Feasibility of a Wireless Implantable Multi-electrode System for High-bandwidth Prosthetic Interfacing: Animal and Cadaver Study

Abstract

Background

Currently used prosthetic solutions in upper extremity amputation have limited functionality, owing to low information transfer rates of neuromuscular interfacing. Although surgical innovations have expanded the functional potential of the residual limb, available interfaces are inefficacious in translating this potential into improved prosthetic control. There is currently no implantable solution for functional interfacing in extremity amputation which offers long-term stability, high information transfer rates, and is applicable for all levels of limb loss. In this study, we presented a novel neuromuscular implant, the the Myoelectric Implantable Recording Array (MIRA). To our knowledge, it is the first fully implantable system for prosthetic interfacing with a large channel count, comprising 32 intramuscular electrodes.

Questions/purposes

The purpose of this study was to evaluate the MIRA in terms of biocompatibility, functionality, and feasibility of implantation to lay the foundations for clinical application. This was achieved through small- and large-animal studies as well as test surgeries in a human cadaver.

Methods

We evaluated the biocompatibility of the system’s intramuscular electromyography (EMG) leads in a rabbit model. Ten leads as well as 10 pieces of a biologically inert control material were implanted into the paravertebral muscles of four animals. After a 3-month implantation, tissue samples were taken and histopathological assessment performed. The probes were scored according to a protocol for the assessment of the foreign body response, with primary endpoints being inflammation score, tissue response score, and capsule thickness in µm. In a second study, chronic functionality of the full system was evaluated in large animals. The MIRA was implanted into the shoulder region of six dogs and three sheep, with intramuscular leads distributed across agonist and antagonist muscles of shoulder flexion. During the observation period, regular EMG measurements were performed. The implants were removed after 5 to 6 months except for one animal, which retained the implant for prolonged observation. Primary endpoints of the large-animal study were mechanical stability, telemetric capability, and EMG signal quality. A final study involved the development of test surgeries in a fresh human cadaver, with the goal to determine feasibility to implant relevant target muscles for prosthetic control at all levels of major upper limb amputation.

Results

Evaluation of the foreign body reaction revealed favorable biocompatibility and a low-grade tissue response in the rabbit study. No differences regarding inflammation score (EMG 4.60 ± 0.97 [95% CI 4.00 to 5.20] versus control 4.20 ± 1.48 [95% CI 3.29 to 5.11]; p = 0.51), tissue response score (EMG 4.00 ± 0.82 [95% CI 3.49 to 4.51] versus control 4.00 ± 0.94 [95% CI 3.42 to 4.58]; p > 0.99), or thickness of capsule (EMG 19.00 ± 8.76 µm [95% CI 13.57 to 24.43] versus control 29.00 ± 23.31 µm [95% CI 14.55 to 43.45]; p = 0.29) were found compared with the inert control article (high-density polyethylene) after 3 months of intramuscular implantation. Throughout long-term implantation of the MIRA in large animals, telemetric communication remained unrestricted in all specimens. Further, the implants retained the ability to record and transmit intramuscular EMG data in all animals except for two sheep where the implants became dislocated shortly after implantation. Electrode impedances remained stable and below 5 kΩ. Regarding EMG signal quality, there was little crosstalk between muscles and overall average signal-to-noise ratio was 22.2 ± 6.2 dB. During the test surgeries, we found that it was possible to implant the MIRA at all major amputation levels of the upper limb in a human cadaver (the transradial, transhumeral, and glenohumeral levels). For each level, it was possible to place the central unit in a biomechanically stable environment to provide unhindered telemetry, while reaching the relevant target muscles for prosthetic control. At only the glenohumeral level, it was not possible to reach the teres major and latissimus dorsi muscles, which would require longer lead lengths.

Conclusion

As assessed in a combination of animal model and cadaver research, the MIRA shows promise for clinical research in patients with limb amputation, where it may be employed for all levels of major upper limb amputation to provide long-term stable intramuscular EMG transmission.

Clinical Relevance

In our study, the MIRA provided high-bandwidth prosthetic interfacing through intramuscular electrode sites. Its high number of individual EMG channels may be combined with signal decoding algorithms for accessing spinal motor neuron activity after targeted muscle reinnervation, thus providing numerous degrees of freedom. Together with recent innovations in amputation surgery, the MIRA might enable improved control approaches for upper limb amputees, particularly for patients with above-elbow amputation where the mismatch between available control signals and necessary degrees of freedom for prosthetic control is highest.

