請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/72282
標題: | 基於患者急診電子病歷利用深度學習之檢傷預測系統 A System for Triage Prediction of Emergency Department Based on Electronic Medical Record Using Deep Learning |
作者: | Li-Hung Yao 姚力宏 |
指導教授: | 傅立成(Li-Chen Fu) |
關鍵字: | 急診室,分級檢傷,住院預測,遞歸神經網絡,卷積神經網路,深度學習, Emergency Department,Triage System,Convolutional Neural Network,Recurrent Neural Networks,Hospital Admission,Deep Learning, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 急診室擁擠已經成為世界各地所需面對的公共醫療問題。根據許多國家的研究,急診患者的數量不斷攀升,這導致急診醫護人員面臨了前所未有的擁擠和護理負擔,更將造成患者的延誤治療。為了解決急診擁擠問題,各國積極的發展不同的嚴重程度分類標準,其中最被廣泛使用的是緊急嚴重程度指標(ESI),該方法將患者分為五級,以利護理人員能夠依據分級來做救治的參考。儘管ESI提升了急診治療的效率,但到目前為止,該方法因過度仰賴於護理人員的主觀判斷,很容易將大多數患者分類到ESI的第三級。因此,一個可以幫助醫生準確地分類患者病情的系統是迫切需要的。在此論文中,我們提出了一個基於患者急診電子健康記錄(EMRs)的患者住院預測系統,此系統透過卷積神經網路結合遞歸神經網絡匹配聚焦機制進行分類,比起過去相關研究僅使用純結構化或純文字資料特徵,本系統可以處理更加完整的不同類型資料,並使用在不同語言當中。為了進行驗證,在美國急診醫療調查資料中,使用118,602名患者資料,精準度和AUROC分別為0.83及0.87。測試在745,441名患者的台灣資料當中,精準度和AUROC則可達到0.83和0.88。此外,為了驗證本系統在醫療領域的有效性,我們亦將其應用在其他的醫療結果預測,包含死亡率及是否需要入住ICU,相比較其他的傳統機器學習方法,我們的結果在精準度及AUROC高了3~5%。 Emergency Department (ED) crowding has become an issue of delayed patient treatment and even a public healthcare problem around the world. According to recent research studies of many countries, the increasing number of patients in the emergency department which has led to unprecedented crowding and delays in care. For that reason, triage into five-level Emergency Severity Index (ESI) has become a major method for improving medical priorities in ED. Although the ESI mitigates the process of ED treatment, so far it still heavily relies on the nurse's subjective judgment and is easy to triage most patients to ESI level 3 in current practice. Therefore, a system that can help the doctors to accurately triage a patient's condition is imperative. In this work, we propose a system based on the patients’ ED electronic medical record to predict hospitalizations after assigned procedures in ED are completed. This system used Convolutional Neural Networks combined with Recurrent Neural Networks together with attention mechanism for classification. Compared with past related research that only used structural or textual data features, our system can process more different types of data and thus to more applicable to different countries with different languages. For verification, the data from an open dataset (National Hospital Ambulatory Medical Care Survey) is used which includes 118,602 patient visits of the United States EDs, and the accuracy and AUROC can achieve 0.83 and 0.87, respectively. On the other hand, validation on our dataset that includes 745,441 patient visits in National Taiwan University Hospital, and the accuracy and AUROC can reach 0.83 and 0.88, respectively. Moreover, in order to effectively evaluate the prediction quality of our proposed system, we also applied the model on other clinical outcomes that contain the mortality and the admission of ICU, the results showed that our method is 3~5% higher in accuracy than other common methods, including three traditional machine learning algorithms. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/72282 |
DOI: | 10.6342/NTU202004205 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊網路與多媒體研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-0309202016460700.pdf 目前未授權公開取用 | 3.34 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。