基于GAN-CNN的心律失常识别

doi:10.16180/j.cnki.issn1007-7820.2022.03.007

摘要/Abstract

摘要：

心电图分析是医生诊断心律失常的重要依据。对心律失常的准确判断有助于患者及时了解身体状况并发现潜在疾病。然而,心电图分析不仅费时费力,而且还依赖于临床经验,因此心电图分析的效率一直受到医生数量和工作效率的限制。深度学习技术的发展为计算机辅助诊断系统的开发提供了基础。文中将一维心电信号转换为二维灰度图像,并采用一种GAN-CNN网络解决心电数据不平衡的问题,可同时实现7类心律失常类型和正常心搏的识别。实验使用MIT-BIH心律失常数据库进行验证,平均准确率达到了99.32%,敏感性和特异性分别为99.69%和98.91%。

关键词: 心电图, 心律失常, 深度学习, 辅助诊断, 心电信号, GAN-CNN, 二维图像, 心电数据不平衡

Abstract:

ECG analysis is an important basis for doctors to diagnose arrhythmia. The judgment of arrhythmia helps patients understand their physical conditions in time and find potential diseases. However, ECG analysis is not only time-consuming and labor-intensive, but also relies on clinical experience. Therefore, the efficiency of ECG analysis has always been limited by the number of doctors and work efficiency. The development of deep learning technology provides a foundation for the development of computer-aided diagnosis systems. In this study, a one-dimensional ECG signal is converted into a two-dimensional gray image, and a GAN-CNN network is used to solve the problem of ECG data imbalance, which can simultaneously realize the recognition of 7 types of arrhythmia and normal heartbeat. The experiment is verified by the MIT-BIH arrhythmia database. The average accuracy rate reaches 99.32%, and the sensitivity and specificity are 99.69% and 98.91%, respectively.

Key words: electrocardiogram, arrhythmia, deep learning, auxiliary diagnosis, electrocardiograph signal, GAN-CNN, two-dimensional image, unbalanced electrocardiogram data

中图分类号:

TP183

陈鹏,刘子龙. 基于GAN-CNN的心律失常识别[J]. 电子科技, 2022, 35(3): 45-50.

Peng CHEN,Zilong LIU. Arrhythmia Recognition Based on GAN-CNN[J]. Electronic Science and Technology, 2022, 35(3): 45-50.