Introduction

Restoring hand function in patients who undergo upper limb amputation poses a great challenge to physicians and orthopaedic technicians. The task of providing a highly capable substitute for the lost limb increases in difficulty with ascending amputation level [3, 30, 40]. Although body-powered systems are still widely used, myoelectric control approaches are now considered the standard of care, offering the advantage of transferring intuitive user intent to a robotic arm [3, 7]. Currently available devices facilitate prosthetic control through transcutaneous electromyography (EMG) recording, using surface electrodes embedded into the socket of the device. The drawbacks of this decades-old approach are well known, including crosstalk between signals, frequent alterations in signal quality through postural changes or sweating, inaccessible deep muscles, and a generally low signal-to-noise ratio [8, 26, 27]. These shortcomings can be addressed through implantation of electrodes, thereby moving them closer to the signal source and limiting noise. As in any implantable biomedical device, biocompatibility is a major concern, particularly regarding long-term use [12]. Although there has been some work on implantable myoelectric interfaces in the setting of functional electrical stimulation after upper motoneuron damage [15, 21], only a few systems intended for prosthetic interfacing have been reported so far [27]. We previously presented a preliminary implant design that uses four epimysial contact sites to acquire muscle signals, showing promising results in an animal study [5]. Most notably, we recently reported on the first long-term implantation of the Implantable Myoelectric Sensors system in three patients with targeted muscle reinnervation, demonstrating improved prosthetic control through the benefits of intramuscular signal pickup [36]. The Implantable Myoelectric Sensors system uses a circumferential coil for communication with and powering of the implanted electrodes, an approach that is not feasible for patients who undergo short transhumeral amputation or those with shoulder disarticulation. The osseointegrated human-machine gateway (OHMG) is another implanted solution for neuromuscular interfacing, which can transfer information into and out of the body via wired communication for prosthesis control and sensory feedback [20, 28, 29]. However, this method currently only uses two EMG signals for prosthetic control [24] and requires a diaphysis for implantation, which also renders it nonapplicable for glenohumeral amputations. Moreover, a permanent transcutaneous port is only accepted by few patients who undergo amputation and poses a constant infection risk, which can result in surgical removal [39, 41]. The mentioned systems use two to six distinct myoelectric electrode sites, which means that only 1° to 3° of freedom can be attained for direct prosthesis control. Regarding the number of available efferent control signals, these systems thus do not provide an advantage compared with currently used surface electrodes.

Although innovations in amputation surgery have translated into clinical practice, the current bottleneck for improving prosthesis control is the biotechnological interface. There is clearly a need for a novel solution that can translate the dense neuromuscular information created by targeted muscle reinnervation into highly dexterous and natural prosthetic hand movement. Such a system should be fully implantable, provide long-term biocompatibility, and offer high-bandwidth myoelectric information transfer from several intramuscular recording sites. Furthermore, it should be compatible with all levels of major upper limb amputation. Here, we evaluate a novel solution regarding these criteria. The Myoelectric Implantable Recording Array (MIRA) consists of a central unit connected to 32 electrode contacts, which are distributed across eight intramuscular leads for gathering EMG signals. The flat-coil design enables flexible placement within the limb and provides wireless communication and power supply.

The aim of this study was to evaluate the MIRA in terms of biocompatibility, functionality, and feasibility of implantation to lay the foundations for clinical application. This was achieved through small- and large-animal studies as well as test surgeries in a human cadaver.

Materials and Methods

Overview

Testing the MIRA consisted of three distinct experiments, gradually evaluating its usability for patients with limb loss (Fig. 1). First, biocompatibility was tested in rabbits, followed by chronic functional evaluations in sheep and dogs. The surgical procedures for implanting the system in humans were developed and assessed in a fresh cadaveric specimen. Primary endpoints were inflammation, tissue response, and capsule thickness for the rabbit study; mechanical stability, telemetric capability, and EMG signal quality after long-term implantation in the large-animal study; and, finally, feasibility to implant relevant target muscles in all major amputation levels during the test surgeries. These experiments were conducted in a stepwise fashion to assess viability and functionality of the MIRA system for human application and for developing appropriate surgical approaches.

Fig. 1.

An overview of the stepwise study design for evaluation of the MIRA is shown. A color image accompanies the online version of this article.

Implant Design

The MIRA system (Ripple Neuro LLC) consists of a central unit (29.6 mm x 23 mm x 5.3 mm) and nine individual silicone leads (Fig. 2). Eight leads carry four stainless steel contacts, with a polypropylene barb at the end to provide intramuscular fixation. Design of the leads was adapted from an intramuscular lead design presented by Memberg et al. [22]. The ninth lead is shorter and serves as a reference. Thirty-two contacts may thus be used to gather raw, intramuscular EMG data. Leads may be manufactured in varying lengths to accommodate different anatomic requirements. The central unit is hermetically enclosed in yttria-stabilized zirconia ceramic, which is also widely used in other medical implants such as joint prostheses [18, 33]. The data gathered from the 32 channels are transmitted to an external transceiver disc via transcutaneous telemetry using infrared light, at a transfer rate of 10.2 Mb/s or more. The implant is powered through inductive coupling.

Fig. 2.

