中国生物医学工程学报..docVIP

AR模型在远程心电诊断中的应用 1，2 浙江科技学院信息与电气工程学系,杭州市310012 ； 2.浙江大学生物医学工程教育部重点实验室,杭州310027） 摘要: 在远程心电诊断中，计算机辅助ECG诊断通常是在接收到ECG信号、进而解压重建后进行的，这样便造成诊断工作的延误。为此，本文提出了一种适于远程心电诊断的基于AR模型的ECG直接分类的方法，它无需等ECG解压重建后再行诊断。特别地，AR模型方法能克服现存ECG分类技术实现困难、处理时间长和只能对2-3类ECG进行分类的不足。具体地，利用AR模型系数及其建模误差作为特征对ECG信号进行压缩和分类，在对信号特征分类时，采用了非线性二次判别函数的形式。利用文中方法对MIT-BIH标准数据库中的NSR、APC、PVC、SVT、VT和VF各200个样本信号进行了测试，获得了93.5%-97.86%的分类精度。 关键词: 远程心电，ECG信号，AR建模，二次判别函数，特征提取 Application of AR Model in Telediagnosis of Cardiac Arrhythmias GE Ding-fei1，XIA Shun-ren2 (1. Department of Information Electrical Engineering, Zhejiang University of Science and Technology, Hangzhou 310012; 2.The Key Laboratory of Biomedical Engineering, Ministry of Education, Zhejiang University, Hangzhou 310027) Abstract: Electrocardiogram (ECG) feature extraction for the classification and diagnosis is performed on the reconstructed ECG signals in telecardiogram diagnosis systems. Computer-assistant automatic diagnosis will be delayed due to the reconstruction and feature extraction. A technique based on autoregressive (AR) modeling is proposed, which is suitable for telecardiogram diagnosis systems. The classification can be performed directly before ECG signal reconstructed. The AR coefficients and modeling errors were used to compress and classify the ECGs. More importantly, AR modeling is easy to compute and implement, and most of other existing techniques classify two or three arrhythmias only or have significantly large processing times. AR technique was utilized for classification into arrhythmias such as NSR, PVC, APC, VT, VF, and SVT. The accuracy of classification is 93.5% to 97.86% using the quadratic discrimination function stage-by-stage. Key words: Telecardiogram, ECG signal, AR modeling, Quadratic discrimination function, Feature extraction 0 引言 随着现代通迅技术的发展，远程医疗的实现已成为可能，并将会越来越流行。心电信号(ECG)是医生进行诊断的重要依据。在远程心电诊断中，ECG的压缩方法对计算机辅助自动诊断的快速性有着直接的影响，ECG的压缩方法可分为三类[1]:（1）直接方法：在空域内处理ECG信号的压缩方法。直接方法因为计算简单和高效率而得到广泛地应用，但在远程心电诊断中，计算机辅助自动诊