Fig. 5

Confusion matrix of the best machine learning algorithm compared to human classification. A Four-muscle classification performance (normalized by rows and rounded). Depicted is the confusion matrix of the RF on raw data for the cross validation during training (left) and the test (middle), compared to the results of the human classification (right). The RF can distinguish all 4 muscles extremely well during training, but has more difficulties classifying lower limb muscles in the test dataset. Due to the reliance on latency as the main distinguishing criterion, the neurophysiologists can confidently differentiate between upper and lower limbs, but have poorer performance on individual muscles. B Limb classification performance (normalized by rows and rounded). The scores from the matrices in (A) are summed across limbs. This shows the good differentiation of upper versus lower extremities by both the ML algorithm and the neurophysiologists