Objective. Electrical stimulation of peripheral nerves has long been a treatment option to restore impaired neural functions that cannot be restored by conventional pharmacological therapies. Endovascular neurostimulation with stent-mounted electrode arrays is a promising and less invasive alternative to traditional implanted electrodes, which typically require invasive implantation surgery. In this study, we investigated the feasibility of endovascular stimulation of the femoral nerve using a stent-mounted electrode array and compared its performance to that of a commercially available pacing catheter. Approach. In acute animal experiments, a pacing catheter was implanted unilaterally in the femoral artery to stimulate the femoral nerve in a bipolar configuration. Electromyogram of the quadriceps and electroneurogram of a distal branch of the femoral nerve were recorded. After retrieval of the pacing catheter, a bipolar stent-mounted electrode array was implanted in the same artery and the recording sessions were repeated. Main Results. Stimulation of the femoral nerve was feasible with the stent-electrode array. Although the threshold stimulus intensities required with the stent-mounted electrode array (at 100–500 µs increasing pulse width, 2.17 ± 0.87 mA–1.00 ± 0.11 mA) were more than two times higher than the pacing catheter electrodes (1.05 ± 0.48 mA–0.57 ± 0.28 mA), we demonstrated that, by reducing the stimulus pulse width to 100 µs, the threshold charge per phase and charge density can be reduced to 0.22 ± 0.09 µC and 24.62 ± 9.81 µC cm−2, which were below the tissue-damaging limit, as defined by the Shannon criteria. Significance. The present study is the first to report in vivo feasibility and efficiency of peripheral nerve stimulation using an endovascular stent-mounted electrode array.
1.Introduction
Neural bionic devices interfacing with the peripheral nervous system have been clinically used for decades to restore lost neural functions secondary to diseases or injuries. Stimulation of the vagus nerve has been used for the treatment of drug-resistant epilepsy and depression [1, 2], and is currently being explored as a treatment option for inflammatory bowel disease and musculoskeletal diseases [3, 4]. Clinical application of stimulation of smaller peripheral nerves includes hypoglossal nerve stimulation for the treatment of obstructive sleep apnoea and tibial nerve stimulation for the management of overactive bladder syndrome [5, 6]. Promising results have also been shown with peripheral nerve stimulation for chronic pain management [7, 8] and giving sensory feedback to prosthetic limbs [9].
Despite the evidence showing the benefits of peripheral neural interfaces, the invasive nature of the implantation of traditional electrode arrays and the open surgery required have limited their clinical incorporation as a long-term treatment option. Nerve cuff electrodes are the most commonly implanted peripheral nerve interface for stimulating peripheral nerves. As an extraneural device, nerve cuff electrodes preserve the structural integrity of peripheral nerves and are therefore generally considered safe and suitable for chronic implantation. However, cuff electrodes are directly secured around the target nerve and are associated with persisting inflammatory foreign body response that may lead to permanent neuromorphological changes at the implanted site, including reduced myelin thickness, reduction in axon diameter, and focal reduction of axon density [10–12]. If the nerve cuff is implanted near a joint, the relative movement between the cuffed nerve and the implant may exacerbate the foreign body response and traction from the lead wires may lead to nerve entrapment and cause further tissue damage [13, 14]. Fibrotic tissue growth between the implant and the nerve as well as into the endoneurial space have been well documented in histological studies following nerve cuff implantation [10, 11, 15, 16].
In contrast, endovascular electrode arrays can be introduced into a blood vessel near the target neural tissue with the aid of guidewires and catheters, negating the need for open or keyhole surgery and extensive dissection to expose neural tissue. One of the most advanced technologies in the development of endovascular peripheral interfaces is the StentrodeTM (Synchron, Inc., Brooklyn, USA) that consists of an electrode array mounted on a self-expanding stent [17–24]. The StentrodeTM is capable of anchoring within a blood vessel immediately after implantation and is further stabilized over time through endothelialization and blood vessel wall incorporation [19, 22]. Previous pre-clinical studies using a sheep model have shown the feasibility and efficacy of stent-mounted electrode arrays both as a chronic recording and stimulating device targeting the motor cortex, when implanted intracortically in the superior sagittal sinus [17, 19, 20]. The first-in-human clinical trial of the Stentrode as an endovascular motor neuroprosthesis is well underway and the results so far demonstrated its safety and feasibility as a minimally invasive brain–computer interface [21, 24].
