Natural Language Processing and Large Language Model-Based Automated AO/OTA Classification of Proximal Femoral Fractures: A MIMIC-IV Retrospective Study
Abstract
Abstract Background Accurate AO/OTA classification of proximal femoral fractures is a critical step in surgical decision-making and prognostic evaluation; however, the complexity of the classification system and the heterogeneity of terminology in radiology reports pose significant challenges for large-scale, standardized application of the classification. In recent years, with the advancement of AI technologies, large language models (LLMs) equipped with retrieval-augmented generation (RAG) capabilities have offered new possibilities for automatically extracting fracture classification information from unstructured radiology reports. Methods This study conducted a retrospective analysis of 23,179 radiology reports from the MIMIC-IV database. Through a multi-stage annotation process involving independent labeling by two orthopedic surgeons and adjudication by a senior expert, a total of 157 acute cases of AO/OTA type 31A (intertrochanteric fractures) and type 31B (femoral neck fractures) were ultimately included. We evaluated three experimental models: zero‑shot prompting, few‑shot learning (5 examples per class), and RAG‑augmented prompting (incorporating a structured AO classification knowledge base). Classification performance was assessed using accuracy, F1‑score, and confusion matrices. All type 31C (femoral head fractures) cases underwent expert re‑adjudication to evaluate annotation quality. Results The RAG‑LLM framework achieved an accuracy of 94.29% (F1 = 94.29%) in…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.