bs_bs_banner
Copyright © 2014 International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc.
Subchronic Stimulation Performance of Transverse Intrafascicular Multichannel Electrodes in the Median Nerve of the Göttingen Minipig *Kristian R. Harreby, *Aritra Kundu, †Ken Yoshida, ‡Tim Boretius, ‡Thomas Stieglitz, and *Winnie Jensen *Department of Health Science and Technology, Aalborg University, Aalborg, Denmark; †Department of Biomedical Engineering, Indiana University–Purdue University Indianapolis, Indianapolis, IN, USA; and ‡Department of Microsystems Engineering (IMTEK), University of Freiburg, Freiburg, Germany
Abstract: This work evaluated the subchronic stimulation performance of an intraneural multichannel electrode (transverse intrafascicular multichannel electrode, TIME) in a large human-sized nerve. One or two TIMEs were implanted in the right median nerve above the elbow joint in four pigs for a period of 32 to 37 days (six TIMEs in total). The ability of the contact sites to recruit five muscles in the forelimb was assessed via their evoked electromyographic responses. Based on these responses, a selectivity index was defined. Four TIMEs were able to selectively recruit a subset of muscles throughout the implantation period. The required recruitment current sig-
nificantly increased, while there was a tendency for the recruitment selectivity to decrease over time. Histological assessment showed that all TIMEs remained inside the nerve and that they were located between fascicles. The average thickness of the encapsulation of the electrode was estimated to be 115.4 ± 51.5 μm (mean ± SD). This study demonstrates the feasibility of keeping the TIME electrodes fixed and functional inside a large polyfascicular human-sized nerve in a subchronic setting. Key Words: Peripheral intraneural interface—Selective stimulation— Interneural stimulation—Neural prosthetics.
Advanced prosthetic limbs with sensing capabilities and multiple degrees of freedom have been developed for amputee subjects (1). However, currently, most advanced prosthetic devices used are sequentially controlled via information extracted from surface electromyographic recordings (EMGs) (1,2). These control schemes fail to provide intuitive control; rather, they typically only provide one degree of freedom at a time, and the user must rely on visual or audio feedback (3,4). A promising method for controlling a prosthetic arm is targeted muscle reinnervation (2,5). This method involves the surgical redirection of nerves in the stump to muscles and skin
in the chest region. Volitional motor commands directed to the stump will then project to the reinnervated muscles, which, via EMG, are mapped to appropriate control commands on the prosthetic arm. In addition, as the sensory nerve fibers of the stump reinnervate the skin of the chest, this skin area can potentially be used for providing sensory input from a prosthetic device (5). Although promising, the method of nerve reinnervation is highly invasive and may primarily be attractive to above-elbow amputees, in whom the loss of residual nerve function is limited (6). An alternative method for interfacing a prosthetic device would be to interface the peripheral nerves directly. The feasibility of this approach was shown by, for example, Dhillon et al. in longtime amputees, who demonstrated that sensations related to the phantom limb could be evoked through selective microstimulation of the nerve stump. In addition, they were able to obtain control signals directly from the nerve to control a prosthetic limb (7–9).
doi:10.1111/aor.12347 Received February 2014; revised April 2014. Address correspondence and reprint requests to Dr. Kristian R. Harreby, Kristian R. Harreby, Aalborg University, Frederik Bajers Vej 7D, 9220 Aalborg, Denmark. E-mail: [email protected] .dk Artificial Organs 2015, 39(2):E36–E48
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES To control a prosthetic hand with multiple degrees of freedom or to provide sensory feedback from artificial limbs in general, a peripheral neural interface should be able to selectively interface with multiple subsets of nerve fibers throughout different fascicles. In general, the more active sites a neural interface has and the closer they are placed to different populations of nerve fibers, the more nerve fibers can be selectively addressed. However, as the active sites are placed in the direct vicinity of nerve fibers, the degree of invasiveness increases and the electrodes tend to pose a greater risk of damaging the nerve (6). The TIME project (CP-FP-INFSO 224012, http:// www.project-time.eu) aims to develop a novel neural prosthesis system to apply multichannel microstimulation to the nerve stump of a volunteer amputee to evoke phantom limb sensations and to explore if sensory feedback can be used as an alternative treatment of phantom limb pain. Within this framework, our group designed the transverse intrafascicular multichannel electrode (TIME). The TIME is a nextgeneration intraneural electrode built upon the principles of the thin-film longitudinal intrafascicular electrode (tfLIFE) (10). As our research work was included in a translational research program, that is, we envisioned a human clinical trial with a duration of 30 days at its very end, design and development included assessment and management of possible risks according to ISO 14971 (Application of Risk Management to Medical Devices). The tfLIFE and TIME were designed to place multiple active sites (>12) inside the nerve while keeping the risk of nerve damage low. Both electrodes are approximately of the same physical size as the electrodes used by Horch et al. (10,11), but instead of having one or two active sites on each electrode, they contain 12 or more active sites distributed along both sides of a thin, flexible polymer substrate. The tfLIFE is implanted in parallel to the nerve fibers, and thus all active sites will redundantly interface with the same subpopulation of nerve fibers, whereas the TIME is implanted transversely through the nerve, thus interfacing with different fiber populations within the nerve over its cross-section (10,11). The use of TIMEs as neural interfaces in humans in the future will depend on the preclinical evaluation of the stability, functionality, and biocompatibility of the electrodes. Recently, the selectivity performances of tfLIFE and TIME were compared during acute experiments in the sciatic nerve of the rat (12) and later in the median nerve of the farm pig (13). These studies both showed that the TIME was able to recruit a higher number of muscles with a higher degree of selectivity than the tfLIFE. Histological
E37
assessment showed that due to differences in nerve morphology, the electrodes had been placed inside the fascicles in the rat model (three fascicles in the sciatic nerve at the implant site), while in the pig model (>30 fascicles in the median nerve at the implant site) the TIMEs had not penetrated the fascicles and thus were located between fascicles. Before the TIME can be applied in subchronic human implants ( 0.4, which corresponded to an average recruitment of nontarget muscles of less than one-fourth of that of the target muscle. Finally, to quantify the selectivity performance of a whole TIME device, SID≥30% was defined as
SI D≥30%
∑ =
NA j =1
SIM ≥30%, j
(5)
NA
Thus, SID≥30% is simply the average of all SIM≥30% for all the monitored muscles. SID≥30% and the number of selectively recruited muscles for the first and last follow-up were compared using the Wilcoxon signed-rank test.
