Prediction of the oxygen uptake patterns during an incremental exercise test using long short - term memory in electromyography

Prediction models of the oxygen uptake (VO2) from electromyograms (EMG) lower limb and respiratory muscles during an incremental exercise test were examined. Healthy male adults (n=15) underwent using a cycle ergometer. To predict patterns VO2, we used type recurrent neural network, long short-term memory. The measured VO2 as training data for deep learning, two prediction input values set: muscle model model. In model, EMGs rectus femoris vastus lateralis input. sternocleidomastoid inspiratory time both predicted increased test. histogram showed peak difference between 0 0.5 mL/kg/min. Bland-Altman plots demonstrated that most distributed within range agreement. root mean square error (RMSE) period was 2.1 ± 0.7 mL/kg/min 2.8 1.1 RMSE with increasing course time. ergometer task, each enable estimation pattern VO2. Mild to moderate intensity suitable by electromyography.