Electrode Array for Reversing the Recruitment Order of ... - Springer Link

C 2006) pp. 152–160 Annals of Biomedical Engineering, Vol. 34, No. 1, January 2006 ( DOI: 10.1007/s10439-005-9012-5

Electrode Array for Reversing the Recruitment Order of Peripheral Nerve Stimulation: Experimental Studies ZENG LERTMANORAT, KENNETH J. GUSTAFSON, and DOMINIQUE M. DURAND Neural Engineering Center, Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH (Received 10 January 2005; accepted 4 August 2005; published online: 2 February 2006)

ameter fibers are always recruited before smaller ones as stimulation amplitude increases. Selective activation of small fibers is desirable in several applications. In motor control applications, a recruitment order from large to small fibers is opposite to the physiological recruitment order of motor fibers during voluntary muscle contraction35 and results in poor grading of muscle force and rapid onset of muscle fatigue.21 In neural prostheses for bladder control, selective activation in the sacral roots of small bladder efferent fibers without activation of larger fibers innervating the urethral sphincter could produce voiding.2,12,26 Several stimulation techniques for selective activation of small fibers have been proposed; high frequency block,1,34 anodic block,2,6,12,26,29 subthresholddepolarizing prepulse,5,9,30,37 and single cathode.33 These techniques, however, require high charge injection, which could lead to electrode corrosion.27 The number of stimulation pulses for the high frequency block (0.6–20 kHz)32 is at least 20 times more than conventional stimulation frequency (25 Hz). Anodic block and single cathode require a long-duration stimulation pulse (>500 µs) and suprathreshold stimulation amplitude (∼3 times). Previous studies in computer simulations showed that selective activation of small myelinated axons could be achieved by manipulating the extracellular voltage profile along the nerve (Ve ).13,14 The technique relied on the fact that excitability of myelinated axons is proportional to the second difference of Ve (so called the activating function)18 and that the internodal distance of myelinated axons is proportional to the axon diameter.11 By reshaping the Ve profile along the axons such that the activating function of small axons became larger than that of large axons, small axons were recruited with lower stimulation amplitude than large axons.13,14 The Ve profile was generated by an array of alternating anodes and cathodes placed along the nerve. Arrays of 5, 7, and 11 contacts were tested. The 5-contact array with 0.75 mm contact separation recruited axons having diameter smaller than 7.5 µm before larger ones, whereas arrays of 7 and 11 contacts with the same contact separation suppressed

Abstract—One of the most challenging problems in peripheral nerve stimulation is the ability to activate selectively small axons without large ones. Electrical stimulation of peripheral nerve activates large diameter fibers without small ones. Currently available techniques for selective activation of small axons before large ones require long-duration stimulation pulses (>500 µs) and large stimulation amplitude, which shorten battery life of the implanted stimulator and could lead to electrode corrosion. In the current study, the hypothesis that small axons can be recruited before large ones with narrow pulsewidth (50 µs) using an electrode array was tested in both computer simulations and experiments in the cat lateral gastrocnemius (LG) model. The LG nerve innervates both LG and soleus muscle groups with axons within 10–13 and 8–12 µm diameter ranges, respectively. A finite element model of LG nerve was constructed and simulations showed that, when activating 40% of LG, a conventional tripolar electrode activated only 9% of soleus whereas the electrode arrays of 5, 7, and 11 contacts activated 39, 46, and 60% of soleus respectively, suggesting that the arrays could activate small axons before fully recruiting large axons. In animal experiments, peak twitch force of LG and soleus were plotted as a function of stimulation amplitude to indicate the recruitment curve. At 40% activation of LG, a conventional tripolar electrode activated only 7% of soleus whereas the electrode arrays of 5, 7, and 11 contacts activated 43, 48, and 72% of soleus respectively. The electrode arrays also decreased significantly the recruitment curve slopes to only 10–20% of the value obtained for the tripolar electrode in both computer simulations and experiments. In conclusion, the 5-, 7-, and 11-contact arrays can be used to reverse the recruitment order of peripheral nerve stimulation with a narrow pulse. Keywords—Neural prostheses, Recruitment order, Selective neural stimulation.

