基于1D CNN-BiLSTM网络联合集成学习的心律失常智能诊断系统  

Arrhythmia intelligent diagnosis system based on 1D CNN-BiLSTM network joint integrated learning

作　　者：张明伟 张天逸[1,2] 程云章 ZHANG Mingwei;ZHANG Tianyi;CHENG Yunzhang(School of Health Sciences and Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China;Shanghai Interventional Medical Device Engineering Technology Research Center,Shanghai 200093)

机构地区：[1]上海理工大学健康科学与工程学院,上海200093 [2]上海介入医疗器械工程技术研究中心,上海200093

出　　处：《生物医学工程研究》2022年第3期259-267,共9页Journal Of Biomedical Engineering Research

基　　金：上海市公共卫生体系建设三年行动计划(2020-2022年)学科带头人计划项目(GWV-10.2-XD32);上海工程技术研究中心资助项目(18DZ2250900)。

摘　　要：为提高心律失常智能诊断的准确率,本研究提出了一种多网络融合模型和Stacking集成学习算法,用于八种心律失常疾病的智能诊断。使用1D CNN-BiLSTM融合网络提取单导联信号的高维特征和时域相关性特征,将十二个导联的心电信号特征融合,得到高维的特征张量,采用Stacking集成学习算法训练得到泛化性更好的诊断模型。通过比较准确性、精确性、召回率、F1-Score四个诊断性能指标,验证了利用十二导联融合特征作为最终诊断特征,准确率有显著提升,且Stacking集成学习算法较单一机器学习算法有更好的性能。本研究通过将机器学习、神经网络、集成学习算法有效结合,训练得到的心律失常智能诊断模型有较高的准确率,为基于心电信号的心律失常智能诊断提供了一种新方法。In order to improve the accuracy of intelligent diagnosis of arrhythmia,we proposed a multi-network fusion model and stacking integrated learning for intelligent diagnosis of eight arrhythmia diseases.The 1D CNN-BiLSTM fusion network was used to extract the high-dimensional features and time-domain correlation features of the single-lead signal,the twelve-leads electrocardrogram(ECG)signal features were fused to obtain high-dimensional feature tensors,and Stacking ensemble learning was used to obtain a diagnostic model with better generalization.By comparing the diagnostic performance indicators of accuracy,precision,recall,F1-score,it was verified that the accuracy of the twelve-lead fusion feature as the final diagnostic feature was greatly improved,and the Stacking integrated learning algorithm had better performance than the single machine learning algorithm.Through the effective combination of machine learning,neural network and ensemble learning algorithm,the intelligent diagnosis model of arrhythmia obtained by training has high accuracy,which provides a new method for intelligent diagnosis of arrhythmia based on ECG signals.

关 键 词：心律失常诊断 多导联信号 小波软阈值去噪 多网络联合 轻量级CNN 集成学习框架 

分 类 号：R318[医药卫生—生物医学工程] TN911.7[医药卫生—基础医学]