独立分量分析在心房纤颤检测中的应用研究

Application of independent component analysis in atrial fibrillation detection

为了从心房纤颤患者的动态心电图记录中抽取心房活动信号并对其进行特征分析,提出了一种无损型心房纤颤自动诊断方法.证明了应用独立分量分析(ICA)必须满足的3个基本条件:源间独立性、瞬时线性混合和至多一个高斯源,并建立了盲源分离的数学模型.采用快速固定点优化算法分析仿真试验和临床数据,计算各被分离分量的峰度值,有效提取了心房纤颤信号,定性和定量地表明了该方法的准确性和鲁棒性.

A non-invasive diagnosis approach for atrial fibrillation （AF） was proposed by extracting atrial activity （AA） signal from real ambulatory electrocardiogram records and analyzing the AA features. Independent component analysis （ICA） theory was used to verify that three fundamental requirements which must be satisfied in ICA are source independence, at most a Gaussian source and instant linear mixture. A mathematical model was formulated for this blind source separation problem. Then a fast and efficient fixed-point ICA algorithm was applied to analyze the simulation and clinical data, and the kurtosis values of separated independent components were also evaluated. The results qualitatively and quantitatively show that the proposed ICA method for AF signal feature extraction and analysis is appropriate and robust.