identification. Latency metrics and error rates were recorded to assess technical performance.
Results: Agentic artificial intelligence system successfully generated clinically relevant treatment recommendations, detailed procedural information, and identified similar cases from the literature. The system demonstrated high accuracy with a 0% error rate across all functions. Response times averaged 1.95 seconds for treatment recommendations, 2.96 seconds for procedure details, and 2.46 seconds for similar case identification.
Conclusion: This proof-of-concept study demonstrates the potential of agentic artificial intelligence systems to support clinical decision-making in craniofacial surgery. We named the resultant system CASPER (Craniofacial AI System for Procedure Evaluation and Recommendations). CASPER's ability to process complex surgical literature and generate contextualized recommendations represents a significant advance in surgical decision support technology. Further validation with clinical outcomes and usability studies are needed before widespread implementation.
Back to 2025 Abstracts
