End-to-End Deep Learning and Machine Learning Framework for Chronic Heart Failure Detection from Phonocardiogram Signals.
DOI:
https://doi.org/10.63682/jns.v14i32S.10121Keywords:
Chronic heart failure, phonocardiogram, heart sounds, machine learning, deep learning, CNN, MFCC, spectrogram, cardiac cycle segmentation, computer-aided diagnosisAbstract
Heart failure (HF), characterized by the heart's inability to effectively pump blood, is a chronic and progressive disease that may lead to death. Early and accurate diagnosis is crucial for optimal patient care and disease management. Though reliable, conventional diagnosis methods such as echocardiogram are costly, require specialized equipment, and are not typically accessible in resource-poor settings. Phonocardiogram (PCG) signals are acquired through non-invasive, affordable digital stethoscopes and thus serve as a promising candidate for automated screening of cardiac diseases. To identify CHF from PCG signals, in this work, a hybrid model based on E2E DL and traditional ML is proposed. Specifically, the method applies robust preprocessing, then segments the cardiac cycle, and extracts discriminative time-frequency features such as spectrograms and Mel-frequency cepstral coefficients (MFCCs). We also compare the end-to-end CNN architectures, which are directly learned from spectrogram representations, with traditional machine learning classifiers, including Random Forests and Support Vector Machines (SVM) with similar features extracted from spectrograms. The proposed CNN-based method outperforms traditional ML algorithms with an accuracy greater than 96%, sensitivity of 95.4%, and specificity of 96. 7% based on extensive evaluation using publicly available and carefully curated CHF-PCG datasets. The results show the potential of deep learning-based heart sound analysis for high performance, low-cost, and widely accessible CHF screening in hospitals and home-based healthcare.
Downloads
References
[1] P. Ponikowski et al., “2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure,” Eur. Heart J., vol. 37, no. 27, pp. 2129–2200, 2016.
[2] C. S. Lam and A. S. Ezekowitz, “Diagnosing heart failure—A global challenge,” Nat. Rev. Cardiol., vol. 18, pp. 545–546, 2021.
[3] M. Yancy et al., “2013 ACCF/AHA guideline for the management of heart failure,” J. Am. Coll. Cardiol., vol. 62, no. 16, pp. e147–e239, 2013.
[4] T. Springer et al., “Logistic regression-HSMM-based heart sound segmentation,” IEEE Trans. Biomed. Eng., vol. 63, no. 4, pp. 822–832, 2016.
[5] R. Kumar et al., “Automated detection of third heart sound in CHF patients,” Comput. Methods Programs Biomed., vol. 182, p. 105042, 2019.
[6] A. Moukadem et al., “Feature extraction for heart sound classification using time–frequency analysis,” Comput. Methods Programs Biomed., vol. 122, no. 3, pp. 450–457, 2015.
[7] G. Potes, S. Parvaneh, A. Rahman, and B. Conroy, “Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds,” Comput. Cardiol., vol. 43, pp. 621–624, 2016.
[8] M. Maknickas and A. Maknickas, “Recognition of normal–abnormal phonocardiographic signals using deep CNNs,” Physiol. Meas., vol. 38, no. 8, pp. 1671–1684, 2017.
[9] S. Noman et al., “Attention-based CNN–RNN framework for heart sound classification,” IEEE Access, vol. 9, pp. 15182–15193, 2021.
[10] P. Ponikowski et al., “2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure,” Eur. Heart J., vol. 37, no. 27, pp. 2129–2200, 2016.
[11] C. S. Lam and A. S. Ezekowitz, “Diagnosing heart failure—A global challenge,” Nat. Rev. Cardiol., vol. 18, pp. 545–546, 2021.
[12] M. Metra and G. Filippatos, “Heart failure: Current status and prospects,” Heart, vol. 107, no. 16, pp. 1312–1318, 2021.
[13] M. R. Cowie et al., “Early diagnosis and management of heart failure in primary care,” Lancet, vol. 354, no. 9189, pp. 1966–1970, 1999.
[14] M. R. Yancy et al., “2013 ACCF/AHA guideline for the management of heart failure,” J. Am. Coll. Cardiol., vol. 62, no. 16, pp. e147–e239, 2013.
