摘要
本文提出了一种集成学习方法以提升室性早搏的识别性能.MIT-BIH两个通道的数据分别经过卷积神经网络进行室性早搏心拍分类,然后按照融合规则对分类结果进行融合决策,其准确率、灵敏度和特异性分别为99.91%、98.76%、99.97%,优于已有算法的室性早搏心拍分类结果.此外,面向临床应用,本文还利用卷积神经网络和诊断规则相结合的方法实现了病人间室性早搏识别实验,在有14万多条记录的数据集上,取得的准确率、灵敏度及特异性分别为97.87%、87.94%、98.02%,验证了该算法的有效性.
In order to improve the recognition performance of premature ventricular contraction（PVC）,this paper reports an algorithm based on ensemble learning.First,the tow-lead ECG signals from the MIT-BIH Arrhythmia database are classified into PVC and non PVC beats using lead convolutional neural network（LCNN） classifier.Then the results are fused with some rules.The accuracy,sensitivity and specificity of the proposed algorithm are 99.91%,98.76%and 99.97%,respectively,which are better than that of other existing algorithms for PVC beats classification.In addition,this paper realizes an inter-patient PVC recognition experiment by combining LCNN and diagnostic rules for clinical application.The effectiveness of the proposed algorithm has been confirmed by the accuracy（97.87%）,sensitivity（87.94%） and specificity（98.02%） with the data set over 140000 ECG records.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2017年第2期501-507,共7页
Acta Electronica Sinica
关键词
室性早搏
卷积神经网络
诊断规则
premature ventricular contraction（PVC）
lead convolutional neural network（LCNN）
diagnosis rules
