Electrocardiogram-Capable Smartwatches: Assessing Their Clinical Accuracy and Application

@article{PhysioNet-ecg-capable-smartwatches-1.0.1,
  author = {Recas, Joaquin and {Buelga Suárez}, Mauro and González-Cabeza, Sergio and Sanz-Guerrero, Mario and Diaz-Vicente, Marian and Rebolleda, Alfonso and {Piñuel Moreno}, Luis and {Alonso Salinas}, Gonzalo Luis},
  title = {{Electrocardiogram-Capable Smartwatches: Assessing Their Clinical Accuracy and Application}},
  journal = {{PhysioNet}},
  year = {2026},
  month = oct,
  note = {Version 1.0.1},
  doi = {10.13026/g59m-vn30},
  url = {https://doi.org/10.13026/g59m-vn30}
}

@article{pollard_physionet_2026, title = {{PhysioNet} as a global platform for biomedical research}, volume = {1}, issn = {3005-0693}, url = {https://doi.org/10.1038/s44360-026-00096-z}, doi = {10.1038/s44360-026-00096-z}, abstract = {Twenty-five years after its first public repository was established, the PhysioNet resource for biomedical and clinical data has enabled tens of thousands of research papers. Further efforts are ongoing to advance secure data sharing, community engagement and scalable research infrastructure.}, number = {8}, journal = {Nature Health}, author = {Pollard, Tom and Moody, Benjamin E. and Lehman, Li-wei H. and Gow, Brian J. and Fernandes, Chrystinne and Xie, Chen and Johnson, Alistair and Mark, Roger G. and Heldt, Thomas}, month = aug, year = {2026}, pages = {792--795},}

Abstract

Wearable technology has progressed significantly, such that today's smartwatches are able to offer advanced health monitoring functions, including the electrocardiogram (ECG). This study presents a dataset that can be used to clinically assess the accuracy of four leading smartwatch models: Apple Watch Series 9, Samsung Galaxy Watch 6, Fitbit Sense 2, and Withings ScanWatch. Using the standardized protocol of the International Electrotechnical Commission (IEC 60601-2-25:2011) and a METRON PS-440 patient simulator, controlled ECG signals were generated and recorded to assess smartwatch performance against a reference electrocardiograph (Philips TC30). Parameters such as heart rate (30 to 300 beats per minute), R-wave amplitude (500 to 2000 µV), and ST-segment elevation/depression (-800 to +800 µV) were configured in the patient simulator and recorded five times across all devices, resulting in a total of 900 records in the dataset. This extensive dataset provides valuable information on the feasibility of smartwatches as diagnostic tools, allowing for the observation of both their strengths and limitations for clinical and remote cardiovascular monitoring.

The recordings from the smartwatches last 30 seconds and correspond to Lead I, with a different sampling rate for each device: 512 Hz for Apple Watch Series 9; 500 Hz for Samsung Galaxy Watch 6; 250 Hz for Fitbit Sense 2; and 300 Hz for Withings ScanWatch. In the case of the Philips TC30 reference device, the recordings last 11 seconds, include 12 leads and have a sampling rate of 500 Hz.

Background

The proliferation of smartwatches—with 165.4 million units sold in 2023 [1]—highlights their potential to transform cardiovascular health monitoring. These devices have demonstrated efficacy in detecting Atrial Fibrillation (AF) [2-5], a leading cause of stroke and heart failure. However, broader applications, such as detecting ischemic heart disease or other arrhythmias, remain underexplored. Conventional tools like 12-lead ECGs and Holter monitors, while accurate, are limited by accessibility challenges and the intermittent nature of their monitoring. This database enables the study of the feasibility of using ECG-capable smartwatches as diagnostic tools, focusing on their compliance with clinical standards and their ability to measure critical ECG parameters beyond AF detection.

Methods

Design of the Evaluation Process

The data acquisition follows the International Electrotechnical Commission Standard IEC 60601-2-25:2011 [6], which specifies requirements for electrocardiographic device accuracy. A METRON PS-440 patient simulator was used to generate synthetic ECG signals representing a wide range of clinically relevant scenarios. These signals were transmitted to both the smartwatches under evaluation and a Philips TC30 electrocardiograph, which served as the gold standard. All measurements were taken in quintuplicate, as specified by the calibration standard.

