Reconnecting the Hand and Arm to the Brain: Efficacy of Neural Interfaces for Sensorimotor Restoration After Tetraplegia

Abstract

BACKGROUND AND OBJECTIVES:

Paralysis after spinal cord injury involves damage to pathways that connect neurons in the brain to peripheral nerves in the limbs. Re-establishing this communication using neural interfaces has the potential to bridge the gap and restore upper extremity function to people with high tetraplegia. We report a novel approach for restoring upper extremity function using selective peripheral nerve stimulation controlled by intracortical microelectrode recordings from sensorimotor networks, along with restoration of tactile sensation of the hand using intracortical microstimulation.

METHODS:

A 27-year-old right-handed man with AIS-B (motor-complete, sensory-incomplete) C3–C4 tetraplegia was enrolled into the clinical trial. Six 64-channel intracortical microelectrode arrays were implanted into left hemisphere regions involved in upper extremity function, including primary motor and sensory cortices, inferior frontal gyrus, and anterior intraparietal area. Nine 16-channel extraneural peripheral nerve electrodes were implanted to allow targeted stimulation of right median, ulnar (2), radial, axillary, musculocutaneous, suprascapular, lateral pectoral, and long thoracic nerves, to produce selective muscle contractions on demand. Proof-of-concept studies were performed to demonstrate feasibility of using a brain-machine interface to read from and write to the brain for restoring motor and sensory functions of the participant's own arm and hand.

RESULTS:

Multiunit neural activity that correlated with intended motor action was successfully recorded from intracortical arrays. Microstimulation of electrodes in somatosensory cortex produced repeatable sensory percepts of individual fingers for restoration of touch sensation. Selective electrical activation of peripheral nerves produced antigravity muscle contractions, resulting in functional movements that the participant was able to command under brain control to perform virtual and actual arm and hand movements. The system was well tolerated with no operative complications.

CONCLUSION:

The combination of implanted cortical electrodes and nerve cuff electrodes has the potential to create bidirectional restoration of motor and sensory functions of the arm and hand after neurological injury.

ABBREVIATIONS:

BMI

brain-machine interface

C-FINEs

composite flat interface nerve electrodes

FES

functional electrical stimulation

IFG

inferior frontal gyrus

ReHAB

Reconnecting the Hand and Arm to the Brain.

Spinal cord injury (SCI) affects more than 250 000 people nationwide with more than 12 000 new cases each year and less than a 1% full recovery rate.¹ 52% of all SCI occurs at high cervical levels, resulting in tetraplegia and the inability to independently perform activities of daily living such as drinking and self-feeding. Many assistive technologies have been developed that provide these individuals with improved abilities to interact with their environment.^2-4 However, control of these devices is limited by inadequate command signals for telling the device what to do, so restoration of dexterous function has been elusive. In the case of upper motor neuron injuries, nerves and muscles distal to the injury are physiologically normal and remain electrically excitable,⁵ which opens the possibility of using functional electrical stimulation (FES) of peripheral nerves and muscles to reanimate the limbs and restore function in people with chronic weakness or paralysis.^5-9

Brain-machine interfaces (BMIs) record neural activity from motor regions of the brain and decipher the cortical patterns into multidimensional movement signals.^10-15 The combination of BMIs (for acquiring movement intention) with FES (for reanimating the paralyzed limb) is an appealing solution because it bypasses the injury and restores function using the most intuitive command source and output system conceivable: the individual's own brain and limbs.^15,16 Although prior BMI-controlled FES studies demonstrated basic feasibility, the achievable functionality was limited in part by the lack of addressable muscles and/or the limited strength of contractions generated by the FES electrode technology used.^15,16 In addition, although prior systems provided movement-related feedback, they did not provide direct somatosensory (touch) feedback during BMI-controlled FES grasping.¹⁷ Touch feedback is critical for dexterous object manipulation, and will likely further enhance upper extremity function provided by the BMI-FES system, as has been observed with robotic systems.^18-20 Here, we describe the surgical approach for implantation of percutaneous multiarray intracortical BMI and peripheral nerve FES systems in the same individual with the goal of restoring brain-controlled movements of the paralyzed upper extremity. We present data demonstrating the efficacy of each system component, including the use of the BMI for restoring touch sensation and the use of the combined BMI-FES system for completing an activity of daily living.

