基于聚类分析的心电节拍分类算法  

ECG beat classification algorithm based on cluster analysis

作　　者：鄢羽[1] 孙成[2] 

机构地区：[1]重庆医科大学附属第一医院信息中心,重庆400016 [2]中国科学院深圳先进技术研究院,广东深圳518055

出　　处：《计算机应用》2014年第7期2132-2135,2139,共5页journal of Computer Applications

基　　金：广东省与中国科学院全面战略合作计划项目(2009B091300160)

摘　　要：为提高计算机辅助心电节拍分类算法的准确率和普适性,提出一种基于聚类分析的心电节拍分类算法,该算法利用心电节拍个体内差异性较小的特性,采用两级聚类分析、抽样代表性心电节拍的方法,结合心电医师的辅助诊断,实现对心电节拍的准确分类。为了验证算法的准确性,采用国际公认的标准数据库——MIT-BIH心律失常数据库,AAMI/ANSI标准规定的心电节拍分类方法及准确率的计算方法进行仿真实验,最终总体分类准确率达到99.07%。与Kiranyaz等(KIRANYAZ S,INCE T,PULKKINEN J,et al.Personalized long-term ECG classification:A systematic approach[J].Expert Systems with Applications,2011,38(4):3220-3226.)的心电节拍分类算法相比,该算法无需进行设定的训练,且S类心电节拍分类灵敏度由40.15%提高到89.82%,显著提高了分类算法的普适性。In order to improve the accuracy and universality of computer-assisted classification algorithm, a Electrocardiography （ECG） beat classification algorithm based on cluster analysis was presented in this paper. The algorithm considered that one patients＇ ECG beats repeated periodically, and used the method of two-stage cluster analysis, and selecting representative ECG beats, combined with the diagnosis of cardiac physicians to achieve accurate ECG beat classification rate. In order to verify the accuracy of the algorithm, using the internationally standard database MIT-BIH arrhythmia database, the ECG beat classification method and the accuracy evaluation method specified by AAMI/ANSI standard were used to perform simulation experiments, the final overall classification accuracy rate is 99. 07%. Compared with Kiranyaz＇ method （KIRANYAZ S, INCE T, PULKKINEN J, et al. Personalized long-term ECG classification： A systematic approach[ J]. Expert Systems with Applications, 2011, 38（4） ： 3220 -3226. ）, this method does not require specific training step, and the sensitivity of the ECG beats which labeled as S raise to 89. 82% from 40. 15%, significantly improving classification algorithm＇s generalization capability.

关 键 词：心电节拍分类 聚类分析 辅助诊断 ANSI/AAMI标准 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]