Harnessing Moderate-Sized Language Models for Reliable Patient Data Deidentification in Emergency Department Records: Algorithm Development, Validation, and Implementation Study.
運用中等規模語言模型於急診部門紀錄中進行可靠的病患資料去識別化:演算法開發、驗證與實作研究
JMIR AI 2025-07-03
Genomic language models (gLMs) decode bacterial genomes for improved gene prediction and translation initiation site identification.
基因體語言模型(gLMs)解碼細菌基因體以提升基因預測與轉譯起始位點鑑定
Brief Bioinform 2025-07-03
GeneLM 是用 transformer 架構(DNABERT)打造的基因體語言模型,能像讀語言一樣學 DNA 序列,提升細菌基因預測準確度。它分兩步驟:先找編碼區,再精準定位起始點。比起傳統和其他深度學習工具,GeneLM 在基因辨識和起始點預測都更厲害,顯示語言模型有機會大幅改變基因體註解方式。
相關文章PubMedDOI推理
Clinical decision support using pseudo-notes from multiple streams of EHR data.
利用來自多重 EHR 資料流的偽筆記進行臨床決策支援
NPJ Digit Med 2025-07-03
MEME 是一個把表格型電子病歷資料轉成類似文字「pseudo-notes」的深度學習模型,能善用語言模型做臨床預測。它用 self-attention 結合不同資料領域,預測急診病人結果時表現比傳統方法和 GPT-4 prompt 更好,也能輕鬆適應新型 EHR 資料集。
相關文章PubMedDOI推理
Large language model trained on clinical oncology data predicts cancer progression.
基於臨床腫瘤學資料訓練的大型語言模型可預測癌症進展
NPJ Digit Med 2025-07-03
Retrieval-augmented generation elevates local LLM quality in radiology contrast media consultation.
檢索增強生成(Retrieval-augmented generation)提升放射學顯影劑諮詢中在地大型語言模型(LLM)的品質
NPJ Digit Med 2025-07-03
Jackalope Plus tool for post-coordination, ontology development, and precise mapping in observational health studies.
Jackalope Plus 工具於觀察性健康研究中的後設協調、本體論開發與精確對應
Sci Rep 2025-07-02
Jackalope Plus 是一款新工具,結合 SNOMED CT 後協調和 GPT-4o mini LLM,能更快更準確地把複雜健康資料對應到 OMOP CDM。正確率超過 77.5%,比 Usagi 更好,準確度媲美人工但速度快很多。免費開放研究用途,不會儲存病患資料,未來會持續優化。
相關文章PubMedDOI推理
Construction and Application of Traditional Chinese Medicine Knowledge Graph Based on Large Language Model.
基於大型語言模型的中醫知識圖譜建構與應用
Interdiscip Sci 2025-07-02