Fig. 4

Reinnervated Muscle Interfaces. a The most common signal used for controlling myoelectric prostheses is sEMG. These voltages are detected from the skin through removable electrode contacts (top). The most accurate signal interpretation is through motor unit decomposition. An sEMG signal is broken down by an algorithm into the individual firing rates of unique motor units (bottom). This algorithm requires high-density surface electrode arrays (Farina et al., 2017). b Implanted electrodes can be used to record eEMG signals for prosthesis control. Currently, the best method for doing this in human subjects is to connect the electrodes with an osseointegrated implant that interfaces with the external prosthesis, a.k.a. the e-OPRA (top). These implants require that there be a permanent opening in the skin. The e-OPRA presents an opportunity for developing stable bi-directional interfaces. eEMG signals can control the robotics, and nerve cuff electrodes implanted in the residual limb can be used to provide stimulation based on signals from artificial touch sensors that are within the prosthesis (bottom) (Mastinu et al., 2020). c RPNI muscle grafts have been shown to be useful for prosthesis control. The only published human trials so far recorded iEMG signals using temporary bipolar electrodes that were implanted percutaneously in the RPNI muscles of 2 upper-limb amputees (top). Simple amplitude-based classification algorithms could be used to reliably operate a prosthetic hand through a series of complex gestures by using the iEMG signals recorded from multiple RPNIs (bottom) (Vu et al., 2020)