Animal and nerve assessment Animals were euthanized after at least 30 days post-implant. Following the terminal follow-up, the location of each TIME was determined, and the nerve, still containing the implanted electrodes, was removed for postmortem histological analysis. The excised implanted nerve was fixed (frozen by liquid nitrogen or embedded in formalin and later fixed in paraffin), and thin 5-μm transverse sections were made. The TIMEs were then visually located, and the thickness of the encapsulation was defined as the distance from the polyimide structure to the rim of the fibrotic capsule. This distance was estimated for each TIME based on the mean of five measurements (AxioVision v. 4.6.30, Carl Zeiss Imaging Solutions GmbH, Oberkochen, Germany).
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES RESULTS Six TIMEs (referred to as P1T1 [Pig 1 TIME 1], P2T1, P2T2, P3T1, P4T1, and P4T2) were implanted in four animals for 33.8 ± 2.4 days (mean ± SD). During the first follow-up, bipolar EMG recordings were performed on all five muscles in all four pigs. However, EMG wires were vulnerable to breakage both close to the Omnetics connector and at the connection between the MP35N wires and the stainless-steel wires in the legs of the pigs. Due to these problems, the majority of the muscles were eventually monitored via monopolar EMG recordings, and some were omitted. During the last followup, EMG was obtained from all five muscles in P1 (needle electrodes, all bipolar), P2 (needle electrodes, all bipolar), and P3 (three bipolar, two monopolar) and from four muscles in P4 (all monopolar). The connections between the MP35N wires and the Omnetics connectors were functional in all TIMEs, except in two cases where postmortem assessment of the cable resistance indicated one of 12 and two of 12 broken connections in P3T1 and P4T2, respectively. Both of these TIMEs gradually ceased to evoke muscle activation during the implantation period; however, the few broken wires could not fully
E41
explain the overall performance decline. These failures were assigned to behavior of the minipigs that would not be expected in human clinical trials, such as scratching and rubbing against the pen wall. Recruitment Stimulation with increasing intensities evoked increasing muscle responses starting from 1.5–3 ms after stimulation onset and lasting 5–10 ms (see Fig. 2A). Typically, the evoked responses maintained the same shape while increasing; however, additional phases occasionally occurred (see Fig. 2A, M2 F PL). Recruitment curves were in general not smooth sigmoidal shapes but increased irregularly, sometimes with intermediate plateaus (Fig. 2B,C). Sometimes we observed a concurrence between these plateaus and increase in recruitment. In the majority of the cases, all muscles were recruited by each individual stimulation configuration, and recruitment started at similar current levels; however, some differences were seen in the slope of the recruitment curve and the maximal EMGRL reached for the individual muscles, meaning that some muscles were more selectively recruited than others. Monopolar stimulation configurations evoked higher muscle recruitment levels, starting at lower
FIG. 2. Examples of evoked EMG responses (A) and recruitment curves (B,C) obtained from P2T1 at day 7. (A) Evoked EMG obtained while stimulating active site 3′ against the subcutaneous stainless steel needle (ground [G]). Bipolar recordings from M2 F PL and M3 F CR were performed, and monopolar recordings were performed for M1 P T, M4 FDS, and M5 F DP. For clarity, only EMGs for stimulation intensities dividable by 100 μA are shown (100 μA, 200 μA . . . 1200 μA). Time 0 indicates the onset of the 100-μs pulse (vertical, black solid line), and the 3–10 ms time window used for calculating the RMS values is indicated (vertical dotted lines). Increasing stimulation intensity resulted in progressively larger and sometimes more complex evoked EMG responses (see M2 F PL). (B) Recruitment curves from the monopolar stimulation configurations for the three active sites on each side of the TIME structure. The dotted vertical line indicates EMGRL30%. Each active site recruits all five muscles to some extent, although M3 F CR does not reach EMGRL30%. (C) Bipolar TIME configurations start recruitment at higher current levels and do not reach as high an EMGRL as monopolar configurations. Artif Organs, Vol. 39, No. 2, 2015
E42
K.R. HARREBY ET AL.
current intensities, than those observed for bipolar configurations (Fig. 2B,C). During the first followup, 27 of 60 bipolar configurations (10 in six TIMEs) were able to produce EMGRL > EMGRL30%. However, the bipolar configurations gradually became unable to produce muscle recruitment, and at the final follow-up only three of 60 bipolar configurations were able to reach EMGRL30%. The morphology of the muscle recruitment curves produced by stimulation on individual contacts tended to change over 2–4 follow-ups (see Fig. 3). There was a general tendency for the curves to gradually shift to the right, indicating that higher current levels were needed to recruit the muscles. Overall, five of six TIMEs were able to recruit at least one muscle to EMGRL30% at all 12 active sites (monopolar stimulation) during the first follow-up. In P3T1, six of the 12 active sites (1–6) were initially able to recruit muscles to >EMGRL30%, whereas only minor recruitment was achieved at the remaining six active sites (1′–6′). The number of recruited muscles gradually decreased at all contact sites, and following day 15, only one muscle could be activated to >EMGRL30%; after day 20, no recruitment could be observed. P4T2 went from recruiting three different muscles to >EMGRL30% on day 17 to being able to activate only one on day 21, which was the last day any recruitment was observed. Interestingly, postmortem histology actually showed that these two specific TIMEs were surrounded by the thinnest encapsulation of all the TIMEs (54.5 μm and 93.7 μm, respectively, compared with the remaining TIMEs [136.0 ± 42.9 μm]). The poor performance of these electrodes was, however, already seen during the first follow-up, as these electrodes needed significantly higher currents to recruit muscles (ID30% of 745 μA and 666 μA, respectively) compared with the other TIMEs (379 ± 86 μA). The remaining four TIMEs (P1T1, P2T1, P2T2, and P4T1) were able to evoke EMGRL > EMGRL30% at nearly all active sites (typically all 12, and always ≥9) throughout the complete timeframe of implantation. Recruitment current Although fluctuations were seen, there was a general trend for the IM30% to increase throughout the duration of the implantation (see Fig. 4A). The trend was stronger for the ID30%, which consistently and significantly increased in all TIMEs from the first follow-up to the last (P = 0.028, Fig. 4B). During the first follow-up, the mean ID30% across all six electrodes was 488 ± 68 μA (mean ± SEM); this increased to 769 ± 128 μA during the last follow-up. However, if limited to the four TIMEs that were Artif Organs, Vol. 39, No. 2, 2015
functional during the whole period, the ID30% went from an initial 379 ± 42 μA to a final 552 ± 48 μA. Selectivity Neighboring active sites on one side of the device or on different sides of the device (e.g., 1 and 2 or 1 and 1′) recruited similar subsets of muscles (Fig. 2B,C), whereas small differences in muscle recruitment often resulted in variations between the maximal SIM values reached. In P2 and P4 the two
FIG. 3. Examples of the recruitment curves of M3 F CR, M4 F DS, and M5 F DP produced when performing stimulation on active site 3′ in P4T1 throughout the whole duration of the implantation. The follow-up and days of implantation are indicated at the left, and the recruitment scale is indicated at the bottom right. Recruitment curves for M3 F CR , M4 F DS, and M5 D DP are based on bipolar recordings until follow-ups (FUs) 12, 10, and 6, respectively. The recruitment curves increase irregularly as a function of the stimulation current. The morphology of the curves changes gradually over periods of 2–4 follow-ups. A general trend is that the curves tend to shift to the right, indicating that more current is needed to recruit the muscles later in the period of implantation.
