Flat Electrode Contacts for Peripheral Nerve Stimulation

The majority of available systems for nerve stimulation use circumferential stimulation electrodes inside an insulating cuff, which produce largely uniform current density within the nerve. Flat stimulation electrodes that contact only one side of the nerve may provide advantages including simpler implantation, ease of production, and more resistance to mechanical failure. However, it is possible that the flat configuration will yield inefficient fiber recruitment due to a less uniform current distribution within the nerve. Here we tested the hypothesis that flat electrodes will require higher current amplitude to achieve effective stimulation than circumferential designs. Computational modeling and in vivo experiments were performed to evaluate fiber recruitment in different nerves and different species using a variety of electrode designs. Initial results demonstrated similar fiber recruitment in the rat vagus and sciatic nerves with a standard circumferential cuff electrode and a cuff electrode modified to approximate a flat configuration. Follow up experiments comparing true flat electrodes to circumferential electrodes on the rabbit sciatic nerve confirmed that fiber recruitment was equivalent between the two designs. These findings demonstrate that flat electrodes represent a viable design for nerve stimulation that may provide advantages over the current circumferential designs for applications in which the goal is uniform activation of the nerve.


28
Peripheral nerve stimulation has emerged as an effective treatment for a variety of disorders. Vagus 29 nerve stimulation (VNS) is one of the most widely used peripheral nerve stimulation strategies and has 30 been employed in over 70,000 patients for control of epilepsy (1). Recent clinical studies demonstrate 31 the potential of VNS for treatment of other neurological disorders, including stroke (2), tinnitus (3), and 32 headache (4). Other emerging nerve stimulation therapies include tibial nerve stimulation for bladder 33 control (5), sacral nerve stimulation for constipation (6), occipital nerve stimulation for migraines (7), 34 and hypoglossal nerve stimulation for sleep apnea (8). Given the broad potential applications for 35 neurological disorders, there is a great deal of interest in identifying optimal stimulation strategies to 36 maximize benefits in patients (9). 37 Implanted cuff electrodes are the gold-standard method for nerve stimulation. The majority of recent 38 developments have been focused on selective stimulation of certain regions or fascicles within the nerve 39 (10-12). Such designs could potentially eliminate side-effects that arise due to stimulation of off-target 40 fibers and could allow for more precise stimulation of target fibers. However, in other applications 41 including VNS, the primary goal based on existing evidence is to generate uniform activation of each 42 region of the nerve. This is typically done with circumferential or helical electrodes that cover the 43 majority of the circumference of the nerve. These designs provide largely uniform stimulation within the 44 nerve yielding a steep recruitment curve and ensuring that any desired fibers are activated with minimal 45 current. 46 Alternative designs may provide better avenues for nerve stimulation and obviate common issues such 47 as a high rate of lead breakage (13). A flat configuration with electrode contacts on one side of the nerve 48 could be built with a simpler and more compact design that would facilitate implantation, provide 4 49 greater resistance to mechanical failure, and reduce cost of production. However, this electrode 50 geometry provides contact with only a portion of the circumference of the nerve, which is likely to 51 produce non-uniform stimulation. This would yield increased activation of axons near the contacts and  single fascicle, perineurium, epineurium, two platinum contacts, an insulating cuff, and ambient 62 medium, similar to previous studies (14,15). In a subset of models, a multi-fascicle nerve containing five 63 fascicles was used (Fig. 9). The nerve had a diameter of either 0.9 mm for the rat sciatic, 0.4 mm for the 64 rat vagus, or 3 mm for the rabbit sciatic (16-18). Perineurium thickness was set to 3% of the fascicle 65 diameter (19). Epineurium thickness was set to 0.13 mm for the rat sciatic, 0.1 mm for the rat vagus, and  cross-sectional area of the nerve and of the cuff lumen was matched between the circumferential and flat 72 electrode models by increasing the inner diameter of the flat cuff by 8.67%. The nerve was modified to 73 take on the shape of the flat cuff ( Fig. 6) (23). Helical electrodes with a width of 0.7mm and thickness of 74 0.01mm had a pitch of 2 mm and completed a 270° arc. The insulation had the same pitch and 75 completed 2.5 turns. The width of the insulation was 1.4 mm and the thickness was 0.9 mm. The empty 76 space in all models was filled with an ambient medium with conductance varying from saline (2 S/m) to 77 fat (0.04 S/m). For the rat models, ambient mediums were 20 mm in length and 4 mm in diameter. For 78 the rabbit models, they were 120 mm in length and 40 mm in diameter. The outer boundaries of all 79 models were grounded. A 1 mA positive current was applied on one contact, and a 1 mA negative 6 80 current on the other. Due to the model being purely resistive, the voltage field only needed to be solved 81 for a single current amplitude. Electrical properties for each material were based on field standards and 82 can be found in table S1 (24-27).