Previous animal studies have shown the feasibility of intraoperative endovascular stimulation of peripheral nerves, including vagal nerve stimulation, using an electrode catheter placed in the superior vena cava [25] or using a basket mapping catheter from within the internal jugular vein [26], and stimulation of the femoral nerve, with a wire electrode placed in the femoral artery [27]. Clinically, a stimulation lead wire placed in the patient's pericardiophrenic vein to deliver unilateral phrenic nerve stimulation, coupled with a fully implantable stimulator, has been used to regulate breathing in patients with central sleep apnea [28, 29]. However, there has been no reported attempt to date to expand the application of the endovascular stent-mounted electrode array technology to the peripheral nervous system.
The present study was undertaken to investigate the feasibility of a stent-mounted electrode array positioned in the femoral artery to deliver bipolar electrical stimulation to the compound femoral nerve and elicit muscle responses in anaesthetized sheep. The femoral artery was chosen as the implantation site for endovascular electrodes due to its superficial location, ease of access, and proximity to the femoral nerve. Chronically implanted medical bionic devices with transvenous leads, such as pacemakers and defibrillators, have shown their long-term safety through many years of clinical outcomes. In contrast, intra-arterial devices or devices with transarterial leads are currently only intended for acute implantation and transient use. Such devices include ablation catheters used in renal nerve denervation [30] and infusion catheters for transarterial chemoembolization [31]. The potential effects of chronic transarterial lead placement to vessel patency remain to be investigated by further pre-clinical and clinical studies. Furthermore, the progressive development of wireless stimulator technology [27, 32] will also facilitate miniaturization of endovascular stimulators.
We also evaluated the stimulation efficacy of the stent-mounted electrode array by comparing, with an endovascular pacing catheter, the evoked compound nerve and muscle action potential amplitudes and the strength-duration curves. Selective stimulation of the proximal femoral nerve with implanted nerve cuff electrodes can activate the knee extensor muscle group to restore standing in individuals suffering lower extremity paralysis after a spinal cord injury [33–36], rendering it a clinically relevant stimulation target. The sheep was chosen as an appropriate large animal model for the implantation of endovascular electrode arrays since their blood vessels are more similar in size to human compared with smaller animal models, and the anatomy of the femoral triangle in sheep bears some similarities to that of humans. Specifically, the diameters of the sheep external iliac artery (7–8 mm) and the femoral artery (5 mm) [37] are slightly smaller than but comparable to the diameters of the human common femoral artery (8.2–9.8 mm) [38] and superficial femoral artery (7.8 mm) [39].
2.Methods
2.1.Surgical procedures
All animal care and experimental protocols were reviewed and approved by the Animal Ethics Committee of the Florey Institute of Neuroscience and Mental Health, Melbourne, Australia (Ethics approval number: 21-065-FINMH). Four adult female sheep (Australian Merino, 30–50 kg, obtained from the Beverley Merino Sheep Stud, Redesdale, Central Victoria) were anesthetized with isoflurane (1.5%–2.0% in oxygen/air mixture) during surgery and data acquisition. Tracheal intubation was performed following anesthesia induction with sodium thiopentone (Jurox Thiobarb, 15 mg kg−1 IV) for mechanical ventilation. Intramuscular injection of analgesic, Flunixin (Ilium Flunixil, 1.1 mg kg−1), and antibiotic/local anesthetic mixture, procaine penicillin (Ilium Propercillin, 900 mg), were administrated at the beginning of the surgery. Heart rate and blood oxygen level were monitored with a pulse oximeter attached to the tongue. The sheep was placed in a supine position and the sartorius muscle was exposed through a 10 cm incision in the proximal medial thigh. The femoral artery was then exposed by blunt dissection through this muscle and catheterized with an 8 F catheter sheath (figure 1(a)). The catheter sheath was sutured to skin to prevent dislocation. Heparin (7500 IU per injection) was administrated through the sheath prior to the introduction of the pacing catheter, and then prior to stent introduction, to prevent blood coagulation.
