{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,18]],"date-time":"2026-04-18T03:00:27Z","timestamp":1776481227745,"version":"3.51.2"},"reference-count":41,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2020,9,9]],"date-time":"2020-09-09T00:00:00Z","timestamp":1599609600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2020,9,9]],"date-time":"2020-09-09T00:00:00Z","timestamp":1599609600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["npj Digit. Med."],"abstract":"<jats:title>Abstract<\/jats:title><jats:p>Wearable devices enable theoretically continuous, longitudinal monitoring of physiological measurements such as step count, energy expenditure, and heart rate. Although the classification of abnormal cardiac rhythms such as atrial fibrillation from wearable devices has great potential, commercial algorithms remain proprietary and tend to focus on heart rate variability derived from green spectrum LED sensors placed on the wrist, where noise remains an unsolved problem. Here we develop DeepBeat, a multitask deep learning method to jointly assess signal quality and arrhythmia event detection in wearable photoplethysmography devices for real-time detection of atrial fibrillation. The model is trained on approximately one million simulated unlabeled physiological signals and fine-tuned on a curated dataset of over 500\u2009K labeled signals from over 100 individuals from 3 different wearable devices. We demonstrate that, in comparison with a single-task model, our architecture using unsupervised transfer learning through convolutional denoising autoencoders dramatically improves the performance of atrial fibrillation detection from a F1 score of 0.54 to 0.96. We also include in our evaluation a prospectively derived replication cohort of ambulatory participants where the algorithm performed with high sensitivity (0.98), specificity (0.99), and F1 score (0.93). We show that two-stage training can help address the unbalanced data problem common to biomedical applications, where large-scale well-annotated datasets are hard to generate due to the expense of manual annotation, data acquisition, and participant privacy.<\/jats:p>","DOI":"10.1038\/s41746-020-00320-4","type":"journal-article","created":{"date-parts":[[2020,9,10]],"date-time":"2020-09-10T11:32:35Z","timestamp":1599737555000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":122,"title":["Multi-task deep learning for cardiac rhythm detection in wearable devices"],"prefix":"10.1038","volume":"3","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0490-4448","authenticated-orcid":false,"given":"Jessica","family":"Torres-Soto","sequence":"first","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9418-9577","authenticated-orcid":false,"given":"Euan A.","family":"Ashley","sequence":"additional","affiliation":[]}],"member":"297","published-online":{"date-parts":[[2020,9,9]]},"reference":[{"key":"320_CR1","doi-asserted-by":"publisher","first-page":"1561","DOI":"10.1016\/j.hrthm.2018.06.037","volume":"15","author":"AD William","year":"2018","unstructured":"William, A. D. et al. Assessing the accuracy of an automated atrial fibrillation detection algorithm using smartphone technology: The iREAD Study. Heart Rhythm 15, 1561\u20131565 (2018).","journal-title":"Heart Rhythm"},{"key":"320_CR2","doi-asserted-by":"crossref","unstructured":"Shcherbina, A. et al. Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. J. Pers. Med. 7, 3 (2017).","DOI":"10.3390\/jpm7020003"},{"key":"320_CR3","doi-asserted-by":"publisher","DOI":"10.1161\/CIRCEP.118.006834","volume":"12","author":"J Wasserlauf","year":"2019","unstructured":"Wasserlauf, J. et al. Smartwatch performance for the detection and quantification of atrial fibrillation. Circ. Arrhythm. Electrophysiol. 12, e006834 (2019).","journal-title":"Circ. Arrhythm. Electrophysiol."},{"key":"320_CR4","doi-asserted-by":"publisher","first-page":"566","DOI":"10.1109\/TBME.2005.869784","volume":"53","author":"BS Kim","year":"2006","unstructured":"Kim, B. S. & Yoo, S. K. Motion artifact reduction in photoplethysmography using independent component analysis. IEEE Trans. Biomed. Eng. 53, 566\u2013568 (2006).","journal-title":"IEEE Trans. Biomed. Eng."