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES
FIG. 4. Current for providing 30% muscle activation over the implantation period. (A) The required current for recruiting each muscle (IM30%) and the average required recruitment current of all muscles activated by a TIME device (ID30%) are shown for P4T1. (B) The ID30% values for all six TIMEs over the whole duration of the implantation. The mean and standard error of the mean of the first and last follow-ups (FFU and LFU) are indicated with black dots and bars. Most TIMEs remain well below the safe upper current limit of 1200 μA at 100 μs; however, P3T1 and P4T2 gradually fail to reach EMGRL30% , which is indicated by the current saturating at 1200 μA. Across all electrodes there is a significant (P = 0.028) increase in ID30% from first to last follow-up.
E43
TIMEs implanted in a single median nerve innervated different subsets of muscles (Fig. 5A,B). The SIM≥30% values of individual muscles could both increase and decrease over time, and no clear trend was observed at the level of individual muscles; however, typically a decreasing trend could be observed for SID≥30% (Fig. 6A). Across all six electrodes; there was an overall tendency for SID≥30% to decrease over time (not statistically significant; P = 0.12). The mean SID≥30% across all electrodes changed from 0.25 ± 0.04 to 0.14 ± 0.05 from the first follow-up to the last (Fig. 6B). Two electrodes (P2T2 and P4T2) were able to recruit three muscles selectively during an early follow-up, whereas typical TIMEs recruited between zero and two muscles selectively (Fig. 6C). The initial mean was 1.17 ± 0.37, which decreased to 0.67 ± 0.38 at the final follow-up, whereas this latter decreasing trend did not reach significance (P = 0.18). During the initial follow-up one-third of the muscles were most selectively activated by bipolar stimulation configurations. However, the SID≥30% increased only by 0.005 when compared with pure monopolar stimulation configurations. Animal and nerve assessment After the initial postoperative swelling had ceased (∼5 days) the animals did not show any discomfort and supported their full body weight on the implanted leg. After 20 days, an infection was observed in P1 at the leadout wires at the back of
FIG. 5. The plots show SIM as a function of EMGRL (based on bipolar needle electrodes) while performing stimulation using a subset of active sites (1–6) in each of the two TIMEs implanted in P2 on day 37. Due to the definition of the SIM, it is prone to large fluctuations at low muscle activity and recruitment, and thus, EMGRL30 days. We found that bipolar stimulation tended to produce less activation than monopolar stimulation, and eventually very few bipolar configurations were able to produce EMGRL > EMGRL30%. Bipolar configurations produce a more focused electrical field than monopolar configurations; this has been shown to improve recruitment selectivity of the TIME in a rat model (25). If insertion methods are developed for placing TIMEs inside fascicles in large polyfascicular nerves, for example via high-velocity insertion (26) or via microdissection of the nerve, bipolar stimulation might help to increase selectivity in this nerve model as well. Similarly to other studies, we assessed the recruitment characteristic of the TIME by measuring the evoked EMG (12,27–29); however, a number of factors should be taken into consideration when using this method. Firstly, repeated activation of muscles, performed at frequencies as low as 2 Hz, may potentiate the evoked muscle force; however,
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES the effect on the measured EMG, which is assessed in the current study, seems limited (30). Secondly, recruitment of sensory nerve fibers may trigger muscle reflexes. To focus on the directly evoked EMGRL, this parameter was calculated based on a time window prior to the occurrence of most evoked reflex activations (3–10 ms after stimulation pulse). Thirdly, cross-talk will occur if a muscle electrode records EMG activity from muscles other than those it was intended for; this will produce a negative bias in the selectivity measures. In our study, the observed coherence between plateaus and the increase in recruitment curves from different muscles (Fig. 2B) could indicate the presence of cross-talk. However, while stimulating single fascicles (following dissection) using a hook electrode in P3 during the terminal follow-up, it was possible to obtain very selective recruitment of single muscles in some cases (e.g., SIM > 0.9 at EMGRL = 80%); in other cases, several muscles were activated by a single fascicle. Obtaining a high selectivity would not have been possible if extensive EMG cross-talk had been present, and therefore we believe that the amount of cross-talk in our study was limited. In some cases monopolar EMG recordings were used, which are more susceptible to cross-talk than bipolar recordings. However, during the last follow-up, bipolar EMG recordings were obtained (with the use of additional needle electrodes) in three of the four functional TIMEs (P1T1, P2T1, P2T2); thus, the overall conclusion of the study will not be affected by this. Recruitment current We found that ID30% consistently increased during the implant period. In the four of six TIMEs that remained functional, the ID30% remained well below the maximal charge capacity of the TIME. The gradual failure to recruit muscles in the other TIMEs may be explained by a gradual formation of connective tissue and by the active sites being located at greater distance from fascicles which innervated the subset of monitored muscles (they had the highest initial ID30%). The required charge for producing muscle activation with the USEA (assessed via palpation) in the chronic cat model was reported to increase during the first month of implant and then plateaued at around 80 μA (bipolar; pulse widths of 200 μs) (23). With the differences in methodology in mind, the stimulation intensity used was similar to that found in the current study (see initial recruitment in Fig. 2B). In contrast, Grill and Mortimer reported that the threshold current for a tripolar cuff decreased after implantation and stabilized at 200–600 μA (pulse
E45
