Selective neural electrical stimulation restores hand and forearm movements in individuals with complete tetraplegia

The incidence of spinal cord injuries (SCIs) in Western Europe and the United States is estimated at 16 and 40 cases per million, respectively [1], with the proportion of cervical injuries steadily increasing. SCIs can have a devastating impact on patient health, autonomy and quality of life. Technical aids (e.g., motorized wheelchairs, orthoses, medical electric beds, transfer boards, home automation, etc.) can restore some independence to people with tetraplegia, but recovery of gripping movements is still the priority [2‐5]. Indeed, most activities of daily living are performed via hand movements and therefore restoration of active forearm, hand and wrist motricity would help patients recover greater autonomy and thus increase their quality of life. In the absence of spinal cord repair, only partial solutions are available today for this population. Functional surgery is mainly based on musculotendinous transfers and has been used to restore partial movements of the hand and wrist for several decades. A set of muscles under voluntary contraction are therefore used, with the only condition being that their initial function is compensated by agonist muscles still under voluntary control [6]. More recently, nerve transfers have been attempted to re-innervate paralyzed muscle to recover voluntary control [7]. However, both methods require a sufficient number of muscles or nerves that are still under voluntary control. The transferred as well as the remaining agonist muscles should be strong enough to ensure efficient recovery [8]. The alternative is to use an implanted or external functional electrical stimulation (FES) device, provided that the sublesional paralyzed muscles are still innervated by intact motoneurons. One of the first applications of FES to recover hand motion was reported by Catton and Backhouse in 1954 [9]. FES was then used to recover grip movements in patients with high tetraplegia as early as 1963 [10‐12]. These devices used either intramuscular or epimysial electrodes, requiring one electrode for each muscle involved in the targeted movement. Surface electrode arrays [13] may be used but are clearly limited in their present form to the laboratory. Moreover, external or percutaneous devices are very limited in terms of acceptability, efficiency and benefits and thus are not used by patients in a daily-living context. The only successful device has been the FreeHand system: more than 250 patients [14] have had it successfully implanted with clear benefits [15, 16]. Up to 12 muscular electrodes have been implanted [17], but a research version attempted to replace several muscle electrodes with a single neural 4-contact electrode [18]. The results on selectivity remained limited during the implanted phase, due to the adopted approach based on a monopolar scanning of the available electrode contacts. The same team also tried neural cuff electrodes to provide movements in the whole upper limb but the setup required 2 implants, 12 to 14 intramuscular electrodes and 6 4-contact cuffs electrodes [19]. The implantation needed 3 (resp. 2) surgeries for the first (resp. second) patient. In both cases, one implant was almost dedicated to the fingers, thumb, wrist movements with 12 intramuscular electrodes. The second implant was mostly dedicated to elbow and shoulder movements with 2 intramuscular electrodes and the use of cuff electrodes limited to the best contact response among the 4 against a global reference leading to a monopolar like stimulation that is not the most selective one. Simplified steering current paradigm was also tested and showed increased selectivity even though limited to the radial and musculocutaneous nerves [20]. The FreeHand device was commercialized but was limited by the complexity of the surgery due to the high number of muscle electrodes implanted over the arm/forearm/hand; the surgery was estimated to last 5 h with an access to all targeted muscles [15] and it is not commercialized since 2001.

A neat solution would consist of activating groups of muscles through a limited number of neural cuff electrodes. Multicontact selective neural stimulation has the advantage of activating several muscles via one electrode, and it requires much less energy than epimysial or intramuscular stimulation as thresholds are at least 10 times lower. Human trials have demonstrated the feasibility of this approach for hand movements [18, 21, 22] but, as it combined multisite neuromuscular stimulation on all targeted muscles, it leaded to a high number of implanted electrodes (more than 10) and the associated wires from the forearm to the chest and therefore was from this point of view not more advantageous than the original FreeHand system. The limited efficacy of the nerve stimulation was due to both the limited selectivity of the electrode that was used and the simplicity of the stimulation paradigm, essentially monopolar configurations. Indeed, there were four contacts with a global reference far from the electrode and a single contact among the four was used during stimulation. More complex multicontact electrodes were successfully used in the upper limb of humans, i.e., the FINE [23, 24] and the TIME [25, 26]. However, the stimulation paradigm remained limited to monopolar stimulation for which a single active contact was used toward a global ground. This approach was also not the most selective. Moreover, TIME was used for sensory feedback recovery, the FINE and the 4-contact electrode showed limited selectivity, which meant that a high number of implanted elements were needed to target all the required muscles.