},{"key":"320_CR5","doi-asserted-by":"publisher","first-page":"522","DOI":"10.1109\/TBME.2014.2359372","volume":"62","author":"Z Zhang","year":"2015","unstructured":"Zhang, Z., Pi, Z. & Liu, B. TROIKA: a general framework for heart rate monitoring using wrist-type photoplethysmographic signals during intensive physical exercise. IEEE Trans. Biomed. Eng. 62, 522\u2013531 (2015).","journal-title":"IEEE Trans. Biomed. Eng."},{"key":"320_CR6","doi-asserted-by":"publisher","DOI":"10.1136\/bmjopen-2016-013535","volume":"7","author":"M Lown","year":"2017","unstructured":"Lown, M., Yue, A., Lewith, G., Little, P. & Moore, M. Screening for Atrial Fibrillation using Economical and accurate TechnologY (SAFETY)\u2014a pilot study. BMJ Open 7, e013535 (2017).","journal-title":"BMJ Open"},{"key":"320_CR7","doi-asserted-by":"crossref","unstructured":"Elgendi, M. Optimal signal quality index for photoplethysmogram signals. Bioengineering 3, 21 (2016).","DOI":"10.3390\/bioengineering3040021"},{"key":"320_CR8","doi-asserted-by":"publisher","DOI":"10.1136\/heartjnl-2018-313147","author":"M-Z Poh","year":"2018","unstructured":"Poh, M.-Z. et al. Diagnostic assessment of a deep learning system for detecting atrial fibrillation in pulse waveforms. Heart https:\/\/doi.org\/10.1136\/heartjnl-2018-313147 (2018).","journal-title":"Heart"},{"key":"320_CR9","first-page":"753","volume":"19","author":"L Krivoshei","year":"2017","unstructured":"Krivoshei, L. et al. Smart detection of atrial fibrillation. Europace 19, 753\u2013757 (2017).","journal-title":"Europace"},{"key":"320_CR10","doi-asserted-by":"publisher","first-page":"787","DOI":"10.1088\/1361-6579\/aa5dd7","volume":"38","author":"VDA Corino","year":"2017","unstructured":"Corino, V. D. A. et al. Detection of atrial fibrillation episodes using a wristband device. Physiol. Meas. 38, 787\u2013799 (2017).","journal-title":"Physiol. Meas."},{"key":"320_CR11","doi-asserted-by":"publisher","DOI":"10.1038\/srep45644","volume":"7","author":"S-C Tang","year":"2017","unstructured":"Tang, S.-C. et al. Identification of atrial fibrillation by quantitative analyses of fingertip photoplethysmogram. Sci. Rep. 7, 45644 (2017).","journal-title":"Sci. Rep."},{"key":"320_CR12","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-019-49092-2","volume":"9","author":"SK Bashar","year":"2019","unstructured":"Bashar, S. K. et al. Atrial fibrillation detection from wrist photoplethysmography signals using smartwatches. Sci. Rep. 9, 15054 (2019).","journal-title":"Sci. Rep."},{"key":"320_CR13","doi-asserted-by":"crossref","unstructured":"Madani, A., Arnaout, R., Mofrad, M. & Arnaout, R. Fast and accurate view classification of echocardiograms using deep learning. NPJ Digit. Med. 1, 6 (2018).","DOI":"10.1038\/s41746-017-0013-1"},{"key":"320_CR14","doi-asserted-by":"publisher","first-page":"1623","DOI":"10.1161\/CIRCULATIONAHA.118.034338","volume":"138","author":"J Zhang","year":"2018","unstructured":"Zhang, J. et al. Fully automated echocardiogram interpretation in clinical practice. Circulation 138, 1623\u20131635 (2018).","journal-title":"Circulation"},{"key":"320_CR15","doi-asserted-by":"publisher","first-page":"2668","DOI":"10.1016\/j.jacc.2018.03.521","volume":"71","author":"KW Johnson","year":"2018","unstructured":"Johnson, K. W. et al. Artificial intelligence in cardiology. J. Am. Coll. Cardiol. 71, 2668\u20132679 (2018).","journal-title":"J. Am. Coll. Cardiol."},{"key":"320_CR16","doi-asserted-by":"publisher","first-page":"3","DOI":"10.1038\/s41746-019-0207-9","volume":"3","author":"T Pereira","year":"2020","unstructured":"Pereira, T. et al. Photoplethysmography based atrial fibrillation detection: a review. NPJ Digit. Med. 3, 3 (2020).","journal-title":"NPJ Digit. Med."