METHODS

ReHAB System Overview

The “Reconnecting the Hand and Arm to the Brain” (ReHAB) system (Figure 1) creates a computerized interface between sensorimotor regions of the brain and the peripheral nerves, circumventing the spinal injury to create an alternative pathway for transmitting activity between the brain and periphery. A motor pathway records neural activity from electrodes distributed across sensorimotor cortex, decodes this information to determine intended movements, and then stimulates subfascicular regions of peripheral nerves to actuate appropriate muscles. In parallel, a sensory pathway provides electrical stimulation to the somatosensory cortex for restored touch perception.

FIGURE 1.

Schematic of the Reconnecting the Hand and Arm to the Brain system. Cortical arrays implanted in motor regions record brain activity associated with intended movements, which is translated into stimulation patterns sent to nerve cuff electrodes surrounding nerves in the contralateral upper limb. Connections between external neurostimulation hardware and the implanted nerve cuff electrodes are made through percutaneous leads exiting the body in the deltoid and cervical regions. Cortical arrays implanted in somatosensory cortex can record neural activity associated with touch stimuli applied to the hand and/or elicit sensations of touch on the hand and arm due to stimulation applied through individual electrode contacts.

Study Participant

A 27-year-old man with C3–C4 AIS-B tetraplegia (motor-complete, sensory-incomplete) from a diving accident 5 years prior was enrolled into the study. An Investigational Device Exemption (ClinicalTrials.gov ID: NCT03898804) was obtained from the US Food and Drug Administration for use of intracortical and peripheral nerve electrodes for research purposes. The protocol was approved by the local Institutional Review Board, which also provided study oversight. Informed consent was obtained for study enrollment before all procedures.

Cranial Electrode Implantation

Six 8 × 8 (64-channel) 1.5-mm iridium oxide microelectrode arrays were implanted into motor and sensory regions of the left (language dominant) hemisphere. Under awake anesthesia, a frontoparietal craniotomy was performed, and central sulcus at the face was identified using somatosensory-evoked potential phase reversal. The hand area on the precentral gyrus was identified through anatomic landmarks and frameless stereotaxy (StealthStation S8 Surgical Navigation System Medtronic PLC).^21,22 The hand-related area of primary somatosensory cortex was identified through awake sensory mapping, in which 4 of the 5 fingers were localized (Figure 2A and 2B). This information was used to allow placement of 2 arrays each in primary motor and sensory hand areas. Additional electrodes were implanted in anterior intraparietal area according to landmarks previously described²³ and in inferior frontal gyrus (IFG) in areas confirmed to be away from language function using awake speech mapping (Figure 2A and 2B). Each array was implanted using a high-speed pneumatic press and implantation of grounding wires.^15,24 The arrays were attached in pairs to 3 percutaneous pedestals (NeuroPort System, Blackrock Microsystems)—1 anterior and 2 posterior to the craniotomy. Connectors were spaced appropriately to allow attachment of headstages for intracortical recording and/or stimulation (Figure 2C and 2D).

FIGURE 2.

