“Long-term stability of stimulating spiral nerve cuff electrodes on human peripheral nerves”

Fourteen human volunteers were implanted with spiral nerve cuff electrodes for use in motor or sensory neuroprostheses at Case Western Reserve University (CWRU) since 2005 (Table 1). All spiral nerve cuff electrodes were fabricated with a single lead according to the same design standards, and constructed from identical materials. At time of implantation, all surgeons measured the diameter of the target nerves intraoperatively and selected spiral cuffs of a corresponding size (2–10 mm in diameter) from an inventory of sterile devices.

Thirteen of these volunteers were paralyzed by SCI and received upper or lower extremity motor system neuroprostheses. Subjects 01, 03, and 05 have reaching and grasping neuroprostheses; Subjects 02, 04, 07, 08, 13, and 14 have standing neuroprostheses; Subject 06 has a stepping neuroprosthesis; and Subjects 09 and 10 have standing and stepping neuroprostheses. Subject 11 has an upper limb amputation and uses the system for somatosensory feedback of interactions between his end effector and the environment. The performance of the spiral cuffs in Subject 12 is unobservable due to explantation of the pulse generator. Specific details of the electrode placement have been described elsewhere [6, 7, 11, 30, 33‐35]. The Institutional Review Boards (IRB) of the MetroHealth Medical Center and/or the Louis Stokes Cleveland Department of Veterans Affairs Medical Center approved each study and the subjects provided consent prior to participation. All studies were conducted under a Food and Drug Administration (FDA) Investigational Device Exemption.

Each cuff electrode was connected to an implanted stimulator–telemeter (IST) developed at Case Western Reserve University [36, 37]. An IST is capable of measuring sensor signals, telemetering the data for signal processing, and receiving power and control information from an external control unit. For the lower extremity motor system users, the control unit communicates with the IST via inductive radiofrequency coils. The user places an external coil on his or her skin directly over the inductive coil that sits within the implanted stimulator. Once the coil is placed, they can send stimulation commands by pressing pre-programmed buttons on the external control unit. For the upper extremity motor system users, EMG signals from implanted electrodes are telemetered out of the body through the coil. The external control unit processes these control signals and appropriate stimulation commands are sent back through the coil to the IST.

The IST can deliver constant-current, charge-balanced biphasic stimulus pulses with pulse widths between 1 and 255 μs and pulse amplitudes between 0.1–20 mA. The titanium case of the IST serves as the common return electrode. For cuffs with multiple embedded contacts, each contact was separately connected to its own independent stimulation channel. For monopolar cuff electrodes where all four contacts were connected in series, only one variable stimulation channel was used. Stimulating current was generally limited to 2.1 mA or less, and the pulse widths were adjusted to maintain safe charge density limits [38]. The stimulation frequency used for the upper extremity motor subjects was 12.5 Hz. The stimulation frequency used when calculating charge threshold for Subject 11 was 20 Hz for the first 2.2 years and 100 Hz for the remaining data sets. The frequency changed in order to match the value typically used in functional testing. A frequency of 20 Hz was used with the remaining participants. Pulse amplitude was varied between 0.3-18 mA (Table 2).

Moments about the knee and ankle joints were measured with a 6-DOF load cell (JR3, Woodland, CA, USA) on a Biodex dynamometer. Knee extension moment was measured in response to varying the stimulation delivered by spiral cuffs on the femoral nerve for six volunteers with low tetraplegia and paraplegia (Subjects 07–10, 13, 14). The knee joint was fixed at 20° of flexion and its center of rotation was aligned with one axis of the load cell. For Subjects 09 and 10, ankle moments were measured in response to varying the stimulation delivered via tibial and fibular spiral cuffs. The ankle was fixed at 0° with the load cell centered about the knee joint. The resulting moment data were sampled at 150 Hz, low-pass filtered at 31.25 Hz, and normalized to body weight.

Pulse width-modulated moment recruitment curves were generated for every independent contact. Joint moment was measured as pulse widths were varied between 0 and 255 μs. For each participant, the lowest pulse amplitude that could achieve full recruitment of the targeted motor population was selected. Pulse width and amplitude were limited such that the charge stayed within safe limits [38].

The primary outcome measures utilized to assess chronic spiral nerve cuff performance were recruitment curves, charge threshold, and percent overlap of the recruited motor unit populations (Fig. 2). When collecting recruitment curves, pulse trains were used to generate “tetanic” moments about the knee and ankle joints. We present the recruitment curves generated in one session that occurred 4.5 years after implantation for Subjects 09 and 10, and 2 years after implantation for Subjects 13 and 14. Femoral nerve recruitment curves are presented for Subjects 09, 10, 13, and 14, and tibial and fibular nerve recruitment curves are presented for just Subjects 09 and 10.