},{"key":"320_CR17","doi-asserted-by":"crossref","unstructured":"Shashikumar, S. P., Shah, A. J., Li, Q., Clifford, G. D. & Nemati, S. A deep learning approach to monitoring and detecting atrial fibrillation using wearable technology. In 2017 IEEE EMBS International Conference on Biomedical Health Informatics (BHI) 141\u2013144 (2017).","DOI":"10.1109\/BHI.2017.7897225"},{"key":"320_CR18","doi-asserted-by":"publisher","first-page":"409","DOI":"10.1001\/jamacardio.2018.0136","volume":"3","author":"GH Tison","year":"2018","unstructured":"Tison, G. H. et al. Passive detection of atrial fibrillation using a commercially available smartwatch. JAMA Cardiol. 3, 409\u2013416 (2018).","journal-title":"JAMA Cardiol."},{"key":"320_CR19","doi-asserted-by":"publisher","first-page":"66","DOI":"10.1016\/j.ahj.2018.09.002","volume":"207","author":"MP Turakhia","year":"2019","unstructured":"Turakhia, M. P. et al. Rationale and design of a large-scale, app-based study to identify cardiac arrhythmias using a smartwatch: the Apple Heart Study. Am. Heart J. 207, 66\u201375 (2019).","journal-title":"Am. Heart J."},{"key":"320_CR20","doi-asserted-by":"publisher","unstructured":"Shen, Y. et al. Ambulatory Atrial Fibrillation Monitoring Using Wearable Photoplethysmography with Deep Learning. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining 1909\u20131916, (Association for Computing Machinery, 2019). https:\/\/doi.org\/10.1145\/3292500.3330657.","DOI":"10.1145\/3292500.3330657"},{"key":"320_CR21","doi-asserted-by":"publisher","unstructured":"Gotlibovych, I. et al. End-to-end deep learning from raw sensor data: atrial fibrillation detection using wearables. arXiv [stat.ML], https:\/\/doi.org\/10.1145\/nnnnnnn.nnnnnnn (2018).","DOI":"10.1145\/nnnnnnn.nnnnnnn"},{"key":"320_CR22","unstructured":"Ochiai, K., Takahashi, S. & Fukazawa, Y. Arrhythmia detection from 2-lead ECG using convolutional denoising autoencoders (2018)."},{"key":"320_CR23","doi-asserted-by":"publisher","first-page":"861","DOI":"10.21105\/joss.00861","volume":"3","author":"L McInnes","year":"2018","unstructured":"McInnes, L., Healy, J., Saul, N. & Gro\u00dfberger, L. UMAP: Uniform Manifold Approximation and Projection. JOSS 3, 861 (2018).","journal-title":"JOSS"},{"key":"320_CR24","unstructured":"Hadley, D., Hsu, D., Pickham, D., Drezner, J. A. & Froelicher, V. F. QT Corrections for Long QT Risk Assessment: Implicationsfor the Preparticipation Examination. Clin. J. Sport Med. 29, 285\u2013291 (2019)."},{"key":"320_CR25","first-page":"625","volume":"11","author":"D Erhan","year":"2010","unstructured":"Erhan, D. et al. Why does unsupervised pre-training help deep learning? J. Mach. Learn. Res. 11, 625\u2013660 (2010).","journal-title":"J. Mach. Learn. Res."},{"key":"320_CR26","unstructured":"Raghu, M., Zhang, C., Kleinberg, J. & Bengio, S. Transfusion: Understanding Transfer Learning for Medical Imaging. In: H. Wallach (ed.) Advances in Neural Information Processing Systems 32, 3347\u20133357 (Curran Associates, Inc., 2019)."},{"key":"320_CR27","doi-asserted-by":"crossref","unstructured":"Li, S., Liu, Z.-Q. & Chan, A. B. Heterogeneous multi-task learning for human pose estimation with deep convolutional neural network. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops 482\u2013489 (2014).","DOI":"10.1109\/CVPRW.2014.78"},{"key":"320_CR28","doi-asserted-by":"publisher","first-page":"1909","DOI":"10.1056\/NEJMoa1901183","volume":"381","author":"MV Perez","year":"2019","unstructured":"Perez, M. V. et al. Large-scale assessment of a smartwatch to identify atrial fibrillation. N. Engl. J. Med. 381, 1909\u20131917 (2019).","journal-title":"N. Engl. J. Med."},{"key":"320_CR29","doi-asserted-by":"publisher","first-page":"669","DOI":"10.1088\/1361-6579\/aa670e","volume":"38","author":"PH Charlton","year":"2017","unstructured":"Charlton, P. H. et al. Extraction of respiratory signals from the electrocardiogram and photoplethysmogram: technical and physiological determinants. Physiol. Meas. 38, 669\u2013690 (2017).","journal-title":"Physiol. Meas."},{"key":"320_CR30","unstructured":"IEEE SP Cup 2015 - IEEE Signal Processing Society. http:\/\/archive.signalprocessingsociety.org\/community\/sp-cup\/ieee-sp-cup-2015\/ (2015)."