深度学习心音信号分类研究进展

Advances in heart sound classification based on deep learning

摘要

心音是反映心脏健康状况的重要生理信号，心音分类可辅助心脏疾病的早期诊断。相比依赖人工特征提取的传统方法，深度学习能够自动学习心音特征，在分类精度和鲁棒性方面具有明显优势。本文通过系统检索国内外相关文献，围绕深度学习在心音分类领域的研究展开综述。首先对心音信号的特点、常用数据集及预处理方法进行梳理，其次归纳卷积神经网络、循环神经网络、Transformer及多模型融合等深度学习方法在心音分类任务中的应用现状，并进一步从模型结构、分类性能和临床适用性等方面进行比较与分析。最后，讨论深度学习在心音分类任务中的应用前景，以期促进深度学习方法在心音信号分类中的应用。

Abstract

Heart sounds are vital physiological signals that reflect cardiac status, and heart sound classification plays a crucial role in the early diagnosis of cardiovascular diseases. Compared with traditional methods that rely on handcrafted feature extraction, deep learning techniques can automatically learn representative features from heart sound signals, achieving superior classification accuracy and robustness. This study provides an overview of recent advances in deep learning-based heart sound classification by retrieving and analyzing relevant domestic and international research. This review systematical introduces the characteristics of heart sound signals, commonly-used datasets, and preprocessing methods, summarizes the application status of deep learning approaches, including convolutional neural networks, recurrent neural networks, Transformer-based models, and hybrid architectures, in heart sound classification, and further compares their model architectures, classification performance, and clinical applicability. Finally, the application prospects of deep learning in heart sound classification is discussed, aiming to promote the application of deep learning techniques in heart sound classification.

关键词

心音信号 / 深度学习 / 分类算法 / 神经网络 / 综述

Key words

heart sound signal / deep learning / classification algorithm / neural network / review

参考文献

[1] 孙文慧, 陈扶明, 张乙鹏, 等 . 心肺音分离方法研究进展[J]. 中国医疗设备, 2024, 39(3): 154-159.

[2] Sun WH, Chen FM, Zhang YP, et al. Research progress of heart—lung sound separation methods[J]. China Medical Devices, 2024, 39(3): 154-159.

[3] 李金萌 . 基于深度学习的心音信号分类方法研究[D]. 济南: 齐鲁工业大学, 2024.

[4] Li JM . Research on heart sound signal classification method based on deep learning[D]. Jinan: Qilu University of Technology, 2024.

[5] 吴延军, 徐泾平, 赵艳 . 心音的产生与传导机制[J]. 生物医学工程学杂志, 1996(3): 280-288.

[6] Wu YJ, Xu JP, Zhao Y . Mechanism of generation and transmission of heart sounds[J]. Journal of Biomedical Engineering, 1996(3): 280-288.

[7] Gomes EF, Bentley PJ, Pereira E, et al, Classifying heart sounds—approaches to the Pascal challenge[C]// Proceedings of the International Conference on Health Informatics—Volume 0BIOSTEC, Barcelona, Spain, 2013: 337-340

[8] Yaseen, Son GY, Kwon S . Classification of heart sound signal using multiple features[J]. Appl Sci, 2018, 8(12): 2344.

[9] Liu CY, Springer D, Li Q, et al. An open access database for the evaluation of heart sound algorithms[J]. Physiol Meas, 2016, 37(12): 2181-2213.

[10] Oliveira J, Renna F, Costa PD, et al. The CirCor DigiScope dataset: from murmur detection to murmur classification[J]. IEEE J Biomed Health Inform, 2022, 26(6): 2524-2535.

[11] Dong FQ, Qian K, Ren Z, et al. Machine listening for heart status monitoring: introducing and benchmarking HSS—the heart sounds Shenzhen corpus[J]. IEEE J Biomed Health Inform, 2020, 24(7): 2082-2092.

