基于混合时频域特征的卷积神经网络心律失常分类方法的研究

doi:10.12263/DZXB.20211181

电子学报 ›› 2023, Vol. 51 ›› Issue (3): 701-711.DOI: 10.12263/DZXB.20211181

基于混合时频域特征的卷积神经网络心律失常分类方法的研究

吕杭¹, 蒋明峰¹, 李杨¹, 张鞠成²(), 王志康²()

^1.浙江理工大学计算机科学与技术学院，浙江杭州 310018
^2.浙江大学医学院附属第二医院，浙江杭州 310009

收稿日期:2021-08-30 修回日期:2022-07-09 出版日期:2023-03-25
- 通讯作者:
- 蒋明峰
- 作者简介:
- 吕杭男， 1997年8月出生于湖北省黄冈市，现为浙江理工大学研究生. 研究领域方向为医学影像处理.
  蒋明峰（通讯作者）男， 1977年5月生于江西丰城，现为浙江理工大学计算机科学与技术学院教授、博士生导师，主要研究方向为计算机医学图像处理、生物医学信号处理．
  李杨男，1986年05月出生于辽宁抚顺，现为浙江理工大学计算机科学与技术学院讲师、硕士生导师，主要研究方向为计算机医学图像处理、计算机视觉. E-mail: dr.yangli@outlook.com
  张鞠成男，1989年10月出生，山东潍坊人，工程师，主要研究方向为MRI关键技术和心脏电生理研究. E-mail: jucheng@zju.edu.cn
  张鞠成男，1989年10月出生，山东潍坊人，工程师，主要研究方向为MRI关键技术和心脏电生理研究. E-mail: jucheng@zju.edu.cn
  王志康男，1970年5月出生，浙江杭州人，研究员，现为浙江大学校医院副院长，主要研究方向为生物医学信号处理. E-mail: 2192009@zju.edu.cn
  王志康男，1970年5月出生，浙江杭州人，研究员，现为浙江大学校医院副院长，主要研究方向为生物医学信号处理. E-mail: 2192009@zju.edu.cn
- 基金资助:
- 浙江省科技厅重点研发项目(2020C03060);国家自然科学基金(62101497);浙江省自然科学基金-数理医学学会联合基金重点项目(LSZ19F010001)

Research on Arrhythmia Classification by Using Convolutional Neural Network with Mixed Time-Frequency Domain Features

LÜ Hang¹, JIANG Ming-feng¹, LI Yang¹, ZHANG Ju-cheng²(), WANG Zhi-kang²()

^1.School of Computer Science and Technology，Zhejiang Sci-Tech University，Hangzhou，Zhejiang 310018，China
^2.The Second Affiliated Hospital，School of Medicine Zhejiang University，Hangzhou，Zhejiang 310009，China

Received:2021-08-30 Revised:2022-07-09 Online:2023-03-25 Published:2023-04-20
- Corresponding author:
- JIANG Ming-feng
- Supported by:
- Key Research and Development Program of Zhejiang Province(2020C03060);National Natural Science Foundation of China(62101497);Joint Fund of Zhejiang Provincial Natural Science Foundation(LSZ19F010001)

摘要/Abstract

摘要：

心律失常是常见的心血管疾病之一，目前很多方法通过计算机辅助系统对心电图进行分析以识别心律失常，但由于大多数心律失常数据样本较少，计算机辅助系统识别心律失常效果不佳.本文提出了一种基于混合时频域分析特征提取的卷积神经网络方法，该方法提取心电图的RR间期时域特征、希尔伯特-黄变换提取的频域特征和连续小波变换提取的时频域联合特征，经过特征融合后输入卷积神经网络训练分类模型，并采用Focal Loss作为网路的损失函数，实现对心律失常的分类.本文使用MIT-BIH（Massachusetts Institute of Technology-Boston's Beth Israel Hospital）心律失常数据库验证本文提出方法对4类心电数据分类的结果，实验结果表明，与现有的分类算法相比，本文所提出的混合时频域特征方法能有效提升心律失常分类的准确性.