A-C The design of the MIRA implant is shown. (A) This schematic drawing depicts the MIRA implant, including its dimensions in mm. (B) This close-up view shows the implant’s central unit. (C) This close-up view shows the intramuscular silicone leads, each with four stainless steel contacts and a polypropylene barb at the end. A color image accompanies the online version of this article.

Biocompatibility

The principal metrics used to assess biocompatibility were inflammation, tissue response, and capsule thickness after 3 months of implantation of the EMG leads in rabbits. These quantitative measures were scored according to histopathological analysis and used to compare the local tissue effects of chronic implantation between the leads and a biologically inert control article (high-density polyethylene).

The in vivo biocompatibility tests were conducted in four female New Zealand white rabbits (age 5.2 months, weight range 2.50 to 2.55 kg). The EMG leads and a biologically inert control material, high-density polyethylene, were implanted into the paravertebral muscles of the animals. Ten pieces each of the distal EMG leads (1 mm x 10 mm) and the high-density polyethylene (1 mm x 10 mm) were implanted into the paravertebral muscles of the animals over 3 months. We chose this duration to allow sufficient time for completion of the major processes of foreign body reaction [2]. The number of samples was determined according to standard group sizes for the evaluation of biomedical implant biocompatibility because a power analysis was not feasible due to lack of previous data [1]. We reduced the number of animals to four because up to six individual leads could be implanted in each rabbit. No randomization of the animals was necessary as control and test substances were both implanted in each animal at corresponding sites (left and right) to reduce confounding factors. We used female animals because of a handling preference at the local institution, and each animal was selected for the study based on a physical examination to ensure good health and physical suitability. All animals were single-housed in cages, with monitored room temperature and lighting as well as food and water ad libitum. Anesthesia was performed with inhaled isoflurane, at a maintenance dose of 1% to 5%. The surgical approach was made through a 3-cm to 5-cm incision along the midline of the back, the fascia was cut and the paravertebral muscles were exposed. Small individual pockets in the muscle were created. EMG lead pieces were placed into the right paravertebral muscles and high-density polyethylene pieces were placed into the left paravertebral muscles at corresponding locations, after which the wounds were closed layer by layer. Postoperative analgesia was provided with tramadol at 11 mg/kg in a drinking water solution, administered twice daily.

After 91 days of follow-up, the animals were euthanized, the implants were extracted, and tissue samples were taken. Gross macroscopic observations were recorded during explantation. Representative blocks of tissue, including the implant, were collected for both the EMG leads and the control article and stained with hematoxylin and eosin. Histopathological scoring was performed according to a specific protocol for biocompatibility evaluation (Supplementary Digital Content 1; http://links.lww.com/CORR/A736). The main outcome parameters were inflammation score, which includes the presence and type of inflammatory cell infiltration; tissue response score, which includes neovascularization; fibrosis and fatty infiltrate; and finally, thickness of the implant capsule. Other evaluated features were the presence of necrosis, granulation tissue, mineralization, granulomas, myocyte degeneration and regeneration, hemorrhage, tissue ingrowth into the device, foreign debris, and pseudo bursal formation.

Functionality

The principal metrics we used to assess chronic in vivo functionality in large-animal models were mechanical stability of the implant, its ability to establish telemetric connection and transfer biological signals as well as the quality of transmitted EMG signals, in particular through analysis of the signal-to-noise ratio. Secondary parameters of interest were electrode impedance and tissue reaction.

For the purpose of functional evaluation, the full implant was tested in two chronic large-animal models, using six dogs (English beagles, male, age 14 months, weight range 8.5 to 12.1 kg) and three sheep (Pecora Alpina Tirolese, male, age 24 months, weight range 65 to 80 kg). In both studies, one MIRA system was implanted in the shoulder region of each animal, with the intramuscular electrodes placed in different agonistic and antagonistic muscles. Since this study does not produce numerical data amenable to sample size calculations, group sizes were similar to prior comparable studies which evaluated neuromuscular implants [1, 5, 23]. We chose only male animals to prevent unplanned breeding in the setting of group housing. All animals were housed in groups, with species-appropriate temperature and lighting as well as water ad libitum and feed according to local species-specific protocols. Anesthesia and perioperative pain management were performed according to the following protocols:

Dogs: premedication with morphine (0.5 to 1.0 mg/kg), atropine (0.02 to 0.04 mg/kg), and acepromazine (0.01 to 0.05 mg/kg); induction with propofol (2 to 10 mg/kg); maintenance with isoflurane (0% to 5% inhalation); sheep: premedication with atropine (1 mg), ketamine (3.0 mg/kg), and detomidine (0.07 mg/kg); induction with propofol (2.5 to 5 mg/kg) and fentanyl (0.1 mg); maintenance with isoflurane (1% to 2% inhalation), fentanyl (0.01 mg/kg/hour), propofol (10 mg/kg/hour), and atracurium (0.3 mg/kg/hour). The animals were mechanically ventilated and body temperature was monitored throughout surgery.

