计算机科学

计算机科学 ›› 2025, Vol. 52 ›› Issue (6A): 240700195-8.doi: 10.11896/jsjkx.240700195

• 智能医学工程 • 上一篇    下一篇

一种通过增强图像编码和非对称卷积网络的心音分类算法

王晟懿1, 杨宏波2, 潘家华2, 王威廉1   

  1. 1 云南大学信息学院 昆明 650504
    2 云南省阜外心血管病医院 昆明 650102
  • 出版日期:2025-06-16 发布日期:2025-06-12
  • 通讯作者: 王威廉(wlwang_47@126.com)
  • 作者简介:(2814604624@qq.com)
  • 基金资助:
    国家自然科学基金(81960067);云南省重大科技专项基金(2018ZF017)

Heart Sound Classification Algorithm Using Enhanced Image Coding and AsymmetricConvolutional Networks

WANG Shengyi1, YANG Hongbo2, PAN Jiahua2, WANG Weilian1   

  1. 1 School of Information Science and Engineering,Yunnan University,Kunming 650504,China
    2 Kunming Medical University Affiliated Cardiovascular Hospital,Kunming 650102,China
  • Online:2025-06-16 Published:2025-06-12
  • About author:WANG Shengyi,born in 1999,postgra-duate.Her main research interests include signal processing and machine learning.
    WANG Weilian,born in 1947,bachelor,professor.His main research interests include signal proces-sing and pattern recognition,biological signal proces-sing,digital-analog hybrid IC and ASIC design.
  • Supported by:
    National Natural Science Foundation of China(81960067) and Major Science and Technology Projects of Yunnan Province in 2018(2018ZF017).

PDF (PC)

35

摘要： 文中提出了一种通过增强图像编码和非对称卷积网络的心音分类算法。与传统的基于统计特征和时频域特征提取心音的方法不同,该算法通过引入分数阶傅里叶变换(FrFT)分别对格拉姆角场(GAF)、马尔可夫场(MTF)、递归图(RP) 3种图像编码方法进行增强,构成FrFT-GAF,FrFT-MTF,FrFT-RP图像编码模块。运用上述图像编码模块将一维心音信号转换为二维编码特征图,并利用计算机视觉技术在分类任务中的优势,采用非对称卷积网络(ACNet)对心音的二维编码特征图进行分析处理,从而实现对心音的有效分类。此外,还分别对上述图像编码模块的性能进行了评估和比较。实验结果表明,在心音二分类任务中,FrFT-RP模块的分类效果最好,在数据集1和数据集2(Physio Net/CinC 2016数据集)上的准确率分别为0.981和0.977,F1分别为0.989和0.974。FrFT-MTF和FrFT-GAF模块的效果次之。使用FrFT增强图像编码特征后较以往方法有明显提升,为心音信号分类提供了新的思路和方法,有望应用于先心病机器辅助诊断。

关键词: 先天性心脏病, 心音, 图像编码, 分数阶傅里叶变换, 非对称卷积网络

Abstract: This paper proposes a heart sound classification algorithm using enhanced image coding and asymmetric convolutional networks.Unlike traditional methods that extract heart sounds based on statistical features and time-frequency domain features,this algorithm enhances three image coding methods—Gramian angular field(GAF),Markov transition field(MTF),and recurrence plot(RP)—by introducing fractional Fourier transform(FrFT),which constitutes the image coding modules of FrFT-GAF,FrFT-MTF,and FrFT-RP,respectively.The one-dimensional heart sound signal is transformed into a two-dimensional encoded feature map using these image coding modules.An asymmetric convolutional network(ACNet) leverages computer vision advantages to analyze and process the two-dimensional encoded feature map for effective heart sound classification.In addition,the performance of the above image coding modules is evaluated and compared respectively.Experimental results demonstrate that the FrFT-RP module achieves the best classification performance in binary heart sound classification tasks,with an accuracy of 0.981 and 0.977,and F1 score of 0.989 and 0.974 on dataset 1 and dataset 2(Physio Net/CinC 2016 dataset),respectively.The FrFT-MTF and FrFT-GAF modules show effective performance in that order.The performance of the method using FrFT to enhance image encoding features has significantly improved compared to previous methods,providing novel approaches and methods for heart sound signal classification,is expected to be applied in machine assisted diagnosis of congenital heart disease.