width 100 μs; tripolar stimulation) after 8–12 weeks of implantation (22). This difference in the development of the recruitment current may be a result of the fact that formation of connective tissue will always increase the distance from contact sites to the target nerve fibers (and thus the required current) with interneural electrodes, whereas the formation of connective tissue outside an implanted cuff (31) may actually help confine the induced charge within the cuff. The TIME was recently evaluated in the smaller sciatic nerve of rats (12), where muscles were activated with less than one-tenth of the charge used in the current study. Our histological analysis showed that the TIMEs in our pig model were placed between fascicles, whereas in the rat model they were inserted directly into the fascicles using visual guiding. The larger charge needed in the pig model can be explained by the extrafascicular location of the electrodes in this model, where both the greater distance from the target nerve fibers and the low conductance of the perineurium (32) tended to hinder recruitment. The tfLIFE was evaluated in a human amputee; two tfLIFEs were implanted in the median and ulnar nerve of the arm stump (33). Using maximal charges of less than one-tenth of that required for reaching EMGRL30% in our experiment, it was possible to evoke tactile sensations during the first 10 days of implantation (33). It was not determined whether the tfLIFE was placed inside a fascicle or between fascicles in this case. Selectivity The multicontact cuff, the USEA, and the TIME were shown to recruit muscles with several degrees of freedom when evaluated in smaller nerve models (cat and rat) (12,22,23). When implanted in the sciatic nerve of the rat, the TIME could produce selective recruitment of the three muscles monitored (12). In our case, the SID≥30% changed from 0.25 ± 0.04 in the first follow-up to 0.13 ± 0.05 during the last follow-up (mean ± SEM), and the number of selectively activated muscles per TIME decreased from 1.17 ± 0.37 to 0.66 ± 0.38. It is not surprising that the selectivity parameters tended to decrease over time due to the formation of a fibrotic capsule around the electrode structure. However, the selectivity level and the number of selectively recruited muscles still appear low compared with the performance of the TIME in the rat model. The differences in results are likely due to differences in the nerve morphology and/or the methodologies applied. Because of the established Artif Organs, Vol. 39, No. 2, 2015
E46
K.R. HARREBY ET AL.
somatotopic organization of motor fibers and the small size (diameter 1–1.5 mm, three fascicles) of the rat nerve, the TIME could be inserted using visual guidance into all the fascicles that innervated the three monitored muscles (12,34). In the polyfascicular median nerve of our pig model (diameter 3.5–4 mm, 25–35 fascicles) (14), it was not possible to distinguish individual fascicles during insertion, nor was the somatotopic organization of the fascicles known, and therefore electrodes could only be inserted “blindly” in our study. Microdissection of the nerve during the implantation procedure would have enabled insertion of the TIMEs into individual fascicles using visual guidance. Such an approach, however, was ruled out due to the increased invasiveness and time consumption it would have required. As the pig nerve contains around 30 fascicles, only a small subset of these will be located just next to an implanted TIME (∼7, unpublished findings) and thus can actually be expected to be selectively recruited. Making the simplified assumption that each of the five monitored muscles was fully innervated by one specific fascicle, the probability of selectively activating the monitored muscles using a single TIME can be calculated using binomial coefficients. Even under the ideal assumption that all seven fascicles neighboring a TIME were selectively recruited, the risk of detecting no selectively recruited muscles or only one is 67%, and the chance of having more than two muscles selectively recruited is a mere 7%. Based on this model, the average number of selectively recruited muscles that can be expected is 1.17 per TIME, which is indeed very similar to the number of selectively recruited muscles (1.17 ± 0.98) obtained during the first follow-up (Fig. 6B). Hence, several fascicles might have been selectively recruited, but only the recruitment of the subset of fascicles that innervated the five monitored muscles would have been detected. Therefore, our results do not rule out that a single TIME could be sufficient for inducing several types of distinct sensory input; however, several TIMEs would be needed if a whole polyfascicular nerve needed to be selectively activated. Figure 5 shows how two TIMEs implanted with a 45° difference were able to selectively recruit two different subsets of muscles. To maximize the benefit of using multiple TIMEs, they should be implanted in parallel relative to each other so that they interface non-overlapping cross-sections of the nerve (35). There are two main differences between the commonly used SI(I)V (12,13,21,36) and the new SI(I)M. Firstly, the mean recruitment of nontarget muscles (penalty) is weighted by a constant number, which in our case is related to the total number of EMG chanArtif Organs, Vol. 39, No. 2, 2015
nels. This prevents the SI(I)M from being further positively biased if EMG cannot be obtained from all five muscles. Secondly, if all muscles are activated to the same degree, this does not provide any functionally selective activation; however, the SI(I)V will take a value of 1/NA, where NA is the number of muscles monitored. This induces a positive bias, which makes it difficult to distinguish the bias from actual functional selectivity. Our SI(I)M therefore includes a correction that removes the 1/NA bias and scales the functionally relevant selectivity to the interval of [0–1]. In general, this makes the SI(I)M more conservative than the SI(I)V; for example, the mean SID≥30% obtained during the first follow-up was 0.25, whereas it would have been 0.40 if SI(I)V had been used. Animal and nerve assessment Following a 3–5-day recovery period, the animals showed no discomfort. Visual assessment ruled out only severe nerve damage (motor fibers); however, these observations were supported by the histological analysis, which showed no sign of necrosis or nerve damage. The thickness of the encapsulation was estimated to be 115.4 ± 51.5 μm. In comparison, the thickness of the encapsulation found following chronic implantation of the TIME (37) and tfLIFE (24) in the sciatic nerves of rats was only around 30–60 μm (tfLIFE encapsulation estimated from illustrations in article). When the USEA was implanted in the sciatic nerve of the cat it resulted in 426 μm of connective tissue between the base of the electrode and the nerve; however, inside the nerve the distance between the individual shanks and the nerve fibers was only 30.4 μm (23). The thicker encapsulation seen in the current study might be due to the different species used or to the differences in the duration of implantation (i.e., 1 month in our study and more than 2 months in the cited studies). The encapsulation and fixation of the implants observed postmortem indicated that the approach of anchoring the TIME (leadout wires, ceramic connector, and polyimide loop) transversely in the nerve is possible and safe. CONCLUSIONS Before the transverse intrafascicular multichannel electrodes may be tested in subchronic human implants (
Copyright © 2014 International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc.
