Extended Data Fig. 8: High Density EMG decomposition in single motor units spike trains.

a. Pipeline to extract single motor units spike trains from high density surface electromyography (HDEMG). We used an 8 × 8 channel flexible grid to record muscle activity from both knee extensor (Rectus Femoris (RF)) and flexor (Biceps Femoris (BF)). Participants performed two sets of three isometric Maximum Voluntary Contractions (MVC) of knee extension and flexion. We used the convolution kernel compensation (CKC) method to decompose the EMG signals in single motor units spike trains. b. Rescued motoneuron identification: a motor unit was defined as rescued if its mean peak firing rate (across trials) was higher than the 99.73 % CI (red line) of mean peak firing rate across motor units recorded during the first session. The 99.73% CI was computed with bootstrap (N = 10,000). c. Example of motor units smoothed discharge rate (SDR) obtained during the 3rd MVC. Each line represents the SDR of every motoneuron. Light and dark colors delineate the data from the first and last session where we performed the experiment d. Circles illustrate the number of identified motoneurons over the different phases of the study. (that is, PI = pre-implant; Wi = i-th week after the implant; PE = post-explant). e, Example (SMA02) of the mean and 95% CI of post-explant single motoneurons firing rates during MVC. In gray regions indicate the start and end windows used to compute firing rate at the beginning and the end of the MVC. f, Percentage change in the mean firing rate between rescued and dysfunctional motoneurons at the start (first 150 ms) and end (last 150 ms) of the MVC for each participant. 95% CI were computed using bootstrapping (n = 10,000). We assessed statistical significance using a two-sided test with bootstrapping (n = 10000). Asterisk *, **, and *** denote significant differences with p-values of p < 0.05, p < 0.01, and p< 0.001, respectively.