用于心衰患者重入院预测的LSTM模型  

LSTM-BASED READMISSION PREDICTION OF PATIENTS WITH HEART FAILURE

作　　者：李臻[1,2] 陈若愚[1,2] 鲁兴华 刘秀磊[1] Li Zhen;Chen Ruoyu;Lu Xinghua;Liu Xiulei(Laboratory of Data Science and Information Studies,Beijing Information Science and Technology University,Beijing 100101,China;Beijing Key Laboratory of Internet Culture and Digital Dissemination Research,Beijing Information Science and Technology University,Beijing 100101,China;Department of Biomedical Informatics,University of Pittsburgh,Pittsburgh 15206,Pennsylvania,USA)

机构地区：[1]北京信息科技大学数据科学与情报分析实验室,北京100101 [2]北京信息科技大学网络文化与数字传播北京市重点实验室,北京100101 [3]匹兹堡大学生物医学信息学系,宾夕法尼亚州匹兹堡15206

出　　处：《计算机应用与软件》2023年第6期343-349,共7页Computer Applications and Software

基　　金：国家重点研发计划项目(2017YFB1400402);网络文化与数字传播北京市重点实验室开放课题项目(ICDDXN006);北京信息科技大学“勤信人才”培育计划项目(QXTCP C202111)。

摘　　要：电子健康记录(EHR)中蕴含着丰富的语义信息,目前对EHR的数据挖掘主要基于传统机器学习方法,涉及大量人工特征工程。但由于EHR数据存在维度大、时间跨度长等特点,传统机器学习方法较难有效地捕捉EHR中的深层语义信息。使用LSTM模型对心力衰竭患者的EHR数据进行建模和训练,以30天内重入院作为预测目标开展研究。实验结果表明,LSTM模型能够有效地捕捉EHR时序数据中的语义信息,与传统机器学习方法相比,在ROC-AUC指标上提升了10.48百分点。Electronic Health Records(EHR)contain rich semantic information.Currently,the mining of EHR is mainly based on traditional machine learning methods and involves lots of manual feature engineering.However,due to the high dimensionality and long span characteristic of EHR,it is difficult for traditional machine learning methods to effectively capture the deep semantic information contained in EHR.LSTM was used to model and train the EHR data of patients with heart failure.The research was carried out on 30-day readmission prediction.The experimental results show that LSTM model can effectively capture the semantic information in the time series of EHR.Compared with the traditional machine learning methods,the ROC-AUC of the proposed method is improved by at least 10.48%.

关 键 词：电子健康记录 深度学习 循环神经网络 心力衰竭 重入院预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP319[自动化与计算机技术—控制科学与工程]