[12] Centracchio J, Parlato S, Esposito D, et al. Accurate localization of first and second heart sounds via template matching in forcecardiography signals [J]. Sensors, 2024, 24(5): 1525.

[13] Shervegar VM . A robust sliding window adaptive filtering technique for phonocardiogram signal denoising[J]. Expert Syst, 2025, 42(1): e13361.

[14] Liu ZC, Shao TP, Zhang X . BCG signal analysis based on improved VMD algorithm[J]. Measurement, 2024, 231: 114631.

[15] Chen YL, Wu DP, Wei KY, et al. Joint VMD and PSSVM optimization algorithm for seizure prediction based on ECG signals[J]. Biomed Signal Process Control, 2025, 103: 107478.

[16] Fang Y, Leng HX, Wang WB, et al. Multi—level feature encoding algorithm based on FBPSI for heart sound classification[J]. Sci Rep, 2024, 14(1): 29132.

[17] Lin MF, Hong YZ, Hong SC, et al. Discrete wavelet transform based ECG classification using gcForest: a deep ensemble method[J]. Technol Health Care, 2024, 32(S1): 95-105.

[18] Huang NE, Shen Z, Long SR, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non—stationary time series analysis[J]. Proc R Soc A, 1998, 454(1971): 903-995.

[19] Ali SN, Shuvo SB, Al—Manzo MIS, et al. An end—to—end deep learning framework for real—time denoising of heart sounds for cardiac disease detection in unseen noise[J]. IEEE Access, 2023, 11: 87887-87901.

[20] Liang MF, Hu JQ, Zhou XB, et al. A GAN based heart sound denoising model[C]// 2022 12th International Conference on Information Technology in Medicine and Education (ITME). Piscataway, NJ, USA: IEEE, 2022: 666-670.

[21] Al—Zaben A, Al—Fahoum A, Ababneh M, et al. Improved recovery of cardiac auscultation sounds using modified cosine transform and LSTM—based masking[J]. Med Biol Eng Comput, 2024, 62(8): 2485-2497.

[22] Dissanayake T, Fernando T, Denman S, et al. A robust interpretable deep learning classifier for heart anomaly detection without segmentation[J]. IEEE J Biomed Health Inform, 2021, 25(6): 2162-2171.

[23] Arjoune Y, Nguyen TN, Doroshow RW, et al. A noise—robust heart sound segmentation algorithm based on Shannon energy[J]. IEEE Access, 2024, 12: 7747-7761.

[24] Xiao PZ, Wang KP . Segmentation of heart sound signals using improved Hilbert transform and wavelet packet transform[J]. Circuits Syst Signal Process, 2025, 44(7): 4752-4773.

[25] Li LY, Huang M, Dao L, et al. Construction and validation of a method for automated time label segmentation of heart sounds[J]. Front Artif Intell, 2024, 6: 1309750.

[26] Liu QS, Wu XM, Ma XJ . An automatic segmentation method for heart sounds[J]. Biomed Eng Online, 2018, 17(1): 106.

[27] Xu XC, Geng XG, Gao ZX, et al. Optimal heart sound segmentation algorithm based on K—mean clustering and wavelet transform[J]. Appl Sci, 2023, 13(2): 1170.

[28] Rabiner LR . A tutorial on hidden Markov models and selected applications in speech recognition[J]. Proc IEEE, 1989, 77(2): 257-286.

[29] Martins ML, Coimbra MT, Renna F . Markov—based neural networks for heart sound segmentation: using domain knowledge in a principled way[J]. IEEE J Biomed Health Inform, 2023, 27(11): 5357-5368.

[30] Sadeghi Nia P, Danandeh Hesar H . Abnormal heart sound detection using time—frequency analysis and machine learning techniques[J]. Biomed Signal Process Control, 2024, 90: 105899.

[31] Enériz D, Rodriguez—Almeida AJ, Fabelo H, et al. Low—cost FPGA implementation of deep learning—based heart sound segmentation for real—time CVDs screening[J]. IEEE Trans Instrum Meas, 2024, 73: 1-16.