Subchronic Stimulation Performance of Transverse Intrafascicular Multichannel Electrodes in the Median Nerve of the Göttingen Minipig *Kristian R. Harreby, *Aritra Kundu, †Ken Yoshida, ‡Tim Boretius, ‡Thomas Stieglitz, and *Winnie Jensen *Department of Health Science and Technology, Aalborg University, Aalborg, Denmark; †Department of Biomedical Engineering, Indiana University–Purdue University Indianapolis, Indianapolis, IN, USA; and ‡Department of Microsystems Engineering (IMTEK), University of Freiburg, Freiburg, Germany
Abstract: This work evaluated the subchronic stimulation performance of an intraneural multichannel electrode (transverse intrafascicular multichannel electrode, TIME) in a large human-sized nerve. One or two TIMEs were implanted in the right median nerve above the elbow joint in four pigs for a period of 32 to 37 days (six TIMEs in total). The ability of the contact sites to recruit five muscles in the forelimb was assessed via their evoked electromyographic responses. Based on these responses, a selectivity index was defined. Four TIMEs were able to selectively recruit a subset of muscles throughout the implantation period. The required recruitment current sig-
nificantly increased, while there was a tendency for the recruitment selectivity to decrease over time. Histological assessment showed that all TIMEs remained inside the nerve and that they were located between fascicles. The average thickness of the encapsulation of the electrode was estimated to be 115.4 ± 51.5 μm (mean ± SD). This study demonstrates the feasibility of keeping the TIME electrodes fixed and functional inside a large polyfascicular human-sized nerve in a subchronic setting. Key Words: Peripheral intraneural interface—Selective stimulation— Interneural stimulation—Neural prosthetics.
Advanced prosthetic limbs with sensing capabilities and multiple degrees of freedom have been developed for amputee subjects (1). However, currently, most advanced prosthetic devices used are sequentially controlled via information extracted from surface electromyographic recordings (EMGs) (1,2). These control schemes fail to provide intuitive control; rather, they typically only provide one degree of freedom at a time, and the user must rely on visual or audio feedback (3,4). A promising method for controlling a prosthetic arm is targeted muscle reinnervation (2,5). This method involves the surgical redirection of nerves in the stump to muscles and skin
in the chest region. Volitional motor commands directed to the stump will then project to the reinnervated muscles, which, via EMG, are mapped to appropriate control commands on the prosthetic arm. In addition, as the sensory nerve fibers of the stump reinnervate the skin of the chest, this skin area can potentially be used for providing sensory input from a prosthetic device (5). Although promising, the method of nerve reinnervation is highly invasive and may primarily be attractive to above-elbow amputees, in whom the loss of residual nerve function is limited (6). An alternative method for interfacing a prosthetic device would be to interface the peripheral nerves directly. The feasibility of this approach was shown by, for example, Dhillon et al. in longtime amputees, who demonstrated that sensations related to the phantom limb could be evoked through selective microstimulation of the nerve stump. In addition, they were able to obtain control signals directly from the nerve to control a prosthetic limb (7–9).
doi:10.1111/aor.12347 Received February 2014; revised April 2014. Address correspondence and reprint requests to Dr. Kristian R. Harreby, Kristian R. Harreby, Aalborg University, Frederik Bajers Vej 7D, 9220 Aalborg, Denmark. E-mail: [email protected] .dk Artificial Organs 2015, 39(2):E36–E48
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES To control a prosthetic hand with multiple degrees of freedom or to provide sensory feedback from artificial limbs in general, a peripheral neural interface should be able to selectively interface with multiple subsets of nerve fibers throughout different fascicles. In general, the more active sites a neural interface has and the closer they are placed to different populations of nerve fibers, the more nerve fibers can be selectively addressed. However, as the active sites are placed in the direct vicinity of nerve fibers, the degree of invasiveness increases and the electrodes tend to pose a greater risk of damaging the nerve (6). The TIME project (CP-FP-INFSO 224012, http:// www.project-time.eu) aims to develop a novel neural prosthesis system to apply multichannel microstimulation to the nerve stump of a volunteer amputee to evoke phantom limb sensations and to explore if sensory feedback can be used as an alternative treatment of phantom limb pain. Within this framework, our group designed the transverse intrafascicular multichannel electrode (TIME). The TIME is a nextgeneration intraneural electrode built upon the principles of the thin-film longitudinal intrafascicular electrode (tfLIFE) (10). As our research work was included in a translational research program, that is, we envisioned a human clinical trial with a duration of 30 days at its very end, design and development included assessment and management of possible risks according to ISO 14971 (Application of Risk Management to Medical Devices). The tfLIFE and TIME were designed to place multiple active sites (>12) inside the nerve while keeping the risk of nerve damage low. Both electrodes are approximately of the same physical size as the electrodes used by Horch et al. (10,11), but instead of having one or two active sites on each electrode, they contain 12 or more active sites distributed along both sides of a thin, flexible polymer substrate. The tfLIFE is implanted in parallel to the nerve fibers, and thus all active sites will redundantly interface with the same subpopulation of nerve fibers, whereas the TIME is implanted transversely through the nerve, thus interfacing with different fiber populations within the nerve over its cross-section (10,11). The use of TIMEs as neural interfaces in humans in the future will depend on the preclinical evaluation of the stability, functionality, and biocompatibility of the electrodes. Recently, the selectivity performances of tfLIFE and TIME were compared during acute experiments in the sciatic nerve of the rat (12) and later in the median nerve of the farm pig (13). These studies both showed that the TIME was able to recruit a higher number of muscles with a higher degree of selectivity than the tfLIFE. Histological
E37
assessment showed that due to differences in nerve morphology, the electrodes had been placed inside the fascicles in the rat model (three fascicles in the sciatic nerve at the implant site), while in the pig model (>30 fascicles in the median nerve at the implant site) the TIMEs had not penetrated the fascicles and thus were located between fascicles. Before the TIME can be applied in subchronic human implants ( 0.4, which corresponded to an average recruitment of nontarget muscles of less than one-fourth of that of the target muscle. Finally, to quantify the selectivity performance of a whole TIME device, SID≥30% was defined as
SI D≥30%
∑ =
NA j =1
SIM ≥30%, j
(5)
NA
Thus, SID≥30% is simply the average of all SIM≥30% for all the monitored muscles. SID≥30% and the number of selectively recruited muscles for the first and last follow-up were compared using the Wilcoxon signed-rank test.