The surgical technique for implantation was similar in sheep and dogs (Fig. 3A-D). The central unit was placed directly on the fascia of the adjacent muscles and secured with nonabsorbable sutures (Supplementary Digital Content 2; http://links.lww.com/CORR/A737). Care was taken to position the central unit in a location of little relative motion during gait, to facilitate a stable telemetry connection to the external transceiver disc. The subcutaneous tissue overlying the central unit was resected to minimize the distance between the implant and the transceiver. After implant placement and fixation, the wounds were closed layer by layer. Postoperative analgesia was provided with carpofen (4.0 mg/kg, 1 to 2 times/day) for all animals with additional morphine (0.5 to 1 mg/kg) or fentanyl skin patches (2 to 3 μg/kg) as needed. The animals were observed until primary wound healing occurred. Afterward, regular EMG and impedance measurements were conducted. For the recording sessions, the animals were fitted with an external receiver apparatus consisting of a transceiver disc, which must be aligned over the implant and is connected via cable to a processor for data storage (Nomad, Ripple Neuro LLC). Therefore, custom-fitted saddles were designed and crafted for the animals, positioning the processor on the back and the transceiver disc over the shoulder, secured in place through tubular bandage (Fig. 3E). For the EMG measurement sessions, the animals were regularly trained to tolerate the saddle and reliably perform walking tests. Five to 6 months postoperatively, the devices of the three sheep and five dogs were explanted; the system remained implanted in one dog. Duration of implantation was chosen to provide functional data that can be classified as long-term or chronic in the evaluation of neuromuscular implants, which has been defined as a minimum of 3 months [1, 5]. Given that these implants are designed to be worn by prosthesis users for years to decades and thus must be able to withstand continuous mechanical stress due to extremity motion, it was decided to double this time period to detect any later mechanical complications. Before explantation, a final test of the telemetric communication was performed; during the procedure, histologic samples were taken for tissue analysis.

Fig. 3.

A-E These images show surgery and follow-up in sheep. (A) First, the target muscles in the shoulder region were accessed via a curvilinear incision. (B) After all relevant muscles were identified, intramuscular leads were inserted with a specialized applicator. Care was taken to avoid lead kinking after final placement. (C) The central unit of the implant was then secured to the muscle fascia with nonabsorbable sutures. (D) The subcutaneous fascia was subsequently closed, taking care to leave a window over the central unit to ease telemetric communication. (E) The setup during a measurement session after successful implantation is shown. A customized saddle was designed to carry a processor for data storage (Nomad, Ripple Neuro LLC), which was connected to the transceiver disc aligned over the implant. The green light on the transceiver disc indicates an active connection to the implant. A color image accompanies the online version of this article.

The signal-to-noise ratio was computed as the ratio (in dB) between the EMG signal power and the power of baseline noise in resting conditions. The signal-to-noise ratio of the EMG activity was used to quantify the quality of each myoelectric signal, using two recording sessions from a dog and a sheep. These two representative recordings were chosen according to the animals’ ability to produce repetitive muscle activations during the measurement sessions. The raw myoelectric signals were recorded at a sampling frequency of 1000 Hz. The signal-to-noise ratio was computed from signal intervals of 500 ms after removal of the offset (DC component). Active and baseline signal intervals were automatically identified as the intervals with the largest and lowest signal power, respectively.

Feasibility of Implantation

The main goal of the test surgeries was to develop suitable surgical approaches for each major amputation level of the upper limb, with the aim to place intramuscular leads in all relevant target muscles at each corresponding level [35]. Secondary parameters of interest were a biomechanically stable environment of the central unit and appropriate placement of the external transceiver disc, which would not be expected to interfere with prosthetic fitting or cause patient discomfort. To test the feasibility of this concept in humans, implantation was tested in a fresh human cadaver (male, 71 years). In an interdisciplinary setting, which included specialists from reconstructive surgery (CG, OCA), engineering (CH, DM), and prosthetic and orthopaedic technology (HD, who is not a study author), the requirements for a successful surgical approach were discussed and then explored in the cadaver. Surgical approaches were developed for the three major amputation levels of the upper limb (transradial, transhumeral, and glenohumeral).

Ethical Approval

All animal experiments received approval by the responsible review boards (protocols IACUC# RPL-1702 and BMBWF-66.009/0036-V/3b/2019). Animal care was in accordance with the respective institutional guidelines.

Statistical Analyses

All quantitative data are presented as mean ± SD. Histological scores of the biocompatibility study were compared between EMG leads and control article using paired t-tests. Signal-to-noise ratio in dB was compared between a sheep and a dog EMG measurement session, also using the paired t-test. All statistical analyses were performed using the Analysis ToolPak for Microsoft Excel (2018). Statistical significance was defined as p < 0.05.

