Abstract 21042: Cardiovascular Risk Stratification Using Off-the-Shelf Wearables and a Multi-Task Deep Learning Algorithm
Abstract
Introduction: We aimed to evaluate whether a novel deep neural network (DNN) can predict cardiovascular risk factors from off-the-shelf wearables with a photoplethysmographic (PPG) heart rate sensor and accelerometer. Longitudinal heart rate variability and activity patterns have previously been associated with incident hypertension, diabetes, and sleep apnea, conditions which are frequently undiagnosed. Methods: Health eHeart, an IRB-approved UCSF study, enrolled 6,115 active users of the Cardiogram app for Apple Watch. Heart rate and step counts were collected for a period of 1 to 53 weeks (mean=8.9). Data from 70% of participants (33,628 person-weeks of data) was used to train a semi-supervised, multi-task DNN with both convolutional and recurrent layers to simultaneously predict prevalent hypertension, sleep apnea, and diabetes. Test performance characteristics were estimated using the remaining 30% of participants. Results: Mean age was 42.3 ± 12.1, 69% male. 2,230 (36.5%) of participants had hypertension, 1,016 (16.6%) had sleep apnea, and 462 (7.6%) had diabetes. In the validation set, the DNN outperformed a baseline logistic regression model incorporating age, sex, and beta blocker use, predicting prevalent hypertension with a c-statistic of 0.819 (95% CI 0.76-0.88; with an optimal operating point yielding 84.8% sensitivity and 63.6% specificity) vs a baseline c-statistic of 0.682 (95% CI 0.60-0.76), and prevalent sleep apnea with a c-statistic of 0.902 (95% CI 0.85-0.95; with an optimal operating point yielding 90.4% sensitivity and 59.8% specificity) vs a baseline c-statistic of 0.459 (95% CI 0.39-0.53). Results were not statistically significant for diabetes. Conclusions: Our DNN demonstrates surprisingly good prediction of hypertension and sleep apnea given that its only inputs are heart rate and step count. Whether such DNNs can provide durable and portable predictions for these conditions in other study samples is worth pursuing.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Researchers are increasingly using consumer wearables or wrist-worn smartwatches and fitness monitors for digital measurement of cardiovascular health and psychophysiological stress because these devices allow continuous, scalable, unobtrusive big data collection of cardiac activity in real-world conditions with large samples.
“Researchers are increasingly using consumer wearables or wrist-worn smartwatches and fitness monitors for the digital measurement of cardiovascular health and psychophysiological stress as these devices allow for continuous, scalable, and unobtrusive, “big data” collection of overall cardiac activity in real-world conditions with large samples25–30.”
From Guidelines for wrist-worn consumer wearable assessment of heart rate in biobehavioral research (2020) · cites this paper as [30]
Cited by
Sorted by citations
Wearables and the Medical Revolution
Dunn, J., Runge, R., & Snyder, M. (2018) · Personalized Medicine · doi:10.2217/pme-2018-0044 · Cited by 654 · see citing works
Hong, S., Zhou, Y., Shang, J., Xiao, C., & Sun, J. (2020) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2020.103801 · Cited by 488 · see citing works
Guidelines for wrist-worn consumer wearable assessment of heart rate in biobehavioral research
Nelson, B. W., Low, C. A., Jacobson, N., Areán, P., Torous, J., & Allen, N. B. (2020) · npj Digital Medicine · doi:10.1038/s41746-020-0297-4 · Cited by 241 · see citing works
Kitt, J., Fox, R., Tucker, K. L., & McManus, R. J. (2019) · Current Hypertension Reports · doi:10.1007/s11906-019-0949-4 · Cited by 147 · see citing works
DeepHeart: Semi-Supervised Sequence Learning for Cardiovascular Risk Prediction
Ballinger, B., Hsieh, J., Singh, A., Sohoni, N., Wang, J., Tison, G., Marcus, G., Sanchez, J., Maguire, C., Olgin, J., et al. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11891 · Cited by 142 · see citing works
“We are currently deploying our algorithm in ambulatory, mobile trials in partnership with major medical centers to show real-world efficacy and build a rigorous clinical evidence base (Tison et al. 2017b).”
MethodInfluentialSelf-citationArtificial intelligence in sleep medicine: background and implications for clinicians
Goldstein, C. A., Berry, R. B., Kent, D. T., Kristo, D. A., Seixas, A. A., Redline, S. S., & Westover, M. B. (2020) · Journal of Clinical Sleep Medicine · doi:10.5664/jcsm.8388 · Cited by 145 · see citing works
Artificial Intelligence and Hypertension: Recent Advances and Future Outlook
Chaikijurajai, T., Laffin, L. J., & Tang, W. H. W. (2020) · American Journal of Hypertension · doi:10.1093/ajh/hpaa102 · Cited by 92 · see citing works
Artificial Intelligence in Hypertension Management: An Ace up Your Sleeve
Visco, V., Izzo, C., Mancusi, C., Rispoli, A., Tedeschi, M., Virtuoso, N., Giano, A., Gioia, R., Melfi, A., Serio, B., et al. (2023) · Journal of Cardiovascular Development and Disease · doi:10.3390/jcdd10020074 · Cited by 81 · see citing works
“In particular, BP can be estimated from a PPG signal obtained from a smartphone or a smartwatch by using DL [54,55].”