Animal and nerve assessment Animals were euthanized after at least 30 days post-implant. Following the terminal follow-up, the location of each TIME was determined, and the nerve, still containing the implanted electrodes, was removed for postmortem histological analysis. The excised implanted nerve was fixed (frozen by liquid nitrogen or embedded in formalin and later fixed in paraffin), and thin 5-μm transverse sections were made. The TIMEs were then visually located, and the thickness of the encapsulation was defined as the distance from the polyimide structure to the rim of the fibrotic capsule. This distance was estimated for each TIME based on the mean of five measurements (AxioVision v. 4.6.30, Carl Zeiss Imaging Solutions GmbH, Oberkochen, Germany).
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES RESULTS Six TIMEs (referred to as P1T1 [Pig 1 TIME 1], P2T1, P2T2, P3T1, P4T1, and P4T2) were implanted in four animals for 33.8 ± 2.4 days (mean ± SD). During the first follow-up, bipolar EMG recordings were performed on all five muscles in all four pigs. However, EMG wires were vulnerable to breakage both close to the Omnetics connector and at the connection between the MP35N wires and the stainless-steel wires in the legs of the pigs. Due to these problems, the majority of the muscles were eventually monitored via monopolar EMG recordings, and some were omitted. During the last followup, EMG was obtained from all five muscles in P1 (needle electrodes, all bipolar), P2 (needle electrodes, all bipolar), and P3 (three bipolar, two monopolar) and from four muscles in P4 (all monopolar). The connections between the MP35N wires and the Omnetics connectors were functional in all TIMEs, except in two cases where postmortem assessment of the cable resistance indicated one of 12 and two of 12 broken connections in P3T1 and P4T2, respectively. Both of these TIMEs gradually ceased to evoke muscle activation during the implantation period; however, the few broken wires could not fully
E41
explain the overall performance decline. These failures were assigned to behavior of the minipigs that would not be expected in human clinical trials, such as scratching and rubbing against the pen wall. Recruitment Stimulation with increasing intensities evoked increasing muscle responses starting from 1.5–3 ms after stimulation onset and lasting 5–10 ms (see Fig. 2A). Typically, the evoked responses maintained the same shape while increasing; however, additional phases occasionally occurred (see Fig. 2A, M2 F PL). Recruitment curves were in general not smooth sigmoidal shapes but increased irregularly, sometimes with intermediate plateaus (Fig. 2B,C). Sometimes we observed a concurrence between these plateaus and increase in recruitment. In the majority of the cases, all muscles were recruited by each individual stimulation configuration, and recruitment started at similar current levels; however, some differences were seen in the slope of the recruitment curve and the maximal EMGRL reached for the individual muscles, meaning that some muscles were more selectively recruited than others. Monopolar stimulation configurations evoked higher muscle recruitment levels, starting at lower
FIG. 2. Examples of evoked EMG responses (A) and recruitment curves (B,C) obtained from P2T1 at day 7. (A) Evoked EMG obtained while stimulating active site 3′ against the subcutaneous stainless steel needle (ground [G]). Bipolar recordings from M2 F PL and M3 F CR were performed, and monopolar recordings were performed for M1 P T, M4 FDS, and M5 F DP. For clarity, only EMGs for stimulation intensities dividable by 100 μA are shown (100 μA, 200 μA . . . 1200 μA). Time 0 indicates the onset of the 100-μs pulse (vertical, black solid line), and the 3–10 ms time window used for calculating the RMS values is indicated (vertical dotted lines). Increasing stimulation intensity resulted in progressively larger and sometimes more complex evoked EMG responses (see M2 F PL). (B) Recruitment curves from the monopolar stimulation configurations for the three active sites on each side of the TIME structure. The dotted vertical line indicates EMGRL30%. Each active site recruits all five muscles to some extent, although M3 F CR does not reach EMGRL30%. (C) Bipolar TIME configurations start recruitment at higher current levels and do not reach as high an EMGRL as monopolar configurations. Artif Organs, Vol. 39, No. 2, 2015
E42
K.R. HARREBY ET AL.
current intensities, than those observed for bipolar configurations (Fig. 2B,C). During the first followup, 27 of 60 bipolar configurations (10 in six TIMEs) were able to produce EMGRL > EMGRL30%. However, the bipolar configurations gradually became unable to produce muscle recruitment, and at the final follow-up only three of 60 bipolar configurations were able to reach EMGRL30%. The morphology of the muscle recruitment curves produced by stimulation on individual contacts tended to change over 2–4 follow-ups (see Fig. 3). There was a general tendency for the curves to gradually shift to the right, indicating that higher current levels were needed to recruit the muscles. Overall, five of six TIMEs were able to recruit at least one muscle to EMGRL30% at all 12 active sites (monopolar stimulation) during the first follow-up. In P3T1, six of the 12 active sites (1–6) were initially able to recruit muscles to >EMGRL30%, whereas only minor recruitment was achieved at the remaining six active sites (1′–6′). The number of recruited muscles gradually decreased at all contact sites, and following day 15, only one muscle could be activated to >EMGRL30%; after day 20, no recruitment could be observed. P4T2 went from recruiting three different muscles to >EMGRL30% on day 17 to being able to activate only one on day 21, which was the last day any recruitment was observed. Interestingly, postmortem histology actually showed that these two specific TIMEs were surrounded by the thinnest encapsulation of all the TIMEs (54.5 μm and 93.7 μm, respectively, compared with the remaining TIMEs [136.0 ± 42.9 μm]). The poor performance of these electrodes was, however, already seen during the first follow-up, as these electrodes needed significantly higher currents to recruit muscles (ID30% of 745 μA and 666 μA, respectively) compared with the other TIMEs (379 ± 86 μA). The remaining four TIMEs (P1T1, P2T1, P2T2, and P4T1) were able to evoke EMGRL > EMGRL30% at nearly all active sites (typically all 12, and always ≥9) throughout the complete timeframe of implantation. Recruitment current Although fluctuations were seen, there was a general trend for the IM30% to increase throughout the duration of the implantation (see Fig. 4A). The trend was stronger for the ID30%, which consistently and significantly increased in all TIMEs from the first follow-up to the last (P = 0.028, Fig. 4B). During the first follow-up, the mean ID30% across all six electrodes was 488 ± 68 μA (mean ± SEM); this increased to 769 ± 128 μA during the last follow-up. However, if limited to the four TIMEs that were Artif Organs, Vol. 39, No. 2, 2015
functional during the whole period, the ID30% went from an initial 379 ± 42 μA to a final 552 ± 48 μA. Selectivity Neighboring active sites on one side of the device or on different sides of the device (e.g., 1 and 2 or 1 and 1′) recruited similar subsets of muscles (Fig. 2B,C), whereas small differences in muscle recruitment often resulted in variations between the maximal SIM values reached. In P2 and P4 the two
FIG. 3. Examples of the recruitment curves of M3 F CR, M4 F DS, and M5 F DP produced when performing stimulation on active site 3′ in P4T1 throughout the whole duration of the implantation. The follow-up and days of implantation are indicated at the left, and the recruitment scale is indicated at the bottom right. Recruitment curves for M3 F CR , M4 F DS, and M5 D DP are based on bipolar recordings until follow-ups (FUs) 12, 10, and 6, respectively. The recruitment curves increase irregularly as a function of the stimulation current. The morphology of the curves changes gradually over periods of 2–4 follow-ups. A general trend is that the curves tend to shift to the right, indicating that more current is needed to recruit the muscles later in the period of implantation.