83
Once the Comsol model was solved, the voltage distribution inside the fascicle was exported and read 84 into a NEURON model consisting of 500 parallel axons uniformly distributed throughout the fascicle.

85
The multi-fascicle nerve had 100 axons in each of the five fascicles. Axons were designed using the 86 model created by McIntyre, Richardson, and Grill (28). All electrical parameters were identical in this 87 study, but the geometric parameters were interpolated using either a 1 st or 2 nd order polynomial. All 88 fitted functions can be found in table S2. Each fiber was set to the length of the corresponding Comsol 89 model, either 20 mm or 120 mm. Diameters were taken from a normal distribution meant to represent A-90 fibers (rat sciatic: 6.87 ± 3.02 µm, rat vagus: 2.5 ± 0.75 µm, rabbit sciatic: 8.85 ± 3.1 µm) (29,30). Rat 91 vagus fiber diameters were estimated based on the conduction velocity of the fibers mediating the 92 Hering-Breuer reflex (31)(32)(33)(34). In both sciatic models, fibers with a diameter less than 2 µm were 93 recreated until their diameter was greater than 2 µm. In the vagus model, the same technique was 94 applied but with a cutoff of 1 µm.

95
After a 0.5 ms delay to ensure all axons had reached a steady baseline, a biphasic pulse of varying 96 current amplitude was applied to the NEURON model. The voltage field calculated in Comsol was 97 linearly scaled to the specified current and applied for 0.1 ms, and then the inverse was applied 98 immediately after for another 0.1 ms. Voltage traces from nodes at the proximal end of the axon were 99 recorded and used to determine whether that axon was activated at the given current amplitude. The  River, 3 to 6 months old, 250 to 500 g) were housed in a 12:12 h reverse light-dark cycle. Six rats were 107 used for sciatic experiments, and six rats were used for vagus experiments. Four New Zealand white 108 male rabbits (Charles River, 3 to 6 months old, 2 to 4 kg) were housed in a 12:12 h light-dark cycle. All    Rats were anesthetized using ketamine hydrochloride (80 mg/kg, intraperitoneal (IP) injection) and 134 xylazine (10 mg/kg, IP) and given supplemental doses as needed. Once the surgical site was shaved, an 135 incision was made on the skin directly above the biceps femoris (16,37). The sciatic nerve was exposed  Rats were anesthetized using ketamine hydrochloride (80 mg/kg, IP) and xylazine (10 mg/kg, IP) and 157 given supplemental doses as needed. An incision and blunt dissection of the muscles in the neck 158 exposed the left cervical vagus nerve, according to standard procedures (38-40). The nerve was placed 159 into the cuff electrode, and leads from the electrode were connected to the programmable stimulator.

160
The cavity was kept full of saline at all times. To assess activation of the vagus nerve, blood oxygen 161 saturation (SpO2) was recorded using a pulse-oximeter (Starr Life Sciences™, MouseOx Plus®) as 162 previously described (32). Data was read into MATLAB® using a Starr Link Plus™ with the outputs 163 connected to analog channels on the Arduino®. Data was sampled at 10 Hz and filtered with a 10 164 sample moving average filter. Stimulation was delivered every 60 seconds, but was delayed if needed to allow the oxygen saturation to 168 return to baseline. Each parameter was repeated twice.

170
Both hind legs of the rabbit were shaved over the incision site the day before surgery. Anesthesia was 171 induced with 3% inhaled isoflurane at 3 L/min. A single intraperitoneal injection of ketamine 172 hydrochloride (35 mg/kg) and xylazine (5 mg/kg) was given after induction. Isoflurane was maintained 173 throughout the experiment. Eye ointment was applied to both eyes to prevent drying. Rectal temperature 174 and breathing were monitored throughout the procedure. The incision sites were cleaned with 70% 175 ethanol, followed by povidone-iodine, followed again by 70% ethanol. An incision site was made along 176 the axis of the femur. The sciatic nerve was exposed with blunt dissection to separate the biceps femoris 177 and quadriceps femoris muscles. Alm retractors were placed to allow cuff implantation. After placing 178 the cuff around the nerve, the retractors were withdrawn.

