<< Back to the abstract archive
Development of a Novel Artificial Intelligence RAG-Enhanced Large Language Model for Microsurgery Clinical Decision Support
Berk B Ozmen, MD; Nishant Singh, MSc; Kavach Shah, MSc; Ibrahim Berber, MSc; Eugene Pinsky, PhD; Graham S Schwarz, MD, MSE, FACS
Cleveland Clinic Department of Plastic Surgery
2025-01-10
Presenter: Berk B Ozmen, MD
Affidavit:
Yes
Director Name: Graham S Schwarz, MD, MSE, FACS
Author Category: Fellow Plastic Surgery
Presentation Category: Clinical
Abstract Category: General Reconstruction
Background: Microsurgical decision-making requires extensive knowledge integration and real-time analysis of complex clinical scenarios. Despite advances in artificial intelligence, current decision support tools lack the ability to process unstructured surgical knowledge and provide context-aware recommendations grounded in peer-reviewed literature.
Methods: We developed a surgical decision support system combining Retrieval-Augmented Generation (RAG) with GPT-4o large language models. The system's knowledge base comprises 4,876 full-text microsurgical papers from PubMed Central, processed through a hierarchical document processing pipeline employing the RAPTOR methodology. System evaluation was conducted using 10 standardized microsurgical queries covering key aspects of reconstructive surgery, including free flap surgery, vascular complications, and technical decision-making.
Results: The system successfully processed and responded to complex microsurgical queries across multiple domains. Responses demonstrated accurate retrieval of relevant surgical literature and generation of clinically appropriate recommendations, with particular strength in technical guidance and evidence-based management protocols. All recommendations maintained clear traceability to peer-reviewed sources, ensuring accountability and verification.
Conclusion: Our RAG-enhanced system provides evidence-based, context-aware decision support for microsurgical applications, offering potential benefits for both clinical practice and surgical education. The successful integration of large language models with peer-reviewed surgical literature represents a significant advance in surgical decision support systems.