Results

Biocompatibility

The EMG leads were deemed biocompatible, since no differences regarding inflammation, tissue response and capsule thickness were found compared with the inert control article (high-density polyethylene) after 3-month intramuscular implantation in rabbits.

Of the four rabbits that were implanted, none showed any adverse reactions during the evaluation period, and no animals were excluded during the study period. Gross macroscopic observations revealed no signs of peri-implant hemorrhage, fluid accumulation, encapsulation, necrosis, discoloration, infection, or draining lymph nodes in any of the specimens. Initial histopathologic observations confirmed the gross macroscopic findings; the implant-myocyte interface appeared normal, and there was no evidence of sepsis. A further histopathologic analysis revealed that inflammation in response to implantation of the EMG lead pieces was composed of low numbers of macrophages and multinucleated giant cells along the implant’s surface, with two instances of lymphocytes in the fibrous capsule surrounding the implant. Implantation of both the high-density polyethylene and the EMG lead pieces resulted in infiltration of macrophages, multinucleated giant cells, and lymphocytes. As a sign of the healing response, some neovascularization was observed, as well as fatty infiltration and fibrous encapsulation of the implant. Neither scores for inflammation (EMG 4.60 ± 0.97 [95% CI 4.00 to 5.20] versus control 4.20 ± 1.48 [95% CI 3.29 to 5.11]; p = 0.51) nor tissue response (EMG 4.00 ± 0.82 [95% CI 3.49 to 4.51] versus control 4.00 ± 0.94 [95% CI 3.42 to 4.58]; p > 0.99) indicated any differences in biocompatibility between the EMG leads and the control article (Fig. 4). Regarding average fibrous capsule thickness, no differences could be identified (EMG 19.00 ± 8.76 µm [95% CI 13.57 to 24.43] versus control 29.00 ± 23.31 µm [95% CI 14.55 to 43.45]; p = 0.29). Additionally, there was no evidence of mineralization, hemorrhage, necrosis, granulation tissue, granuloma formation, myocyte necrosis or degeneration, or tissue ingrowth into the implants in either of the groups. Overall, it was considered that the inflammatory response to chronic implantation of the EMG leads was low grade and would not be expected to inhibit healing of the implantation site (Supplementary Digital Content 1; http://links.lww.com/CORR/A736).

Fig. 4.

This figure shows histopathologic results 3 months after implantation of the intramuscular leads. The biocompatibility of the novel EMG leads was compared with that of the inert control substance after chronic intramuscular implantation. Scoring was performed according to a standardized protocol for assessing tissue responses (Supplementary Digital Content 1; http://links.lww.com/CORR/A736); the mean capsule thickness is shown in µm. Similar results were observed across all areas of interest, demonstrating low irritancy of the EMG lead. HDPE: high-density polyethylene. A color image accompanies the online version of this article.

Functionality

Telemetrical communication with all implanted devices remained unrestricted throughout the testing period of 5 to 6 months. The system was able to transmit intramuscular EMG data in all dogs and one sheep throughout the evaluation period. In the remaining two sheep, the implant dislocated shortly after surgery and was thus not able to transmit EMG activity, however, with unrestricted telemetric functionality. No major adverse reactions were observed in any dogs or sheep after implantation of the system, and no animals were excluded from the study.

All six implants tested in dogs maintained 32 fully functional channels with low impedances (< 5 kΩ) and had continued ability to transfer EMG signals from contracting muscles. The system was explanted from five animals after 6 months, but it remained implanted in one animal for continued testing. The remaining system continues to be fully functional after being implanted for more than 2 years.

Regular long-term evaluations were also performed in one sheep, in which the system was consistently able to transmit EMG data from all channels over the entire implantation period of 23 weeks. EMG signals from the measurements in the sheep demonstrated low crosstalk between channels of agonist and antagonist muscles (Fig. 5A). Implant impedance during the study period showed an initial increase after implantation and a plateau at values below 1 kΩ (Fig. 5B). Histologic samples of implanted muscles after electrode explantation demonstrated a thin implant capsule and unharmed muscular architecture (Fig. 5C). The implant dislocated shortly after surgery in two sheep. This was due to failure of the central unit fixation on the underlying fasica. These implants remained functional, except for a single broken lead in one of the systems, which resulted from dislocation. Overall, this was the only lead failure in our large-animal studies.

Fig. 5.

A-C This figure shows functional and histologic results of chronic implantation of the MIRA system (at 23 weeks). (A) EMG activity from all 32 channels is shown. In the EMG sample on the right, muscle activation phases are highlighted, displaying the selective acquisition of agonistic and antagonistic muscle signals without crosstalk, as well as low noise during inactivity. The implanted muscles for each set of eight intramuscular electrodes are indicated, and the corresponding motor activity of a sheep (elbow in extension versus elbow in flexion) is shown. (B) The mean impedance across all channels for the duration of implantation, with standard deviation bars for each measurement is shown. (C) A histologic slide from a muscle sample taken from the triceps during device explantation is shown. Masson trichrome staining reveals a thin implant capsule around an intramuscular lead (in blue), surrounded by otherwise undamaged muscular architecture.