关键词: 时频域分析, 连续小波变换, 希尔伯特-黄变换, 心律失常分类, Focal Loss, 卷积神经网络

Abstract:

Arrhythmia is one of cardiovascular diseases, and many methods are used to analyze electrocardiogram by computer-aided system to identify arrhythmia. However, most of the data samples of arrhythmia are small, and the computer-aided system is not effective in identifying arrhythmia. In this paper, a mixed time-frequency domain feature extraction method is proposed for arrhythmia classification by using convolution neural network method. The fused features consist of the time domain characteristics from the RR interval, frequency domain characteristics from hilbert-huang transform, and joint time-frequency domain features extracted from continuous wavelet transform. Then the fused features are used as an input to the convolution neural network for training classification model, and the focal loss is used as the loss function of the training model, so as to realize the arrhythmias classification. In addition, the MIT-BIH （Massachusetts Institute of Technology-Boston's Beth Israel Hospital） arrhythmia database is used to verify the performances of the proposed method for arrhythmias classification of four types of ECG (Electrocardiograph) data. Experimental results show that compared with the existing classification algorithms, the proposed method improves the F1 of class obviously.

Key words: time-frequency domain analysis, continuous wavelet transform, hilbert-huang transform, arrhythmia classification, focal loss, convolutional neural network

中图分类号:

TP391.41

吕杭, 蒋明峰, 李杨, 等. 基于混合时频域特征的卷积神经网络心律失常分类方法的研究[J]. 电子学报, 2023, 51(3): 701-711.

Hang Lü, Ming-feng JIANG, Yang LI, et al. Research on Arrhythmia Classification by Using Convolutional Neural Network with Mixed Time-Frequency Domain Features[J]. Acta Electronica Sinica, 2023, 51(3): 701-711.