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES
FIG. 4. Current for providing 30% muscle activation over the implantation period. (A) The required current for recruiting each muscle (IM30%) and the average required recruitment current of all muscles activated by a TIME device (ID30%) are shown for P4T1. (B) The ID30% values for all six TIMEs over the whole duration of the implantation. The mean and standard error of the mean of the first and last follow-ups (FFU and LFU) are indicated with black dots and bars. Most TIMEs remain well below the safe upper current limit of 1200 μA at 100 μs; however, P3T1 and P4T2 gradually fail to reach EMGRL30% , which is indicated by the current saturating at 1200 μA. Across all electrodes there is a significant (P = 0.028) increase in ID30% from first to last follow-up.
E43
TIMEs implanted in a single median nerve innervated different subsets of muscles (Fig. 5A,B). The SIM≥30% values of individual muscles could both increase and decrease over time, and no clear trend was observed at the level of individual muscles; however, typically a decreasing trend could be observed for SID≥30% (Fig. 6A). Across all six electrodes; there was an overall tendency for SID≥30% to decrease over time (not statistically significant; P = 0.12). The mean SID≥30% across all electrodes changed from 0.25 ± 0.04 to 0.14 ± 0.05 from the first follow-up to the last (Fig. 6B). Two electrodes (P2T2 and P4T2) were able to recruit three muscles selectively during an early follow-up, whereas typical TIMEs recruited between zero and two muscles selectively (Fig. 6C). The initial mean was 1.17 ± 0.37, which decreased to 0.67 ± 0.38 at the final follow-up, whereas this latter decreasing trend did not reach significance (P = 0.18). During the initial follow-up one-third of the muscles were most selectively activated by bipolar stimulation configurations. However, the SID≥30% increased only by 0.005 when compared with pure monopolar stimulation configurations. Animal and nerve assessment After the initial postoperative swelling had ceased (∼5 days) the animals did not show any discomfort and supported their full body weight on the implanted leg. After 20 days, an infection was observed in P1 at the leadout wires at the back of
FIG. 5. The plots show SIM as a function of EMGRL (based on bipolar needle electrodes) while performing stimulation using a subset of active sites (1–6) in each of the two TIMEs implanted in P2 on day 37. Due to the definition of the SIM, it is prone to large fluctuations at low muscle activity and recruitment, and thus, EMGRL30 days. We found that bipolar stimulation tended to produce less activation than monopolar stimulation, and eventually very few bipolar configurations were able to produce EMGRL > EMGRL30%. Bipolar configurations produce a more focused electrical field than monopolar configurations; this has been shown to improve recruitment selectivity of the TIME in a rat model (25). If insertion methods are developed for placing TIMEs inside fascicles in large polyfascicular nerves, for example via high-velocity insertion (26) or via microdissection of the nerve, bipolar stimulation might help to increase selectivity in this nerve model as well. Similarly to other studies, we assessed the recruitment characteristic of the TIME by measuring the evoked EMG (12,27–29); however, a number of factors should be taken into consideration when using this method. Firstly, repeated activation of muscles, performed at frequencies as low as 2 Hz, may potentiate the evoked muscle force; however,
TRANSVERSE INTRAFASCICULAR MULTICHANNEL ELECTRODES the effect on the measured EMG, which is assessed in the current study, seems limited (30). Secondly, recruitment of sensory nerve fibers may trigger muscle reflexes. To focus on the directly evoked EMGRL, this parameter was calculated based on a time window prior to the occurrence of most evoked reflex activations (3–10 ms after stimulation pulse). Thirdly, cross-talk will occur if a muscle electrode records EMG activity from muscles other than those it was intended for; this will produce a negative bias in the selectivity measures. In our study, the observed coherence between plateaus and the increase in recruitment curves from different muscles (Fig. 2B) could indicate the presence of cross-talk. However, while stimulating single fascicles (following dissection) using a hook electrode in P3 during the terminal follow-up, it was possible to obtain very selective recruitment of single muscles in some cases (e.g., SIM > 0.9 at EMGRL = 80%); in other cases, several muscles were activated by a single fascicle. Obtaining a high selectivity would not have been possible if extensive EMG cross-talk had been present, and therefore we believe that the amount of cross-talk in our study was limited. In some cases monopolar EMG recordings were used, which are more susceptible to cross-talk than bipolar recordings. However, during the last follow-up, bipolar EMG recordings were obtained (with the use of additional needle electrodes) in three of the four functional TIMEs (P1T1, P2T1, P2T2); thus, the overall conclusion of the study will not be affected by this. Recruitment current We found that ID30% consistently increased during the implant period. In the four of six TIMEs that remained functional, the ID30% remained well below the maximal charge capacity of the TIME. The gradual failure to recruit muscles in the other TIMEs may be explained by a gradual formation of connective tissue and by the active sites being located at greater distance from fascicles which innervated the subset of monitored muscles (they had the highest initial ID30%). The required charge for producing muscle activation with the USEA (assessed via palpation) in the chronic cat model was reported to increase during the first month of implant and then plateaued at around 80 μA (bipolar; pulse widths of 200 μs) (23). With the differences in methodology in mind, the stimulation intensity used was similar to that found in the current study (see initial recruitment in Fig. 2B). In contrast, Grill and Mortimer reported that the threshold current for a tripolar cuff decreased after implantation and stabilized at 200–600 μA (pulse
E45