Electromyographic (EMG) needle electrodes were inserted into the vastus medialis muscle along the muscle fiber orientation to record quadriceps EMG responses. The superficial branch of the femoral nerve on the medial surface of the femoral artery that distally forms the saphenous nerve was placed on a pair of hook electrodes to record femoral nerve electroneurogram (ENG). A skin pocket was created on the chest and a stainless-steel ground reference plate was placed in the pocket. A g.USBamp RESEARCH amplifier (16 ADC channels with 24-bit resolution, independent sampling, 87nV sensitivity, and unity gain) and the software g.RECORDER (G.Tec, Schiedlberg, OOE/AUT) were used for EMG and ENG acquisition at 38.4 kHz. At the end of the recording session, tracing the femoral artery proximally, a large femoral nerve branch dorsolateral to the artery that innervated the vastus medialis muscle was exposed through careful dissection (figure 1(b)). The compound nerve slightly proximal to the branching point was transected and a suprathreshold stimulus was delivered through the stent-mounted electrodes afterwards to confirm that the EMG responses were not due to direct stimulation of the muscles. At the end of the experiment, each animal was euthanized with an IV injection of sodium pentobarbitone (100 mg kg−1).
2.2.Endovascular electrode placement
A 5F bipolar pacing catheter (PACEL, St. Jude Medical, Saint Paul, IT) was introduced into the femoral artery through the catheter sheath. While the pacing catheter was advanced proximally in the artery, bipolar, cathodic, and monophasic stimulation (1 mA, 10 Hz, pulse width = 500 µs) were applied to the femoral nerve through the catheter, and the optimal catheter electrode position was determined by visually observed maximal contraction of the medial quadriceps and recorded with fluoroscopy (figures 2(a) and (b)). Stimulation trials were then performed to determine the threshold amplitude and strength-duration curve of endovascular femoral nerve stimulation with the pacing catheter electrodes.
Upon completion of the pacing catheter stimulation sessions, the pacing catheter was removed and a stent with an electrode array attached was introduced into the femoral artery through the catheter sheath. The stent-electrode arrays were constructed in-house using a modified methodology [20]. The stent used as the scaffold for electrode attachment was a commercially available self-expanding nitinol stent (4 mm diameter, Solitaire SAB, Covidien, USA or Trevo NXT ProVue, Stryker, Kalamazoo, USA), and an electrode array with four 750 µm diameter laser-cut platinum electrodes was mounted on the stent structs in a bipolar configuration (figure 2(c)). When fully deployed, the inter-electrode distance between proximal and distal pair of electrodes was approximately 10 mm. The proximal end of the stent was soldered to a stainless-steel stylus to aid stent insertion and deployment. Electrodes of the same polarity were electrically connected to an insulated platinum contact (two contacts in total) at the distal end of the stylus with 33 µm diameter insulated platinum-tungsten wires. As with the catheter electrodes, the proximal electrodes actively delivered the stimuli, whereas the distal ones were used as return electrodes. Under fluoroscopic guidance, the stent was advanced to a depth so that the proximal electrodes approximately coincided with the recorded pacing catheter proximal electrode position (figures 2(a) and (b)). The same stimulation sessions were then performed again with the stent-electrode array.
