基于时域特征的房颤识别方法

Atrial fibrillation recognition method based on time domain characteristics

针对便携式心电采集设备采集的手部心电信号质量较差、阵发性房颤识别困难问题,提出一种房颤自动识别方法,利用信息熵和连续小波变换(CWT,continuous wavelet transform)筛选奇异波形,准确识别心电信号中的R波,并利用R波信息提取心电信号的时域特征,利用BP神经网络构建阵发性房颤识别模型.在PCinCC2017和AFDB数据集上的实验表明,本文方法的房颤识别的灵敏度和特异性分别高于96%和98%,对失常10秒左右的短时阵发房颤的识别灵敏度和特异性均高于94%,可以应用于家庭便携式房颤监测.

Aiming at the poor quality of hand electrocardiogram(ECG)collected by portable ECG acquisition equipment and difficulty in identifying paroxysmal atrial fibrillation(AF),an automatic recognition method for AF is proposed.This method,with the help of information entropy and continuous wavelet transform(CWT),identify the R peak of the ECG,and calculate the time domain characteristics as the input parameters of the BP neural network to train a diagnostic model automatically recognition AF.The model is validated in the PCinCC2017 and MIT-BIH AFDB databases,sensitivity and specificity are higher than 96%and 98%respectively,The recognition sensitivity and specificity of paroxysmal AF with about 10 seconds are both higher than 94%,which can be applied to the portable home AF monitoring.