法學期刊
  • 社群分享
論著名稱:
利用機器學習精準搜尋並智慧分案以提升癌登個案篩選效能(Enhancing Cancer Case Screening Efficiency through Machine Learning for Accurate Search and Intelligent Allocation)
文獻引用
編著譯者: 黃圓婷沈怡妏李佳鴻游淑蓉劉曄霞李季樺黃志仁
出版日期: 2023.11
刊登出處: 台灣/醫療品質雜誌第 17 卷 第 6 期/36-42 頁
頁  數: 7 點閱次數: 20
下載點數: 28 點 銷售明細: 權利金查詢 變更售價
授 權 者: 財團法人醫院評鑑暨醫療品質策進會 授權者指定不分配權利金給作者)
關 鍵 詞: 癌症癌症登記篩選預測機器學習
中文摘要: 目的:癌症登記資料庫是癌症醫療品質改善的實證根本,目前依賴著人工逐筆檢視篩選,但符合申報條件僅佔 50.4%。希冀透過機器學習自然語言處理擷取病歷資訊等關鍵字,能更精準地篩選出需申報的癌症個案並同時正確分類癌別。
材料與方法:利用南部某醫學中心 2017 年及 2018 年已分類的 3,000 筆個案含 21,994 份病歷資料、影像報告及病理報告進行機器訓練學習。利用多元分類模型(ML.NET Multiclass Classification SDCA Maximum Entropy),並依 30 癌別進行關鍵字標註,建立智慧系統預測模組。
結果:篩選結果分為「需申報」、「不需申報」、「疑似個案」三組。智慧系統預測個案申報平均正確率為 89.7%及癌別分類平均正確率為 89.5%。
結論:智慧預測系統協助癌登個案篩選以提升篩選效能,讓癌症登記師專注於摘錄資料的完整性及正確性,未來期可導入圖文辨識,強化預測系統判讀能力,提供各臨床團隊更高的分析價值。
英文關鍵詞: Cancercancer registryscreening predictionmachine learning
英文摘要: Objective: Cancer registration registries serve as the empirical foundation for improving the quality of cancer care. Unlike current methods, which rely on manual review and screening and yield only a 50.4% reporting eligibility, this study leverages machine learning and natural language processing to extract key medical record information, thus enhancing the precision in selecting cases for reporting and in classifying cancer types.
Materials and Methods: The study utilized 3,000 categorized cases from 2017 and 2018, accompanied by 21,994 medical records, imaging reports, and pathology reports from a medical center in southern Taiwan, for machine learning training. A multiclass classification model, ML.NET Multiclass Classification SDCA Maximum Entropy, was employed, and keywords were annotated for 30 types of cancer to construct a smart prediction module.
Results: The screening results were categorized into three groups: “to be reported”, “not to be reported”, and “suspected cases.” The intelligent system achieved an average accuracy rate of 89.7% in case reporting and 89.5% in cancer-type classification.
Conclusion: This smart predictive system enhances the efficiency of cancer case screening, allowing registry staff to focus on the completeness and accuracy of data extraction. Future iterations could incorporate image and text recognition to strengthen the predictive capabilities of the system, thereby providing higher analytical value to clinical teams.
目  次: 前言
材料與方法
開發機器預測系統
研究結果和討論
結論
參考文獻
相關法條:
相關判解:
相關函釋:
相關論著:
黃圓婷、沈怡妏、李佳鴻、游淑蓉、劉曄霞、李季樺、黃志仁,利用機器學習精準搜尋並智慧分案以提升癌登個案篩選效能,醫療品質雜誌,第 17 卷 第 6 期,36-42 頁,2023年11月。
返回功能列