2.3.Endovascular stimulation trials
Endovascular electrodes were connected to a constant-current stimulator (TWISTER® MM, Dr Langer Medical, Waldkirch, DE). Trains of bipolar, monophasic, cathodic stimulation (frequency = 10 Hz, duration = 20 s) were delivered through the endovascular electrodes with pulse widths from 100 µs to 500 µs at 100 µs increments and amplitudes ranging from 0.1 mA up to 5 mA. The visually detectable EMG response threshold was first determined by increasing the stimulus intensity until contractions of the quadriceps were observed. To record the full range of response amplitudes, stimulation amplitudes were then further increased at small increments until intense muscle contractions occurred, at which point the stimulation amplitude was gradually reduced to below the visual threshold until both EMG and ENG responses disappeared in processed recordings. The quadriceps were allowed to rest for a 20 s interval between stimulation trials to prevent muscle fatigue, trials repeated up to two times with the same parameters were combined during data analysis.
2.4.Data postprocessing
Data postprocessing was performed using MATLAB (version R2023a, The Mathworks, Inc., Natick, USA). Prior to filtering, stimulus artifacts were removed by linearly interpolating between 0.77 ms before and 2 ms after the initiation of a stimulation current pulse. EMG and ENG signals were bandpass filtered with the passband 5-2000 Hz, and 2-1000 Hz (zero-phase Butterworth, 4th order), respectively. Due to the presence of 50 Hz interference in the ENG recordings, a series of notch filters (zero-phase Butterworth, 2nd order) were applied to remove mains interference at 50 Hz and its harmonics up to 1000 Hz. Filtered recordings were segmented into stimulus time-locked epochs and averaged from 200 epochs per 20 s trial. From averaged epochs, threshold stimulus intensity was determined as the minimal intensity that evoked a response of amplitude at least 10 µV larger than three times of the baseline standard deviation, which corresponded to an activity level distinguishable from baseline EMG and slight twitching of the quadriceps muscle. To exclude the stimulus artifact and interference from the digital filters, EMG peak-to-peak amplitude within 3–20 ms latency and ENG peak-amplitude within 2–14 ms latency were considered in further analysis. Threshold intensities derived from repeated trials of the same stimulus intensity were averaged. Supplementary figure 1 shows an example of the data postprocessing pipeline.
Rheobase, the minimal stimulus intensity required for excitation with a pulse of 'infinite' duration, and chronaxie, the pulse duration required for excitation when the stimulus intensity is twice that of the rheobase, are two important constants that characterize neuronal excitability by external electrical stimuli. From the strength-duration curves, chronaxie and rheobase were estimated by fitting each data trace to equation (1) [34],
where X is the pulse duration, Y is the minimal stimulus intensity to evoke excitation, Y0 is the threshold response when the duration of the pulse is close to zero, NS is the rheobase, and chronaxie is calculated from the time constant K as 
[34].
All data are reported as mean ±95% confidence interval (CI), and ranges are reported as minima-maxima. Due to the small number of samples, the distribution of the threshold stimulus intensity data is unclear; therefore, the non-parametric Friedman test was used to test the main effect. Comparisons of chronaxie and rheobase values estimated from strength-duration curves were performed between the pacing catheter and stent-electrode array within subject using paired t-tests. P-values below 0.05 were accepted as statistically significant.
2.5.Histology examination
To confirm the close spatial positions of the femoral nerve and artery, histology examination was performed on tissue samples harvested from two animals. Immediately following euthanasia of one animal, the femoral artery, femoral nerve, and surrounding connective and muscle tissue were harvested and fixed in 10% neutral buffered formalin for at least two weeks. The fixed tissue samples were embedded in paraffin and axial slices of 5 µm thickness were obtained on a microtome at the estimated electrode positions. The stained slices were stained with hematoxylin and eosin and examined and digitized using a ZEISS Axioscan 7 slide scanner microscope (ZEISS Group, Oberkochen, DE).