[32] Park C, Shin K, Seo J, et al. Enhancement of phonocardiogram segmentation using convolutional neural networks with Fourier transform module[J]. Biomed Eng Lett, 2025, 15(2): 401-413.

[33] Mcdonald A, Gales MJF, Agarwal A . A recurrent neural network and parallel hidden Markov model algorithm to segment and detect heart murmurs in phonocardiograms[J]. PLOS Digit Health, 2024, 3(11): e0000436.

[34] Chen Y, Sun YN, Lü JC, et al. End—to—end heart sound segmentation using deep convolutional recurrent network[J]. Complex Intell Syst, 2021, 7(4): 2103-2117.

[35] Guan TX, Xu DY, Cai SS, et al. A deep—learning—based cardiac sound segmentation method for smart auscultation applications[C]// 2023 4th International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering (ICBAIE). Piscataway, NJ, USA: IEEE, 2023: 171-175.

[36] 卢官明, 蔡亚宁, 卢峻禾, 等 . 基于Bi—LSTM和时序注意力的异常心音检测[J]. 南京邮电大学学报(自然科学版), 2025, 45(1): 12-20.

[37] Lu GM, Lu JH, Cai YN, et al. Abnormal heart sound detection using Bi—LSTM and temporal attention mechanism[J]. Journal of Nanjing University of Posts and Telecommunications(Natural Science Edition), 2025, 45(1): 12-20.

[38] Deng EQ, Jia YB, Zhu GB, et al. Heart sound signals classification with image conversion employed[J]. Electronics, 2024, 13(7): 1179.

[39] Jiang ZH, Song WH, Yan YH, et al. Automated valvular heart disease detection using heart sound with a deep learning algorithm[J]. Int J Cardiol Heart Vasc, 2024, 51: 101368.

[40] Abbas S, Ojo S, Al Hejaili A, et al. Artificial intelligence framework for heart disease classification from audio signals[J]. Sci Rep, 2024, 14(1): 3123.

[41] Elola A, Aramendi E, Oliveira J, et al. Beyond heart murmur detection: automatic murmur grading from phonocardiogram[J]. IEEE J Biomed Health Inform, 2023, 27(8): 3856-3866.

[42] Ranipa K, Zhu WP, Swamy MNS . A novel feature—level fusion scheme with multimodal attention CNN for heart sound classification[J]. Comput Methods Programs Biomed, 2024, 248: 108122.

[43] Kumar SS, Ramachandran P . Nested compressive sensing for ECG signals using novel hybrid two—level approach of DWT and DCT sensing and sparsification[J]. Discover Appl Sci, 2024, 7(1): 18.

[44] Chen X, Guo XM, Zheng YN, et al. Heart function grading evaluation based on heart sounds and convolutional neural networks[J]. Phys Eng Sci Med, 2023, 46(1): 279-288.

[45] Katal N, Garg H, Sharma B . A comparative analysis of CNNs and LSTMs for ECG—based diagnosis of arrythmia and congestive heart failure[J]. Comput Methods Biomech Biomed Engin, 2026, 29(7): 1608-1636.

[46] Telceken M . A new feature extraction method for AI based classification of heart sounds: dual—frequency cepstral coefficients (DFCCs)[J]. Eur Phys J Spec Top, 2025, 234(15): 4699-4710.

[47] Patwa A, Mahboob Ur Rahman M, Al—Naffouri TY . Heart murmur and abnormal PCG detection via wavelet scattering transform and 1D—CNN[J]. IEEE Sens J, 2025, 25(7): 12430-12443.

[48] Yang WR, Xu JJ, Xiang JH, et al. Diagnosis of cardiac abnormalities based on phonocardiogram using a novel fuzzy matching feature extraction method[J]. BMC Med Inform Decis Mak, 2022, 22(1): 230.