179
Stimulation consisted of 0.5 second trains of biphasic pulses (100 µs pulse width) at 10 Hz with varying 180 current amplitudes ranging from 20-1600 µA. Stimulation using the circumferential cuff electrode was 181 delivered using the same system described above for the rat sciatic. Stimulation with the glass-182 encapsulated electrode was delivered directly from the PCB. The on board stimulation circuit had a 183 resolution of 33µA, which was too large to accurately fit sigmoid functions to the fiber recruitment 184 curve in most cases. Values for current were randomly interleaved. Stimulation was delivered every 5 185 seconds and each parameter was repeated in triplicate. Data was sampled at 500 Hz using the same load 186 cell collection system described above. were fitted with a sigmoid function (Fig. 1c). Restrictions were placed on the fitted curve such that the 193 point at 1% of Y max could not be at a negative current intensity. For each curve, the slope was calculated then the data were compared using a two tailed two-sample t-test with either equal or unequal variance 208 depending on the F-test. The statistical test used for each comparison is noted in the text. All 209 calculations were performed in MATLAB.

212
One-sided and circumferential electrodes provide equivalent recruitment 213 of rat sciatic nerve 214 Flat electrode contacts that do not surround the entire nerve may yield less efficient fiber recruitment 215 than circumferential electrode contacts. We tested recruitment efficacy using computational modeling 216 and in vivo experiments on the rat sciatic nerve. To represent flat electrodes, we used a modified 217 circumferential electrode that only provided 60° of coverage compared to the standard 270° (Fig. 3b).

218
Fiber recruitment functions were created using the 60° and 270° designs as well as an intermediate 120° with a shorter arc just as it does a standard electrode (Fig. S1).

240
To confirm modeling predictions, we evaluated nerve recruitment in the rat sciatic nerve using the 60°, 241 120°, and 270° electrodes. In vivo data closely resembled data derived from the model, with flat and 242 circumferential contacts demonstrating comparable fiber recruitment. No significant differences were 243 found between recruitment thresholds for any of the electrode configurations ( Fig. 3d; 60° These results confirm that the one-sided electrodes and circumferential electrodes yield equivalent nerve 253 recruitment across a range of stimulation intensities.

255
One-sided electrodes recruit more efficiently than circumferential 256 electrodes in the rat vagus nerve 257 We next tested recruitment using the same 60° and 270° cuff electrodes on the rat vagus nerve, which 258 has a smaller diameter and different fascicular organization.   We next sought to confirm these findings in vivo. To evaluate activation of vagus nerve fibers, we 279 measured rapid stimulation-dependent reduction in oxygen saturation, a well-described biomarker of Reducing the angle has no effect. c) Recruitment curves generated using a cuff with a 1 mm inner diameter, but with the nerve on the opposite side of the cuff lumen from the contacts. Reducing the angle decreases recruitment. d) Recruitment curves generated by modeling cuff electrodes with various angles of completion around the rat vagus. Instead of a 1 mm inner diameter, the cuff diameter was set to 0.44 mm to keep the ratio of the cuff diameter to nerve the same as in the sciatic model. When the cuff is sized to fit the nerve, reducing the angle has little effect on fiber recruitment. 17 280 vagus nerve stimulation ascribed to activation of the Hering-Breuer reflex (32). Stimulation of vagal A-281 fibers, including the pulmonary stretch receptors, temporarily prevents inhalation and causes blood 282 oxygen saturation to fall (Fig. 5a) (31). As a result, measurement of oxygen saturation provides a simple 283 means to assess vagal A-fiber recruitment.

284
The 60° electrodes recruited fibers more effectively than the 270° electrodes, corroborating findings 285 from the model. A trend toward reduced threshold was observed with the 60° electrode, although this  The results presented above support the notion that flat electrodes provide at least as effective fiber 299 recruitment as circumferential electrodes. However, whereas the 60° electrodes used in the above 300 experiments contact only a single side of the nerve similar to a flat electrode, they are not truly flat and 301 thus do not capture all the features of the geometry that may influence fiber recruitment. Therefore, we 302 sought to confirm these results using a true flat electrode. The electrode was manufactured on a printed 303 circuit board (PCB), encapsulated in glass, and inserted into a silicone sleeve that acted as an insulating 304 cuff (Fig. 8c). These electrodes were tested on the rabbit sciatic nerve, which is an order of magnitude 305 larger than the rat sciatic nerve (16,18).

307
We performed modeling to evaluate recruitment using flat contacts and circumferential contacts. The similar in all cases (Fig. 7b). A helical cuff electrode, like those used in clinical VNS applications, 319 provides similar recruitment to the flat electrode design (Fig. S2). On a multi-fascicle nerve, flat and 320 circumferential electrodes provided similar recruitment of the whole nerve despite each fascicle being 321 recruited differently. These data suggest that flat electrodes recruit fibers similarly to currently used 322 electrode designs.  In this study, we examined the viability of flat electrodes for nerve stimulation. Circumferential 336 electrode contacts were compared to flat electrode contacts on multiple nerves and in multiple species. 337 We find that in all cases tested, recruitment is either equivalent or the flat contacts have a steeper 338 recruitment function and lower saturation current.