图/表 10

参考文献 32

1	BARMPOUTIS P, DIMITROPOULOS K, APOSTOLIDIS A, et al. Multi-lead ECG signal analysis for myocardial infarction detection and localization through the mapping of Grassmannian and Euclidean features into a common Hilbert space[J]. Biomedical Signal Processing and Control, 2019, 52: 111-119.
2	LIN X J, GREEN J C, XIAN H, et al. Holiday and weekend effects on mortality for acute myocardial infarction in Shanxi, China: A cross-sectional study[J].International Journal of Public Health, 2020, 65(6): 847-857.
3	YE C, KUMAR B V, COIMBRA M T. Heartbeat classification using morphological and dynamic features of ECG signals[J]. IEEE Transactions on Biomedical Engineering, 2012, 59(10): 2930-2941.
4	KUTLU Y, KUNTALP D. A multi-stage automatic arrhythmia recognition and classification system[J]. Computers in Biology and Medicine, 2011, 41(1): 37-45.
5	JOHNSON K W, TORRES SOTO J, GLICKSBERG B S, et al. Artificial intelligence in cardiology[J]. Journal of the American College of Cardiology, 2018, 71(23): 2668-2679.
6	DAQROUQ K, ALKHATEEB A, AJOUR M N, et al. Neural network and wavelet average framing percentage energy for atrial fibrillation classification[J]. Computer Methods and Programs in Biomedicine, 2014, 113(3): 919-926.
7	WANG Z Z, LI H Y, HAN C, et al. Arrhythmia classification based on multiple features fusion and random forest using ECG[J]. Journal of Medical Imaging and Health Informatics, 2019, 9(8): 1645-1654.
8	CHEIKHROUHOU O, MAHMUD R, ZOUARI R, et al. One-dimensional CNN approach for ECG arrhythmia analysis in fog-cloud environments[J]. IEEE Access, 2021, 9: 103513-103523.
9	POURBABAEE B, ROSHTKHARI M J, KHORASANI K. Deep convolutional neural networks and learning ECG features for screening paroxysmal atrial fibrillation patients[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2018, 48(12): 2095-2104.
10	NURMAINI S, TONDAS A E, DARMAWAHYUNI A, et al. Robust detection of atrial fibrillation from short-term electrocardiogram using convolutional neural networks[J]. Future Generation Computer Systems, 2020, 113: 304-317.
11	ÇıNAR A, TUNCER S A. Classification of normal sinus rhythm, abnormal arrhythmia and congestive heart failure ECG signals using LSTM and hybrid CNN-SVM deep neural networks[J]. Computer Methods in Biomechanics and Biomedical Engineering, 2021, 24(2): 203-214.
12	ZHANG Y T, LI J Y, WEI S S, et al. Heartbeats classification using hybrid time-frequency analysis and transfer learning based on ResNet[J]. IEEE Journal of Biomedical and Health Informatics, 2021, 25(11): 4175-4184.
13	SABUT S, PANDEY O, MISHRA B S P, et al. Detection of ventricular arrhythmia using hybrid time-frequency-based features and deep neural network[J]. Physical and Engineering Sciences in Medicine, 2021, 44(1): 135-145.
14	XU Y, ZHANG S, CAO Z, et al. Extreme learning machine for heartbeat classification with hybrid time-domain and wavelet time-frequency features[J]. Journal of Healthcare Engineering, 2021, 2021: 6674695.
15	WANG T, LU C, SUN Y, et al. Automatic ECG classification using continuous wavelet transform and convolutional neural network[J]. Entropy (Basel, Switzerland), 2021, 23(1): E119.
16	郑近德, 潘海洋, 戚晓利, 等. 基于改进经验小波变换的时频分析方法及其在滚动轴承故障诊断中的应用[J]. 电子学报, 2018, 46(2): 358-364.
	ZHENG J D, PAN H Y, QI X L, et al. Enhanced empirical wavelet transform based time-frequency analysis and its application to rolling bearing fault diagnosis[J]. Acta Electronica Sinica, 2018, 46(2): 358-364. (in Chinese)
17	HUANG J S, CHEN B Q, YAO B, et al. ECG arrhythmia classification using STFT-based spectrogram and convolutional neural network[J]. IEEE Access, 2019, 7: 92871-92880.
18	GRAMATIKOV B, GEORGIEV I. Wavelets as alternative to short-time Fourier transform in signal-averaged electrocardiography[J].Medical and Biological Engineering and Computing, 1995, 33(3): 482-487.
19	杨向林, 严洪, 许志, 等. 基于Hilbert-Huang变换的ECG消噪[J]. 电子学报, 2011, 39(4): 819-824.
	YANG X L, YAN H, XU Z, et al. ECG de-noising based on Hilbert-Huang transform[J]. Acta Electronica Sinica, 2011, 39(4): 819-824. (in Chinese)
20	MOODY G B, MARK R G. The impact of the MIT-BIH arrhythmia database[J]. IEEE Engineering in Medicine and Biology Magazine, 2001, 20(3): 45-50.
21	LIU F F, LIU C Y, ZHAO L N, et al. An open access database for evaluating the algorithms of electrocardiogram rhythm and morphology abnormality detection[J]. Journal of Medical Imaging and Health Informatics, 2018, 8(7): 1368-1373.
22	MAR T, ZAUNSEDER S, MARTÍNEZ J P, et al. Optimization of ECG classification by means of feature selection[J]. IEEE Transactions on Biomedical Engineering, 2011, 58(8): 2168-2177.
23	MONDÉJAR-GUERRA V, NOVO J, ROUCO J, et al. Heartbeat classification fusing temporal and morphological information of ECGs via ensemble of classifiers[J]. Biomedical Signal Processing and Control, 2019, 47: 41-48.
24	DE CHAZAL P, O'DWYER M, REILLY R B. Automatic classification of heartbeats using ECG morphology and heartbeat interval features[J]. IEEE Transactions on Biomedical Engineering, 2004, 51(7): 1196-1206.
25	HE J Y, RONG J, SUN L, et al. A framework for cardiac arrhythmia detection from IoT-based ECGs[J].World Wide Web, 2020, 23(5): 2835-2850.
26	ZHANG Z C, DONG J, LUO X Q, et al. Heartbeat classification using disease-specific feature selection[J]. Computers in Biology and Medicine, 2014, 46: 79-89.
27	GUO M F, ZENG X D, CHEN D Y, et al. Deep-learning-based earth fault detection using continuous wavelet transform and convolutional neural network in resonant grounding distribution systems[J]. IEEE Sensors Journal, 2018, 18(3): 1291-1300.
28	WU Z Q, LAN T J, YANG C W, et al. A novel method to detect multiple arrhythmias based on time-frequency analysis and convolutional neural networks[J]. IEEE Access, 2019,7: 170820-170830.
29	HUANG N E, SHEN Z, LONG S R, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings of the Royal Society of London Series A: Mathematical, Physical and Engineering Sciences, 1998, 454(1971): 903-995.
30	周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017, 40(6): 1229-1251.
	ZHOU F Y, JIN L P, DONG J. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017, 40(6): 1229-1251. (in Chinese)
31	LU Y, JIANG M F, WEI L Y, et al. Automated arrhythmia classification using depthwise separable convolutional neural network with focal loss[J]. Biomedical Signal Processing and Control, 2021, 69: 102843.
32	KAPLAN BERKAYA S, UYSAL A K, SORA GUNAL E, et al. A survey on ECG analysis[J]. Biomedical Signal Processing and Control, 2018, 43: 216-235.