Surgical implantation of the cortical electrodes. A, Awake stimulation mapping was performed to identify implant locations. Nonspeech regions of IFG (striped dark blue) were identified posterior to regions that resulted in speech arrest during stimulation (dark blue). Hand regions of primary somatosensory cortex were located by reports of perceived sensations on individual fingers (red, orange, light blue, purple) during stimulation. Locations selected for the implanted arrays are indicated as black squares. B, Locations of implanted electrodes on the brain, superimposed on preoperative structural MRI. S1 arrays targeted index and ring fingertip locations from A. M1 arrays were placed directly across the central sulcus, targeting hand and arm area. The IFG array targeted the border of area 44 and ventral premotor cortices, and the AIP array targeted the medial junction of parietal and postcentral sulci. C, Computed tomography image of array pedestal locations with associated cabling to implanted arrays. D, Postoperative image of healed array pedestal exit sites with and without caps (in place when system not in use). AIP, anterior intraparietal area; IFG, inferior frontal gyrus; ReHAB, Reconnecting the Hand and Arm to the Brain.

Peripheral Nerve Electrode Implantation

In separate operations under general anesthesia, 9 16-channel composite flat interface nerve electrodes (C-FINEs) were implanted around nerves responsible for dexterous function of the right upper extremity. The C-FINE is an extraneural electrode designed to reshape nerves to more selectively recruit interior fascicles.²⁵ Standard surgical approaches were used to expose the infraclavicular brachial plexus (for median, musculocutaneous, lateral pectoral, proximal ulnar, and long thoracic nerves); distal upper arm (distal ulnar nerve); and posterior upper arm/shoulder (radial, axillary, and suprascapular nerves). Ultrasound was used to identify and evaluate the anatomy of exposed nerve segments. A 16-channel C-FINE was secured around each nerve²⁶ (Figure 3A) and individual contacts were stimulated intraoperatively to confirm muscle contractions. Percutaneous in-line connectors (Figure 3B) were connected and tunneled to percutaneous exit sites located in the lateral arm and shoulder (Figure 3C and 3D). A total of 138 double-wound percutaneous leads passed through 71 exit sites, including a triceps intramuscular electrode inserted during a subsequent surgery to further enhance arm extension.

FIGURE 3.

Surgical implantation of C-FINEs and tunneling of externalized leads. A, C-FINE around the exposed radial nerve (not yet sutured closed, arrow). B, C-FINE on suprascapular nerve and inline connector (arrow). C, Inline connectors in nerve exposure incision and tunneled externalized leads, which were subsequently individually tunneled to separate percutaneous exit sites. D, X-ray of implanted nerve cuffs (outlined in red) and inline connectors (outlined in blue), with percutaneous exit sites (circled in yellow). C-FINEs, composite flat interface nerve electrodes.

Postoperative System Validation

Surgical outcome was assessed in monthly visits to evaluate wound healing, side effects, and complications. Chronic recording (all) and stimulation (S1 only) capabilities of the implanted microelectrode arrays were assessed, including measurement of the number of individually recordable neurons and their signal-to-noise ratios to quantify the efficacy of the recordings over a 1-year period. To assess movement-related cortical modulation, the study participant observed movements of different body segments while attempting to perform those movements on cue.^27,28 Neural signals were recorded and analyzed to determine whether the attempted movements were differentiable based on neural activity alone. To assess the efficacy of the S1 arrays at producing sensation, manual microstimulation of individual contacts was delivered as 100 Hz trains of biphasic, cathode-first, charge-balanced pulses of 200 µs duration and 80 µA amplitude (Cerestim R96, Blackrock Neurotech). Stimulation of individual contacts of each C-FINE was used to determine the pattern of muscle activations achievable, as well as the muscle recruitment order. Muscle recruitment was determined by analyzing normalized electromyography signals recorded through fine wire electrodes inserted under ultrasound guidance into target muscles. Joint movement produced by varying pulse amplitude and pulse width of several C-FINE contacts was analyzed, and simultaneous stimulation of multiple C-FINE contacts was used to produce functional hand grasps. Finally, we assessed the participant's ability to modulate brain activity in real time, first to command movements of a virtual arm and second to perform a self-feeding task using his own arm, reanimated through FES. In both tasks, the cortical recordings were deciphered through a mathematical algorithm (the neural decoder) to translate brain signal patterns into movement intent (eg, hand open/close or elbow flexion/extension). The movement intent was converted to changing joint angles in the virtual arm task or to nerve cuff stimulation patterns in the self-feeding task. We quantified his proficiency in commanding the virtual arm to reach specified targets and observed his ability to successfully perform the self-feeding task.

