Seismocardiography-based estimation of hemodynamic parameters during submaximal ergometer test

Abstract Objective. To evaluate the feasibility of seismocardiography (SCG)-based estimation of hemodynamic parameters during submaximal cycle ergometer exercise across different body mass index (BMI) groups. Approach. Sixty healthy adults ( n = 15 per BMI group: underweight, normal weight, overweight, obese) performed a YMCA submaximal cycling test while SCG signals were recorded using a chest-mounted accelerometer. Transthoracic bioimpedance (PhysioFlow) served as reference. Time-domain features from tri-axial SCG signals were used in subject-specific random forest regressors to estimate stroke volume (SV), heart rate (HR), cardiac output (CO), and cardiac index. Performance was evaluated across baseline, exercise, and post-exercise phases using the mean absolute percentage error (MAPE) and coefficient of determination ( R 2 ). Main results. While SCG signals were successfully acquired across all phases, estimation performance varied significantly by physiological state. Models achieved MAPEs below 8% for all parameters overall. However, model reliability was condition-dependent, with optimal performance during post-exercise recovery (median R 2 = 0.75 for HR and CO; 0.42 for SV) with reduced reliability during active cycling. SCG features demonstrated limited sensitivity to BMI variations compared to reference hemodynamic parameters, which may limit personalized estimation accuracy across diverse body compositions. Significance. SCG acquisition is technically viable during exercise, but reliable hemodynamic estimation under high-motion conditions remains limited due to motion artifacts and physiological variability. Post-exercise recovery provides optimal conditions for SCG-based monitoring. SCG shows promise as a lightweight approach for cardiovascular assessment in recovery or low-motion scenarios rather than during active exercise. Further validation using gold-standard methods is warranted.