- 學位論文
基於機器學習的多中心菌血症預測模型
Machine Learning Models for Bacteremia Prediction in Multicenter Study
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
菌血症為血液中存在細菌的狀況。菌血症的患者若缺乏及時的治療可能會進一步引發敗血症，嚴重者甚至導致休克。不幸的是，除了發燒以外，菌血症通常沒有任何其他明顯症狀，因此難以與其他病症區分。目前為止，血液培養是唯一診斷菌血症的方法。然而，此方法通常需要等待大約24-48小時才能得知結果，因此可能無法及時給予患者治療而導致後續更為嚴重的症狀。近年來，由於機器學習用於醫療數據的分析變得越來越普及，許多醫院使用機器學習來協助醫師進行診斷，讓病患得到更及時的治療。本論文的目標是透過機器學習來建立多中心之菌血症急診患者的預測模型，本研究使用了國立臺灣大學醫學院附設醫院 (NTUH) 的病人資料，其中涵蓋了台北、新竹和雲林分院自2009年至2018年的資料集。本論文主要分為三個部分，於每部都會對預測模型進行詳細的分析。 本研究比較在靜態環境中機器學習模型的預測能力，以及探討模型更新的方法。在第一部分中，本研究合併了三個數據集來訓練基準模型，然後將其與分開訓練的資料集進行比較。於第二部分中，本研究探討模型更新的方法，以進一步了解特徵重要性在不同年份的變化，並總結出相對有效的更新方法。 總結而言，本論文探索了機器學習模型在靜態環境中對菌血症的預測能力，以及更新模型的方法。透過文中機器學習模型的探討，可以進一步減少患者等待治療的時間、提高病患的存活率、同時也可以奠定未來對於菌血症研究的堅固基礎。
並列摘要
Bacteremia is a serious and potentially life-threatening condition caused by the presence of bacteria in the bloodstream. It is a type of blood stream infection that can lead to sepsis, which is potentially fatal if not treated promptly. Unfortunately, aside from fever, there are typically no other specific symptoms that can be used to identify the occurrence of bacteremia. This makes diagnosis reliant on blood culture, a process that can take up to 24-48 hours to produce results. In this thesis, we aim to develop machine learning models for predicting multicenter bacteremia in emergency department patients. The study utilizes 10 years of data collected from National Taiwan University Hospital (NTUH) branches in Taipei, Hsinchu, and Yunlin. The research is divided into two parts to conduct a detailed investigation of the development of these models. Our goal is to investigate the predictive power of machine learning models in a static environment and the best practices for updating these models on a yearly basis. The experiments conducted in this thesis contribute to a deeper understanding of the nature of bacteremia, leading to a reduction in the waiting time for patients to receive appropriate treatment. By exploring the predictive power of machine learning models in a static environment and best practices for updating these models on a yearly basis, this study provides insights that can ultimately improve patient outcomes.
參考文獻
L. M. Bush. Bacteremia - Infectious Diseases. [Online]. Available: https://www.msdmanuals.com/professional/infectious-diseases/biology-of-infectious-disease/bacteremia
J. H. P. Dong Hyun Choi, Ki Jeong Hong, “Prediction of bacteremia at the emergency department during triage and disposition stages using machine learning models,” Am J Emerg Med, pp. 53:86–93, 2022 Mar.
L. J. Vorvick. Blood culture. [Online]. Available: https://www.mountsinai.org/health-library/tests/blood-culture
M. P. Weinstein, M. L. Towns, S. M. Quartey, S. Mirrett, L. G. Reimer, G. Parmigiani, and L. Barth Reller, “The Clinical Significance of Positive Blood Cultures in the 1990s: A Prospective Comprehensive Evaluation of the Microbiology, Epidemiology, and Outcome of Bacteremia and Fungemia in Adults,” Clinical Infectious Diseases, vol. 24, no. 4, pp. 584–602, 04 1997. [Online]. Available: https://doi.org/10.1093/clind/24.4.584
S. B. W. Nathan I Shapiro, Richard E Wolfe, “Who needs a blood culture? a prospectively derived and validated prediction rule,” The Journal of emergency medicine, pp. 35(3):255–264, 2008.