图/表 6

表1

图1

图2

图3

图4

表2

参考文献 20

[1]	Ebrahimi Z, Loni M, Daneshtalab M, et al. A review on deep learning methods for ECG arrhythmia classification[J]. Expert Systems with Applications, 2020,10(33):1-40.
[2]	Chen A Y, Wang F, Liu W H, et al. Multi-information fusion neural networks for arrhythmia automatic detection[J]. Computer Methods and Programs in Biomedicine, 2020,193(1):1-11.
[3]	张晓军, 吴芝路. 基于改进差分阈值算法的心电检测技术研究[J]. 电子科技, 2020,33(2):6-13.
	Zhang Xiaojun, Wu Zhilu. Research of ECG detection technology based on improved differential threshold algorithm[J]. Electronic Science and Technology, 2020,33(2):6-13.
[4]	Ceylan R, Özbay Y. Comparison of FCM,PCA and WT techniques for classification ECG arrhythmias using artificial neural network[J]. Expert Systems with Applications, 2007,33(2):286-295.
[5]	Jun T J, Park H J, Yoo H,et al. GPU based cloud system for high-performance arrhythmia detection with parallel k-NN algorithm[C]. Orlando:Proceedings of the Thirty-eighth Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2016.
[6]	Melgani F, Bazi Y. Classification of electrocardiogram signals with support vector machines and particle swarm optimization[J]. IEEE Transactions on Information Technology in Biomedicine, 2008,12(5):667-677.
[7]	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.
[8]	Mohapatra S K, Srivastava G, Mohanty M N. Arrhythmia classification using deep neural network[C]. Bhubaneswar:Proceedings of the International Conference on Applied Machine Learning, 2019.
[9]	Yildirim Ö. A novel wavelet sequence based on deep bidirectional LSTM network model for ECG signal classification[J]. Computers in Biology and Medicine, 2018,96(2):189-202.
[10]	Baloglu U B, Talo M, Yildirim O, et al. Classification of myocardial infarction with multi-lead ECG signals and deep CNN[J]. Pattern Recognition Letters, 2019,122(4):23-30.
[11]	Hammad M, Zhang S Z, Wang K Q. A novel two-dimensional ECG feature extraction and classification algorithm based on convolution neural network for human authentication[J]. Future Generation Computer Systems, 2019,101(7):180-196.
[12]	Wasimuddin M, Elleithy K, Abuzneid A,et al. ECG signal analysis using 2-D image classification with convolutional neural network[C]. Las Vegas:Proceedings of the International Conference on Computational Science and Computational Intelligence, 2019.
[13]	Pan J, Tompkins W J. A real-time QRS detection algorithm[J]. IEEE Transactions on Biomedical Engineering, 1985,32(3):230-236.
[14]	Ramachandran P, Zoph B, Le Q V. Searching for activation functions[J].ArXiv Preprint, 2017(10):1-10.
[15]	Dang H, Sun M Y, Zhang G H,et al. A novel deep convolutional neural network for arrhythmia classification[C]. Kusatsu:Proceedings of the International Conference on Advanced Mechatronic Systems, 2019.
[16]	Martis R J, Acharya U R, Ray A K,et al. Application of higher order cumulants to ECG signals for the cardiac health diagnosis[C]. Boston:Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2011.
[17]	Homaeinezhad M R, Atyabi S A, Tavakkoli E, et al. ECG arrhythmia recognition via a neuro-SVM-KNN hybrid classifier with virtual QRS image-based geometrical features[J]. Expert Systems with Applications, 2012,39(2):2047-2058.
[18]	张异凡, 黄亦翔, 汪开正, 等. 用于心律失常识别的 LSTM 和 CNN并行组合模型[J]. 哈尔滨工业大学学报, 2019,51(10):76-82.
	Zhang Yifan, Huang Yixiang, Wang Kaizheng, et al. Arrhythmia classification using parallel combination of LSTM and CNN[J]. Journal of Harbin Institute of Technology, 2019,51(10):76-82.
[19]	Acharya U R, Oh S L, Hagiwara Y, et al. A deep convolutional neural network model to classify heartbeats[J]. Computers in biology and medicine, 2017,89(5):389-396.
[20]	Izci E, Ozdemir M A, Degirmenci M,et al. Cardiac arrhythmia detection from 2D ECG images by using deep learning technique[C]. xIzmir:Proceedings of the Medical Technologies Congress, 2019.

类型	记录	数量
FVN	108,109,114,124,200,201,202,203,205,208,210 213,214,215,219,223,233	803
LBBB	109,111,207,213	8 072
NOR	100,101,103,105,108,112,113,114,115,117,121 122,123,202,205,219,230,234	75 052
PAB	102,104,107,217	7 024
PVC	106,116,119,200,201,203,208,210,213,215,221 228,233	7 130
RBBB	118,124,212,231	7 256
VEB	207	106
VFW	207	472

方法	类别	数据集大小	精确度 /%	敏感性 /%	特异性 /%	阳性 /%
GAN-CNN (DA+Swish)	8	44 000	99.32	99.69	98.91	99.84
GAN-CNN (Swish)	8	28 881	97.95	98.82	98.55	99.65
GAN-CNN (DA+ReLU)	8	44 000	99.17	99.61	98.55	99.79
VGG16(DA)	8	44 000	98.69	99.29	98.27	99.75
MFCNN^[15]	5	109 449	95.48	96.53	87.74	-
SVM^[16]	2	1 247	98.48	98.90	98.04	98.13
SVM+KNN^[17]	7	7 185	98.06	-	-	-
LSTM+CNN^[18]	5	94 131	99.11	97.27	99.44	-
CNN^[19]	5	452 960	94.03	96.71	91.54	97.86
2-DCNN^[20]	5	27 789	97.42	-	-	-