Restoring Fine Motor Control with Regenerative Peripheral Nerve Interfaces (RPNIs) and Implanted EMG Electrodes
Alex K Vaskov, PhD1, Dylan M Wallace, BS2, Theodore A Kung, MD3, Cynthia A Chestek, PhD4, Paul S Cederna, M.D.5 and Stephen WP Kemp, Ph.D.1, 1University of Michigan, Ann Arbor, MI, 2Univeristy of Michigan, Ann Arbor, MI, 3Section of Plastic & Reconstructive Surgery, University of Michigan, Ann Arbor, MI, 4Biomedical Engineering, Univeristy of Michigan, Ann Arbor, MI, 5Plastic Surgery, University of Michigan, Ann Arbor, MI
Introduction: Advanced prosthetic limbs have the potential to increase functionality for patients with upper limb amputations. However, this potential is often unrealized due to an inability to record strong and reliable control signals for multiple hand functions. Regenerative Peripheral Nerve Interface (RPNI) surgery is commonly performed to treat residual limb pain and also amplifies efferent motor action potentials to produce control signals in lieu of missing muscles. This study examined the signal strength and control benefits of implanted EMG electrodes in RPNIs.
Materials and Methods: An RPNI consists of a free muscle graft reinnervated by a transected peripheral nerve. Three participants with transradial amputations (P1, P2, P3) had bipolar electrodes surgically implanted into previously created median and ulnar nerve RPNIs and residual innervated muscles. A fourth participant with a transradial amputation (P4) had bipolar electrodes surgically implanted into newly created median, ulnar, and radial RPNIs and residual innervated muscles during the same operation. Signal-to-noise ratios (SNRs) were measured monthly across 267 days for P1, 1569 days for P2, 335 days for P3, and 251 days for P4.
For each participant, two movement classifiers were trained to predict thumb opposition, finger abduction, finger adduction, thumb flexion, index flexion, and rest. The first classifier used signals from RPNIs and residual muscles to predict movement intent. The second classifier only used signals from residual muscles to predict movement intent. The impact of RPNIs signals on predicting movement intent was determined by comparing the prediction accuracy of the two classifiers.
Results: Across all four participants, the implanted RPNIs produced large amplitude EMG with a median SNR of 32.2 dB. P1-P3’s RPNIs remained stable with no decline in SNR over time (p > 0.05, F-test). P4’s RPNIs first had a median SNR of 25.9 dB measured 103 days after simultaneous RPNI creation and electrode implantation. By day 188, P4’s RPNI signal strength increased to 35.3 dB and remained stable. Removing RPNIs as a control input decreased prediction accuracy of thumb opposition by a median of 20.5% and finger adduction by 18.9%. Finger abduction and thumb flexion decreased by 10.8 and 8.4% respectively. Prediction of index flexion was largely unaffected with less than a 1% decrease.
Conclusion: RPNIs produced large amplitude motor signals with no signal deterioration when electrodes were implanted into previously created RPNIs or at the time of RPNI creation. At a transradial level, RPNIs enhanced the control of fine thumb and finger movements.
Back to 2024 Abstracts