},{"key":"320_CR31","doi-asserted-by":"publisher","first-page":"1946","DOI":"10.1109\/TBME.2013.2246160","volume":"60","author":"W Karlen","year":"2013","unstructured":"Karlen, W., Raman, S., Ansermino, J. M. & Dumont, G. A. Multiparameter respiratory rate estimation from the photoplethysmogram. IEEE Trans. Biomed. Eng. 60, 1946\u20131953 (2013).","journal-title":"IEEE Trans. Biomed. Eng."},{"key":"320_CR32","doi-asserted-by":"crossref","unstructured":"Gondara, L. Medical image denoising using convolutional denoising autoencoders. In 2016 IEEE 16th International Conference on Data Mining Workshops (ICDMW) 241\u2013246 (2016).","DOI":"10.1109\/ICDMW.2016.0041"},{"key":"320_CR33","doi-asserted-by":"crossref","unstructured":"He, K., Zhang, X., Ren, S. & Sun, J. Delving deep into rectifiers: surpassing human-level performance on ImageNet classification. In 2015 IEEE International Conference on Computer Vision (ICCV) 1026\u20131034 (2015).","DOI":"10.1109\/ICCV.2015.123"},{"key":"320_CR34","unstructured":"Kingma, D. P. & Ba, J. Adam: a method for stochastic optimization. arXiv [cs.LG] (2014)."},{"key":"320_CR35","doi-asserted-by":"crossref","unstructured":"Zeiler, M. D. & Fergus, R. Visualizing and understanding convolutional networks. In Computer Vision \u2013 ECCV 2014 818\u2013833 (Springer International Publishing, 2014).","DOI":"10.1007\/978-3-319-10590-1_53"},{"key":"320_CR36","unstructured":"Yosinski, J., Clune, J., Bengio, Y. & Lipson, H. In Advances in Neural Information Processing Systems Vol. 27 (eds Ghahramani, Z., Welling, M., Cortes, C., Lawrence, N. D. & Weinberger, K. Q.) 3320\u20133328 (Curran Associates, Inc., 2014)."},{"key":"320_CR37","unstructured":"Maas, A. L., Hannun, A. Y. & Ng, A. Y. Rectifier nonlinearities improve neural network acoustic models. In Proc. icml Vol. 30, 3 (2013)."},{"key":"320_CR38","unstructured":"Ioffe, S. & Szegedy, C. Batch normalization: accelerating deep network training by reducing internal covariate shift. In Proceedings of the 32nd International Conference on International Conference on Machine Learning - Volume 37, 448\u2013456 (JMLR.org, 2015)."},{"key":"320_CR39","unstructured":"Bergstra, J., Yamins, D. & Cox, D. D. Making a science of model search: hyperparameter optimization in hundreds of dimensions for vision architectures. In Proceedings of the 30th International Conference on International Conference on Machine Learning - Volume 28, I-115\u2013I-123 (JMLR.org, 2013)."},{"key":"320_CR40","unstructured":"O\u2019Donoghue, J. Introducing: Blocks and Fuel \u2013 Frameworks for Deep Learning in Python \u2013 KDnuggets. https:\/\/www.kdnuggets.com\/2015\/10\/blocks-fuel-deep-learning-frameworks.html."},{"key":"320_CR41","unstructured":"Abadi, M. et al. Tensorflow: Large-scale machine learning on heterogeneous distributed systems. arXiv 2016: 1603. 04467 (2019)."}],"container-title":["npj Digital Medicine"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41746-020-00320-4.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41746-020-00320-4","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41746-020-00320-4.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,12,7]],"date-time":"2022-12-07T02:33:39Z","timestamp":1670380419000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41746-020-00320-4"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,9,9]]},"references-count":41,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2020,12]]}},"alternative-id":["320"],"URL":"https:\/\/doi.org\/10.1038\/s41746-020-00320-4","relation":{},"ISSN":["2398-6352"],"issn-type":[{"value":"2398-6352","type":"electronic"}],"subject":[],"published":{"date-parts":[[2020,9,9]]},"assertion":[{"value":"24 January 2020","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"22 July 2020","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"9 September 2020","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"E.A.A. reports advisory board fees from Apple, DeepCell, Myokardia, and Personalis, outside the submitted work. J.T.S. declares no competing interests.","order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"116"}}