RESULTS

Surgical Outcome

All surgical procedures were well tolerated by the participant with no significant procedural or long-term complications. There was no discomfort associated with the cranial or peripheral exit sites. The participant did not experience any new motor, sensory, or speech deficits.

Motor Recording

Each channel on each microelectrode array could detect up to 5 individually identifiable neurons, based on their distinct wave shapes^29,30 (Figure 4A). There were significant changes over time in the number of isolatable single neurons on each array, with decreases in neuronal counts seen on arrays implanted in somatosensory and primary motor areas and increases seen in arrays implanted in anterior intraparietal and IFG areas (Figure 4B). By contrast, the average signal-to-noise ratio remained consistent over time (P > .05, regression slope test) for all electrode arrays (Figure 4B). Hence, the chronic health of the electrode-tissue interface remained generally stable up to 1 year postimplantation.

FIGURE 4.

Cortical recording of individual neurons and stability over time. A, Neural yields of the implanted arrays. Each 8 × 8 plot represents an implanted 64-channel microelectrode array, with each box insert showing the neural yield from a single channel. Channels could record multiple isolatable neural spikes (represented as different colors, data shown from postimplant day 123). Higher yields were observed on S1, M1, and IFG arrays, with lower yield observed on the AIP array. B, Measurements of the chronic health of the electrode-tissue interface. Top plot: Number of isolatable neurons on each pedestal (n = 128 channels per pedestal) over the first year of the study. Unit counts decreased over time on arrays implanted in S1 and M1 while they increased on the pedestal recording IFG and AIP (increases were mostly seen in IFG). The average SNR ratio was relatively constant (P > .05, regression slope test) across 1 year of neural recordings (bottom plot). AIP, anterior intraparietal area; IFG, inferior frontal gyrus; SNR, signal-to-noise.

The cued movement task revealed that the normalized multiunit spiking activity from motor arrays modulated differentially to specific attempted movements. Attempting different grasp pattern resulted in a different level of activation within individual electrodes (Figure 5A, sample electrode channel in IFG). This differential modulation to grasp pattern occurred about a second before the onset of the attempted movement and typically within a few hundred milliseconds after the cue. Second, the participant was asked to attempt to move 8 different body parts. A classification algorithm correctly identified which body part was moved 79% of the time from recorded signals from the IFG array alone (Figure 5B). This diversity of body representations in what is conventionally thought of as “hand” area is consistent with a previous study in the human hand motor area after paralysis.²⁸

FIGURE 5.

Modulation of neural signals to various attempted movements. A, The participant attempted performing 3 different hand grasp patterns, which resulted in differentiable modulation in the neural activity. Shown is a sample electrode's normalized mean activity time course (solid lines) and standard error (shaded regions) to imagination of power, palmar, and lateral grasp patterns. Imagination of grasp patterns resulted in distinguishable cortical responses both during the visually cued premovement phase (−1500 to 0 ms) and the imagined movement phase (after 0 ms). B, Using cross-validated linear discriminant analysis, attempted movements of 8 different body parts could be accurately classified in 79% of the trials from the neural activity on the IFG array alone. IFG, inferior frontal gyrus.

Sensory Stimulation

Stimulation of individual electrodes in the somatosensory arrays reliably produced sensation of touch on the index, middle, and ring fingers (Figure 6), typically on the anterior surface of the fingertips and occasionally the posterior side of the distal phalange(s). Microelectrodes eliciting reliable sensation tended to be along the posterior edge of the arrays. The locations of sensation elicited by intracortical microstimulation in each region of cortex (Figure 6) matched the sensory locations reported during intraoperative stimulation mapping of the same region (Figure 2A). Sensations were reported to feel like “pressure” applied to the skin or “light squeezing” of the fingertip.