[49] Zheng YN, Guo XM, Yang Y, et al. Phonocardiogram transfer learning—based CatBoost model for diastolic dysfunction identification using multiple domain—specific deep feature fusion[J]. Comput Biol Med, 2023, 156: 106707.

[50] Bahreini M, Barati R, Kamali A . Cardiac sound classification using a hybrid approach: MFCC—based feature fusion and CNN deep features[J]. EURASIP J Adv Signal Process, 2025, 2025(1): 2.

[51] Li F, Liu M, Zhao YJ, et al. Feature extraction and classification of heart sound using 1D convolutional neural networks[J]. EURASIP J Adv Signal Process, 2019, 2019(1): 59.

[52] Khan SUR, Khan Z . Detection of abnormal cardiac rhythms using feature fusion technique with heart sound spectrograms[J]. J Bionic Eng, 2025, 22(4): 2030-2049.

[53] Ren Z, Cummins N, Pandit V, et al. Learning image—based representations for heart sound classification[C]// Proceedings of the 2018 International Conference on Digital Health. New York, NY, USA: ACM, 2018: 143-147.

[54] Chowdhury TH, Poudel KN, Hu YT . Time—frequency analysis, denoising, compression, segmentation, and classification of PCG signals[J]. IEEE Access, 2020, 8: 160882-160890.

[55] Riccio D, Brancati N, Sannino G, et al. CNN—based classification of phonocardiograms using fractal techniques[J]. Biomed Signal Process Control, 2023, 86: 105186.

[56] Xiao F, Liu HQ, Lu J . A new approach based on a 1D + 2D convolutional neural network and evolving fuzzy system for the diagnosis of cardiovascular disease from heart sound signals[J]. Appl Acoust, 2024, 216: 109723.

[57] Xiong WH, Zhang G, Yan DM, et al. Multichannel feature fusion network—based technique for heart sound signal classification and recognition[J]. Expert Syst Appl, 2025, 273: 126839.

[58] Mahmood A, Dhahri H, Alhajlah M, et al. Enhanced classification of phonocardiograms using modified deep learning[J]. IEEE Access, 2024, 12: 178909-178916.

[59] Li F, Tang H, Shang S, et al. Classification of heart sounds using convolutional neural network[J]. Appl Sci, 2020, 10(11): 3956.

[60] Bao XQ, Xu YJ, Lam HK, et al. Time—frequency distributions of heart sound signals: a comparative study using convolutional neural networks[J]. Biomed Eng Adv, 2023, 5: 100093.

[61] Qiao LH, Li ZX, Xiao B, et al. QDRJL: quaternion dynamic representation with joint learning neural network for heart sound signal abnormal detection[J]. Neurocomputing, 2023, 562: 126889.

[62] Al—Shannaq MA, Nasrawi A, Bsoul AARK, et al. Abnormal heart sound recognition using SVM and LSTM models in real—time mode[J]. Sci Rep, 2025, 15(1): 9129.

[63] Latifi SA, Ghassemian H, Imani M . Multi—branch convolutional network and LSTM—CNN for heart sound classification[J]. Phys Eng Sci Med, 2026, 49(1): 263-287.

[64] 王慧 . 基于最小门控单元的心音分类研究[J]. 智能计算机与应用, 2024, 14(3): 81-85.

[65] Wang H. Classification of heart sounds based on minimal gated unit[J]. Intelligent Computer and Applications, 2024, 14(3): 81-85.

[66] Roy TS, Roy JK, Mandal N . Design of ear—contactless stethoscope and improvement in the performance of deep learning based on CNN to classify the heart sound[J]. Med Biol Eng Comput, 2023, 61(9): 2417-2439.

[67] Rizal A, Fauzi H, Wulandari DR, et al. Classification of normal and abnormal heart sounds using Hjorth features and LSTM—RNN algorithm[J]. J Phys Conf Ser, 2023, 2673(1): 012003.

