Abstract 24: Successful Control of Virtual and Robotic Hands using Neuroprosthetic Signals from Regenerative Peripheral Nerve Interfaces in a Human Subject

PURPOSE: Regenerative Peripheral Nerve Interfaces (RPNIs) show promise in controlling neuroprosthetic devices. We have implanted and recorded from RPNIs in 3 human subjects. Here, we present the results from our longest implanted subject with a distal transradial amputation. METHODS: An RPNI consists of a muscle graft that is neurotized by the distal end of a transected peripheral nerve. Once revasularized and reinnervated, the RPNI muscle graft serves as a stable bioelectric amplifier for efferent nerve action potentials and produces recordable electromyography (EMG) signals. The subject was implanted with RPNIs on the residual median, ulnar, and dorsal radial nerves. Using ultrasound, RPNIs were located, and percutaneous fine-wire bipolar electrodes were inserted for acute EMG recordings. Temporal features of the EMG waveforms (100-500Hz) were used for decoding algorithms. RESULTS: Eight months post-surgery, we recorded 300–400µV EMG signals from the median RPNI with signal-to-noise ratio (SNR) of 24.2 and 100–120µV EMG signal from the ulnar RPNI with SNR of 5.84. Additionally, EMG from residual muscles was obtained including the flexor digitorum superficialis with 100–120µV signals, SNR of 6.30, and flexor pollicis longus with ~1mV signals, SNR of 47.8. With these signals, the subject controlled a virtual robotic hand in real time with 96% accuracy, choosing 1 of 4 movements within 212 trials. Importantly, the subject controlled a physical Touch Bionics iLimb neuroprosthetic hand with 100% accuracy, choosing 1 of 3 movements within 100 trials. CONCLUSION: RPNIs harness neural signals from transected peripheral nerves with sufficient amplitude and fidelity to control an advanced neuroprosthetic limb. K.A. Carrie: None. P.P. Vu: None. Z.T. Irwin: None. P.T. Henning: None. T.A. Kung: None. C.A. Chestek: None. P.S. Cederna: None. S.W. Kemp: None.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.