基于BP神经网络的老年男性保健人群缺血性心脑血管病预测模型研究  

Construction and evaluation of predictive model for ischemic cardiovascular diseases of senior men based on BP neural network

作　　者：陈金宏[1] 吴海云[2] 何耀[2] 秦银河[2] 

机构地区：[1]武警总医院医务部,北京100039 [2]解放军总医院老年医学研究所,北京100853

出　　处：《第三军医大学学报》2011年第8期797-799,共3页Journal of Third Military Medical University

基　　金：中央保健专项基金(06H050);"十一五"国家科技支撑计划(2009BAI86B01);全军医学科学技术研究"十一五"计划课题(06L037);首都医学发展基金重点项目(2007-2039)~~

摘　　要：目的建立基于BP神经网络的老年男性保健人群缺血性心脑血管病(ischemic cardiovascular disease,ICVD)的发病预测模型。方法以某保健医院所有老年男性保健对象2003年5月的体检资料为基线数据,剔除基线前已患有ICVD者。收集历年住院资料、问卷调查资料、电话回访资料,观察截止到2010年10月的ICVD发病及死亡情况。按照4∶1的比例将基线人群划分为训练组和测试组,利用MATLAB 7.0软件的神经网络工具箱构建训练组的BP神经网络模型,用测试组加以验证,并与传统的COX比例风险回归模型进行比较。结果 BP人工神经网络模型的判别能力、群体水平的预测能力方面优于COX比例风险回归模型。结论利用BP神经网络进行疾病预测,能获得更好的预测效果,并且对资料的类型、分布不作任何要求,是一种较好的疾病预测方法。Objective To construct and evaluate a predictive model for ischemic cardiovascular diseases（ICVDs） of senior men based on back-propagation（BP） neural network.Methods The physical examination data of senior men excluding ICVD patients,recorded in a health-care hospital in May 2003,were used as baseline data.The hospitalization data,questionnaire data,and telephone follow-up survey data of these senior men up until Oct.2010 were collected to analyze ICVD morbidity and ICVD-caused death rate.The baseline population was divided into a training group and a test group at a ratio of 4∶1.The data set of the training group was employed to construct a BP neural network predictive model with the neural network toolbox of MATLAB 7.0,as well as a conventional COX proportional-hazards regression model.The two models were assessed with the data set of the test group and were then compared.Results The distinguishing capability and predicating capability of the BP neural network model were higher than those of the COX proportional-hazards regression model.Conclusion The BP neural network model exhibits good effects on disease predication without restrictions on type and distribution of data.

关 键 词：老年保健人群 缺血性心脑血管病 BP神经网络 

分 类 号：R181.37[医药卫生—流行病学] R54[医药卫生—公共卫生与预防医学]