BackgroundLim, W. K., Davila, S., Teo, J. X., Yang, C., Pua, C. J., Blöcker, C., Lim, J. Q., Ching, J., Yap, J. J. L., Tan, S. Y., et al. (2018) · PLOS Biology · doi:10.1371/journal.pbio.2004285 · Cited by 70 · see citing works
Quantitative detection of sleep apnea with wearable watch device
Hayano, J., Yamamoto, H., Nonaka, I., Komazawa, M., Itao, K., Ueda, N., 田中, 春., & Yuda, E. (2020) · PLoS ONE · doi:10.1371/journal.pone.0237279 · Cited by 67 · see citing works
Advances in Non-Invasive Blood Pressure Measurement Techniques
Panula, T., Sirkiä, J., Wong, D., & Kaisti, M. (2022) · IEEE Reviews in Biomedical Engineering · doi:10.1109/rbme.2022.3141877 · Cited by 70 · see citing works
Sameh, A., Rostami, M., Oussalah, M., Korpelainen, R., & Farrahi, V. (2024) · Artificial Intelligence Review · doi:10.1007/s10462-024-11009-5 · Cited by 57 · see citing works
“For example, passive non-invasive signals can be also used for analysis of sleep quality, activity behaviors and other health related parameters (Tison et al. 2017a , b ; Nahavandi et al. 2022 ).”
Zhou, W., Chan, Y. E., Foo, C.-S., Zhang, J., Teo, J. X., Dávila, S., Huang, W., Yap, J. J. L., Cook, S. A., Tan, P. B. O., et al. (2022) · Journal of Medical Internet Research · doi:10.2196/34669 · Cited by 25 · see citing works
“A few recent studies have used black-box deep neural networks to relate high-resolution heart rate and step count time series recorded using wearables to the risk of developing atrial fibrillation, sleep apnea, and hypertension [31,32].”
Wu, S., Chen, M., Wei, K., & Liu, G. (2021) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2021.106442 · Cited by 23 · see citing works
Digital Health: New Approaches in Hypertension Management
Kitt, J., Fox, R., & Tucker, K. L. (2020) · Journal of Cardiology and Cardiovascular Sciences · doi:10.29245/2578-3025/2020/1.1186 · Cited by 11 · see citing works
Adikari, D., Gharleghi, R., Zhang, S., Jorm, L. R., Sowmya, A., Moses, D. A., Ooi, S., & Beier, S. (2022) · BMJ Open · doi:10.1136/bmjopen-2021-054881 · Cited by 6 · see citing works
Digitalization of hypertension management: a paradigm shift
Kaur, S., Gulati, H. K., & Baldi, A. (2024) · Naunyn-Schmiedeberg s Archives of Pharmacology · doi:10.1007/s00210-024-03229-x · Cited by 4 · see citing works
“For instance, the efficacy of Cardiogram® app on the Apple® watch has undergone evaluation, utilizing deep learning algorithms to predict hypertension based on heart rate and step count data (Tison et al. 2017).”
Review of IoT-based AI analysis method for heart monitoring and heart disease prediction
Mun, E., & Cho, J. (2022) · International Conference on Information and Communication Technology Convergence (ICTC) · doi:10.1109/ictc55196.2022.9952501 · Cited by 3 · see citing works
Sleep Tracker and Smartphone: Strengths and Limits to Estimate Sleep and Sleep-Disordered Breathing
Romano, S., & Insalaco, G. (2020) · Springer eBooks · doi:10.1007/978-3-030-42998-0_23 · Cited by 2 · see citing works
- Authors
- Tison, G. H., Singh, A. C., Ohashi, D. A., Hsieh, J. T., Ballinger, B. M., Olgin, J. E., Marcus, G. M., & Pletcher, M. J.
- Affiliations
- University of California, San Francisco (US)
- Engineering, Cardiogram, Inc., San Francisco, CA
- Venue
- Circulation
- Published
- 2017-11-14
- Volume
- 136
- Issue
- suppl_1
- ISSN
- 0009-7322 (ISSN-L)
- 1524-4539
- OpenAlex
- W2770013197
- Cited by
- 19
- References
- 0
- Citations by year
- 2024: 2 · 2023: 1 · 2022: 4 · 2021: 1 · 2020: 7 · 2019: 1 · 2018: 3
- Type
- conference abstract
- Language
- en
- Open access
- closed
- Locations
- Circulation (publishedVersion)
- Topics
- Physical Sciences › Engineering › Biomedical Engineering › Non-Invasive Vital Sign Monitoring
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Keywords
- cardiovascular risk prediction, wearable sensors, photoplethysmography, heart rate variability, accelerometer, step count, hypertension, sleep apnea, diabetes mellitus, multi-task deep learning, semi-supervised learning, Apple Watch, activity patterns, mobile health, cardiovascular risk stratification
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
