摘要
心脏疾病是威胁人类健康的最大病因,ECG信号的复杂性使得人工检测需要大量时间且容易误诊,因此基于心电图心拍数据实现计算机辅助ECG判断具有重要意义。提出基于QRS波群的心拍特征提取方法,以Pan-Tompkins算法实现QRS波群定位,设计心拍截取规则;构建一维卷积神经网络(CNN)模型,实现ECG四类心拍数据(正常搏动、左束支传导阻滞、右束支传导阻滞、室性早搏)的自动分类检测。为验证提出心拍截取方法的有效性,以MIT-BIH心率失常数据库45条数据进行验证,结果显示其灵敏度为99.1%、特异性为99.4%。采用截取的四类心拍数据验证一维CNN自动ECG分类检测模型的可用性,结果显示模型总体分类准确率为98.95%。
Heart disease is the biggest threat to human health,the complexity of ECG signals makes manual detection taking a lot of time and easy to misdiagnose,Therefore,it is of great significance to realize computer-aided ECG judgment based on ECG data.In this paper,a method of feature extraction based on QRS wave group is proposed,QRS wave group localization is implemented by using Pan-Tompkins algorithm,the heart beat interception rule was designed;A one-dimensional convolutional neural network(CNN) model was constructed to realize automatic classification detection of four types of ECG beat data(Normal beating,Left bundle branch block,Right bundle branch block,Premature ventricular beats).To verify the effectiveness of this heart beat intercept method,experiments were carried out based on 45 data of MIT-BIH Arrhythmia Database which show that the sensitivity is 99.1% and specificity is 99.4%.The four types of cardiac beat data are used to verify the availability of one-dimensional CNN automatic classification detection model,and the results show that the overall classification accuracy of the model is 98.95%.
作者
张俊飞
毕志升
王静
吴小玲
ZHANG Junfei;WANG Jing;BI Zhisheng;WU Xiaoling(Department of Biomedical Engineering,Guangdong Medical University,Guangzhou 511436,China)
出处
《自动化与仪器仪表》
2019年第12期71-75,共5页
Automation & Instrumentation
基金
国家自然科学基金青年科学基金项目“高维多目标进化算法的研究与应用”(No.61603106)
2018年广州市高校创新创业教育项目“云空间学习共同体支持下的翻转课堂设计与实践”(No.201709k56)
2017年广州市教育局市属高校教育教学改革项目“互联网+教学质量评估与监控教学形态建设研究”(No.2017A05)
