基于一维卷积神经网络与循环神经网络串联的心音分析方法

doi:10.12263/DZXB.20200713

摘要/Abstract

摘要：

面向心脏疾病计算机辅助诊断，本文提出一种基于一维卷积神经网络和循环神经网络混合深度学习结构的心音分析方法.本结构首先利用卷积神经网络学习心脏病症在心音信号上的表征，然后通过循环神经网络处理心音信号中的时序信息进行分类，在提升心音分类正确率的同时，大幅度降低了网络参数.为验证本深度学习结构所学特征的有效性，除已有的成人心音数据集外，本文还专门构建了一个面向婴幼儿先天性心脏病的心音数据集，并通过端到端的类别响应图证明了本方法在室缺诊断时学习到的心音信号特征符合临床医师的心音听诊经验.实验结果表明，本文方法能在3 153例成人心音数据分类上达到92.56%的正确率，在528例婴幼儿心音数据分类上达到97.48%正确率，模型参数仅有0.05 M.

关键词: 心音听诊, 一维卷积神经网络, 循环神经网络, 类别响应图

Abstract:

For the computer-aided heart disease diagnosis, this paper proposes a method of heart sound analysis based on the mixed structure of one-dimensional convolutional neural network and recurrent neural network. The proposed structure uses the convolutional neural network to learn the representation of heart disease on the heart sound signal, and then processes the time sequence information in the heart sound signal through the recurrent neural network for classification, which greatly reduces the network parameters while improving the accuracy of heart sound classification. The experimental results show that the proposed method can achieve the accuracy of 92.56% on the classification of 3153 cases of normal and abnormal heart sounds for adults, and 97.48% on the classification of 528 cases of normal and abnormal heart sounds for infants and children. The parameter of the proposed method is 0.05M, which is suitable for portable application situations.

Key words: heart sound analysis, one-dimensional convolutional neural network, recurrent neural network, class activation map

中图分类号:

TP391

肖斌, 陈嘉博, 毕秀丽, 张俊辉, 李伟生, 王国胤, 马旭. 基于一维卷积神经网络与循环神经网络串联的心音分析方法[J]. 电子学报, 2022, 50(10): 2425-2432.

XIAO Bin, CHEN Jia-bo, BI Xiu-li, ZHANG Jun-hui, LI Wei-sheng, WANG Guo-yin, MA Xu. A Method of Heart Sound Analysis Based on One-Dimensional Convolutional Neural Network and Recurrent Neural Network[J]. Acta Electronica Sinica, 2022, 50(10): 2425-2432.