The analysis of the representative sample recordings from dog and sheep implants showed an average signal-to-noise ratio across all channels of 22.2 dB ± 6.2 dB, which is above the accepted thresholds for good signal quality and indicates high-quality detection of myoelectric signals [16, 20, 38]. The implanted electrodes in the sheep exhibited a higher signal-to-noise ratio (25.1 ± 6.6 dB [95% CI 22.8 to 27.4]) than did those in dogs (19.3 ± 4.2 dB [95% CI 17.9 to 20.8]; p < 0.001).

Feasibility of Implantation

We found it was possible to implant the MIRA at all major amputation levels of the upper limb in a human cadaver; specifically, it was successfully implanted at the transradial, transhumeral, and glenohumeral levels. In an interdisciplinary setting, the following locations for implantation were determined to reach relevant target muscles, provide a stable environment for the central unit, and facilitate unhindered prosthetic fitting.

In transradial amputation, placement of the central unit over the muscle belly of the flexor carpi radialis muscle provided a mechanically stable environment, which is also feasible in very short stumps. From this position, leads were able to reach four muscles from the flexor compartment and, via tunneling through the interosseous membrane, four of the extensor compartments (Fig. 6A-B). This enables up to eight individual myosignals that can be used for direct myoelectric control or that can be combined with clinically available control algorithms such as pattern recognition [34].

Fig. 6.

The MIRA was implanted at different levels of major upper limb amputation. (A-B) Implant placement during transradial amputation, as tested on the left forearm of a fresh cadaver is shown. In (A), the positioning of the central implant over the belly of the flexor carpi radialis muscle is demonstrated; (B) the leads can reach various muscles of the flexor compartment. Four leads were tunneled through the interosseous membrane and inserted into the superficial and deep extensor muscles. (C) The implant configuration during glenohumeral amputation, as determined during test surgery is shown. Lead lengths can be adapted to fit individual anatomic conditions. Placement was tailored to accommodate potential target muscles after targeted muscle reinnervation: Leads 1 to 3 are placed in one head of the major pectoral muscle and Lead 4 in the minor pectoral muscle. Leads 5 and 6 are positioned in the latissimus dorsi, Lead 7 in the teres major, and Lead 8 in the deltoid muscle. Published with permission from Aron Cserveny. A color image accompanies the online version of this article.

At the transhumeral level, the central unit of the implant was placed at the lateral intermuscular septum, since medial placement may cause discomfort to the patient during full arm adduction, considering the need for an external transceiver disc aligned over the implant. In that position, all muscles used for prosthetic control with or without targeted muscle reinnervation were reached; that is, the two heads of the biceps, the brachialis, brachioradialis, and all heads of the triceps muscle.

At the glenohumeral level, the central unit was placed in the infraclavicular fossa, from there the leads were inserted into each of the three heads of the major pectoral, the minor pectoral, and the deltoid. Reaching the teres major and the latissimus dorsi muscles at the same time was not feasible using the standard implant but would require an adapted version with longer leads (Fig. 6C).

Discussion

Currently available solutions in upper limb prosthetics are limited by information transfer rates of neuromuscular interfacing. Implantable approaches with high channel count and stable signal acquisition are necessary to improve functionality of prosthetic devices. With this study, we set out to present and evaluate the MIRA, a novel, fully implantable neuromuscular implant with 32 intramuscular channels. Conducted in a stepwise fashion, small- and large-animal studies and test implantations in a human cadaver were performed to assess biocompatibility, functionality, and feasibility of human use. We found the EMG channels of the MIRA to be fully biocompatible and long-term in vivo testing revealed continued functionality and mechanical stability of the implant. EMG signal quality was above accepted thresholds for prosthetic control and demonstrated low crosstalk between channels. It was possible to implant the system at all levels of major upper limb amputation, while providing biomechanical stability and access to relevant target muscles.

Limitations

This study has limitations. Our study’s limitations include the relatively low sample sizes in the small- (n = 10) and large-animal (n = 9) studies. Although such group sizes are commonly used in the evaluation of neuromuscular implants [1], this may limit the validity of our findings. Also, the use of a mixed large-animal model may increase confounding factors and reduce comparability of results. However, for the purpose of functional long-term evaluation, it was deemed favorable to test the implant in different animals with different biomechanical forces during gait. The use of single-sex animals in each study may result in sex-related bias of our results. However, regarding foreign body reaction and neuromuscular interfacing, no sex-specific differences are expected which would be of interest to explore in an animal model. Regarding the use of a single cadaver, a male cadaver was chosen since the musculoskeletal system of males is on average larger than females and transfer of our results to a smaller-sized anatomy would not be a limitation regarding length of the leads. Still, the use of a single cadaver may limit the validity of our results.

