Prediction of oxygen uptake kinetics during heavy-intensity cycling exercise by machine-learning analysis.

Hedge ET; Amelard R; Hughson RL;

Journal of applied physiology (Bethesda, Md. : 1985) [J Appl Physiol (1985)] 2023 May 18.
Date of Electronic Publication: 2023 May 18.

Non-intrusive estimation of oxygen uptake (V̇O 2 ) is possible with wearable sensor technology and artificial intelligence. V̇O 2 kinetics have been accurately predicted during moderate exercise using easy-to-obtain sensor inputs. However, V̇O 2 prediction algorithms for higher intensity exercise with inherent nonlinearities are still being refined. The purpose of this investigation was to test if a machine learning model can accurately predict dynamic V̇O 2 across exercise intensities, including slower V̇O 2 kinetics normally observed during heavy- compared to moderate-intensity exercise. Fifteen young healthy adults (7 females; peak V̇O 2 : 42±5 mL·min -1 ·kg -1 ) performed three different pseudorandom binary sequence (PRBS) exercise tests ranging in intensity from low-to-moderate, low-to-heavy, and ventilatory threshold-to-heavy work rates. A temporal convolutional network was trained to predict instantaneous V̇O 2 , with model inputs including heart rate, percent heart rate reserve, estimated minute ventilation, breathing frequency, and work rate. Frequency domain analyses between V̇O 2 and work rate were used to evaluate measured and predicted V̇O 2 kinetics. Predicted V̇O 2 had low bias (-0.017 L·min -1 , 95% limits of agreement: [-0.289, 0.254]), and was very strongly correlated ( r rm =0.974, p <0.001) with the measured V̇O 2 . The extracted indicator of kinetics, mean normalized gain (MNG), was not different between predicted and measured V̇O 2 responses (main effect: p =0.374, η p 2 =0.01), and decreased with increasing exercise intensity (main effect: p <0.001, η p 2 =0.64). Predicted and measured V̇O 2 kinetics indicators were moderately correlated across repeated measurements (MNG: r rm =0.680, p <0.001). Therefore, the temporal convolutional network accurately predicted slower V̇O 2 kinetics with increasing exercise intensity, enabling non-intrusive monitoring of cardiorespiratory dynamics across moderate- and heavy-exercise intensities.