Key words: Congenital heart disease, Heart sound, Image coding, Fractional Fourier transform, Asymmetric convolutional network

中图分类号: 

  • TN912
[1]MA L Y,WANG Z W,FAN J,et al.Highlights of the China Cardiovascular Health and Disease Report 2022 [J].Chinese Family Medicine,2023,26(32):3975-3994.
[2]SUN S P,WANG H B,JIANG Z W,et al.Segmentation-based heart sound feature extraction combined with classifier models for a VSD diagnosis system[J].Expert Systems with Applications,2014,41(4):1769-1780.
[3]CHOWDHURY M,POUDEL K,HU Y.Detecting abnormalPCG signals and extracting cardiac information employing deep learning and the shannon energy envelope[C]//2020 IEEE Signal Processing in Medicine and Biology Symposium(SPMB).IEEE,2020:1-4.
[4]KAMSON A P,SHARMA L N,DANDAPAT S.Enhancement of the heart sound envelope using the logistic function amplitude moderation method[J].Computer Methods and Programs in Biomedicine,2020,187:105239.
[5]VARGHEES V N,RAMACHANDRAN K I.Effective HeartSound Segmentation and Murmur Classification Using Empirical Wavelet Transform and Instantaneous Phase for Electronic Stethoscope[J].IEEE Sensors Journal,2017,17(12):3861-3872.
[6]ZHANG W J,HAN J Q,DENG S W.Abnormal heart sound detection using temporal quasi-periodic features and long short-term memory without segmentation[J].Biomedical Signal Processing and Control,2019,53:101560.
[7]IBRAHIM N,JAMAL N,SHA’ABANI A H,et al.A Comparative Study of Heart Sound Signal Classification Based on Temporal,Spectral and Geometric Features[C]//2020 IEEE－EMBSConference on Biomedical Engineering and Sciences(IECBES).2021:24-29.
[8]ESLAMIZADEH G,BARATI R.Heart murmur detection based on wavelet transformation and a synergy between artificial neural network and modified neighbor annealing methods[J].ArtificialIntelligence in Medicine,2017,78:23-40.
[9]BHATIKAR S R,DEGROFF C,MAHAJAN R L.A classifier based on the artificial neural network approach for cardiologic auscultation in pediatrics[J].Artificial Intelligence in Medicine,2005,33(3):251-260.
[10]ZENG W,LIN Z X,YUAN C Z,et al.Detection of heart valve disorders from PCG signals using TQWT,FA-MVEMD,Shannon energy envelope and deterministic learning[J].Artificial Intelligence Review,2021(7):1-38.
[11]CHEN J X,GUO Z H,XU X,et al.A Robust Deep Learning Framework Based on Spectrograms for Heart Sound Classification[J].IEEE/ACM Transactions on Computational Biology and Bioinformatics,2024,21(4):936-947.
[12]JAMIL S,ROY A M.An efficient and robust Phonocardio-graphy(PCG)-based Valvular Heart Diseases(VHD) detectionframework using Vision Transformer(ViT)[J].Computers in Biology and Medicine,2023,158:106734.
[13]NOGUEIRA D M,FERREIRA C A,GOMES E F,et al.Classifying Heart Sounds Using Images of Motifs,MFCC and Temporal Features[J].Journal of Medical Systems,2019,43(6):168.
[14]KUI H R,PAN J H,ZONG R,et al.Heart sound classification based on log Mel-frequency spectral coefficients features and convolutional neural networks[J].Biomedical Signal Processing and Control,2021,69:102893.
[15]OBAIDAT M S.Phonocardiogram signal analysis:Techniquesand performance comparison[J].Journal of Medical Engineering &Technology,Taylor & Francis,1993,17(6):221-227.