[68] Malik H, Bashir U, Ahmad A . Multi—classification neural network model for detection of abnormal heartbeat audio signals[J]. Biomed Eng Adv, 2022, 4: 100048.

[69] Zhang HB, Zhang P, Wang ZW, et al. Multi—feature decision fusion network for heart sound abnormality detection and classification[J]. IEEE J Biomed Health Inform, 2024, 28(3): 1386-1397.

[70] Choudhary RR, Singh MR, Jain PK . Heart sound classification using a hybrid of CNN and GRU deep learning models[J]. Procedia Comput Sci, 2024, 235: 3085-3093.

[71] Choudhary RR, Rani M, Kaur R, et al. Heart signal analysis using multistage classification denoising model[J]. J Electr Comput Eng, 2024, 2024(1): 1502285.

[72] Li SZ, Sun J, Yang HB, et al. Interpretable end—to—end heart sound classification[J]. Measurement, 2024, 237: 115113.

[73] Ma KY, Lu JB, Lu BZ . Parameter—efficient densely connected dual attention network for phonocardiogram classification[J]. IEEE J Biomed Health Inform, 2023, 27(9): 4240-4249.

[74] Lin JY, Wang SK, Xuan WP, et al. Portable ECG and PCG wireless acquisition system and multiscale CNN feature fusion Bi—LSTM network for coronary artery disease diagnosis[J]. Comput Biol Med, 2025, 191: 110202.

[75] Lin TY, Wang YX, Liu XY, et al. A survey of transformers[J]. AI Open, 2022, 3: 111-132.

[76] Cheng JW, Sun KX . Heart sound classification network based on convolution and transformer[J]. Sensors (Basel), 2023, 23(19): 8168.

[77] Liu ZH, Liu W, Gu Z, et al. Efficient anomaly detection algorithm for heart sound signal[J]. IEEE Access, 2024, 12: 139225-139236.

[78] Han J, Shaout A . ENACT—heart—ENsemble—based assessment using CNN and transformer on heart sounds[EB/OL]. (2025—02—24). https://arxiv.org/abs/2502.16914.

[79] Wang RS, Duan YF, Li YK, et al. PCTMF—Net: heart sound classification with parallel CNNs—transformer and second—order spectral analysis[J]. Vis Comput, 2023, 39(8): 3811-3822.

[80] Liu ZY, Jiang H, Zhang FW, et al. Heart sound classification based on bispectrum features and vision transformer mode[J]. Alexandria Eng J, 2023, 85: 49-59.

[81] Wang SY, Hu JH, Du YW, et al. WCFormer: an interpretable deep learning framework for heart sound signal analysis and automated diagnosis of cardiovascular diseases[J]. Expert Syst Appl, 2025, 276: 127238.

[82] Ren Z, Chang Y, Nguyen TT, et al. A comprehensive survey on heart sound analysis in the deep learning era[J]. IEEE Comput Intell Mag, 2024, 19(3): 42-57.

[83] Barnawi A, Boulares M, Somai R . Simple and powerful PCG classification method based on selection and transfer learning for precision medicine application[J]. Bioengineering (Basel), 2023, 10(3): 294.

[84] Almanifi ORA, Ab Nasir AF, Mohd Razman MA, et al. Heartbeat murmurs detection in phonocardiogram recordings via transfer learning[J]. Alexandria Eng J, 2022, 61(12): 10995-11002.

[85] Ding SJ, Ding H, Kan MF, et al. A computer—aided heart valve disease diagnosis system based on machine learning[J]. J Healthc Eng, 2023, 2023(1): 7382316.

[86] Harimi A, Ameri MA, Sarkar S, et al. Heart sounds classification: application of a new CyTex inspired method and deep convolutional neural network with transfer learning[J]. Smart Health, 2023, 29: 100416.

[87] Jumphoo T, Phapatanaburi K, Pathonsuwan W, et al. Exploiting data—efficient image transformer—based transfer learning for valvular heart diseases detection[J]. IEEE Access, 2024, 12: 15845-15855.