3.Results
Figure 3 shows a representative histology image of the cross-sections of the femoral neurovascular bundle sectioned approximately at the axial position of the proximal stent electrode, from animal FS04. The presence of the femoral nerve and artery as well as their spatial proximity could be confirmed through the cross-sections. Figure 4(a) shows the averaged EMG and ENG responses to increasing stimulus intensities, whereas figure 4(b) shows the responses to 2 mA current pulses with pulse width 500 µs, before and after proximal femoral nerve transection. With an intact femoral nerve, a high stimulus intensity of 2 mA evoked visible and intense muscle contractions and strong EMG and ENG responses, which ceased after transection of the compound femoral nerve. The disappearance of muscle response after proximal nerve transection confirmed that the muscle response was due to nerve activation rather than direct muscle activation from the current pulses.
Figure 5 shows response amplitudes at pulse widths of 100 µs, 300 µs, and 500 µs of all stimulation trials of all four animal subjects. To compare between different animals, each data point has been normalized on the x-axis by a factor of µoverall/µsubject, where µoverall is the overall average threshold stimulus intensity at each pulse width across all animals and µsubject is the threshold stimulus intensity of an individual subject at the same pulse width. To achieve a similar level of response amplitude, the stimulus intensity required with the stent-mounted electrodes was notably higher than the pacing catheter electrodes, which was consistent across all animals and all tested pulse widths (supplementary figure 2). The peak amplitudes of EMG and ENG responses demonstrated consistent exponential increase in response to incrementing stimulus intensity. The data points were fitted to a generic monotonic exponential model 
using non-linear regression for clearer comparison between pulse widths. Goodness of fit was evaluated using root mean squared error (RMSE). The evoked response amplitudes showed overall good fit to the exponential approximation and good consistency between animals for both electrode types. The exponential growth in evoked response indicated that the current range tested was much lower than the threshold current to saturation.
Figure 6 summarizes the threshold stimulus intensities and total injected charge per phase at the thresholds for the two endovascular electrode types in response to stimulus pulse width varying from 100–500 µs. The minimal stimulus intensity of endovascular femoral nerve stimulation through the stent-mounted electrodes required to evoke muscle and nerve activation decreased with longer pulse widths. The mean threshold intensity of the pacing catheter electrodes decreased from 1.05 ± 0.48 mA to 0.57 ± 0.28 mA as the pulse width increased, whereas for stent-mounted electrodes it decreased from 2.17 ± 0.87 mA to 1.00 ± 0.11 mA. The threshold intensities with the stent-mounted electrodes were consistently higher than with the pacing catheter electrodes across all tested pulse widths (p < 0.0001 for both EMG and ENG data, Friedman test, repeated measure), on average by 2.1 times for both EMG and ENG data measured from the same animal subject. For both types of endovascular electrodes, reducing stimulus pulse width led to significant reduction in total charge injected (p < 0.004 for both pacing catheter and stent-electrode array, Friedman test, main effect), from 0.28 ± 0.14 µC to 0.11 ± 0.05µC (mean ±95% CI, EMG) and 0.33 ± 0.19 µC to 0.11 ± 0.07 µC (ENG) for the pacing catheter at 500 µs and 100 µs pulse widths, respectively. Similarly, for the stent-electrode array, total injected charge reduced from 0.50 ± 0.06 µC to 0.22 ± 0.09 µC (EMG) and 0.51 ± 0.12 µC to 0.24 ± 0.11 µC (ENG).
Fitting strength-duration curves to the exponential equation of X (pulse width) and Y (threshold stimulus intensity) yielded chronaxies of 74 µs (minima to maxima, 61–95 µs, rheobase 0.38–0.76 mA) and 94 µs (41–139 µs, rheobase 0.74–1.14 mA) for EMG responses with the pacing catheter and stent-mounted electrodes, respectively (figures 6(a) and (b)). For ENG responses, the chronaxies were derived as 59 µs (43–68 µs, rheobase 0.34–0.93 mA) and 88 µs (60–120 µs, rheobase 0.76–1.12 mA). For each animal subject, chronaxie values did not differ significantly between the two electrode types (p = 0.50 for EMG data, p = 0.12 for ENG data, paired t-test, n = 4). Rheobase values also did not differ significantly within subject (p = 0.14 for EMG and p = 0.19 for ENG, paired t-test, n = 4).