Our first hypothesis was that multicontact electrodes implanted above the nerve bifurcations would enable the selective activation of several fascicles from the same nerve. Indeed, upper limb fascicles tend to anastomose and separate over a large part of their length but are somatotopically organized distally from the spinal cord [27]. This selective activation might potentially activate various functions or muscles independently. Thus, depending on the subject, it might be possible to selectively activate agonist groups of muscles to obtain functional outcomes.

Second, we hypothesized that using only two multicontact cuff electrodes around the radial and median nerves in association with optimal current spreading to all the contacts would result in much higher selectivity and thus isolated or functional movements [28]. This approach aims at replacing the use of a single individual cuff with monopolar-like stimulation paradigms or / one muscular electrode per targeted muscle, by multi-contact neural electrodes implanted proximally with 3D current shaping to achieve a high precision in activating the targeted zone within the nerve linked to functional desired movements.

To validate both hypotheses, we carried out intraoperative trials on either the median or radial nerve in eight subjects with tetraplegia during already-scheduled surgeries. This paper presents the very encouraging and innovative results of the study that open the path to the full restoration of hand movements with a minimally invasive medical device.

A Pearson’s Chi-squared test was performed for the configurations, which reached the SIR to assess the independence between the configurations, the movements and the participants.

The eight participants presented tetraplegia and were candidates for tendon transfer surgery. Depending on the surgery, either the radial (subjects P1 to P4) or median (subjects P5 to P8) nerve could easily be exposed and stimulated. During surgery, a limited time window of about 30 min was dedicated to the entire experimental procedure, including cuff electrode placement around the nerve. The objective was to avoid increasing the duration of general anesthesia and prolonged lying. Thus, the 3D-shaping current configurations were pre-computed and only the scale factor of the current was adjusted for each subject individually based on threshold detection of muscle contraction through a classical tripolar configuration. Indeed, in a theoretical study combined with a preclinical trial we showed that several 3D-shapings were optimal, depending on the size and location of the targeted area within the nerve, and this was almost independent of the subject, except the global scale [28]. A 3-ring composed of three contacts was used for the radial nerve and a 3-ring of four contacts was used for the median nerve as its diameter was larger and the required selectivity was higher. During surgery, we determined the current intensity threshold for each subject and then the predetermined configurations were automatically scanned with increasing intensities following pre-computed and subject-specific ranges. Both a physician and a trained engineer systematically analyzed all videos and extracted individual forearm movements described in the following.

We implanted a single multicontact cuff electrode around the median or radial nerve in 8 subjects with complete tetraplegia and obtained almost all the movements needed, either combined (Tables 2 and 4) or isolated (Table S2 in Additional file 1, Additional file 2), to restore functional grasping of objects. Indeed, missing key grip for P8 is due to the fact that configuration STR(B2) induced not only a key grip but also some wrist flexion and pronation that could lead to a non-functional movement in the end. For P5 power and hook grips using a combination of fingers flexion TTR(B3) alone or with thumb opposition TTR(B4) was possible but we did not report these movements as functional as the MRC score was lower than 3 and do not provide a powerful grip. Training could probably provide the needed force. Concerning opening the hand, mostly fingers and thumb extension; the movement was obtained with P1. For the others, fingers extension was combined with wrist dorsiflexion – possibly increased by mechanical coupling. For P3, an additional thumb flexion using interleaved TT (C1) can be used to completely open the hand. For P2 and P4, we did not achieve thumb extension probably due to the fact that the cuff has a too limited number of contacts (#3) that should be changed for future trials (Fig. S9, Additional file 1). Indeed, the nerve organization differs from patient to patient and we obviously had not enough contacts within the cuff electrode: on patient P3, at the end of the session, the cuff around the radial nerve was turned about half an inter contact distance and a limited scan to TT configurations was performed. It shows, for instance, isolated fingers flexion that was not observed with the initial position (data excluded as it is a single case without accurate rotation measurement). An augmented number of contacts together with a fine-tuning of all stimulation parameters with some extra configurations tested outside the operating room may greatly extend the number of obtained movements. Finally, all the patients were at the same level of spinal cord injury but not the same Giens score (Table 1) that may lead to muscles that cannot be activated by electrical stimulation.