339
Flat electrodes that only contact a single side of the nerve could provide multiple advantages over the 340 currently used circumferential designs which wrap around the majority of the circumference of the 341 nerve. However, flat electrodes may require more current to achieve the same amount of fiber  Additionally, injection current density was higher with the 60° electrodes given their reduced surface 356 area compared to the 270° electrodes (41). Due to the small size of the nerve relative to the cuff, its 357 position, and the increased current density near the contacts present with the 60° design, the current 358 density within the nerve was higher with the smaller contact angle. Model results suggest that this is 359 only true when the nerve is at the bottom of the cuff and the cuff is significantly larger than the nerve. If 360 the nerve was moved to the opposite side of the cuff, far away from the contacts, the opposite 361 relationship was seen (Fig. 4c), and if the cuff was sized appropriately for the vagus, the 60° contacts did 362 not appear significantly different from the 270° contacts (Fig. 4d). Regardless, the modeling and 363 empirical data support the notion that flat electrodes provide at least equivalent fiber recruitment. and circumferential electrodes to investigate how each design is affected by the orientation of the nerve 383 inside the cuff. The steepness of the recruitment curve for single fascicles was similar in all cases, but 384 the variance in thresholds was greater with the flat electrodes (Fig. 9), which provides an explanation for 385 the variance in thresholds present empirically. Fascicles on the opposite side of the nerve from the flat 386 electrodes will have a higher threshold of activation, but the threshold is still comparable to 23 387 circumferential electrodes, which suggests that clinical efficacy will not be reduced for applications in 388 which whole nerve recruitment is desired.

390
All modeling and in vivo experiments in this study measured A-fiber recruitment, but many applications 391 of nerve stimulation rely on B-and C-fibers as well (43). It is not guaranteed that flat and 392 circumferential electrodes will recruit these other fiber types equivalently, but models of smaller 393 diameter fibers suggest that the increase in fiber recruitment threshold would scale proportionally 394 between the two electrode designs. Thus, while more current is required to activate smaller diameter 395 fibers, we predict that the increase in current is likely to be similar comparing flat and circumferential  Additionally, there is a lack of data on larger diameter nerves (>3 mm) that would be found in humans.

408
Our modeling studies suggest that flat and circumferential electrodes are equivalent on a range of 24 409 clinically applicable nerve sizes. Moreover, comparison of recruitment in the rat sciatic and rabbit sciatic 410 suggests that larger nerves require more current to achieve the same level of activation, but in both cases 411 recruitment is comparable between flat and circumferential electrodes.

412
Our finding that larger nerves require more current to achieve similar levels of activation is intuitive. finding, which appears to lie in the use of tight-fitting stimulating electrodes for human studies and 418 poorly fitting, oversized cuff electrodes for rat studies. When we modeled these configurations, we 419 confirmed that identical VNS parameters can equivalently activate nerves of very different diameters 420 under these conditions (Fig. 10). This is a novel result that could substantially impact both preclinical 421 and clinical stimulation parameters. Follow up studies comparing small diameter nerves in animals to 422 large diameter nerves in humans should be done to confirm this finding.

423
Fig 10: Recruitment of vagus nerve is similar in humans and rats due to cuff electrode design. Larger nerves require more current to recruit, but the therapeutic range of vagus nerve stimulation is similar in rats and humans (Fig. 7b). This phenomenon can be explained by the use of tight-fitting stimulating electrodes for human studies and poorly fitting, oversized cuff electrodes for rat studies. Cuff electrodes used in rats are significantly larger than the nerve which leads to inefficient recruitment and brings the two curves into alignment. If rat cuff electrodes were reduced in size, recruitment would be greatly increased. This is consistent with the importance of the ratio of cuff inner diameter to nerve diameter (Fig. 2a).
If a flat electrode design is used to stimulate the vagus nerve for epilepsy, it is not initially clear whether 425 the stimulation parameters would be different from current clinical parameters, as a helical electrode 426 design is used in clinical applications rather than a cuff electrode (49). We modeled the helical electrode 427 design and compared vagus nerve recruitment to recruitment using a flat electrode design. The helical 428 electrode and flat electrode demonstrated comparable recruitment. The narrow insulating structure used 429 by the helical cuff allows some current to bypass the nerve, which increases the amount of stimulation 430 needed compared to a complete cuff electrode (Fig. S2). The open architecture of the helical cuff is 431 equivalent to having very little cuff overhang, which decreases recruitment compared to a full cuff (Fig.   432 2c).