基于LSTM的冠心病危险因素抽取方法研究  

Study on the Method of Extracting Risk Factors of Coronary Heart Disease Based on LSTM

作　　者：杨晓辉 毕雪华[2] 陈曲[3] 高颖 张琳琳[4] YANG Xiao-hui;BI Xue-hua;CHEN Qu(School of Information,Xinjiang University,Urumqi 830046,Xinjiang Uygur Autonomous Region,P.R.C.)

机构地区：[1]新疆大学软件学院,新疆维吾尔自治区乌鲁木齐市830046 [2]新疆医科大学医学工程技术学院,新疆维吾尔自治区乌鲁木齐市830011 [3]新疆医科大学第一附属医院,新疆维吾尔自治区乌鲁木齐市830054 [4]新疆大学信息学院,新疆维吾尔自治区乌鲁木齐市830046

出　　处：《中国数字医学》2020年第3期92-94,共3页China Digital Medicine

基　　金：新疆维吾尔自治区自然科学基金(编号:2017D01C232)。

摘　　要：电子病历中冠心病危险因素是临床医生对患者进行风险评估的基础和关键。针对非结构化的中文电子病历中冠心病危险因素自动化抽取的实际需求,将危险因素分为两类,分别使用长短期记忆网络LSTM与规则的方法,并对500例临床科室住院患者的电子病历进行实验测试,正确率达85.1%,验证了方法的可行性。The risk factors for coronary heart disease in electronic medical records are the basis and key for clinicians to assess patients’risk.In view of the actual needs of automated extraction of risk factors for coronary heart disease in unstructured Chinese electronic medical records,risk factors were divided into two categories,using long-and short-term memory network LSTM and regular methods,and electronic medical records of 500 clinical inpatients.The experimental test was carried out with a correct rate of 85.1%,which verified the feasibility of the method.

关 键 词：冠心病 危险因素 LSTM 规则 电子病历 

分 类 号：R319[医药卫生—基础医学] TP391[自动化与计算机技术—计算机应用技术]