N	S	V	F	Q
正常搏动(N)	房性早搏(A)	室性早搏(V)	心室和正常搏动的融合(F)	起博心跳
左束支传导阻滞搏(L)	异常房性早搏(a)	室性逸搏(E)		起搏和正常搏动的融合(f)
右束支传导阻滞搏(R)	交界性早搏(J)			未分类搏动(Q)
心房逸搏(e)	室上性早搏或异位性搏动(S)
交界性逸搏(j)

N	S	V	F	Q
正常搏动(N)	房性早搏(A)	室性早搏(V)	心室和正常搏动的融合(F)	起博心跳
左束支传导阻滞搏(L)	异常房性早搏(a)	室性逸搏(E)		起搏和正常搏动的融合(f)
右束支传导阻滞搏(R)	交界性早搏(J)			未分类搏动(Q)
心房逸搏(e)	室上性早搏或异位性搏动(S)
交界性逸搏(j)

数据集/标签	N	S	V	F	总
DS1	45 824	3 788	943	414	50 969
DS2	44 218	3 219	1 836	388	49 661

数据集/标签	N	S	V	F	总
DS1	45 824	3 788	943	414	50 969
DS2	44 218	3 219	1 836	388	49 661

网络层名	核尺寸	滤波器	填充	步长	输出形状	参数
输入层1	-	-	-	-	100×250×1	-
卷积层	7×7	16	vaild	1	94×244×16	800
批归一化	-	-	-	-	94×244×16	64
最大池化	5×5	-	-	-	18×48×16	-
卷积层	3×3	32	vaild	1	16×46×32	4 640
批归一化	-	-	-	-	16×46×32	128
最大池化	3×3	-	-	-	5×15×32	-
卷积层	3×3	64	vaild	1	3×13×64	18 496
批归一化	-	-	-	-	3×13×64	256
最大池化	3×3	-	-	-	1×4×64	-
全连接层	-	-	-	-	1×4×16	1 040
展开层	-	-	-	-	64	-
输入层2	-	-	-	-	4	-
输入层3	-	-	-	-	8×248×1	-
卷积层	7×7	16	same	1	8×248×16	800
批归一化	-	-	-	-	8×248×16	64
最大池化	2×5	-	-	-	4×49×16	-
卷积层	3×3	32	same	1	4×49×32	4 640
批归一化	-	-	-	-	4×49×32	128
最大池化	2×3	-	-	-	2×16×32	-
卷积层	3×3	64	same	1	2×16×64	18 496
批归一化	-	-	-	-	2×16×64	256
最大池化	2×3	-	-	-	1×5×64	-
全连接层	-	-	-	-	1×5×16	1 040
展开层	-	-	-	-	80	-
融合层	-	-	-	-	148	-
全连接层	-	-	-	-	32	4 768
softmax层	-	-	-	-	4	132

基于混合时频域特征的卷积神经网络心律失常分类方法的研究

Research on Arrhythmia Classification by Using Convolutional Neural Network with Mixed Time-Frequency Domain Features