Signal Transmission and Setup

• Patient Simulator Configuration: The METRON PS-440 was programmed to output calibrated ECG signals, including heart rates from 30 to 300 beats per minute (bpm), R-wave amplitudes from 500 to 2000 µV, and ST-segment elevation/depression ranging from -800 to +800 µV. The simulator’s output connectors followed the AHA and IEC color-coding standards to facilitate connection with both the electrocardiograph and the smartwatch sensors.
• Smartwatch Signal Acquisition: Each smartwatch was mounted on an adjustable stand to ensure stable contact between its sensors and the simulator output. Two flexible Holter cables (1.5 mm DIN) were utilized:
  • One cable connected the simulator’s R output (right arm) to the smartwatch’s crown or digital button, simulating the negative electrode.
  • The other cable connected the simulator’s L output (left arm) to the back of the smartwatch, simulating the positive electrode.
• Each smartwatch generated two types of output files for each test:
  • PDF files: Graphical representations of the recorded ECG traces, formatted as single-lead (Lead I) electrocardiograms. These files are NOT included in the database.
  • Data files: Raw numerical data, such as time-series voltage measurements, heart rate calculations, and other derived parameters. Only the raw data in WFDB is included in the database.
• Electrocardiograph Signal Acquisition: The Philips TC30 was connected to the simulator using standard 12-lead ECG cables. Signals were recorded for 11-second intervals per lead, capturing all standard ECG derivations. Calibration was verified prior to each test to ensure the electrocardiograph met accuracy requirements (±5% or ±40µV).

Data Description

This database contains raw data, transformed to WFDB format, generated using a Patient Simulator, the METRON PS-440, and measured with 5 different devices: (1) hospital electrocardiograph model Philips TC30; (2) Apple Watch Series 9; (3) Samsung Galaxy Watch 6; (4) Fitbit Sense 2; and (5) Withings ScanWatch.

The database is organised using a directory for each device:

1. philips_tc30 directory: Raw data in WFDB format for the Philips TC30 hospital electrocardiograph. The recordings last 11 seconds, include 12 leads and have a sampling rate of 500 Hz.
2. applewatch_serie9 directory: Raw data in WFDB format for Apple Watch Series 9. The recordings last 30 seconds and correspond to Lead I, with a sampling rate of 512 Hz.
3. samsunggalaxy6 directory: Raw data in WFDB format for the Samsung Galaxy Watch 6. The recordings last 30 seconds and correspond to Lead I, with a sampling rate of 500 Hz.
4. fitbitsense2 directory: Raw data in WFDB format for the Fitbit Sense 2. The recordings last 30 seconds and correspond to Lead I, with a sampling rate of 250 Hz.
5. withingsscanwatch directory: Raw data in WFDB format for Withings ScanWatch. The recordings last 30 seconds and correspond to Lead I, with a sampling rate of 300 Hz.

For each device the database contains the following records, all experiments in quintuplicate:

• amp_test: 4 experiments
  • amp500: records amp500_0 to amp500_4 corresponding to an ECG signal of 500µV amplitude.
    ...
  • amp2000: records amp2000_0 to amp2000_4 corresponding to an ECG signal of 2000µV amplitude.

• freq_test: 15 experiments
  • f30: records f30_0 to f30_4 corresponding to a 30 BPM ECG signal.
    ...
  • f300: records f300_0 to f300_4 corresponding to a 300 BPM ECG signal.

• sqr-2hz: sqr-2hz_0 to sqr-2hz_4 registers corresponding to a square wave signal with a frequency of 2 Hz.

• st-segment: 16 experiments
  • st-m8: st-m8_0 to st-m8_4 records corresponding to an ECG signal with ST-segment depression of -800µV.
    ...
  • st-p8: st-p8_0 to st-p8_4 records corresponding to an ECG signal with ST-segment elevation of +800µV.

The database contains a total of 900 records.

Usage Notes

This dataset is intended to support the technical and research-oriented evaluation of smartwatches capable of recording electrocardiograms under controlled experimental conditions. It includes electrocardiographic recordings obtained with four commercially available smartwatch models: Apple Watch Series 9, Samsung Galaxy Watch 6, Fitbit Sense 2, and Withings ScanWatch. All selected devices are CE-marked and FDA-accredited for atrial fibrillation detection. The recordings were generated using standardized electrocardiographic signals from a METRON PS-440® patient simulator (Fluke Corporation, Washington, USA) and can be compared with recordings obtained using a 12-lead clinical electrocardiograph, Philips TC30® (Koninklijke Philips N.V., Amsterdam, The Netherlands), used as the gold standard.