INTRODUCTION Extracellular electrical stimulation of peripheral nerves has been used to restore lost functions such as hand grasping, standing, walking, breathing, and bladder emptying to paralyzed individuals.22 One of the most challenging problems in peripheral nerve stimulation is that large diAddress correspondence to Dominique M. Durand, Neural Engineering Center, Department of Biomedical Engineering, Case Western Reserve University, Wickenden Bldg, #112, Cleveland, OH, 44106. Electronic mail: [email protected]

152 0090-6964/06/0100-0152/0

C 2006

Biomedical Engineering Society

Electrode Array for Reversing the Recruitment Order of Peripheral Nerve Stimulation

153

the excitability of axons within 15 ± 4 and 15 ± 2 µm diameter ranges, respectively.14 The diameter selectivity was independent of stimulation pulse width. However, the effectiveness of the electrode array has never been validated experimentally. In the present study, the effect of electrode array stimulation on axonal excitation was examined in computer simulations and animal experiments. Recruitment orders of conventional tripolar electrode and electrode arrays of 5, 7 and 11 contacts were compared. Computer simulations were performed first to predict the recruitment characteristics for the electrode configurations. The recruitment curves were then measured experimentally to test the ability of the electrode array to recruit selectively small diameter axons. The preliminary results of this manuscript have been published in abstract form.15,16 METHODOLOGY Animal Model The lateral gastrocnemius (LG) branch of cat sciatic nerve was used to test the effect of electrode array stimulation on axonal excitation. The LG branch innervates both LG and soleus muscles [Fig. 1(a)]. The cross-section of LG branch is shown in Fig. 2(a) and contains four fascicles. Three small fascicles (L1 –L3 ) innervate LG muscle and the large fascicle innervates soleus. The LG muscle has fast twitch fibers innervated by large axons,3 whereas soleus has mainly slow twitch fibers innervated by smaller axons.20 The fiber diameter distribution of the LG and soleus

Figure 2. Cat Lateral Gastrocnemius (LG) nerve anatomy and fiber diameter distributions. (a) Cross-section of LG nerve after reshaping. LG consists of 4 fascicles. (b) Fiber diameter distribution in LG fascicles. Motor fibers of LG are in 10–13 µm diameter range. (c) Fiber diameter distribution in soleus fascicle. Motor fibers of soleus are in 8–12 µm diameter range.

Figure 1. Experimental setup: An electrode array was placed on the lateral gastrocnemius branch (LG) that innervates both LG and soleus muscles. Tendons of LG and soleus were separated and attached to two force transducers.

nerve branches was obtained and used in the computer simulations. The LG and Soleus nerve branches were taken bilaterally from a cat and fixed with 10% formalin for 3 days. The nerves were then stained with methylene blue, and one-µm thick nerve cross-sections were obtained. The fiber diameter distributions of each LG and soleus fascicle were automatically measured28 and are shown in Fig. 2(b) and (c). The distribution is consistent with that measured by Romero28 in that the fiber distribution of cat sciatic nerve is bimodal with peaks at 6 and 12 µm. Gamma efferent and secondary afferent fibers that do not generate force are ∼6 µm in diameter, whereas alpha efferent and primary afferent fibers are ∼12 µm in diameter. The alpha efferent fibers of LG (9–14 um) are larger than those of soleus (8–12 um) with some overlapping.

154

LERTMANORAT et al.