FIGURE 6.

Sensory projected fields resulting from intracortical microstimulation applied to primary somatosensory cortex. Top left: the positions of the microelectrode arrays relative to the perceived locations reported during intraoperative mapping. Note that the locations reported were whole fingers rather than specific phalanges. Middle: The projected fields for each electrode in the medial (top) and lateral (bottom) arrays. The arrays are oriented to match their positions in the image of the cortex. Bottom right: Color legend for projected fields shown on the array diagrams. Each segment of the finger corresponds to a different shade. Gray electrodes may have elicited sensation in some trials, but the locations were not consistently reported for at least 60% of trials.

Functional Electrical Stimulation

Stimulation of each C-FINE contact resulted in activation of muscles innervated by the stimulated nerve, with additional muscles recruited with increasing stimulation amplitude. Although many contacts within each C-FINE recruited similar groups of muscles, the order of recruited muscles differed across contacts (Figure 7). Recruitment order was more similar for neighboring contacts, and recruitment patterns across contacts demonstrated the spatial clustering of muscle-specific efferents within the nerve fascicles (Figure 7). For example, stimulation of the radial C-FINE resulted in activation of wrist (extensor carpi radialis longus) and finger extensors (extensor digitorum communis/extensor indicis proprius), thumb actuators (extensor pollicis longus/abductor pollicis longus), elbow flexion (brachioradialis), and wrist supination (supinator). Brachioradialis activation was more prominent from contacts on the left half of the cuff while supinator activation was more prominent from the right half; all other muscle activations were evenly distributed. Ulnar nerve stimulation produced activation of thumb flexion and adduction (flexor pollicis brevis and adductor pollicis, from the bottom half of the cuff), as well as index finger flexion (first dorsal interosseous, from bottom half of cuff), with other functions (pinky abduction = abductor digiti minimi, finger flexion = flexor digitorum profundus, and wrist flexion = flexor carpi ulnaris) distributed across the cuff contacts. The median nerve C-FINE produced wrist flexion (flexor carpi radialis, predominantly top half), forearm pronation (pronator teres, predominantly right half), with finger (flexor digitorum superficialis, flexor digitorum profundus) and thumb (flexor pollicis longus) flexion distributed across the contacts. The musculocutaneous (biceps and brachialis), lateral pectoralis (upper and lower pectoralis), and axillary (anterior, middle, and posterior deltoid) nerve cuffs also produced spatially organized recruitment of their respective muscles.

FIGURE 7.

Muscle recruitment order associated with stimulation of individual contacts in the C-FINEs.⁴⁶ Muscles represented by a given color act on that color's associated body part (eg, all blue muscles activate finger flexion or extension). Several contacts exhibited spatial organization of muscle activation (eg, left contacts of radial nerve more readily recruited brachioradialis, bottom contacts of ulnar nerve more readily recruited thumb muscles (adductor pollicis and flexor pollicis brevis). Six of the 8 implanted C-FINES resulted in activation of the expected set of muscles. C-FINEs, composite flat interface nerve electrodes.

Nerve stimulation patterns involving concurrent stimulation of contacts on multiple nerves resulted in coordinated movements of the participant's own limb, resulting in functional reaching and grasping. For example, a combined stimulation pattern of contacts on the median, radial, and ulnar nerves resulted in open/close of a lateral hand grasp, and the desired hand aperture was controlled by modulating the stimulation pulse width (Figure 8A). Other patterns of stimulation allowed for execution of a handshake grasp, hand supination/pronation, wrist flexion/extension, elbow flexion/extension, shoulder horizontal flexion, and shoulder abduction. Cortical signals were decoded into command signals to control several hand grasp and arm movements independently. These command signals (0%-100% for each movement) were first used to volitionally control movements of a virtual arm and hand. The participant was able to successfully modulate his brain signals, deciphered through the neural decoder, to simultaneously control virtual arm reaching and hand grasping to reach target postures (Figure 8B). Second, the command signals were used to control movements of the participant's own arm and hand reanimated through FES. Each decoded command signal was mapped to a corresponding multicontact stimulation pattern for FES to generate the target arm and/or hand movement with the participant's own limb. When the output of the neural decoder commanded the stimulation patterns delivered through the C-FINE cuffs, the participant was able to volitionally move and control his arm and hand to bring an object to his mouth for self-feeding (Figure 8C).