Charge threshold was determined in three ways due to the subjects’ original participation in different studies. For Subject 01, charge threshold was defined as the minimum charge needed to evoke a visible muscle contraction of the upper extremity musculature. Since the number of channels on the implanted stimulator was limited, one radial nerve cuff channel and two musculocutaneous cuff channels were not connected and those thresholds were not measured. Charge thresholds for Subject 01 were repeatedly measured in eight sessions spanning 10.4 years. To determine charge thresholds for Subjects 07 and 08, Gompertz models were fit to the recruitment curves [35]. We set the charge threshold as the charge (pulse amplitude * pulse width) required to obtain 10% of the maximum moment produced, as estimated by the Gompertz model (Fig. 3). Over a period of 5+ years after implantation, we conducted ten sessions for measuring charge threshold in Subject 07 and 11 sessions for Subject 08. For Subject 11, charge threshold was determined by performing an unbiased adaptive procedure called “single-interval adjustment matrix” (SIAM) [30, 39, 40]. The subject was asked whether or not he felt a stimulation-induced sensation, and the stimulation parameters were adjusted based on his response. Stimulation parameters were set for a target performance of 50% and true stimulation was provided 50% of the time. The threshold search continued until there were approximately 12–16 “reversals” between when sensation was or was not perceived. Charge thresholds for Subject 11 were repeatedly measured in 13 sessions spanning 4.6 years. For each electrode contact in the four subjects, we evaluated the relationship between time and charge threshold by generating a Pearson’s correlation coefficient using MATLAB’s “corrcoef” function.

To assess whether multi-contact cuffs remain selective for independent populations of motor units for multiple years post-implantation, we measured the percent overlap of the recruited motor unit populations [35]. We first measured the tetanic moment produced by each individual contact (M_i, M_j…). We then measured the tetanic moment when two contacts were stimulated within 2 ms of each other (M _i ∪ j ), while the units recruited by the first contact were still in their refractory periods when the second pulse was delivered. If two contacts were completely independent and did not have any overlap in the motor unit populations they activated, their individual moments added linearly when the contacts were stimulated 2 ms apart. If there was some overlap, their combined moment was less than the sum of their individual moments because some of the motor units were still in their refractory period and would not respond to additional stimulation. Percent overlap (O_i,j) was defined as:

The lower the percent overlap (O_i,j), the more selective the spiral cuff was in recruiting independent motor unit populations. For this analysis, we focused on Subject 07’s right femoral cuff during one session that occurred 6.3 years after implantation, and Subject 08’s left femoral cuff during a session that occurred after 4.2 years. Only two cuffs were tested due to time constraints. The pulse widths for each channel were chosen so that they generated approximately equal moments. For Subject 07, pulse amplitude was set to 1.4 mA and the pulse widths were 82 μs for Channel 1, 100 μs for Channel 2, 95 μs for Channel 3, and 143 μs for Channel 4. For Subject 08, amplitude was set to 0.8 mA and the pulse widths were 114 μs for Channel 1, 89 μs for Channel 2, and 73 μs for Channel 4.

The 13 SCI participants in this study varied by injury level (C3-T11), nerve cuff location (various upper and lower extremity nerves), and the number of independent stimulation contacts (one to four) per cuff (Table 1). One individual with trans-radial limb loss was included and implanted with a four-contact spiral cuff on his radial nerve. Across all 14 recipients, the aggregate experience amounts to more than 349 cuff-years. The performances of the spiral cuffs implanted in Subjects 01, 07–11, 13, and 14 were verified within the last year (referred to as the eight “active participants” in ongoing research studies). Subject 03 was last tested in 2013; Subjects 02, 04, and 05 were last tested in 2014; and Subject 06 was last tested in 2015. The implanted stimulator was explanted from Subject 12 in 2013 due to a perioperative infection that did not involve the spiral nerve cuffs, and the contacts can no longer be tested. Out of the eight active participants, there are 27 cuffs that have a total of 45 independent contacts. One contact within a multi-contact spiral cuff in Subject 08 does not evoke a motor or sensory response and therefore may not be near excitable nerve fibers. Forty-four out of 45 contacts (98%) were reported as operational and evoking a motor or sensory response in the active research participants.

The six remaining participants no longer participate in research studies. They were implanted with 23 spiral cuffs with a total of 56 independent contacts. One monopolar spiral cuff in Subject 03 pulled off of the right thoracodorsal nerve. This could be due to reports of the subject hooking his right arm around the handles of his wheelchair; this could have applied tension to the lead, since a lack of redundant lead length for strain relief was noted during surgery. As previously mentioned, the stimulator in Subject 12 was explanted after 3 months; all electrode contacts were tested and operational at that time. His nerve cuff electrodes remain intact and have caused no clinical issues, although they are inaccessible to stimulation. No other adverse events related to the operation of the spiral cuff electrodes have been reported.