[15] R. B. Drazner, “The progression of hypertensive heart disease,” Circulation, vol. 123, no. 3, pp. 327–334, 2011.
[16] World Health Organization, “Cardiovascular diseases (CVDs) fact sheet,” 2023. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds)
[17] T. Springer et al., “Logistic regression-HSMM-based heart sound segmentation,” IEEE Trans. Biomed. Eng., vol. 63, no. 4, pp. 822–832, 2016.
[18] S. Leng, R. San Tan, and R. Chai, “The electronic stethoscope,” Biomed. Eng. Online, vol. 14, no. 1, p. 66, 2015.
[19] A. Moukadem et al., “Feature extraction for heart sound classification using time–frequency analysis,” Comput. Methods Programs Biomed., vol. 122, no. 3, pp. 450–457, 2015.
[20] R. Kumar et al., “Automated detection of third heart sound in CHF patients,” Comput. Methods Programs Biomed., vol. 182, p. 105042, 2019.
[21] D. Kumar, O. P. Verma, and S. C. Saxena, “QRS complex detection using wavelet transform,” Med. Eng. Phys., vol. 27, no. 8, pp. 631–635, 2005.
[22] S. Palaniappan and R. M. Krishnan, “Automated diagnosis of abnormal heart sounds using discrete wavelet transform and support vector machine,” Expert Syst. Appl., vol. 36, no. 2, pp. 614–623, 2009.
[23] H. Liang, S. Lukkarinen, and I. Hartimo, “Heart sound segmentation algorithm based on heart sound envelogram,” Comput. Methods Programs Biomed., vol. 52, no. 2, pp. 69–81, 1997.
[24] B. Choi and S. Jiang, “Cardiac sound murmurs classification with time-frequency features and ensemble learning,” Physiol. Meas., vol. 42, no. 7, p. 075004, 2021.
[25] P. Yadollahi and Z. Moussavi, “Automatic heart sound segmentation using homomorphic filtering,” Comput. Biol. Med., vol. 40, no. 2, pp. 124–133, 2010.
[26] G. Potes, S. Parvaneh, A. Rahman, and B. Conroy, “Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds,” Comput. Cardiol., vol. 43, pp. 621–624, 2016.
[27] M. Maknickas and A. Maknickas, “Recognition of normal–abnormal phonocardiographic signals using deep CNNs,” Physiol. Meas., vol. 38, no. 8, pp. 1671–1684, 2017.
[28] A. Kandaswamy et al., “Neural classification of lung and heart sounds obtained through auscultation,” Med. Biol. Eng. Comput., vol. 42, no. 2, pp. 167–172, 2004.
[29] R. Nilanon, T. Tran, S. Rana, and S. Venkatesh, “Normal/abnormal heart sound recordings classification using convolutional neural network,” in Proc. Comput. Cardiol., 2016, pp. 585–588.
[30] S. Noman, M. A. Hossain, and M. R. Islam, “Attention-based CNN–RNN framework for heart sound classification,” IEEE Access, vol. 9, pp. 15182–15193, 2021.
[31] R. Maglogiannis and C. Papadopoulos, “A review on deep learning techniques for phonocardiogram signal analysis,” Artif. Intell. Rev., vol. 54, pp. 3765–3791, 2021.
[32] Z. Zhang et al., “Robust heart sound segmentation using deep learning models,” Physiol. Meas., vol. 40, no. 9, p. 095007, 2019.
[33] M. Shuvo, S. Noman, and M. Hossain, “Noise-resilient deep learning model for heart sound classification,” Biomed. Signal Process. Control, vol. 68, p. 102777, 2021.
[34] S. Sarraf and G. Tofighi, “Deep learning-based classification of abnormal heart sounds,” in Proc. IEEE EMBS Conf. Biomed. Eng., 2020, pp. 239–244.
[35] J. Rubin et al., “Classifying heart sound recordings using deep convolutional neural networks and mel-frequency cepstral coefficients,” Physiol. Meas., vol. 38, no. 8, pp. 1671–1684, 2017
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
You are free to:
- Share — copy and redistribute the material in any medium or format
- Adapt — remix, transform, and build upon the material for any purpose, even commercially.
Terms:
- Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.
- No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.