As seen in the cadaver study results, the standard implant design presented may not be able to reach the dorsal muscles of the shoulder girdle in glenohumeral amputees, due to limited lead lengths. Also, the use of fixed leads without connectors necessitates an open surgical approach, which increases the wound size and risk for postoperative complications such as seroma (such as, the buildup of serous fluid in the surgical cavity), as we saw in the sheep study. This may have been a factor leading to the dislocation of two implants in sheep. However, in humans, compared with sheep, postoperative immobilization and tight bandaging of the residual limb can be considered an effective means to prevent seroma and early implant dislocation. Furthermore, we deliberately decided to not include connectors into the implant design, to limit the likelihood for wire breakage and other mechanical complications.

Biocompatibility

The EMG leads of the MIRA demonstrated good biocompatibility in a rabbit model, as assessed in comparison to a known, biologically inert control substance. The overall histopathological results after 3 month of implantation exhibited well-known features of foreign body reaction, including the development of a fibrous capsule around the implant and presence of inflammatory cells such as macrophages and multinucleated giant cells [1]. Implantation of the EMG leads resulted in no detectable damage to the surrounding muscle tissue and yielded an overall low-grade foreign body response. Compared with the control group, scores for tissue response and inflammation as well as thickness of the implant capsule were not increased. Given the long experience with implantation of the materials used in the tested leads, that is, silicone, stainless steel, and polypropylene, these results are in agreement with other reports and support the suitability for long-term application [19, 32, 43]. The presence of a foreign body reaction has shown to increase tissue impedance, which leads to a decreased amplitude of detected signals [13]. In particular, when directly interfacing peripheral nerves, this effect can prevent the detection and discrimination of meaningful biodata and ultimately lead to device failure [17]. Muscular interfacing, on the other hand, has the advantage of providing much larger compound potentials, which has been demonstrated in successful long-term clinical use of EMG-based systems [29, 36].

Functionality

The system retained its ability to transfer EMG signals from contracting muscles for the full duration of the study (5 to 6 months) in all the dogs and one sheep we studied, including one animal in which we have retained the device for more than 2 years. Impedance of the electrode channels remained stable and low throughout the study period, facilitating the detection of reliable signals with high amplitudes. The implant became dislocated early after surgery in two sheep. As discussed above, we believe this to be a specific complication of the sheep model, which can be effectively prevented in human use. Analysis of the EMG signal quality revealed little crosstalk between agonist and antagonist channels and high signal-to-noise ratio. The amount of signal crosstalk is a major factor influencing the control of myoelectric prostheses. As has been reported previously, the increase of muscle signals after targeted muscle reinnervation may paradoxically lead to a decrease in functional outcome scores when using surface EMG-based control [36]. This is owed to undesired detection of EMG activity of adjacent muscles (that is, crosstalk), which results in faulty activation of movement commands, hindering efficient prosthetic control. The implantation of EMG electrodes has been shown to decrease this effect [20, 27], which is confirmed by our current study. Furthermore, signal-to-noise ratios of EMG measurements in dogs and sheep were roughly between 19 dB and 25 dB, which is comparable to other systems using implanted EMG acquisition and indicates a more effective use of muscle signals than through surface EMG [20].

Feasibility of Implantation

We found that it was possible to implant the MIRA at all levels of major upper limb amputation in a human cadaver (transradial, transhumeral, and glenohumeral levels). Regarding placement of the implant’s central unit, an environment of biomechanical stability was chosen to limit relative motion between skin and prosthesis, which is necessary for stable telemetric connection between the central unit and the externally aligned transceiver disc. For both the glenohumeral and transhumeral levels, locations were chosen where there is no direct muscular fixation to prevent implant movement during contraction. At the transradial level, placement on the proximal muscle belly of the flexor carpi radialis was chosen because there would be little relative muscle excursion during contraction. Another factor we considered was that inclusion of the external transceiver disc in the final prosthetic socket should not cause irritation to the patient. For this reason, we avoided bony areas such as the ulna shaft at the forearm, which would increase the risk for pressure ulcers. Also, we avoided the medial intermuscular septum of the upper arm, since a bulky socket including the transceiver would limit full adduction in the shoulder. From the determined positions for the central unit, it was possible to reach all target muscles that are used for prosthetic control (and targeted muscle reinnervation) in each level, except for the latissimus dorsi and teres major at the glenohumeral level (with placement of the central unit in the infraclavicular fossa). If these muscles are to be included in the prosthetic control algorithm of a patient, a custom implant with longer leads may be necessary. In contrast to neuromuscular implants for prosthetic control which have reached clinical application [28, 31], the MIRA also permits use in short stumps as well as in glenohumeral amputation. Due to its flat coil telemetry, anatomic positioning of the implant is flexible, while the other systems rely on a diaphysis with sufficient length for placement of an external coil or implantation of a titan screw.