Figures 6(c) and (d) depict the increasing total charge injected at the threshold current with pulse width. The pacing catheter electrodes had more than 11 times the geometrical surface area than the 750 µm disk electrodes attached to the stent. Consequently, the charge density at the disk electrode surface was estimated to be notably higher than at the surface of the pacing lead electrodes. At 500 µs pulse width, according to the EMG threshold charge values, the disk electrode charge density reached 56.59 ± 6.37 µC cm−2 compared to 2.85 ± 1.41 µC cm−2 at the pacing catheter electrode surface. The disk electrode charge density was reduced to 24.62 ± 9.81 µC cm−2 at 100 µs pulse width. The Shannon equation, derived from cortex stimulation data using surface electrodes, can be used to estimate whether a certain electrical stimulation modality and paradigm can lead to tissue damage [40]. Substituting the threshold charge values of the stent-electrode array EMG into the Shannon equation, 
, where 
is the charge density per phase in µC cm−2 phase−1 and 
is the total charge in µC phase−1, yielded 
values that increased with pulse width from 0.70 ± 0.32–1.45 ± 0.10. As a 
value lower than 1.5 typically indicates absence of tissue damage, threshold currents at all pulse widths tested are under the Shannon limit and can be predicted to have low risk of causing tissue damage.
4.Discussion
This study demonstrated, for the first time, the feasibility of endovascular stimulation of a peripheral nerve using a stent-mounted electrode array. Implanted in the easily accessible femoral artery, the stent-mounted electrodes were able to elicit quadriceps motor responses in all four animal subjects. They showed comparable performance to a clinically used, commercially available pacing catheter, in terms of threshold stimulus intensity and the magnitude of evoked muscle and nerve responses.
Within the tested parameters, pacing catheter electrodes consistently yielded stronger muscle responses than the stent-mounted electrodes with the same stimulation parameters, as well as lower threshold stimulus intensities in all animal subjects. The low threshold of the pacing catheter could be attributed to its curved tip that ensured good contact between the electrodes and blood vessel wall and, more importantly, the insulative silicone catheter that prevented excessive shorting between the bipolar electrodes through blood. However, in contrast to the intravascular stent mounted array, the silicone catheter obstructs blood flow and is not suitable for long-term implantation.
The maximal EMG threshold current with the stent-mounted electrodes at a minimal pulse width of 100 µs was up to 1.7–3.2 mA, and the threshold currents reduced with increasing pulse width. However, total charge injected per phase increased with pulse width, from 0.22 ± 0.09 µC at 100 µs to 0.50 ± 0.06 µC to at 500 µs pulse width. The threshold charge values were comparable to those of intraoperative stimulation thresholds of a multi-contact nerve cuff electrode implanted around the human ulnar nerve (0.1–0.25 µC at 100–500 µs pulse widths) [41], which is similar to the sheep femoral nerve in terms of nerve size and axon diameter distribution [42]. Limiting the current pulse width to lower pulse widths close to the chronaxie 94 µs could benefit endovascular neurostimulation by both saving power and reducing the risk of tissue damage. Indeed, out of the pulse widths investigated, only 100 µs reduced charge density to under the generally recommended non-damaging threshold of 30 µC cm−2 [43].
EMG threshold currents with the stent-mounted electrodes yielded 
-values ranging over 0.70–1.45, indicating that tissue damage is unlikely. However, it must be noted that the Shannon equation was derived from limited brain stimulation data and it does not take into consideration the specific geometry and location of the stimulating electrodes. Additionally, to achieve a functional stimulation outcome, the stimulus intensity required is usually much higher than the threshold to evoke a detectable response. Nevertheless, endovascular stimulation with the stent-mounted electrode array should gravitate towards lower pulse widths close to the chronaxie and, therefore, lower charge density for safety considerations but, in scenarios where recruitment of smaller axons is desired, longer pulse widths should also be considered [44]. Electrodes larger in quantity and surface area should also be incorporated into the array to limit the charge density while allowing for higher stimulus intensities to be delivered.
