Back to Annual Meeting Program
Real-Time Proportional Control of a Neuroprosthetic Hand by a Rodent Regenerative Peripheral Nerve Interface
Christopher M. Frost, BSE, Daniel Ursu, MS, Andrej Nedic, MSE, Cheryl A. Hassett, BS, Jana D. Moon, BS, Brent Gillespie, PhD, Nicholas Langhals, PhD, Paul S. Cederna, MD, Melanie G. Urbanchek, PhD.
University of Michigan, Ann Arbor, MI, USA.
Purpose: Regenerative peripheral nerve interfaces (RPNIs) are implantable bioartifical interfaces designed to transduce signals between peripheral nerves and prosthetic limbs. RPNIs implanted into rats have shown long-term stability and viability for up to 2 years. To date, control algorithms for translating RPNI signals into real-time control of a neuroprosthetic limb have not been demonstrated. The purpose of this study was to: a) design and validate a system for translating RPNI signals into real-time control of a neuroprosthetic hand; and b) use the RPNI system to demonstrate proportional control of a neuroprosthetic hand.
Methods: Three experimental groups were created in a rat model: 1) Control (n=2); 2) Denervated (n=1); and 3) RPNI (n=3). For the Control groups, the soleus muscle was denervated and a proximal and distal tenotomy with repair was performed. For RPNI and denervated groups, a free soleus muscle was transferred to the lateral compartment of the ipsilateral thigh. In the RPNI group the transferred muscle was reinnervated with the divided tibial nerve. In all groups, bipolar stainless steel wire electrodes were positioned on the muscle. Evaluation was performed 4-5 months after implantation. Voluntary movements were evoked in response to Von Frey monofilament stimulation on the lateral ankle. Using a peak detection algorithm in LabView, RPNI activity was scanned in 300-msec windows and integrated in real-time (Fig-1). Rat movements were videographed in high speed (120-fps). In total, 1040 control and 876 RPNI prosthesis activations were analyzed.
Results: Voluntary rat movement activated the prosthesis in Control and RPNI groups reliably throughout the testing period of up to 15 continuous minutes with no observable instrumentation failure or biological fatigue. As expected, leg movement in the denervated group did not activate the prosthesis further validating the system and indicating minimal signal contamination from surrounding muscle groups within the RPNI. Signal to noise ratio between resting iRPNI and iRPNI after leg movement was excellent across control and RPNI groups (3.55 and 3.81, respectively). “Sensitivity” to accurately detect activation after stimulation and “specificity” to prevent unwanted activation during rest were excellent across RPNI, control, and denervated groups (Table 1). Both RPNI and control groups showed a logarithmic increase in iRPNI with increasing Von Frey filament size (R2=0.758 and R2=0.802, respectively). Higher iRPNI increased output voltage to the prosthesis giving graded control of hand speed.
|Table 1: Validation of RPNI Translation System|
|Values are means±SD. * Indicates significant difference from resting iRPNI. Sensitivity and specificity were excellent across all three groups. iRPNI is the area under the curve measured during each 300 msec RPNI windows during testing. Activated iRPNI is recorded during rat movement while resting iRPNI is during rest. Denervated group did not show activity during rat movement as expected; therefore no activated iRPNI or sensitivity were calculated.|
This study demonstrates for the first time that an RPNI can be used to directly control a prosthetic arm. Signal contamination from muscles adjacent to the RPNI is minimal. Further, the RPNI can provide reliable proportional control of prosthesis hand speed. Acknowledgements:
This work was sponsored by the Defense Advance Research Projects MTO under the auspices of Dr. Jack Judy through the Space and Naval Warfare Systems Center, Pacific Grant/Contract No.N66001-11-C-4190
Back to Annual Meeting Program