Computer Simulation Finite Element Model of LG Branch A three-dimensional finite element model of the LG nerve was implemented using a finite element package (Maxwell 3D-field simulator, Ansoft) and ∼100,000 tetrahedra at 2% energy error (Fig. 3). The model consisted of four fascicles (the endoneurium surrounded by the perineurium), epineurium, stimulating contacts, insulating electrode, and saline (volume conductor) [Fig. 3(a)]. The nerve had an oval shape with a thickness of 0.4 mm. L1 –L3 and S represented the fascicles for LG and soleus respectively. The perineurium thickness was set at 10 µm for all four fascicles based on measurements from the crosssections. Four electrode configurations were implemented; tripolar electrode and electrode arrays of 5, 7, and 11 contacts. All electrodes had stimulating contacts located on both the top and the bottom with an opening of 0.5 mm to

Table 1.

Resistivities of the compartments in the volume conductor model.

Compartments Endoneurium Longitudinal Transverse Perineurium Epineurium Stimulating contacts Insulating electrode Volume conductor

Resistivity

Ohm × cm

Cat dorsal column23

Frog38 Transverse-endoneurium4 Platinum Silicone 1% Saline (0.9 M NaCl)7

1.75 × 102 1.21 × 103 4.78 × 104 1.21 × 103 1.00 × 10−5 1.00 × 109 5.00 × 101

limit the range of axon-contact distance within 0.25 mm [Fig. 3(b)]. For the tripolar electrode, the contact separation was set equal to 3.25 mm and the total length of the insulating electrode was 7.8 mm. For the electrode arrays, the contact separation was set to 0.65 mm and the length of insulating electrode depended on the number of stimulating contacts (3.9–7.8 mm). The 0.65 mm contact separation was chosen in this particular study in order to suppress the excitability of axons having diameter around 11–15 µm in LG.13,14 The stimulating contacts were modeled as voltage source with an exposed area of 0.1 mm × 3 mm. The size of the volume conductor was 50 mm long and 30 mm in diameter. The conductivities of compartments within the model are listed in Table 1. Axon Population The axon diameter distribution was based on that measured from the cross-section of LG branch [8–14 µm range in Fig. 2(b) and (c)] consisting of 123, 222, 125, and 283 axons for L1 –L3 and soleus respectively. All axons were used in the simulations since the ratio of alpha efferent and primary afferent fibers in this axon population is unknown and percent recruitment was the output measure. All axons were positioned randomly within the fascicles. The positions of nodes of Ranvier with respect to the contacts were distributed uniformly. The voltage within the fascicles calculated with the finite element analysis was used in a numerical integration to calculate the membrane potential of each axon. The electrical properties of axons were obtained from the human node of Ranvier first reported by Schwarz et al.31 and modified by McIntyre and Grill.24 The numerical integration was performed using a first order Backward Euler method with a time step equal to 0.5 µs. All equations are described in a previous study.13

Figure 3. Finite element model of LG nerve and stimulation electrode. (a) The LG nerve model. The model consisted of four fascicles (endoneurium surrounded by perineurium), epineurium, stimulating contacts, silicone cuff, and saline volume conductor. (b) Cross-section of the electrode: Stimulating contacts were located on both top and bottom sides, limiting the axon-contact distance within 0.25 mm.

Recruitment Order The recruitment curve generated in the model was defined as the numbers of recruited axons as a function of stimulation amplitude. To evaluate the effect of electrode configuration on axonal excitability, recruitment curves


