{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T06:09:04Z","timestamp":1790834944125,"version":"4.1.0"},"reference-count":34,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,8,26]],"date-time":"2026-08-26T00:00:00Z","timestamp":1787702400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100004242","name":"Princess Nourah Bint Abdulrahman University","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100004242","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Physiol."],"abstract":"<jats:p>Heart disease continues to be the world\u2019s number one killer, making low-cost, early screening a priority. A phonocardiogram (PCG), a simple recording of sounds made by the heart, provides valuable clues to conditions such as murmurs and valve defects and can be recorded cheaply using a digital stethoscope. With less accessibility to specialists, the automatic analysis of heart sounds is preferred for screening and tele-health environments. We compare five deep convolutional neural networks in this work, namely LeNet5, AlexNet, VGG16, ResNet-50 and a hybrid LSTM-CNN, on the task of separating normal from abnormal heart sounds. Before entering the network, each recording is converted to a log-Mel spectrogram, and each model is trained at five different learning rates (1e-2, 1e-3, 5e-4, 1e-4 and 5e-5). The PhysioNet\/CinC Challenge 2016 database is common to all experiments and, after duplicate removal, provides 3, 240 recordings. A leakage-free protocol is enforced throughout: exact duplicates are detected by audio content hashing and the split is patient-independent, which on this dataset is sufficient to prevent several networks from reaching a misleading 100% accuracy. We report accuracy, AUC, sensitivity, specificity, F1 score and the Matthews correlation coefficient. Under five-fold cross-validation VGG16 attains the highest mean accuracy (91.94% \u00b1 1.59, AUC 0.968, MCC 0.745), marginally ahead of ResNet-50 (90.93% \u00b1 1.09); McNemar\u2019s test confirms that the difference between the leading architectures is not statistically significant (p = 0.31), so the two are reported as jointly strongest. An ablation study shows that spectrogram augmentation, rather than ImageNet pre-training, is the decisive design choice, and every model fails at a learning rate of 1e-2. External validation on the CirCor DigiScope 2022 database shows that zero-shot cross-dataset transfer performs at chance level, while the retrained pipeline exceeds the majority baseline, underlining how strongly single-dataset results can overstate real-world generalisation.<\/jats:p>","DOI":"10.3389\/fphys.2026.1927399","type":"journal-article","created":{"date-parts":[[2026,8,26]],"date-time":"2026-08-26T07:55:39Z","timestamp":1787730939000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["An empirical comparison of deep CNN architectures for heart sound classification from phonocardiogram signals"],"prefix":"10.3389","volume":"17","author":[{"given":"Arar","family":"Al Tawil","sequence":"first","affiliation":[{"name":"Department of Computer Science, Faculty of Information Technology, Applied Science Private University","place":["Amman, Jordan"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bodor","family":"Bin Sheeha","sequence":"additional","affiliation":[{"name":"Department of Rehabilitation Sciences, College of Health and Rehabilitation Sciences, Princess Nourah Bint Abdulrahman University","place":["Riyadh, Saudi Arabia"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Aseel","family":"Aburub","sequence":"additional","affiliation":[{"name":"Department of Physiotherapy, Faculty of Allied Medical Sciences, Applied Science Private University","place":["Amman, Jordan"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,8,26]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"5443","DOI":"10.32604\/cmc.2022.025526","article-title":"Arabic music genre classification using deep convolutional neural networks (CNNs)","volume":"72","author":"Almazaydeh","year":"2022","journal-title":"Computers Materials Continua"},{"key":"B2","doi-asserted-by":"publisher","first-page":"123099","DOI":"10.1109\/ACCESS.2024.3451660","article-title":"A hybrid framework of transformer encoder and residential conventional for cardiovascular disease recognition using heart sounds","volume":"12","author":"Al-Tam","year":"2024","journal-title":"IEEE Access"},{"key":"B3","doi-asserted-by":"publisher","DOI":"10.3390\/diagnostics15182379","article-title":"Improving the detection performance of cardiovascular diseases from heart sound signals with a new deep learning-based approach","volume":"15","author":"Ay","year":"2025","journal-title":"Diagnostics"},{"key":"B4","doi-asserted-by":"crossref","first-page":"437","DOI":"10.1007\/978-3-642-35289-8_26","article-title":"Practical recommendations for gradient-based training of deep architectures","volume-title":"Neural Networks: Tricks of the Trade","author":"Bengio","year":"2012"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.3390\/e23060667","article-title":"Deep learning methods for heart sounds classification: A systematic review","volume":"23","author":"Chen","year":"2021","journal-title":"Entropy"},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.3389\/frai.2024.1434022","article-title":"A review on deep learning methods for heart sound signal analysis","volume":"7","author":"Chen","year":"2024","journal-title":"Front. Artif. Intell."},{"key":"B7","first-page":"609","article-title":"Classification of normal\/abnormal heart sound recordings: The PhysioNet\/Computing in Cardiology Challenge 2016","author":"Clifford","year":"2016"},{"key":"B8","doi-asserted-by":"publisher","first-page":"357","DOI":"10.1109\/TASSP.1980.1163420","article-title":"Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences","volume":"28","author":"Davis","year":"1980","journal-title":"IEEE Trans. Acoustics Speech Signal. Process."