FIGURE 8.

Simultaneous stimulation of sets of composite flat interface nerve electrodes contacts resulted in functional movements of the upper limb. A, The command signal (0%-100%) corresponding to the desired hand state (x-axis) modulated the pulse width of electrical stimulation (y-axis) delivered to specific contacts on multiple nerves to produce the desired hand aperture. Nerve stimulation using multiple contacts resulted in varying degrees of opening (left) to closing (right) of the participant's own hand. Top insets: The resulting hand position created through multicontact nerve stimulation at several command signal levels. B, The participant successfully performed brain control of coordinated virtual hand and elbow movements to reach specified positions of the hand. Blue and red boxes represent targets of elbow flexion/extension and hand open/close, respectively. The lines between boxes show the path taken between the target joint positions, noting that movements were made simultaneously, rather than sequentially. C, The participant successfully performed brain control of his own arm and hand movements, reanimated through functional electrical stimulation, to complete a self-feeding task. Both the participant and the researcher shown in this photograph consented to the publication of their images.

DISCUSSION

In this report, we describe a system for brain-controlled stimulation of peripheral nerves to restore voluntary hand and arm movement in a person with chronic motor-complete tetraplegia. We also demonstrate feasibility of a sensory cortical interface that provides touch percepts of the fingertips that can be used for sensory feedback during brain-controlled restored movement.

The most important advantage of the ReHAB system compared with other BMI systems, which have provided control of computer cursors or robotic limbs, is that it restores movement of the user's own arm and hand using FES. FES has been used for many years to activate paralyzed muscles in people with neurological deficits,^5,6 both in the lower (standing/transfer/walking functions^8,9) and upper (reaching/grasping functions^6,7,31) extremities. However, command options to allow the user to tell the system what movements to make have been limited in the high tetraplegia population. The ReHAB system overcomes this challenge by using intracortical microelectrodes to provide a source of complex movement commands in this population.

The ReHAB system advantageously uses nerve cuff electrodes to activate motor efferents, rather than directly stimulating muscles with electrodes placed in individual muscles. Multicontact cuffs can selectively activate different fascicles within a nerve^26,31,32 and thereby selectively recruit one or more muscles innervated by a given nerve. Because a single cuff can recruit all muscles innervated by a given nerve, fewer interfaces need to be implanted to achieve full hand and arm function, thus reducing the necessary number of incisions and the complexity of the implantation procedure. Furthermore, because axons responsible for activation of synergistic muscles tend to be grouped into distinct nerves, peripheral nerve stimulation makes it possible to leverage the natural anatomic organization of the nerves to produce coordinated joint movements. Nerve stimulation can also achieve muscle activation at lower currents than direct muscle stimulation. By using multicontact peripheral nerve stimulating electrodes rather than direct muscle stimulation, we were able to produce coordinated movement across a variety of muscles using relatively few electrode arrays and low stimulation intensities. Note that this study used simple single-contact stimulation of electrodes placed very proximally in the arm to be able to achieve both proximal and distal movement restoration through the same interfaces. Finer movements, for example of individual fingers, are presently being explored using methods such as multicontact stimulation and electrical field steering.³³ Finally, note that the same nerve cuffs used in the present application have been used extensively for lower extremity movement restoration applications in people with chronic paraplegia,^32,34 as well as sensory restoration applications in people with amputations.^35,36 These, along with the percutaneous leads and the intracranial array assembly, have proven durable and safe over many years, with few instances of infection and no adverse events reported during the few instances of explantation.^37-39 Furthermore, the nerve cuff electrode itself has been shown to have no significant impacts on blood nerve barrier permeability, myelination, axon density, and nerve conduction when the pressure exerted by the electrode on the nerve was below 30-40 mm Hg,^25,40,41 a design constraint that was adhered to in this study.