[16]LUBIS C,GONDAWIJAYA F.Heart Sound Diagnose System with BFCC,MFCC,and Backpropagation Neural Network[J].IOP Conference Series:Materials Science and Engineering,2019,508(1):012119.
[17]WANG Y L,SUN J,YANG H B,et al.Classification Model of Heart Sounds in Pulmonary Hypertension Based on Time-Frequency Fusion Features[J].Computer Science,2024,51(S1):387-393.
[18]ECKMANN J P,KAMPHORST S O,RUELLE D.Recurrence Plots of Dynamical Systems[J].Europhys Lett,1987,4(9):973-977.
[19]WANG Z G,OATES T.Imaging Time-Series to Improve Classification and Imputation[J].arXiv.org e-Print archive,2015,arXiv:1506.00327.
[20]ZHOU G,CHIEN C,CHEN J,et al.Identifying pediatric heart murmurs and distinguishing innocent from pathologic using deep learning[J].Artificial Intelligence in Medicine,2024,153:102867.
[21]RICCIO D,BRANCATI N,SANNINO G,et al.CNN-basedclassification of phonocardiograms using fractal techniques[J].Biomedical Signal Processing and Control,2023,86:105186.
[22]NGUYEN M T,LIN W W,HUANG J H.Heart Sound Classification Using Deep Learning Techniques Based on Log-mel Spectrogram[J].Circuits,Systems,and Signal Processing,2023,42(1):344-360.
[23]RANIPA K,ZHU W P,SWAMY M N S.A novel feature-level fusion scheme with multimodal attention CNN for heart sound classification[J].Computer Methods and Programs inBiomedicine,2024,248:108122.
[24]DING X H,GUO Y C,DING G G,et al.ACNet:Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blocks[C]//2019 IEEE/CVF International Conference on Computer Vision(ICCV).2019:1911-1920.
[25]GE B B,YANG H B,MAP Y,et al.Detection of pulmonary arterial hypertension associated with congenital heart disease based on time-frequency domain and deep learning features[J].Biomedical Signal Processing and Control,2023,81:104451.
[26]ABDUH Z,NEHARY E A,WAHED M A,et al.Classification of heart sounds using fractional fourier transform based mel-frequency spectral coefficients and traditional classifiers[J].Elsevier,2020,57:101788.
[27]FAN Q L,YANG H B,GUO T,et al.FrFT-Bark domain feature extraction and CNN residual shrinkage network heart sound classification algorithm[J].Journal of Yunnan University:Natural Science Edition,2023,45(3):564-574.
[28]WANGP F,GONG X G,GUO Q,et al.Children’s Expression Recognition Based on Multi-Scale Asymmetric Convolutional Neural Network[J].International Journal of Advanced Computer Science and Applications(IJACSA),2024:15(7):437.
[29]HOU C Q,LI J S,WANG W,et al.Second-order asymmetric convolution network for breast cancer histopathology image classification[J].Journal of Biophotonics,2022,15(5):e202100370.
[30]WU J,WANG Y X,ZHANG X G.Lightweight AsymmetricConvolutional Distillation Network for Single Image Super-Resolution[J].IEEE Signal Processing Letters,2023,30:733-737.
[31]WANG R S,DUAN Y F,LI Y K,et al.PCTMF-Net:heart sound classification with parallel CNNs-transformer and second-order spectral analysis[J].The Visual Computer,2023,39(8):3811-3822.
[32]MAITY A,PATHAK A,SAHA G.Transfer learning basedheart valve disease classification from Phonocardiogram signal[J].Biomedical signal Processing and Control,2023,85:104805.
[33]ZHANG H B,ZHANG P,WANG Z W,et al.Multi-Feature Decision Fusion Network for Heart Sound Abnormality Detection and Classification[J].IEEE Journals & Magazine,2023:1386-1397.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发