[88] Maity A, Pathak A, Saha G . Transfer learning based heart valve disease classification from phonocardiogram signal[J]. Biomed Signal Process Control, 2023, 85: 104805.

[89] Wang ZH, Qian K, Liu HG, et al. Exploring interpretable representations for heart sound abnormality detection[J]. Biomed Signal Process Control, 2023, 82: 104569.

[90] 王林琳 . 基于卷积循环神经网络的肺音识别研究[D]. 武汉: 华中师范大学, 2023.

[91] Wang LL . Research on lung sound recognition based on convolutional recurrent neural network[D]. Wuhan: Central China Normal University, 2023.

[92] 张乙鹏 . 基于MFCC和深度学习的肺音分类方法研究[D]. 兰州: 甘肃中医药大学, 2024.

[93] Zhang YP . Research on lung sound classification method based on MFCC and deep learning[D]. Lanzhou: Gansu University of Chinese Medicine, 2024.

[94] 陈茜茹, 钟清华, 张涵, 等 . 基于DSP的心肺音分离系统研究[J]. 现代电子技术, 2019, 42(6): 92-95.

[95] Chen QR, Zhong QH, Zhang H, et al. Research on heart and lung sound separation system based on DSP[J]. Modern Electronics Technique, 2019, 42(6): 92-95.

[96] Zang JB, An Q, Li B, et al. A novel wearable device integrating ECG and PCG for cardiac health monitoring[J]. Microsyst Nanoeng, 2025, 11(1): 7.

[97] De Fazio R, Spongano L, De Vittorio M, et al. Machine learning algorithms for processing and classifying unsegmented phonocardiographic signals: an efficient edge computing solution suitable for wearable devices[J]. Sensors (Basel), 2024, 24(12): 3853.

[98] 丁思吉, 丁皓, 阚孟菲, 等 . 心音信号分类识别算法综述[J]. 软件导刊, 2022, 21(12): 272-278.

[99] Ding SJ, Ding H, Kan MF, et al. A review of classification methods of heart sound signal[J]. Software Guide, 2022, 21(12): 272-278.

[100] Gulati A, Qin J, Chiu CC, et al. Conformer: convolution—augmented transformer for speech recognition[C]// Interspeech 2020. Baixas, France: ISCA, 2020: 5036-5040.

[101] Calzoni A, Savardi M, Silvestri M, et al. Bimodal ECG and PCG cardiovascular disease detection: exploring the potential and modality contribution[J]. J Med Syst, 2025, 49(1): 113.

[102] Li JQ, Wang C, Chen J, et al. Explainable CNN with fuzzy tree regularization for respiratory sound analysis[J]. IEEE Trans Fuzzy Syst, 2022, 30(6): 1516-1528.

[103] Wang A, Chen H, Lin ZJ, et al. LSNet: see large, focus small[C]// 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2025: 9718-9729.

[104] Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: transformers for image recognition at scale[EB/OL]. (2021—06—03). https://arxiv.org/abs/2010.11929.

[105] He HY, Zhang JN, Cai YX, et al. MobileMamba: lightweight multi—receptive visual mamba network[C]// 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2025: 4497-4507.

[106] Chen YP, Dai XY, Chen DD, et al. Mobile—former: bridging MobileNet and transformer[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2022: 5260-5269.

[107] Wu K, Zhang JN, Peng HW, et al. TinyViT: fast pretraining distillation for small vision transformers[C]// Computer Vision—ECCV 2022. Cham: Springer Nature Switzerland, 2022: 68-85.

[108] Chakraborty S, Kochhar P, Patil S, et al. Generative adversarial network augmented data for improved heart sound abnormality detection[J]. Comput Biol Med, 2025, 195: 110623.

基金资助

国家自然科学基金(61901515)

国家自然科学基金(62361038)

兰州市科技发展指导性计划(2025-5-243)

兰州市青年科技人才创新项目(2025-QN-145)