In this study, bipolar electrode arrays were only connected to external stimulators through two channels. Practical design of electrode arrays on stents should separate electrode contacts into individual channels to maximize the flexibility in controlling the orientation of the focal stimulation. Acute measurement of threshold stimulus intensity of each proximal-distal electrode pair should be carried out intraoperatively during implantation, so that all electrodes with tolerable small misalignments can be identified and used for further functional stimulation to minimize the overall charge density and improve the safety profile. Future experiments should also consider incorporating an implantation technique that grants reasonable control over stent orientation to maximize the number of electrodes oriented towards the target nerve.
Only acute experiments were performed as the long-term stability of the platinum disk electrodes was not the focus of this study. Nevertheless, it must be noted that charge density must be maintained under the charge injection capacity of the electrodes to prevent the occurrence of irreversible Faradaic reactions and subsequently electrode corrosion and tissue damage in a chronic implantation scenario. In the present study, the threshold charge density of the four-electrode array on a stent was estimated to range from 24.62 µC cm−2 to 56.59 µC cm−2 at 100–500 µs. A previous study reported that in vivo charge injection capacity of implanted smooth platinum disk electrodes of a similar size varied from 3.84 to 16.6 µC cm−2 at pulse widths 100–3200 µs, increasing with the pulse width [45], well under the recommended limit of 30 µC cm−2 for brain and intracochlear stimulation [43]. Methods to lower the risk of irreversible Faradaic reactions include using larger electrodes and incorporating more electrodes in the array to lower charge density as well as increasing the roughness of the electrode surface with porous coatings [46]. Moreover, a good alignment of the electrode array towards the target nerve can also reduce the total charge required [26, 47]. It should also be noted that the blood provides an abundant source of electrolyte for endovascular electrodes, which may lead to increased charge injection capacity compared to electrodes emerged in interstitial fluid. Future studies should also evaluate in vivo charge injection capacities of various electrode materials to provide better insights on safety limits of endovascular neurostimulation.
Other concerns associated with intra-arterial electrodes include the effects of artery wall pulsation and potential vasoconstriction caused by electrical stimulation. In this study, the pulsation of the femoral artery did not generate detectable variations in response amplitudes. This could be attributed to the close proximity between the femoral artery and nerve, the small scale of movement of the artery vessel wall during pulsation and buffering from the surrounding fatty and connective tissue. Further animal studies using neurovascular bundles are required to fully characterize the effects of pulsation on endovascular neurostimulation, as well as the effects of endovascular electrical stimulation on vasoconstriction.
A major limitation of the present study is the small sample size, which renders it difficult to explain the contribution of anatomical variances to the variances in measured threshold currents and response amplitudes. The orientation of the stent-mounted electrode arrays was also not measured for every subject. Changes in stent orientation across animals could be observed from the post-mortem microCT scans of preserved tissue samples and intraoperative fluoroscope images. As predicted by our previous computational modeling study, orienting the endovascular disk electrodes away from the target nerve can lead to up to 33% reduction in axon recruitment level at 90° and up to 72% at 180° with a stimulation current of 5 mA [47]. A previous study investigating endovascular stimulation of the vagus nerve also reported clear effects of the circumferential position of the endovascular electrodes [26]. Future studies should investigate the effects of electrode array orientation, potentially through post-implantation computed tomography (CT) [21] or post-mortem micro-CT imaging of ex vivo tissue samples [22]. Our prior modeling study has also predicted the selectivity of endovascular stimulation to be intrinsically constrained by the neurovascular anatomy and fascicular organization of the target nerve [47]. Although beyond the scope of this study, future studies should aim to address not only the effects of electrode positioning on stimulation threshold but also on selectivity.
