Author(s): MOUFAWAD EL ACHKAR, C., ADAM, K., SOLTANI, R., BONNIER, G., GERBER, M., RENEVEY, P., LEMKADDEM, A., FERRARIO, D., LEMAY, M., Institution: CSEM SA, Country: SWITZERLAND, Abstract-ID: 2421

INTRODUCTION:
VO2Max, the maximal workout oxygen uptake, is a crucial measure of cardiorespiratory fitness (Grivas, 2020). VO2Max is traditionally evaluated with a metabolic cart, a device that measures gas exchange with a mask tightly covering the face connected to a stationary computing unit (Wiecha et al., 2022). While highly accurate, such devices are complicated to use for regular evaluation, as they are cumbersome and bulky, expensive, and require routine calibrations. Wearable systems that measure heart rate (HR) and speed may be suitable candidates to measure VO2Max during standard workouts without overburdening athletes and without the need for performing a maximal exercise (Mežan, 2025). In this work, we present a new method to calculate VO2Max from a running workout and showcase the accuracy of HR, speed, and VO2Max calculations.
METHODS:
Data from a diverse cohort of 20 subjects was collected (11 male, 9 female, age [26-55 years old], weight [55-82 kg], height [157-182 cm]). The study was approved by CER-VD, BASEC-ID: 2024-01496. Subjects performed an exercise session during which they ran on a treadmill for 3min bouts with increments of 1.8km/h in-between, until exhaustion, starting with a speed of 1.8km/h. VO2Max reference was measured as the maximal oxygen uptake volume at exhaustion with a Cosmed K5 cardiopulmonary testing device. HR reference data was collected with a Polar H10 chest strap. Speed reference was obtained directly from the treadmill. Photoplethysmography and 3D acceleration were measured from the wrist with CSEM's proprietary Delta device, built to maximize optical signal quality and cardiovascular parameter accuracy. HR and speed were extracted using real-time, embedded CSEM-proprietary algorithms (Delgado-Gonzalo et al., 2015), with a built-in HR quality index. VO2Max was estimated at each speed by exploiting the linear relationship between maximal aerobic speed and VO2Max (Léger & Mercier, 1984), and extrapolating it based on maximal HR.
RESULTS:
HR root mean squared error (RMSE) was 4.8 ± 2.0 beats per minute overall with a mean absolute percent error (MAPE) of 2.8% without any rejection, and 1.8% if only high data quality points were considered. User-calibrated speed achieved MAPE of 4.4%. VO2Max achieved MAPE of 8.51, 6.78, 10.36, 11.76 % at speeds of 14.4, 12.6, 10.8, and 9km/h, respectively.
CONCLUSION:
VO2Max MAPE at higher speeds is similar to research results on Garmin at 7.2-7.9% (Engel et al., 2025) and compares favorably to Apple watch at 13.31% (Lambe et al., 2025). HR MAPE is comparable to Apple Watch at 2.3-2.9% (Hajj-Boutros et al., 2023). Treadmill speed MAPE of 4.4% (with calibration) provides an accurate benchmark as commercial wearables seldom provide speed accuracies. In conclusion, we show the viability of measuring VO2Max with a wearable sensor simultaneously recording highly accurate HR and speed. Future research is planned to ensure validity in a larger cohort, outdoor and real-life workout conditions, and different wearable settings.