An evaluation framework for ambient digital scribing tools in clinical applications.
臨床應用中環境型數位紀錄工具的評估框架
NPJ Digit Med 2025-06-13
這項研究提出一套完整評估架構,專門檢測 AI 驅動的環境數位紀錄工具,協助醫師減輕紀錄負擔。架構結合人工和自動化評分,評估轉錄、說話者辨識及病歷品質。實測顯示工具在流暢度和清晰度表現佳,但事實正確性和新藥物紀錄仍有待加強,顯示導入前需嚴謹評估與治理。
PubMedDOI♡
站上相關主題文章列表
Assessing the Effectiveness of Automatic Speech Recognition Technology in Emergency Medicine Settings: A Comparative Study of Four AI-powered Engines.
評估自動語音識別技術在急診醫學環境中的有效性:四個人工智慧驅動引擎的比較研究。
Res Sq 2024-08-26
Evaluation Framework of Large Language Models in Medical Documentation: Development and Usability Study.
大型語言模型在醫療文檔中的評估框架:開發與可用性研究。
J Med Internet Res 2024-11-20
這項研究旨在開發和驗證一個框架,以評估大型語言模型(LLM)生成的急診室記錄的準確性和臨床適用性。研究中,52名參與者使用HyperCLOVA X LLM創建了33份記錄,並採用雙重評估方法進行分析。臨床評估顯示評估者之間的可靠性高,而定量評估則識別出七種主要錯誤類型,其中無效生成錯誤最為常見。研究結果顯示該框架在臨床可接受性上具備潛力,為未來的研究和應用提供了方向。
PubMedDOI
Ambient artificial intelligence scribes: utilization and impact on documentation time.
環境人工智慧書記:在文檔時間上的利用與影響。
J Am Med Inform Assoc 2024-12-17
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
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
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