The hook electrodes used in the present study for ENG recording were secured using a stereotaxic frame, thereby immobilizing the targeted nerve. Substantial muscle contractions had the potential to cause pulling of the nerve and possibly nerve damage. The range of current amplitudes investigated in the current study was limited to avoid excessive movement of the leg. Future studies can consider appropriate fixation of the leg or compliant recording electrodes to investigate higher current amplitudes and produce complete muscle recruitment curves. Additionally, determining activation threshold using a fixed potential value could result in underestimated threshold current values. The threshold potential for detecting evoked responses should be set as a fixed percentage of the maximal evoked potential derived from a complete recruitment curve. Moreover, the integrity of the mounted electrode array was verified only through impedance measurement prior to implantation and during the experiments, between the bipolar electrode pairs at stimulation frequency (10 Hz), in addition to visual confirmation via the intraoperative fluoroscopic images. In future experiments, electrochemical impedance spectroscopy should be carried out over a wider frequency band immediately following acute implantation to characterize electrode and tissue impedance.
In this study, the stimulation amplitudes were not randomized which could conceivably lead to a facilitation effect in the response. However, visual inspection did not identify a facilitation effect in the neural response. Future work could consider utilizing randomized stimulus. In this study, the stent-electrode array was always introduced into the same blood vessel after the removal of the electrode catheter. This may contribute to biased outcomes due to changes in the surrounding tissue resulting from endovascular intervention or electrical stimulation in sequence (i.e. Catheter stimulation followed by endovascular stimulation). Future experiments may consider repeating the stent-electrode trials in the contralateral and naive blood vessel of the same animal. Monophasic electrical stimulation has been used in prior animal studies to demonstrate the feasibility of stimulation modalities [20, 27]. However, charge-balanced biphasic stimulation should be adopted in future studies to prevent time-dependent effects on stimulation response as well as thermal tissue damage.
5.Conclusion
The present study is the first to report the in vivo feasibility and efficacy of endovascular electrical stimulation of a peripheral nerve using a stent-mounted electrode array. The results show that it is feasible to stimulate peripheral nerves using a stent-mounted electrode array from within a nearby blood vessel, and the stimulation efficacy was comparable to a commercially available pacing catheter clinically used for intraoperative neurostimulation. Within the parameters tested, endovascular stimulation using the stent-mounted electrodes can be deemed safe according to the Shannon limit. However, the strength-duration curve indicated that a shorter pulse width is desired to lower the threshold charge and further reduce the risk of tissue damage. Overall, these in vivo findings suggest that the endovascular stent-mounted electrode array is a feasible, minimally invasive alternative as a peripheral nerve interface. Future studies are required to investigate the chronic stability of stimulation efficacy, the effects of chronic implantation on local histology and hemodynamics, as well as to explore the feasibility of recording peripheral neural signals with the device.
Acknowledgments
We thank Mr Dain Maxwell for performing the histology (sectioning, staining, imaging). We acknowledge the expert technical assistance of Mr Anthony Dornom, Mr Tom Vale and Mr Quan Nguyen.
This research was supported by The University of Melbourne's Research Computing Services and the Petascale Campus Initiative.
This research was funded by the Australian Government through the National Health and Medical Research Council of Australia (NHMRC) Project Grant 1158912, Development Grant 2000153, US Defense Advanced Research Projects Agency (DARPA) Microsystems Technology Office contract; Office of Naval Research (ONR) Global; US Department of Defence, Epilepsy Research Program CDMRP EP170058.
Data availability statement
The data cannot be made publicly available upon publication because they are not available in a format that is sufficiently accessible or reusable by other researchers. The data that support the findings of this study are available upon reasonable request from the authors.
Conflict of interest
DBG, SEJ, and CNM have patents related to the Stentrode™ technology similar to the device used in this study. SEJ was previously employed at Synchron Australia (previously SmartStent Pty Ltd) and DBG was previously a member of the Scientific Advisory Committee of Synchron Inc. DBG is the director of and SEJ is an investigator in the ARC Training Centre in Cognitive Computing for Medical Technologies for research in brain-computer interface technologies (ARC CCMT); Synchron Inc. is a partner providing co-funding and in-kind support to the ARC CCMT. The work described in this manuscript was not funded or supported by Synchron Inc.