155

of LG and soleus for the four electrode configurations were compared. Ten simulations were performed for each electrode configuration with random axon positions. In each simulation, random variations (max = 20%) were added to the stimulation voltage at each of contacts in the arrays to represent the errors in the voltage amplitude caused by factors such as electrode impedance and tissue inhomogeneities as expected in an in-vivo preparation. Stimulation pulses were asymmetrically charge-balanced biphasic; a 50 µs cathodic pulse followed by a 250 µs anodic pulse with 20% stimulation amplitude of the cathodic phase.14 Experimentation Animal Preparation Acute experiments were conducted on five cats (n = 8 legs) bilaterally weighing between 3.5 and 4.2 kg. The animals were anesthetized with ketamine (35 mg/kg i.m.) and maintained with sodium pentobarbital as needed (5 mg/kg i.v. per dose). Tendons of the LG and soleus muscles were exposed, separated, detached from the calcaneus and attached to two individual force transducers [Fig. 1(a)]. The femur was fixed in a rigid frame. A stimulation electrode was placed on the LG branch that was approximately 10 mm in length. The tibial and medial gastrocmenius nerve branches were cut to prevent interference from other muscles. The exposed nerve and muscle were bathed in NaCl solution (0.9%) and kept warm with a heating lamp. Electrode Fabrication A bipolar and electrode arrays of 7 and 11 contacts were fabricated using silver wire (Medwire: AG 5T) embedded in medical-grade silicone (Nusil: MED-4210). The bipolar electrode had stimulating contacts only on one side (8 mm separation) with an opening of 2 mm × 2 mm and 10 mm in length. Contact separation of the arrays was set equal to 0.65 ± 0.05 mm. The electrode arrays had stimulating contacts located on both top and bottom with an opening of 0.5 mm × 3 mm (Fig. 3). All contacts were made of silver wire coated by AgCl layer to minimize the capacitive component of the contact impedance. Impedance of all contacts was approximately 500 . Experiment Procedure Muscle twitches were induced by stimulation pulses with varying voltage amplitude (Fig. 4). Maximum twitch peaks for LG and soleus were around 15N and 5N respectively. For each stimulation amplitude, five successive twitches with 4-s interval were recorded to obtain an average twitch peak. The voltage amplitude was adjusted to achieve approximately 15–20% increment step of the twitch force and the order of voltage amplitude was randomly selected.

Figure 4. Profiles of twitch forces for LG and soleus from one of the experiments. LG has fast twitch muscle properties (narrow profile), whereas soleus has slow twitch muscle properties (wide profile). The twitch force was normalized and the peak force is indicated.

Recruitment curves were obtained by measuring the twitch force as a function of stimulation amplitude. A bipolar electrode with large opening height (2 mm) was first used to estimate the maximum twitch force before the nerve was reshaped by the array [fascicles oriented as shown in Fig. 2(a)]. The array electrode with small opening height (0.5 mm) was then placed on the nerve. The tripolar stimulation was first applied using contacts in the middle and at the ends of an electrode array as a cathode and reference grounds respectively. Electrode array tests were then carried out, and the tripolar stimulation was performed again to confirm the consistency of the data. Stimulation pulses were asymmetrically charge-balanced biphasic: a narrow cathodic pulse (50 µs) followed by a long anodic pulse (250 µs) with 20% stimulation amplitude of the cathodic phase.14 Before each recruitment curve, LG and soleus muscle lengths were adjusted to produce maximum twitch force in response to suprathreshold stimulation pulses. At the end of each recruitment curve, suprathreshold stimulation pulses were applied to verify that the recorded data were not affected by muscle fatigue or nerve damage. Statistical Analysis The recruitment order was determined by comparing the recruitment levels of LG and soleus. The percent activations of LG and soleus were compared statistically when LG was activated at about 40% (p40 ) and 80% (p80 ). The recruitment orders for tripolar and electrode array configurations were compared (Table 2, Figs. 5 and 6). RESULTS Recruitment Order The effect of electrode configuration on axonal excitability was analyzed by comparing calculated recruitment

156

LERTMANORAT et al. Table 2.

