Decoding Regenerative Peripheral Nerve Interface Signals Contributes to Dexterous Fine Motor Control
Carrie A Kubiak, MD1, Philip Vu, MSE1, Zachary T Irwin, PhD1, Alex K Vaskov, BSc2, Chrono Nu, BSc 1, Deanna Gates, PhD1, Richard B. Gillespie, PhD3, Theodore A Kung, MD4, Paul S Cederna, MD1, Cynthia Chestek, PhD; Stephen WP Kemp, PhD1
1University of Michigan, Ann Arbor, MI, 2The University of Michigan, Ann Arbor, MI, 3Mechanical Engineering, University of Michigan, Ann Arbor, MI, 4Section of Plastic & Reconstructive Surgery, University of Michigan, Ann Arbor, MI
Background: Peripheral nerves provide a promising source for neuroprosthetic control given that they are functionally selective and easy to access. However, current interface methods, such as penetrating electrodes, are limited either by low signal amplitude or interface instability. In contrast, Regenerative Peripheral Nerve Interfaces (RPNIs) are constructed by suturing a graft of devascularized, denervated muscle to the residual end of a severed nerve. The graft then regenerates, revascularizes, and is reinnervated by the nerve, creating a stable bioamplifier that produces high amplitude electromyography (EMG) signals. In addition, nerves can be surgically divided into fascicles to construct multiple independent RPNIs. Here, we demonstrate the extraction of hand level prosthetic control signals from RPNIs. We also demonstrate for the first time control of the DEKA Luke Prosthetic Arm from RPNIs with indwelling wires.
Materials and Methods: We implanted RPNIs in two individuals with upper limb amputations. P1 had a proximal transradial amputation and underwent implantation of 9 RPNIs implanted, with the median, ulnar, and radial nerves subdivided into four, three, and two branches, respectively. P2 had a distal transradial amputation and was implanted with a single graft on each of the median, ulnar, and radial nerves (3 total RPNIs). During acute recording sessions, we used ultrasound to locate and visualize RPNI motor contractions, and implant percutaneous fine-wire bipolar electrodes into each RPNI. P1 was subsequently implanted with 8 indwelling electrodes from Synapse Biomedical (IDE #:G160229).
Results: P1's ulnar RPNIs produced signals with an average maximum voluntary contraction (MVC) of 211ÁV and signal-to-noise ratio (SNR) of 3.48. P2's ulnar RPNI produced EMG signals with an average MVC of 49.2ÁV and SNR 7.45. Likewise, the median RPNI produced signals with an average MVC of 145ÁV and SNR 16.8. Using a combination of RPNI and available residual muscle signals, subjects successfully controlled a virtual prosthesis in real-time. P1 could hit targets with a virtual small finger using a position/velocity Kalman filter with 98.2% accuracy over 56 trials. Similarly, P2 could hit targets with a virtual thumb with 80.6% accuracy over 108 trials. All decoders used temporal features of the EMG waveform within 300-1500Hz and binned at 50ms. Implanted electrodes in P1 enabled real-time control of the DEKA Luke prosthetic arm.
Conclusions: Overall, we have demonstrated that RPNIs produce motor specific contractions and high amplitude signals. Implanted electrodes enable an amputee to control numerous independent degrees of freedom in real-time with an advanced prosthetic arm.
Back to 2019 Absteracts