Novel Method for Accuracy Assessment of Individual Motor Unit Firing Identification from High-Density Surface Electromyograms

We describe a novel method for automatic assessment of the accuracy, with which the individual motor unit firings are identified from high-density surface electromyograms (hdEMG) by Convolution Kernel Compensation (CKC) technique. This method builds on recently introduced Pulse-to-Noise Ratio (PNR), is fully automatic and computationally efficient. We tested the developed methodology on a set of synthetic hdEMG signals with experimentally recorded motor unit action potentials and showed that correctly identified motor unit firings contribute significantly higher values to PNR than false negatives and false positives, respectively. This suggest that we may efficiently distinguish the motor unit firings that are correctly and incorrectly identified by CKC method. The introduced method presents the basis for the precise control of quality in neural codes estimation from hdEMG and can be applied to each individual motor unit firing.