图/表 9

参考文献 22

1	LIU C, SPRINGER D, LI Q, et al. An open access database for the evaluation of heart sound algorithms[J]. Physiological Measurement, 2016, 37(12): 2181-2213.
2	LIANG H, LUKKARINEN S, HARTIMO I. Heart sound segmentation algorithm based on heart sound envelogram[C]//Computers in Cardiology. Lund: IEEE, 1997: 105-108.
3	MOUKADEM A, DIETERLEN A, HUEBER N, et al. A robust heart sounds segmentation module based on S-transform[J]. Biomedical Signal Processing and Control, 2013, 8(3): 273-281.
4	CHOI S, JIANG Z W. Comparison of envelope extraction algorithms for cardiac sound signal segmentation[J]. Expert Systems With Applications, 2008, 34(2): 1056-1069.
5	YAN Z H, JIANG Z W, MIYAMOTO A, et al. The moment segmentation analysis of heart sound pattern[J]. Computer Methods and Programs in Biomedicine, 2010, 98(2): 140-150.
6	ARI S, HEMBRAM K, SAHA G. Detection of cardiac abnormality from PCG signal using LMS based least square SVM classifier[J]. Expert Systems With Applications, 2010, 37(12): 8019-8026.
7	SPRINGER D B, TARASSENKO L, CLIFFORD G D. Logistic regression-HSMM-based heart sound segmentation[J]. IEEE Transactions on Biomedical Engineering, 2016, 63(4): 822-832.
8	CHEN T N, YANG S I, HO L T, et al. S1 and S2 heart sound recognition using deep neural networks[J]. IEEE Transactions on Biomedical Engineering, 2017, 64(2): 372-380.
9	MESSNER E, ZOHRER M, PERNKOPF F. Heart sound segmentation-an event detection approach using deep recurrent neural networks[J]. IEEE Transactions on Biomedical Engineering, 2018, 65(9): 1964-1974.
10	RENNA F, OLIVEIRA J, COIMBRA M T. Convolutional neural networks for heart sound segmentation[C]//2018 26th European Signal Processing Conference(EUSIPCO). Rome: IEEE, 2018: 757-761.
11	ZHANG W J, HAN J Q, DENG S W. Heart sound classification based on scaled spectrogram and tensor decomposition[J]. Expert Systems With Applications, 2017, 84: 220-231.
12	SOETA Y, BITO Y. Detection of features of prosthetic cardiac valve sound by spectrogram analysis[J]. Applied Acoustics, 2015, 89: 28-33.
13	NIVITHA VARGHEES V, RAMACHANDRAN K I. Effective heart sound segmentation and murmur classification using empirical wavelet transform and instantaneous phase for electronic stethoscope[J]. IEEE Sensors Journal, 2017, 17(12): 3861-3872.
14	徐韵秋. 基于深度卷积神经网络的心音智能分析[D]. 重庆: 重庆邮电大学, 2019.
	XU Y Q. Heart Sound Intelligent Analysis Based on Deep Convolutional Neural Networks[D]. Chongqing: Chongqing University of Posts and Telecommunications, 2019. (in Chinese)
15	DOMINGUEZ-MORALES J P, JIMENEZ-FERNANDEZ A F, DOMINGUEZ-MORALES M J, et al. Deep neural networks for the recognition and classification of heart murmurs using neuromorphic auditory sensors[J]. IEEE Transactions on Biomedical Circuits and Systems, 2018, 12(1): 24-34.
16	RUBIN J, ABREU R, GANGULI A, et al. Classifying heart sound recordings using deep convolutional neural networks and mel-frequency cepstral coefficients[C]//2016 Computing in Cardiology Conference(CinC). Vancouver: IEEE, 2016: 813-816.
17	NILANON T, YAO J, HAO J. Normal/abnormal heart sound recordings classification using convolutional neural network[C]//2016 Computing in Cardiology Conference(CinC). Vancouver: IEEE, 2016: 585-588.
18	POTES C, PARVANEH S, RAHMAN A, et al. Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds[C]//2016 Computing in Cardiology Conference(CinC). Vancouver: IEEE, 2016: 621-624.
19	THOMAE C, DOMINIK A. Using deep gated RNN with a convolutional front end for end-to-end classification of heart sound[C]//2016 Computing in Cardiology Conference(CinC). Vancouver: IEEE, 2016: 625-628.
20	XIAO B, XU Y Q, BI X L, et al. Follow the sound of children's heart: A deep-learning-based computer-aided pediatric CHDs diagnosis system[J]. IEEE Internet of Things Journal, 2020, 7(3): 1994-2004.
21	王磊, 徐涛, 宋传东, 等. 基于深度学习的miRNA与疾病相关性预测算法[J]. 电子学报, 2020, 48(5): 870-877.
	WANG L, XU T, SONG C D, et al. Prediction algorithm of association between miRNAs and diseases based on deep learning[J]. Acta Electronica Sinica, 2020, 48(5): 870-877. (in Chinese)
22	周涛, 霍兵强, 陆惠玲, 等. 残差神经网络及其在医学图像处理中的应用研究[J]. 电子学报, 2020, 48(7): 1436-1447.
	ZHOU T, HUO B Q, LU H L, et al. Research on residual neural network and its application on medical image processing[J]. Acta Electronica Sinica, 2020, 48(7): 1436-1447. (in Chinese)

