基于级联分类器的心律失常检测

Arrhythmia detection based on Cascade classifier

常见的心律失常如室性早博(PVC)和左束支传导阻滞(LBBB)在心血管疾病诊断和预后中具有重要的临床价值。本文提出一种用于PVC和LBBB自动检测的级联分类器,通过提取时域和形态特征,采用支持向量机区分PVC和非PVC,再采用加权最小距离分类器(W-MDC)将非PVC分为正常(N)和LBBB。用MIT-BIH心律失常数据库进行算法验证,对N、LBBB和PVC三分类的总体正确率为96.28%,N、LBBB、PVC各类的灵敏度和特异性分别为98.59%、97.15%,81.41%、91.89%和89.22%、84.87%,验证该算法的泛化能力及对不同病人的心拍分类有效性。此外,本文还证明多导联信息综合对LBBB分类性能的提高。

As common cardiac arrhythmias, premature ventricular contraction（PVC） and left bundle branch block（LBBB）have great significance in the diagnosis and prognosis of cardiovascular diseases. For the automatic detection of PVC and LBBB is proposed. By extracting the time-domain and morphological features, support vector machine（SVM） is utilized to distinguish PVC and non-PVC. The labeled non-PVC is then divided into normal（N） and LBBB using weighted minimum distance classifier（W-MDC）. The proposed algorithm is evaluated using MIT-BIH arrhythmia database. The overall accuracy of N, LBBB and PVC classification is 96.28%. The sensitivity and specificity are 98.59% and 97.15% for N class,81.41% and 91.89% for LBBB, 89.22% and 84.87% for PVC, respectively, which inter-patient heartbeat classification and the generalization ability of the proposed algorithm among different individuals. In addition, the synthesis of multileads information is also proved to be able to improve the LBBB detection performance.