This dataset may be used as a benchmark for studying the acquisition, processing, and representation characteristics of electrocardiographic signals obtained using smartwatch-based systems. The heart-rate tests allow assessment of how these devices estimate rhythm across a broad range of frequencies, from marked bradycardia to high heart rates. Overall, these data make it possible to characterize the operating range of smartwatch-based systems and to identify conditions in which internal algorithms may classify recordings as inconclusive or generate less precise estimates. This information may be useful for optimizing automatic rhythm-detection systems and for defining the technical limits of these devices in portable single-lead electrocardiographic monitoring.

The dataset also enables morphological analysis of standardized electrocardiographic signals, including R-wave amplitude measurement, ST-segment assessment, and J-point measurement. These tests may be used to evaluate the extent to which smartwatch-based systems preserve relevant electrocardiographic features and to quantify differences with respect to a clinical reference electrocardiograph. The availability of signals with different levels of ST-segment elevation and depression may be particularly useful for developing and validating tools aimed at the automatic identification of electrocardiographic patterns compatible with myocardial ischemia.

In addition, the dataset includes square-wave and sinusoidal signals, which allow assessment of the dynamic response of the devices and the effect of their internal acquisition and filtering systems. These tests may be used to evaluate signal continuity, potential morphological distortion, frequency response, and the effective passband of smartwatch-based electrocardiographic systems. Overall, these data may help optimize future algorithms for filtering, noise reduction, baseline correction, and preservation of clinically relevant electrocardiographic morphology.

Several limitations should be considered when using this dataset. First, the recordings were mainly obtained under controlled laboratory conditions using standardized simulated signals, which enables reproducible technical assessment but does not fully reproduce the variability, artefacts, and acquisition challenges of recordings obtained in real-world conditions. Second, recordings obtained with smartwatches correspond to a single-lead electrocardiographic configuration and should not be interpreted as equivalent to a conventional 12-lead electrocardiogram. Third, the internal filters, signal-processing algorithms, and calibration procedures implemented by each manufacturer are not fully described, which may influence signal morphology and should be considered when interpreting the data.

In summary, this resource may serve as a reference platform for validating signal-processing methods, developing electrocardiographic analysis algorithms, and guiding future technological improvements in smartwatches capable of recording electrocardiograms. Since clinical electrocardiographs are subject to specific technical standards, such as IEC 60601-2-25:2011, the procedures and data included in this dataset may contribute to the development of objective precision and signal-quality standards for future portable single-lead electrocardiographic devices.

Release Notes

Version 1.0.0

Initial release

Ethics

Data collected from synthetic sources

Acknowledgements

This study is partially funded by the Instituto Ramón y Cajal de Investigación Sanitaria.

Conflicts of Interest

The authors declare that they have no conflict of interest.

References

1. Ubrani, J., Reith, R., & Llamas, R. L. T. (2026). Wearable devices market insights. Retrieved from https://www.idc.com/promo/wearablevendor/
2. Perez, M. V., Mahaffey, K. W., Hedlin, H., Rumsfeld, J. S., Garcia, A., Ferris, T., … Turakhia, M. P. (2019). Large-scale assessment of a smartwatch to identify atrial fibrillation. New England Journal of Medicine, 381(20), 1909–1917. doi:10.1056/nejmoa1901183
3. Lubitz, S. A., Faranesh, A. Z., Selvaggi, C., Atlas, S. J., McManus, D. D., Singer, D. E., … Foulkes, A. S. (2022). Detection of atrial fibrillation in a large population using wearable devices: The Fitbit Heart Study. Circulation, 146(19), 1415–1424. doi:10.1161/circulationaha.122.060291
4. Campo, D., Elie, V., de Gallard, T., Bartet, P., Morichau-Beauchant, T., Genain, N., … Nahum, J. (2022). Atrial fibrillation detection with an analog smartwatch: Prospective clinical study and algorithm validation. JMIR Formative Research, 6(11). doi:10.2196/37280
5. Marsili, I. A., Biasiolli, L., Masè, M., Adami, A., Andrighetti, A. O., Ravelli, F., & Nollo, G. (2020). Implementation and validation of real-time algorithms for atrial fibrillation detection on a wearable ECG device. Computers in Biology and Medicine, 116, 103540. doi:10.1016/j.compbiomed.2019.103540
6. IEC 60601-2-25:1993. (2011). Retrieved from https://webstore.iec.ch/en/publication/16851