层类型	输出大小	配置
输入	1×6 000×1	—
卷积层1	1×2 993×24	卷积核大小15,步长2
池化层1	1×1 495×24	最大池化大小3,步长2
卷积层2	1×744×48	卷积核大小7,步长2
池化层2	1×371×48	最大池化大小3,步长2
卷积层3	1×182×96	卷积核大小7,步长2
LSTM层	1×182×20	5层,时间步大小20
分类器	2	批标准化,非线性激活,全连接层

层类型	输出大小	配置
输入	1×6 000×1	—
卷积层1	1×2 993×24	卷积核大小15,步长2
池化层1	1×1 495×24	最大池化大小3,步长2
卷积层2	1×744×48	卷积核大小7,步长2
池化层2	1×371×48	最大池化大小3,步长2
卷积层3	1×182×96	卷积核大小7,步长2
LSTM层	1×182×20	5层,时间步大小20
分类器	2	批标准化,非线性激活,全连接层

评价指标	1-D CNN	MFCC	PSD-CNN	AdaBoost	1-D Clique	DGRE	SS-PLSR	SS-TD	Proposed
精确分割	否	是	否	是	否	否	是	是	否
ACC	0.893 3	0.933 1	0.890 5	—	0.932 8	—	—	—	0.929 6
Pr(正常)	0.936 4	0.935 8	0.949 4	—	0.962 7	—	0.95	0.92	0.946 7
Pr(异常)	0.739 6	0.853 6	0.710 3	—	0.829 0	—	0.82	0.88	0.857
Se	0.760 8	0.826 6	0.815 0	0.88	0.862 1	0.83	—	—	0.799 2
Sp	0.928 2	0.951 6	0.910 2	0.82	0.951 6	0.96	—	—	0.963 7
Score	0.844 5	0.889 1	0.862 6	0.85	0.906 9	0.90	—	—	0.881 4
Params	0.19 M	12.41 M	0.24 M	—	0.19 M	—	—	—	0.05 M

评价指标	1-D CNN	MFCC	PSD-CNN	AdaBoost	1-D Clique	DGRE	SS-PLSR	SS-TD	Proposed
精确分割	否	是	否	是	否	否	是	是	否
ACC	0.893 3	0.933 1	0.890 5	—	0.932 8	—	—	—	0.929 6
Pr(正常)	0.936 4	0.935 8	0.949 4	—	0.962 7	—	0.95	0.92	0.946 7
Pr(异常)	0.739 6	0.853 6	0.710 3	—	0.829 0	—	0.82	0.88	0.857
Se	0.760 8	0.826 6	0.815 0	0.88	0.862 1	0.83	—	—	0.799 2
Sp	0.928 2	0.951 6	0.910 2	0.82	0.951 6	0.96	—	—	0.963 7
Score	0.844 5	0.889 1	0.862 6	0.85	0.906 9	0.90	—	—	0.881 4
Params	0.19 M	12.41 M	0.24 M	—	0.19 M	—	—	—	0.05 M

评价指标	1-D CNN	MFCC	PSD-NN	1-D lique	Proposed
精确分割	否	是	否	否	否
ACC	0.911 0	0.923 3	0.856 6	0.916 8	0.974 8
Pr(正常)	0.925 8	0.911 4	0.888 2	0.889 6	0.973 5
Pr(异常)	0.909 2	0.932 2	0.842 2	0.936 6	0.985 4
Se	0.960 9	0.955 8	0.959 7	0.934 0	0.994 8
Sp	0.810 6	0.882 5	0.664 4	0.891 7	0.951 4
Score	0.885 8	0.919 2	0.812 1	0.912 9	0.973 1
Params	0.19 M	12.41 M	0.24 M	0.19 M	0.05 M