基于参数模型的VCG信号计算机辅助检测研究

Study of computer assisted VCG signal detection based on parametric model

在心脏病学上,2种最基本和流行的心电导联体系是标准心电(ECG)导联体系和法兰克心电向量(VCG)体系(简称:法兰克心电导联体系)。多数现存的计算机辅助心电分析是在标准心电导联体系上展开的。实际上,与标准心电导联体系比较而言,法兰克心电导联体系由于其相互间正交的原因与解剖学有着更为紧密的联系。为此,本文提出从法兰克心电导联体系中提取心电特征的研究,并以心肌梗死(MI)心电信号为研究对象,其数据取自PTB诊断ECG数据库。利用多变量回归建模技术对VCG进行建模,并将多变量回归模型系数作为VCG分类特征。实验结果表明,在本文的研究条件下,法兰克心电导联体系比标准心电导联体系能取得更好的分类效果。

There are two most popular and basic lead systems used in cardiology, namely, standard electrocardiogram （ECG） lead system and Frank vectorcardiogram （VCG） lead system. Most of the existing computer assisted analyses of ECG signals are based on the standard ECG leads. In practice, Frank VCG leads that are orthogonal each other are more correlated with anatomy compared with the standard ECG leads. The ECG feature extraction for Frank VCG leads is studied in this research. Myocardial infarction （MI） VCG signals taken from PTB diagnostic ECG database were employed for the analysis in this study. Multivariate autoregressive （AR） modeling technique was performed on the VCGs, and the multivariate AR coefficients were used as VCG features for the classification. Experimental results show that Frank VCG leads can obtain better classification effect than standard ECG leads under current condition.