文章摘要

文章信息/Info

Title:: Automatic detection algorithm of paroxysmal atrial fibrillation based on multi-feature fusion

Author(s):: LIU Xuesen1; ZHANG Junshi2; 1.Zhumadian Central Hospital，Zhumadian 463000,China；2.Xinjiang Medical University，Urumchi 830011,China

Keywords:: Multifeature fusion; Paroxysmal atrial fibrillation; Electrocardiosignal; Automatic detection; Feature extraction; Database

摘要:: 阵发性房颤自动检测在背景噪声干扰环境下提取的心电特征过于单一，无法满足患者的医学检测需求。为此，本研究提出基于多特征融合的阵发性房颤自动检测算法研究。心电信号采集过程中易受背景噪声的干扰，故基于小波包分解算法降噪处理心电信号，并利用小波变换技术自动检测R波波峰，依据熵理念，分别在P波缺失与RR间期不规则提取心电信号特征，采用决策级融合算法将提取的心电信号特征进行融合，通过支撑向量机自动检测阵发性房颤心电片段，实现基于多特征融合的阵发性房颤自动检测算法的运行。实验结果显示，本研究提出算法敏感度为94.6%，特异度为93.7%，准确率为94.0%，满足现今阵发性房颤的检测需求，表明该算法可行。

Abstract:: The ECG features extracted by automatic detection of paroxysmal atrial fibrillation(PAF) in the background noise interference environment are too simple to meet the medical detection needs of patients.Therefore, we proposed an automatic detection algorithm for PAF based on multi-feature fusion.ECG signals would be disturbed by background noise in the process of acquisition, so wavelet packet decomposition algorithm was used to denoise ECG signals，and wavelet transform technology was used to detect R wave peak automatically.According to entropy concept, ECG signal features were extracted on P-wave missing and RR interval irregularity.Decision level fusion algorithm was used to fuse the extracted ECG signal features.Through the support vector machine (SVM) automatic detection of PAF ECG segments，the automatic detection algorithm of paroxysmal atrial fibrillation based on multi-feature fusion was realized.The results of numerical experiments showed that the sensitivity, specificity and accuracy of the proposed algorithm were 94.6%, 93.7% and 94.0%.The proposed algorithm can meet the needs of PAF detection and it is feasible.