An open-access dataset reveals how sharply smartwatch coronary heart price accuracy can decline underwater, offering researchers with a brand new basis for bettering wearable monitoring in swimming and aquatic well being.
Study: Wearable Heart Rate Monitoring during Swimming. Image Credit: Dean Drobot / Shutterstock
In a latest research printed within the journal Scientific Data, researchers in Italy described the event and metrological characterization of an open-access dataset designed to guage the efficiency of economic wearable coronary heart price displays in aquatic environments.
The research systematically evaluated three business sensors (a Polar chest strap and Polar and Garmin smartwatches) in aggressive swimmers beneath each dry (metrological management) and moist situations.
Study findings revealed that the wrist-worn sensors carried out considerably higher on land, whereas their metrological efficiency degraded underwater. The authors famous that water can attenuate photoplethysmography (PPG) optical indicators and introduce hydrodynamic noise, offering potential mechanisms for this decline in efficiency.
This dataset gives an open-access benchmark that permits researchers to research machine studying (ML) approaches, sensor fusion, signal-quality evaluation, and improved wearable monitoring in aquatic environments.
Background
Decades of cardiovascular and sports activities science analysis have established steady coronary heart price (HR) monitoring as an important instrument for assessing athletes’ baselines and monitoring adjustments in a number of key physiological measures, together with bodily exertion, nervous system exercise, and psychophysical stress.
The customary strategy to HR measurement is electrocardiography (ECG). However, standard ECG monitoring is commonly restricted to brief measurement durations, and extended monitoring requires specialised transportable methods comparable to Holter displays, thereby limiting its routine use throughout many sporting actions.
The latest speedy proliferation of shopper wearables, significantly smartwatches and HR-monitoring chest straps, has considerably mitigated this limitation, enabling extended coronary heart price monitoring throughout day by day routines and athletic coaching.
However, aquatic environments create distinctive challenges for wrist-based PPG monitoring as a result of water acts as a dense medium that attenuates photoplethysmography (PPG) optical indicators and introduces distinct hydrodynamic noise that differs considerably from dry-land movement artifacts.
These challenges spotlight the necessity for well-characterized aquatic datasets that enable researchers to differentiate the consequences of water from movement artifacts and examine data-processing approaches particularly beneath water-based situations.
About the research
The current research aimed to deal with this limitation and inform future wearable sensor design and software program growth by creating a synchronized multi-sensor dataset throughout dry and aquatic experimental situations.
The research pattern cohort comprised 10 wholesome feminine aggressive swimmers (17 ± 3 years previous; Fitzpatrick pores and skin varieties I and II) who had been required to concurrently put on three business water resistant units: 1. Polar H10 chest strap (ECG-based reference system), 2. Polar Vantage V2 smartwatch (worn on the left wrist; PPG-based), and three. Garmin Venu Sq smartwatch (worn on the correct wrist; PPG-based).
Data assortment was divided right into a “dry condition,” which served because the metrological management, and a “wet condition” representing the aquatic take a look at setting. During the dry situation, researchers measured baseline (resting state), adopted by three treadmill strolling trials (4 km/h) and three working trials (6 km/h) to assist distinguish motion-related results from these launched by water.
In the moist situation, researchers measured resting state in water adopted by 4 swimming actions (freestyle, butterfly stroke, backstroke, and breaststroke; 4 laps every in a 25-meter pool) carried out at self-perceived regular and quick paces.
The remaining dataset compiled 18,622 synchronized 1 Hz timestamps, representing 97.61% of the 19,078 anticipated samples, which had been used to statistically consider the accuracy and precision of the wrist wearable units towards these of the reference chest strap utilizing Bland-Altman settlement plots, Mean Absolute Percentage Error (MAPE), and Pearson’s correlation coefficients (r).
Study findings
Statistical analyses revealed that in dry situations, each smartwatches demonstrated considerably higher settlement with the reference chest strap. The Garmin Venu Sq recorded a imply residual of 1 beat per minute (bpm; MAPE = 4.05%, r = 0.95), whereas the Polar Vantage V2 exhibited a imply residual of -5 bpm (MAPE = 8.00%, r = 0.85).
In stark distinction, in moist situations, the Garmin Venu Sq exhibited substantial coronary heart price underestimation, with a imply residual of -44 bpm (MAPE = 29.95%, r = 0.26). While the Polar Vantage V2 demonstrated notably superior accuracy (imply residual = -14 bpm; MAPE = 17.17%; r = 0.59), its efficiency was nonetheless considerably poorer than beneath dry situations.
Furthermore, Bland-Altman plots revealed distinct “looping” error patterns within the analyses of PPG-based units beneath dry and moist situations, which the authors attributed to physiological and algorithmic lag in PPG optical sensors throughout speedy transitions in bodily exertion.
Conclusions
The authors describe the dataset as a particular useful resource for characterizing how aquatic environments have an effect on the metrological efficiency of wrist-worn shopper optical HR displays and observe that, to their data, no comparable dataset with these traits is at the moment accessible.
The research was restricted by a demographically homogeneous pattern of 10 younger, light-skinned feminine athletes. The Polar H10 reference was not medical-grade, and the dataset contains manufacturer-generated heart-rate sequence relatively than uncooked ECG or PPG waveforms, which limits signal-quality evaluation and the event of customized algorithms. Even so, the open-access dataset is a helpful useful resource for preliminary machine-learning research, sensor-fusion algorithm growth, and refining wearable expertise for athletic efficiency and probably aquatic rehabilitation.