Assessing Retrieval-Augmented Large Language Model Performance in Emergency Department ICD-10-CM Coding Compared to Human Coders.
評估檢索增強大型語言模型在急診部門 ICD-10-CM 編碼中的表現,與人類編碼員相比。
medRxiv 2024-11-01
這項研究探討增強檢索生成(RAG)的大型語言模型(LLMs)在急診科臨床紀錄中生成ICD-10-CM代碼的有效性,並與醫療提供者進行比較。研究基於Mount Sinai Health System的500次急診就診數據,發現RAG增強的LLMs在準確性和特異性上均優於醫療提供者,且GPT-4的表現尤為突出。即使是較小的模型如Llama-3.1-70B,經過RAG後也顯示出顯著提升。這顯示生成式人工智慧在改善醫療編碼準確性及減少行政負擔方面的潛力。
PubMedDOI
Improving large language model applications in biomedicine with retrieval-augmented generation: a systematic review, meta-analysis, and clinical development guidelines.
利用檢索增強生成技術改善生物醫學中的大型語言模型應用:系統性回顧、統合分析及臨床發展指導方針。
J Am Med Inform Assoc 2025-01-15
Retrieval augmented generation for 10 large language models and its generalizability in assessing medical fitness.
10 種大型語言模型的檢索增強生成及其在評估醫療適應性中的普遍性。
NPJ Digit Med 2025-04-04
Comparative Analysis of ChatGPT-4 for Automated Mapping of Local Medical Terminologies to SNOMED CT.
ChatGPT-4 用於自動對應在地醫學術語至 SNOMED CT 的比較分析
Stud Health Technol Inform 2025-05-17