Artificial Intelligence Scribe and Large Language Model Technology in Healthcare Documentation: Advantages, Limitations, and Recommendations.
人工智慧書記與大型語言模型技術在醫療文件中的應用:優勢、限制與建議。
Plast Reconstr Surg Glob Open 2025-01-17
Application of unified health large language model evaluation framework to In-Basket message replies: bridging qualitative and quantitative assessments.
統一健康大型語言模型評估框架在 In-Basket 訊息回覆中的應用:橋接定性與定量評估。
J Am Med Inform Assoc 2025-03-10
Automating Evaluation of AI Text Generation in Healthcare with a Large Language Model (LLM)-as-a-Judge.
以大型語言模型(LLM)作為評審自動化評估醫療領域 AI 文字生成
medRxiv 2025-05-02
Development and validation of the provider documentation summarization quality instrument for large language models.
大型語言模型醫療提供者文件摘要品質評估工具的開發與驗證
J Am Med Inform Assoc 2025-05-05
A framework to assess clinical safety and hallucination rates of LLMs for medical text summarisation.
用於醫學文本摘要之大型語言模型(LLMs)臨床安全性與幻覺率評估框架
NPJ Digit Med 2025-05-13
Assessment and Integration of Large Language Models for Automated Electronic Health Record Documentation in Emergency Medical Services.
大型語言模型於緊急醫療服務自動化電子病歷紀錄之評估與整合
J Med Syst 2025-05-17
Evaluating the Application of Artificial Intelligence and Ambient Listening to Generate Medical Notes in Vitreoretinal Clinic Encounters.
人工智慧與環境聆聽技術於玻璃體視網膜門診紀錄生成之應用評估
Clin Ophthalmol 2025-06-10
這項研究比較了 ChatGPT 3.5 和 Google Gemini 1.0 Pro 在生成視網膜門診紀錄的表現。結果顯示,ChatGPT 3.5 不論在轉錄準確度還是紀錄品質都優於 Gemini。不過,兩者偶爾還是會出現資訊錯誤。整體來說,這類 AI 有助減輕醫師紀錄負擔,但還需要再優化才能更安心使用。
PubMedDOI