基于改进射线拟合法的多导联心电图神经网络分类系统  

A Multi-lead ECG Classification Network System Based on Modified LADT

作　　者：冯俊[1] 邱雅竹[1] 莫智文[1] 

机构地区：[1]四川师范大学数学与软件科学学院

出　　处：《生物医学工程学杂志》2006年第5期956-959,共4页Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(60074014);四川省青年科学基金资助

摘　　要：我们基于心电图特征提取与神经网络分类,提出了一种模拟现实的心电图分类方法,首先应用改进的心电图射线拟合理论,得到心电图多导联的特征;并根据心电图疾病分析的原理与实际诊断的特点,得到一个能对多导联心电图数据疾病类型进行分类的神经网络,为心电图自动分析提出了一个新的思路。经M IT-B IH心电数据库部分波形试验证明,该方法具有较高准确率,对学习过的波形分类正确率为100%,未学习过的78.2%。An electrocardiogram （ECG） classify system based on the features of the ECG and neural network classification, which is the simulation of the real world situation, was present. First, a modified approach of the linear approximation distance thresholding （LADT） algorithm was studied and the features of the ECG were obtained. Then a neural network which can classify the multi-lead ECG data was trained with these features along the theory of the ECG diagnosis and the situation of ECG diagnosis in practice. Thus take a new idea for the ECG automatic analysis. The algorithm was tested using several ECG signals of MIT-BIH, and the performance was good. The correct rate of the trained wave is 100%, untrained is 78.2%.

关 键 词：多导联 心电图分类 特征 神经网络 射线拟合 

分 类 号：TH772.2[机械工程—仪器科学与技术]