Within the past 6 months, pulse width-modulated recruitment curves were generated for four subjects with monopolar spiral nerve cuff electrodes on their femoral nerves (Fig. 4). The knee extension moments produced by all eight cuffs exceeded the 0.135 Nm/kg required to keep the knees locked during standing [41], demonstrating that the devices remain functional for at least 2–4.5 years after implantation. Figure 5 shows the most recent recruitment curves gathered for two subjects (09 & 10) with monopolar nerve cuff electrodes implanted on the tibial and fibular nerves below the knee. All four monopolar fibular nerve cuffs were able to produce a dorsiflexion moment greater than 0.01 Nm/kg, the level needed to prevent foot drop [42]. Plantarflexion moment during push-off at a normal walking speed is 1.6 Nm/kg [43], but the moments produced by the tibial nerve cuffs did not reach this value. When stimulating Subject 10’s left fibular and right tibial nerve cuffs, knee flexion and reflexive responses also occurred, which could influence the moment measured about the ankle. Subject 10 did not regularly use her system, so it is possible that her muscles deconditioned.

Charge thresholds remained stable in the majority of contacts within the 11 spiral nerve cuff electrodes in the four subjects who received the implanted devices between 5 and 11 years ago (Fig. 6). The acute stabilization in the months directly after implantation is discussed in more detail elsewhere for these participants [4, 29, 33, 39]. Subject 01 has two multi-contact cuffs and four monopolar cuffs; Subjects 07 and 08 have multi-contact spiral nerve cuff electrodes on their bilateral femoral nerves; Subject 11 has one multi-contact cuff on his radial nerve. Two contacts in Subject 08 are not included because they did not elicit discernable motor responses. Of the 27 contacts, only eight had a significant increase in charge threshold over time (linear regression, p < 0.05). The charge threshold of the second contact on Subject 01’s multi-contact radial nerve cuff (Rad. Ch2) increased by 370.9nC between the first and last sessions. Rad. Ch3 increased by 120.3nC, Musc. Ch2 increased by 14.4nC, and the suprascapular cuff increased by 36.2nC. For Subject 07, the fourth contact in the left femoral cuff (L4) increased by 175nC, R1 increased by 40.6nC, and R3 increased by 67.2nC. The first contact on Subject 08’s right femoral cuff (R1) increased by 50.4nC. These gradual increases over time or increases during the two most recent sessions likely indicate a permanent change in charge threshold. Even with these increases in stimulation parameters, the stimulation levels were within safe limits. The non-significant fluctuations in charge threshold over time, for example the increase in Subject 08’s R3 threshold at the 5.2-year session and decrease at 5.3 years, could be a result of leg posture during measurement. Even slight differences in the knee angle could affect the measured moment.

Figure 7 shows each contact’s average change in charge threshold per year. The majority of contacts show excellent stability with average changes in threshold close to 0nC/year, with 20 out of the 27 contacts (74%) exhibiting less than a 5nC average increase in charge threshold per year.

Multi-contact cuffs on the femoral nerves of Subject 07 and 08 were still selective for different motor populations after being implanted for several years (Table 3, Fig. 8). Stimulating currents were applied to contacts individually and in pairs within 2 ms of each other. The percent overlap (O_i,j) ranged from 5.5% to 47.4%, with an average overlap of 20.2%. The contacts on opposite sides of the nerve (1&3, 2&4) had 14.6% overlap on average. The contacts on adjacent sides of the nerve (1&2, 1&4, 2&3, 3&4) had 23% overlap on average, which was expected due to their closer proximity. Contacts 1 and 4 in Subject 08 were on adjacent sides of the femoral nerve yet they had the least amount of overlap (5.5%). While there was some overlap in the recruited motor populations after 4.2 years post-implantation, the resulting moment was 0.134 Nm/kg, which is approximately equal to the 0.135 Nm/kg needed to stand. It is possible that these contacts had the least amount of overlap because neither contact generated a strong enough motor response to spill over to another motor unit population. The two contacts with the second smallest percent overlap were Contacts 1 and 3 on opposite sides of Subject 07’s right femoral nerve cuff. When the two contacts were stimulated within 2 ms, the resulting moment was 0.33 Nm/kg, which easily exceeds the 0.135 Nm/kg needed to stand. In 2013, Fisher et al. performed overlap tests with these same two participants [35]. In that study, our Subject 07 is their Subject 1 and our Subject 08 is their Subject 2. There are four pairs of contacts that are presented in both this and the Fisher et al. study. For Subject 07, they measured the overlap of one contact pair, Right 2&4, during four sessions over 27–53 weeks after implantation. After 6.3 years post-implantation, the measured overlap was greater than the original standard deviation range. The percent overlap we measured was 26.8%, which is close to the 25% overlap needed to best prevent knee buckling [44]. For Subject 08, they measured the overlap for three contact pairs during three sessions over 14–37 weeks after implantation. After 4.2 years, two out of those three pairs still fell within the standard deviation range. The pair that fell outside of this range, Left 1&2, had 11.3% overlap. Percent overlap is below 25%, so this contact pair is still highly functional in preventing knee buckling. These results demonstrate that motor unit selectivity is maintained for up to 6.3 years after implantation.