Several implantable myoelectric systems for prosthetic control are currently in development, including the MIRA, an osseointegrated human-machine gateway, and the Implantable Myoelectric Sensors system (Table 1). As has been shown with the Implantable Myoelectric Sensors system, implantation of only five to six myoelectric electrodes combined with targeted muscle reinnervation leads to better results across all functional evaluations when compared with surface EMG-based control [36]. Furthermore, implantable muscle electrodes are less susceptible to signal alterations because of postural changes, and they eliminate varying electrode positions after donning and doffing [28, 42]. Because the signals are registered at their biological source, there is little noise, and detection thresholds are thus much lower. As stated above, the signal-to-noise ratio values observed in the current study are of similar magnitude as has been reported for the osseointegrated human-machine gateway [20]. This translates into improved signal stability and allows for the detection of very small levels of muscle activity. Practically, this results in improved proportional control because differences in muscle activity can be easily registered, as well as overall less strenuous actuation for the patient. Moreover, if electrodes are implanted during targeted muscle reinnervation, the first signals of reinnervated muscles can be registered earlier than by surface electrodes, which markedly accelerates the rehabilitation process [36]. Furthermore, the MIRA may be employed at any amputation level. This is particularly relevant for patients with glenohumeral amputation or short transhumeral/transradial residual limbs because these patients are not eligible for other systems presented to date. Additionally, the high number of individual electrode sites has several implications for advanced control algorithms. As has been shown in recent studies, nerve transfers lead to changes in muscle architecture and function [4, 14]. When a nerve is transferred to a muscle different from its physiologic innervation, the muscle fiber’s composition and the number of individual motor units adapt to the nerve’s properties, which facilitates hyper-reinnervation of the target muscle. Current solutions such as conventional surface electrodes and the previously reported implantable systems for EMG-based control are limited to a maximum of six individual electrodes in patients who undergo above-elbow amputation, which corresponds to 3° of freedom using direct control [5, 28, 36, 37]. Our group has shown that multichannel EMG recorded from reinnervated muscles can be decoded to decipher the underlying neural drive from the spinal cord [6, 9, 10, 14]. These algorithms have also been applied to intramuscular electrodes in an acute setting [25]. Intramuscular recordings are more selective than surface signals; therefore, a substantially smaller number of detection sites is needed for decomposition of the signals into individual motor neuron activity. Intramuscular EMG signals recorded from four to eight detection sites in a muscle have been decoded in the activity of a large number of motor units (for example, the intramuscular electrodes developed by Farina et al. [11]). Therefore, the proposed implantable system of 32 channels may provide a means to decode spinal motor neuron activity from four to eight muscles to establish a direct neural interface with the output circuitries of the spinal cord, thus allowing access to an unprecedented amount and depth of information in prosthetic interfacing.

Table 1.

Overview and characteristics of implantable interfaces for myoelectric prosthesis control

Parameter	MIRA	OHMG [20, 28]	IMES [31, 36, 42]
Type of muscle interface	Intramuscular	Epimysial	Intramuscular
Number of electrodes	32 (monopolar)	Six (two bipolar, four monopolar)	Five to six (bipolar)
Data gathered	Filtered raw EMG	Amplified and filtered EMG	Rectified and averaged EMG
Sampling rate	1000 Hz-2000 Hz	500-1000 Hz	1000 Hz
Data communication and power  transfer	Telemetry (flat coil)	Wired (osseointegration port)	Telemetry (circumferential coil)
Anatomic requirements	Flexible	Diaphysis with sufficient length for an intramedullary implant	Sufficient residual limb length for external coil placement
Compatible with osseointegration	Yes	Yes	No

OHMG = osseointegrated human-machine gateway; IMES = implantable myoelectric sensors.

Conclusion

Our study has confirmed biocompatibility of the MIRA implant as well as its ability to detect high-quality intramuscular EMG in an animal model. As demonstrated in a cadaver, implantation is possible at all levels of major upper limb amputation. In clinical application, patients are expected to benefit from the large channel count of the system with high signal quality, which will translate into more natural and efficient prosthetic control. Future studies in humans should initially aim to evaluate the MIRA in patients with above-elbow amputations, where the clinical need for improved interfacing strategies is highest. Targeted muscle reinnervation should be performed in these patients to provide more intuitive signals for transmission. Further, the combination of the 32 intramuscular channels with decoding algorithms should be investigated in future research, to assess the possibility to detect spinal motor neuron activity for real-time prosthetic control. Finally, future development of the MIRA should aim to include efferent signal transmission to provide sensory feedback to the patient, such as through cuff nerve electrodes.