19. Regenerative Peripheral Nerve Interfaces (RPNIs) and Implanted Electrodes Enable Kinematic Control of Multiple Degrees of Freedom

PURPOSE: Prosthetic limbs and pattern recognition systems can enhance functionality for people with upper limb amputation. Commercially available systems rely on surface electromyography (EMG) and are limited to sequential control of single movements. Intramuscular EMG electrodes have been shown to record stable and independent control signals, enabling reliable control of multiple DOF (Dewald et al., 2019, Lukayenko et al., 2021). However, following amputation there may not be sufficient residual innervated muscles available to provide complex movements of the fingers and wrist. Regenerative Peripheral Nerve Interfaces (RPNIs) surgically provide new muscle targets to the peripheral nerve and amplify efferent nerve activity to compensate for absent muscles needed for prosthetic control (Vu et al., 2020). Here we demonstrate the performance of simultaneous control of multiple continuous degrees-of-freedom (DoF) for hand and wrist rotation. METHODS: Two participants with transradial amputation (P1, P2) had intramuscular bipolar electromyography (EMG) electrodes placed in their RPNIs and residual muscles. For P1, electrodes were placed in 3 previously created RPNIs (1 median, 2 ulnar nerve) and 5 residual muscles. For P2, electrodes were implanted at the time of RPNI creation (4 median and 1 radial nerve) and in 7 residual muscles (incl. wrist supinator & pronator). Percutaneous electrodes were then simultaneously connected to a neural signal processor (Blackrock Neurotech) and real-time computer. Both participants were instructed to follow a virtual hand moving in 2 DoF: hand open/close and wrist rotation. EMG signals were collected from implanted electrodes and used to calibrate continuous Kalman Filters (KFs). Participants then used these KFs in closed-loop control of the 2 DoF hand/wrist to acquire targets simultaneously. Performance was evaluated for total trial time (TTT) and orbiting time (OT). P1 also calibrated and used a 4 DoF decoder (index, middle-ring-small fingers, 2 DoF thumb) using a recurrent neural network (RNN) architecture. RESULTS: During online control, P1 achieved an average TTT of 3.68 seconds and an OT of 2.45 seconds. For P2, who had implanted electrodes targeted toward wrist rotation, TTT was 2.24 seconds and OT was 1.77 seconds. During preliminary online control of the 4 DoF RNN decoder, P1 was able to control all 4 DoF to achieve different hand postures, although no online target metrics have yet been collected. CONCLUSION: RPNIs and implanted electrodes have enabled people with transradial amputations to control simultaneous degrees of freedom continuously. Simultaneous hand and wrist control is necessary for prosthesis users to perform coordinated movements. Furthermore, providing patients with individual finger control may increase hand function far beyond opening and closing pre-programmed grips.