Continuous respiratory rate estimation with a wearable temperature sensor during cycling exercise: a feasibility study

Respiratory rate (RR) measurement has numerous implications for testing and training monitoring both in research and applied settings. Cycling is among the exercise modalities that may benefit the most from the continuous RR monitoring. The substantial advances in technological development recently observed in the field of respiratory monitoring offer new opportunities for RR measurement in applied settings. This study assessed the performance of a measurement system based on a temperature (i.e., a thermistor) sensor embedded in a printed circuit board in estimating RR during a cycling incremental exercise test. The printed circuit board was physically installed at the end of the face mask of a metabolic cart (reference system) commonly used in exercise testing. The reference device was based on airflow measurement and used with the purpose of validating RR estimated from the temperature sensor. A frequency domain analysis was implemented to estimate RR from a windowed signal with an update time of 1 s. Five window sizes have been tested (i.e., 5 s, 10 s, 20 s, 30 s, and 60 s). Per each condition the mean absolute error (MAE), the mean absolute percentage error (MAPE) and the Bland-Altman analysis were used as comparison metrics. Results demonstrated lower MAE and MAPE in the case of higher window size, with the worst results found analyzing data with 5 s window size (MAE < 1.2 bpm, MAPE < 2.8%). The promising results encourage additional tests and refinement of the measuring system for its use in sports applications.