The ReHAB system expands on prior FES studies by providing pathways for direct recording of brain signals to control the patterns of peripheral nerve stimulation to restore voluntary movement. In addition to the potential for high density of information transfer, recording directly from motor areas of the brain may tap into native motor intentions, thus producing more intuitive control. In contrast to implantation only in primary motor areas of the brain, there are many potential advantages to recording from multiple upstream and downstream sensorimotor regions. In nonhuman primates, recording from ventral premotor regions have demonstrated neural activity that tunes to goal-directed activity rather than the more body-centric activity seen in primary cortex.^42,43 Similarly, in monkey and human experiments, single neurons from the anterior intraparietal area have been found to be selective for a variety of imagined hand postures that can be used to control prosthetic hands.^23,44 Combination of information from multiple distinct brain regions may allow for more intuitive and reliable motor control, and we have established that implantation in multiple brain regions is safe and well-tolerated.

Finally, the ReHAB system combines brain-controlled arm movements with sensory feedback provided via microstimulation of primary somatosensory cortex. Sensory feedback is critically important for performing naturalistic movements involving object grasp and manipulation. Previous studies have used focal brain stimulation to produce artificial sensory percepts for improving control of brain-controlled robotic limbs and BMI user experience,^18,19,45 but these and other¹⁷ studies did not combine sensory feedback with brain-controlled FES. Notably, brain stimulation in our participant conveyed a naturalistic sense of touch on regions of the fingers critical for grasp function. Our results confirm that microelectrode stimulation can achieve sensory percepts at the level of individual fingertips and demonstrate that sensory locations evoked by microelectrode stimulation are spatially congruous with those reported during intraoperative awake mapping. When fully integrated into a bidirectional sensorimotor system, the ReHAB technology may provide some degree of functional independence to people living with upper extremity paralysis.

Limitations

It should be noted that only upper motor neuron injuries can be addressed using peripheral nerve stimulation because denervation atrophy associated with peripheral nerve injury (including at the level of SCI) will preclude muscle activation. In addition, it is necessary that the brain cortex be intact for recording of motor function, which prevents application when pathology involves damage to motor neurons (such as stroke) or progressive neurodegenerative conditions. This report involves a single participant; however, follow-up studies are underway to assess generalizability of the approach. Finally, we note that in this early investigational version of the system, the high number of implanted components and externalized leads may be of concern. The choice of hardware in this study allowed for examination of multiple configurations of recording and stimulation to determine necessary requirements for functionality. For future clinical translation, we envision reducing the implanted hardware to only the minimum set of cortical and peripheral electrodes necessary to achieve the desired functions for a given user.

CONCLUSION

Brain-controlled peripheral nerve stimulation has the potential to allow restoration of arm and hand function after neurological injury. Although further development is necessary, this strategy may allow restoration of naturalistic movements of the reanimated upper extremity to perform simple activities of daily living. A complete bidirectional system, fully integrating motor restoration with touch feedback provided by brain stimulation, has the potential to significantly improve user independence and quality of life.

Funding

This study was supported by Congressionally Directed Medical Research Program—Spinal Cord Injury Research Program, Clinical Trial Award SC180308; NIH R01NS119160 (National Institute of Neurological Disorders and Stroke). E.C.C., K.E.A., and P.B. were supported by 5T32EB004314 (National Institute of Biomedical Imaging and Bioengineering). C.A.H. was supported by NIH T32AR007505 (National Institutes of Arthritis and Musculoskeletal and Skin and Diseases).

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials or devices described in this article. Dustin Tyler is a consultant for Barologics, LLC.