This approach is the most economical to date in terms of the number of implantable parts, as only two electrodes above the elbow would be needed to restore the main functional movements of the hand for individuals with complete tetraplegia. Besides, on another application, namely lower limb movement restoration on subject with complete paraplegia [35‐37], we showed that neural stimulation is more reliable and efficient on a long term (9 years) than epimysial, further confirming the interest of a full neural system [38]. We confirmed the hypothesis derived from theoretical studies that the spatial selectivity obtained with both multicontact electrodes and complex current spreading can provide selective movements. In particular, to our knowledge, we provide for the first time, the evidence that a stimulation of the median nerve can provide a selective activation of the fingers, thumb and wrist flexors. This means that despite anastomosis, which is known to occur in the proximal part of upper limb nerves (in our case above the elbow), fascicularization is functionally relevant even though it differs from individual to individual. We further confirmed, for the first time in human, that complex 3D-current shaping is efficient and that the prior theoretical studies without a priori knowledge of a generic nerve are relevant. Indeed, as demonstrated in our technical paper [28], only the scale factor of the current intensity needs to be adjusted but the exact knowledge of the fascicule organization is not mandatory and in any case is not available in individuals.

These results were obtained with a significant group of eight participants. The next step should be a clinical trial with awake participants. This proof of concept was mandatory from an ethical viewpoint, and it clearly showed the possibility of combining isolated movements through interleaved stimulation paradigms or through deeper investigation of the recruitment obtained with the most relevant configurations, which provide even richer and more complex movements. During the intraoperative testing, automatic scanning showed that recruitment was always progressive for both the amplitude of the movement and the number of primary generated movements. For instance, in subject P7 with configuration TT(A4), we first see the isolated flexion of the fingers with an increasing movement amplitude (233–467 μA), followed by the activation of thumb flexion providing a key grip (> 467 μA), and finally wrist pronation. This illustrates that the selective configurations targeted increasing areas as the intensity increased.

Last, owing to pure neural stimulation, no diffusion to antagonist muscles was observed. The opening function through the stimulation of the radial nerve was more difficult to assess for several reasons. Notably, some subjects were spastic so the hand was almost closed and the stimulation of the radial would have needed more reinforcement and rehabilitation or a surgical elongation of the flexion muscles. But even in this case, we obtained hand opening. Finger extension may induce wrist extension so the co-activation of the median nerve would have been beneficial, but it was not possible in the frame of this trial. However, the system will be able to provide this combined stimulation for future trials.

We did not report the results of the pre-surgical mapping as it was an inclusion criterion for the subjects, but we found that the contraction obtained with surface electrical stimulation through accurate mapping was always less efficient, less selective and unable to activate some of the deep and small muscles. Moreover, we did not stimulate the ulnar nerve so the fifth finger was not activated, but this does not seem necessary to obtain strong grasping.

The present work is a proof of concept of restoring functional movements through a multi-contact cuff electrodes placed around the median or radial nerve owing to complex current configurations over the 12 contacts. We thus aim at proposing the less invasive solution to date that could restore useful hand grasping for subjects with complete tetraplegia. Our ultimate goal is to offer a device to the patients who cannot benefit from musculotendinous surgery to help them recover grasp movements. We hope to do so using this technology in combination with EMG from sublesional muscles or movements [39] to control nerve stimulation. The next step will be to implant two neural cuff electrodes and let patients control the device themselves through the EMG-based interface.