width 100 μs; tripolar stimulation) after 8–12 weeks of implantation (22). This difference in the development of the recruitment current may be a result of the fact that formation of connective tissue will always increase the distance from contact sites to the target nerve fibers (and thus the required current) with interneural electrodes, whereas the formation of connective tissue outside an implanted cuff (31) may actually help confine the induced charge within the cuff. The TIME was recently evaluated in the smaller sciatic nerve of rats (12), where muscles were activated with less than one-tenth of the charge used in the current study. Our histological analysis showed that the TIMEs in our pig model were placed between fascicles, whereas in the rat model they were inserted directly into the fascicles using visual guiding. The larger charge needed in the pig model can be explained by the extrafascicular location of the electrodes in this model, where both the greater distance from the target nerve fibers and the low conductance of the perineurium (32) tended to hinder recruitment. The tfLIFE was evaluated in a human amputee; two tfLIFEs were implanted in the median and ulnar nerve of the arm stump (33). Using maximal charges of less than one-tenth of that required for reaching EMGRL30% in our experiment, it was possible to evoke tactile sensations during the first 10 days of implantation (33). It was not determined whether the tfLIFE was placed inside a fascicle or between fascicles in this case. Selectivity The multicontact cuff, the USEA, and the TIME were shown to recruit muscles with several degrees of freedom when evaluated in smaller nerve models (cat and rat) (12,22,23). When implanted in the sciatic nerve of the rat, the TIME could produce selective recruitment of the three muscles monitored (12). In our case, the SID≥30% changed from 0.25 ± 0.04 in the first follow-up to 0.13 ± 0.05 during the last follow-up (mean ± SEM), and the number of selectively activated muscles per TIME decreased from 1.17 ± 0.37 to 0.66 ± 0.38. It is not surprising that the selectivity parameters tended to decrease over time due to the formation of a fibrotic capsule around the electrode structure. However, the selectivity level and the number of selectively recruited muscles still appear low compared with the performance of the TIME in the rat model. The differences in results are likely due to differences in the nerve morphology and/or the methodologies applied. Because of the established Artif Organs, Vol. 39, No. 2, 2015
E46
K.R. HARREBY ET AL.
somatotopic organization of motor fibers and the small size (diameter 1–1.5 mm, three fascicles) of the rat nerve, the TIME could be inserted using visual guidance into all the fascicles that innervated the three monitored muscles (12,34). In the polyfascicular median nerve of our pig model (diameter 3.5–4 mm, 25–35 fascicles) (14), it was not possible to distinguish individual fascicles during insertion, nor was the somatotopic organization of the fascicles known, and therefore electrodes could only be inserted “blindly” in our study. Microdissection of the nerve during the implantation procedure would have enabled insertion of the TIMEs into individual fascicles using visual guidance. Such an approach, however, was ruled out due to the increased invasiveness and time consumption it would have required. As the pig nerve contains around 30 fascicles, only a small subset of these will be located just next to an implanted TIME (∼7, unpublished findings) and thus can actually be expected to be selectively recruited. Making the simplified assumption that each of the five monitored muscles was fully innervated by one specific fascicle, the probability of selectively activating the monitored muscles using a single TIME can be calculated using binomial coefficients. Even under the ideal assumption that all seven fascicles neighboring a TIME were selectively recruited, the risk of detecting no selectively recruited muscles or only one is 67%, and the chance of having more than two muscles selectively recruited is a mere 7%. Based on this model, the average number of selectively recruited muscles that can be expected is 1.17 per TIME, which is indeed very similar to the number of selectively recruited muscles (1.17 ± 0.98) obtained during the first follow-up (Fig. 6B). Hence, several fascicles might have been selectively recruited, but only the recruitment of the subset of fascicles that innervated the five monitored muscles would have been detected. Therefore, our results do not rule out that a single TIME could be sufficient for inducing several types of distinct sensory input; however, several TIMEs would be needed if a whole polyfascicular nerve needed to be selectively activated. Figure 5 shows how two TIMEs implanted with a 45° difference were able to selectively recruit two different subsets of muscles. To maximize the benefit of using multiple TIMEs, they should be implanted in parallel relative to each other so that they interface non-overlapping cross-sections of the nerve (35). There are two main differences between the commonly used SI(I)V (12,13,21,36) and the new SI(I)M. Firstly, the mean recruitment of nontarget muscles (penalty) is weighted by a constant number, which in our case is related to the total number of EMG chanArtif Organs, Vol. 39, No. 2, 2015
nels. This prevents the SI(I)M from being further positively biased if EMG cannot be obtained from all five muscles. Secondly, if all muscles are activated to the same degree, this does not provide any functionally selective activation; however, the SI(I)V will take a value of 1/NA, where NA is the number of muscles monitored. This induces a positive bias, which makes it difficult to distinguish the bias from actual functional selectivity. Our SI(I)M therefore includes a correction that removes the 1/NA bias and scales the functionally relevant selectivity to the interval of [0–1]. In general, this makes the SI(I)M more conservative than the SI(I)V; for example, the mean SID≥30% obtained during the first follow-up was 0.25, whereas it would have been 0.40 if SI(I)V had been used. Animal and nerve assessment Following a 3–5-day recovery period, the animals showed no discomfort. Visual assessment ruled out only severe nerve damage (motor fibers); however, these observations were supported by the histological analysis, which showed no sign of necrosis or nerve damage. The thickness of the encapsulation was estimated to be 115.4 ± 51.5 μm. In comparison, the thickness of the encapsulation found following chronic implantation of the TIME (37) and tfLIFE (24) in the sciatic nerves of rats was only around 30–60 μm (tfLIFE encapsulation estimated from illustrations in article). When the USEA was implanted in the sciatic nerve of the cat it resulted in 426 μm of connective tissue between the base of the electrode and the nerve; however, inside the nerve the distance between the individual shanks and the nerve fibers was only 30.4 μm (23). The thicker encapsulation seen in the current study might be due to the different species used or to the differences in the duration of implantation (i.e., 1 month in our study and more than 2 months in the cited studies). The encapsulation and fixation of the implants observed postmortem indicated that the approach of anchoring the TIME (leadout wires, ceramic connector, and polyimide loop) transversely in the nerve is possible and safe. CONCLUSIONS Before the transverse intrafascicular multichannel electrodes may be tested in subchronic human implants (