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 32

相关文章 15

编辑推荐

Metrics

方法	类别	Precision	Recall	F1	Accuracy
RR CWT	N	0.981 5	0.945 7	0.963 3	0.931 1
	S	0.589 8	0.749 5	0.660 1
	V	0.690 8	0.946 6	0.798 7
	F	0.000 0	0.000 0	0.000
RR CWT Focal Loss	N	0.980 7	0.959 9	0.970 2	0.944 7
	S	0.827 7	0.800 7	0.814 0
	V	0.773 0	0.932 0	0.845 1
	F	0.000 0	0.000 0	0.000 0
RR HHT	N	0.973 2	0.960 8	0.966 9	0.921 6
	S	0.539 9	0.324 6	0.405 4
	V	0.742 5	0.833 2	0.785 2
	F	0.000 0	0.000 0	0.000 0
RR HHT Focal Loss	N	0.960 2	0.959 8	0.975 5	0.933 0
	S	0.601 0	0.377 5	0.499 4
	V	0.775 4	0.819 1	0.789 7
	F	0.000 0	0.000 0	0.000 0
RR CWT HHT	N	0.977 6	0.926 8	0.951 5	0.936 5
	S	0.498 3	0.648 7	0.563 7
	V	0.843 5	0.939 1	0.888 7
	F	0.002 3	0.010 3	0.003 7
RR CWT HHT Focal Loss	N	0.984 3	0.947 3	0.965 5	0.946 7
	S	0.547 3	0.802 8	0.650 9
	V	0.928 9	0.978 3	0.952 9
	F	0.003 9	0.010 3	0.005 7

[1]	黄赟, 张帆, 郭威, 陈立, 羊光. 一种基于数据标准差的卷积神经网络量化方法[J]. 电子学报, 2023, 51(3): 639-647.
[2]	许新征, 李杉. 基于特征膨胀卷积模块的轻量化技术研究[J]. 电子学报, 2023, 51(2): 355-364.
[3]	丁琪, 田萱, 孙国栋. 基于注意力增强的热点感知新闻推荐模型[J]. 电子学报, 2023, 51(1): 93-104.
[4]	张永梅, 孙捷. 基于动静态特征双输入神经网络的咳嗽声诊断COVID-19算法[J]. 电子学报, 2023, 51(1): 202-212.
[5]	袁海英, 成君鹏, 曾智勇, 武延瑞. Mobile_BLNet：基于Big-Little Net的轻量级卷积神经网络优化设计[J]. 电子学报, 2023, 51(1): 180-191.
[6]	吴靖, 叶晓晶, 黄峰, 陈丽琼, 王志锋, 刘文犀. 基于深度学习的单帧图像超分辨率重建综述[J]. 电子学报, 2022, 50(9): 2265-2294.
[7]	毛国君, 王者浩, 黄山, 王翔. 基于剪边策略的图残差卷积深层网络模型[J]. 电子学报, 2022, 50(9): 2205-2214.
[8]	袁海英, 曾智勇, 成君鹏. 面向灵活并行度的稀疏卷积神经网络加速器[J]. 电子学报, 2022, 50(8): 1811-1818.
[9]	王相海, 赵晓阳, 王鑫莹, 赵克云, 宋传鸣. 非抽取小波边缘学习深度残差网络的单幅图像超分辨率重建[J]. 电子学报, 2022, 50(7): 1753-1765.
[10]	李明爱, 张圆圆. 基于连续小波变换和符号传递熵的脑功能网络构建方法[J]. 电子学报, 2022, 50(7): 1600-1608.
[11]	张志文, 刘天歌, 聂鹏举. 基于实景数据增强和双路径融合网络的实时街景语义分割算法[J]. 电子学报, 2022, 50(7): 1609-1620.
[12]	丁毅, 沈薇, 李海生, 钟琼慧, 田明宇, 李洁. 面向CNN的区块链可信隐私服务计算模型[J]. 电子学报, 2022, 50(6): 1399-1409.
[13]	张波, 陆云杰, 秦东明, 邹国建. 一种卷积自编码深度学习的空气污染多站点联合预测模型[J]. 电子学报, 2022, 50(6): 1410-1427.
[14]	廖勇, 李玉杰. 一种轻量化低复杂度的FDD大规模MIMO系统CSI反馈方法[J]. 电子学报, 2022, 50(5): 1211-1217.
[15]	张云, 化青龙, 姜义成, 徐丹. 基于混合型复数域卷积神经网络的三维转动舰船目标识别[J]. 电子学报, 2022, 50(5): 1042-1049.