},{"key":"B9","first-page":"248","article-title":"ImageNet: A large-scale hierarchical image database","author":"Deng","year":"2009"},{"key":"B10","doi-asserted-by":"publisher","DOI":"10.1016\/j.bspc.2020.102211","article-title":"Heart sound classification with signal instant energy and stacked autoencoder network","volume":"64","author":"Deperlioglu","year":"2021","journal-title":"Biomed. Signal Process. Control"},{"key":"B11","volume-title":"Deep Learning","author":"Goodfellow","year":"2016"},{"key":"B12","first-page":"770","article-title":"Deep residual learning for image recognition","author":"He","year":"2016"},{"key":"B13","doi-asserted-by":"publisher","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"Hochreiter","year":"1997","journal-title":"Neural Comput."},{"key":"B14","doi-asserted-by":"publisher","DOI":"10.3389\/fphys.2025.1596150","article-title":"Heart sound classification based on convolutional neural network with convolutional block attention module","volume":"16","author":"Huai","year":"2025","journal-title":"Front. Physiol."},{"key":"B15","article-title":"Adam: A method for stochastic optimization","author":"Kingma","year":"2015"},{"key":"B16","first-page":"1097","article-title":"ImageNet classification with deep convolutional neural networks","author":"Krizhevsky","year":"2012"},{"key":"B17","doi-asserted-by":"publisher","first-page":"2278","DOI":"10.1109\/5.726791","article-title":"Gradient-based learning applied to document recognition","volume":"86","author":"LeCun","year":"1998","journal-title":"Proc. IEEE"},{"key":"B18","doi-asserted-by":"publisher","DOI":"10.3389\/fphys.2022.1084420","article-title":"Heart sound classification based on improved mel-frequency spectral coefficients and deep residual learning","volume":"13","author":"Li","year":"2022","journal-title":"Front. Physiol."},{"key":"B19","doi-asserted-by":"publisher","first-page":"2181","DOI":"10.1088\/0967-3334\/37\/12\/2181","article-title":"An open access database for the evaluation of heart sound algorithms","volume":"37","author":"Liu","year":"2016","journal-title":"Physiol. Meas."},{"key":"B20","article-title":"SGDR: Stochastic gradient descent with warm restarts","author":"Loshchilov","year":"2017"},{"key":"B21","doi-asserted-by":"publisher","first-page":"1671","DOI":"10.1088\/1361-6579\/aa7841","article-title":"Recognition of normal-abnormal phonocardiographic signals using deep convolutional neural networks and mel-frequency spectral coefficients","volume":"38","author":"Maknickas","year":"2017","journal-title":"Physiol. Meas."},{"key":"B22","first-page":"18","article-title":"librosa: Audio and music signal analysis in Python","author":"McFee","year":"2015"},{"key":"B23","first-page":"585","article-title":"Normal\/abnormal heart sound recordings classification using convolutional neural network","author":"Nilanon","year":"2016"},{"key":"B24","doi-asserted-by":"publisher","first-page":"2524","DOI":"10.1109\/JBHI.2021.3137048","article-title":"The CirCor DigiScope dataset: from murmur detection to murmur classification","volume":"26","author":"Oliveira","year":"2022","journal-title":"IEEE J. Biomed. Health Inf."},{"key":"B25","first-page":"2613","article-title":"SpecAugment: A simple data augmentation method for automatic speech recognition","author":"Park","year":"2019"},{"key":"B26","first-page":"621","article-title":"Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds","author":"Potes","year":"2016"},{"key":"B27","doi-asserted-by":"publisher","DOI":"10.1080\/21642583.2024.2420912","article-title":"LightCardiacNet: light-weight deep ensemble network with attention mechanism for cardiac sound classification","volume":"12","author":"Rong","year":"2024","journal-title":"Syst. Sci. Control Eng."},{"key":"B28","first-page":"813","article-title":"Classifying heart sound recordings using deep convolutional neural networks and mel-frequency cepstral coefficients","author":"Rubin","year":"2016"},{"key":"B29","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2015"},{"key":"B30","first-page":"464","article-title":"Cyclical learning rates for training neural networks","author":"Smith","year":"2017"},{"key":"B31","article-title":"A disciplined approach to neural network hyper-parameters: Part 1 - learning rate, batch size, momentum, and weight decay","volume-title":"Arxiv:1803.09820","author":"Smith","year":"2018"},{"key":"B32","article-title":"Do not decay the learning rate; increase the batch size","author":"Smith","year":"2018"},{"key":"B33","volume-title":"Cardiovascular Diseases (Cvds)","year":"2023"},{"key":"B34","doi-asserted-by":"publisher","first-page":"153","DOI":"10.1016\/j.neucom.2018.09.101","article-title":"Heart sounds classification using a novel 1-D convolutional neural network with extremely low parameter consumption","volume":"392","author":"Xiao","year":"2020","journal-title":"Neurocomputing"}],"container-title":["Frontiers in Physiology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fphys.2026.1927399\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,26]],"date-time":"2026-08-26T07:55:41Z","timestamp":1787730941000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fphys.2026.1927399\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,26]]},"references-count":34,"alternative-id":["10.3389\/fphys.2026.1927399"],"URL":"https:\/\/doi.org\/10.3389\/fphys.2026.1927399","relation":{},"ISSN":["1664-042X"],"issn-type":[{"value":"1664-042X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,26]]},"article-number":"1927399"}}