基于卷积神经网络的心律失常分类研究
Classification of Arrhythmia Based on Convolutional Neural Networks
郭宇昊 1王大为1
作者信息
- 1. 山西师范大学 物理与信息工程学院,山西 太原 030031
- 折叠
摘要
心律失常是引起心肌梗塞、突发性心脏死亡等严重疾病的重要原因,常借助心电图进行早期诊断.然而,传统的心电信号分类方法有着复杂的特征提取任务,计算量大、费时费力.为此通过在经典卷积神经网络的基础上加入Dropout层,设计了一种改进的卷积神经网络模型,该模型可以进一步提升模型的泛化能力、提高准确率.其利用CNN自动提取特征,将经过预处理的心电信号直接作为模型的输入,自动识别5种不同类型的心拍.在MIT-BIH心律不齐数据库上进行实验,准确率达到99.68%,特异性为98.94%,灵敏度为99.76%.实验结果表明,相较于经典卷积神经网络,本文提出的方法能够精确、高效地识别不同类型的心律失常疾病.
Abstract
Arrhythmia is a significant cause of serious conditions such as myocardial infarction and sudden cardiac death,and is of-ten diagnosed early using electrocardiogram.However,the traditional ECG signal classification method has complicated task of fea-ture extraction,which is time-consuming and laborious.To address this,we design an improved Convolutional Neural Network(CNN)model by incorporating Dropout layers into the classic CNN framework.This enhancement aims to improve the model's gen-eralization ability and increase accuracy.By using CNN to automatically extract features,the preprocessed ECG signal is directly used as the input of the model to automatically identify 5 different types of heart beats.Experiments conducted on the MIT-BIH Ar-rhythmia Database achieve the accuracy of 99.68%,the specificity of 98.94%,and the sensitivity of 99.76%.The experimental re-sults show that compared with the classic CNN,the proposed method can accurately and efficiently identify different types of ar-rhythmia diseases.
关键词
心律失常/心电信号/卷积神经网络/Dropout层/自动识别Key words
Arrhythmia/Electrocardiogram/Convolutional Neural Network/Dropout layer/automatic identification引用本文复制引用
基金项目
国家自然科学基金(62201332)
山西省基础研究计划(20210302124028)
山西省高等学校科技创新项目(2021L268)
山西师范大学2024年大学生创新创业训练计划项目(2024DCXM-27)
出版年
2024
NETL
NSTL
万方数据