Automatic classification of motor unit potentials in surface EMG recorded from thenar muscles paralyzed by spinal cord injury

Involuntary electromyographic (EMG) activity has only been analyzed in the paralyzed thenar muscles of spinal cord injured (SCI) subjects for several minutes. It is unknown if this motor unit activity is ongoing. Longer duration EMG recordings can investigate the biological significance of this activity. Since no software is currently capable of classifying 24 hours of EMG data at a single motor unit level, the goal of this research was to devise an algorithm that would automatically classify motor unit potentials by tracking the firing behavior of motor units over 24-hours. Two-channels of thenar muscle surface EMG were recorded over 24-hours from 7 SCI subjects with a chronic cervical level injury using a custom data logging device with custom software. The automatic motor unit classification algorithm developed here employed multiple passes through these 24-hour EMG recordings to segment, cluster, form global templates and classify motor unit potentials, including superimposed potentials. The classification algorithm was able to track an average of 19 global classes in 7 24-hour recordings with a mean ( ± SE) accuracy of 89.9 % ( ± 0.98%) and classify potentials from these individual motor units with a mean accuracy of 90.3 % ( ± 0.97%). The algorithm could analyze 24 hours of data in 2–3 weeks wit