摘要
阵发性房颤是一种最为常见的心律失常,发作时心电图主要表现为RR间期绝对不规则以及P波缺失。已有阵发性房颤的自动检测方法仅依赖于心电信号的时域信息,从而使得检测准确率不高。鉴于小波相干分析可以同时展现信号的时域和频域信息,文中提出了一种基于小波相干分析的阵发性房颤自动检测方法。首先,对所有的心电信号预处理;其次,对模板与待测心电信号分别进行小波相干分析得到其小波相干图;进而,计算小波相干值均值、比率和交叉小波相位角方差构成房颤心电特征;最后,将上述特征结合超限学习机完成阵发性房颤的自动检测。文中通过MIT-BIH房颤数据集验证所提算法的有效性与可行性,数值实验结果达到准确率97. 81%,敏感性98. 54%,特异性98. 61%。
Paroxysmal atrial fibrillation(PAF)is a kind of typical arrhythmia,whose characteristics appearing on electrocardiogram(ECG)are two points:RR interval irregularity and P wave absence.Most existing automatic PAF detection methods only rely on the time domain information of ECG signal,so as to bring the low accuracy.In view of that the wavelet coherence analysis can display signal information in both time and frequency domain,this paper proposes an automatic PAF detection method using ECGs based on wavelet coherence analysis.Firstly,ECG signals are preprocessed where denoising,segmenting and template selection are included.Secondly,wavelet coherent maps between the template ECG and the present ECG are drawn according to the wavelet coherence analysis.And then,the mean value of wavelet coherence,the ratio and the variance of cross wavelet phase angle are calculated to be the extracted features.Finally,the features were fed into extreme learning machine(ELM)to achieve the automatic detection of PAF.Performance of the proposed method is verified on MIT-BIH atrial fibrillation database in simulations.The accuracy,sensitivity and specificity of the method reach 97.81%,98.54%,and 98.61%.
作者
王迪
李强
孙亚楠
韦杰英
张瑞
WANG Di;LI Qiang;SUN Yanan;WEI Jieying;ZHANG Rui(Medical Big Data Research Center,Northwest University,Xi′an 710127,China)
出处
《西北大学学报（自然科学版）》
CAS
CSCD
北大核心
2019年第1期27-34,共8页
Journal of Northwest University（Natural Science Edition）
基金
国家自然科学基金面上项目(61473223)
陕西省创新人才推进计划基金资助项目(2018TD-016)
关键词
阵发性房颤
心电图
小波相干分析
特征提取
超限学习机
paroxysmal atrial fibrillation(PAF)
electrocardiogram(ECG)
wavelet coherence analysis
feature extraction
extreme learning machine(ELM)