Comparison of recruitment order among different electrode configurations. LG = 40%

Simulations Tripolar (%) Array-5 (%) Array-7 (%) Array-11 (%) Experiments Tripolar (%) Array-5 (%) Array-7 (%) Array-11 (%)

LG = 80%

n

9.0 39.1 43.5 60.7

± ± ± ±

1.1 18.4 21.2 18.7

50.5 77.7 85.2 93.4

± ± ± ±

3.2 7.6 8.3 3.2

10 10 10 10

7.6 43.7 48.7 72.8

± ± ± ±

7.4 11.6 10.3 25.7

46.9 77.7 92.7 97.0

± ± ± ±

20.5 13.8 6.4 4.3

8 3 3 4

Note. Percent recruitment levels of soleus at stimulation amplitudes recruiting around 40 and 80% of LG are indicated. In both computer simulations and experiments, the tripolar electrode activated LG before soleus, whereas the 5-contact array shifted the soleus recruitment curve to overlap with LG and the 7- and 11-contact arrays reversed the recruitment order.

curves for four electrode configurations: tripolar electrode and electrode arrays of 5, 7, and 11 contacts (Fig. 5). The tripolar electrode recruited 9.0 ± 1.1 and 50.5 ± 3.2% of soleus while recruiting 40 and 80% of LG, respectively as shown in Fig. 5(a) (p40 and p80 < 0.00001). Therefore, LG was easier to excite than soleus as expected due to the larger fiber diameters. For the 5-contact array, the soleus and LG recruitment curves overlapped [Fig. 5(b)], suggesting that the array recruited some small axons of soleus (8–12 µm) at lower voltage levels than required to fully recruit larger axons of LG (10–13 µm). For the 7-contact array, soleus was slightly more excitable than LG [Fig. 5(c)]. At 40 and 80% LG activation levels, the 7-contact array recruited 43.5 ± 21.2 and 85.2 ± 8.3% of soleus. The 11-contact array recruited small axons in soleus before larger axons in LG. At 40 and 80% LG activation levels, the 11-contact array recruited 60.7 ± 18.7 and 93.4 ± 3.2% of soleus [Fig. 5(d)], respectively (p40 and p80 < 0.005). These results indicate that electrode arrays could induce a recruitment

Figure 5. Mean and standard deviation of recruitment of LG and soleus induced by four electrode configurations in computer simulations. (a) Tripolar electrode: LG was recruited at lower stimulation amplitude than soleus. (b) Array of 5 contacts: LG and soleus recruitment curves were overlapped. (c) Array of 7 contacts: soleus was slightly more excitable than LG. (d) Array of 11 contacts: Soleus was excitable than LG. Electrodes arrays could activate some small axons (in soleus) before large ones (in LG) and the effectiveness of the array increased as the number of stimulating contacts increased (n = 10).


157

Figure 6. Mean and standard deviation of recruitment of LG and soleus induced by four electrode configurations from two experiments: “Tripolar and Array-11” and “Array-5 and Array-7” were obtained from different experiments. (a) Tripolar electrode: LG was recruited at lower stimulation amplitude than soleus. (b) Array of 5 contacts: LG and soleus recruitment curves were overlapped. (c) Array of 7 contacts: soleus was slightly more excitable than LG. (d) Array of 11 contacts: Soleus was excitable than LG. Electrodes arrays could activate some small axons (in soleus) before fully recruiting large ones (in LG) and the array-7 was more effective than the array-5.

order from small to large diameters and that the 7- and 11-contact arrays are more effective than the 5-contact array. Experimental results were consistent with the computer simulations. Examples of experimental recruitment curves are showed in Fig. 6. “Array-5 and Array-7” and “Array11” were obtained from two different experiments. At 40 and 80% of LG recruitment levels, the percent recruitments of soleus and LG were compared for all experiments in Table 2. Tripolar stimulation recruited LG before soleus [Fig. 6(a)]. At 40 and 80% LG activation levels, tripolar stimulation recruited only 7.6 ± 7.4 and 46.9 ± 20.5% of soleus (p40 < 0.0001 and p80 = 0.0015). The 5-contact array shifted the soleus recruitment curve to overlap with that of LG [Fig. 6(b)]. For the 7- and 11-contact arrays, soleus was more excitable than LG [Fig. 6(c) and (d)]. At 40 and 80% LG activation levels, the 7-contact array recruited 48.7 ± 10.3 and 92.7 ± 6.4% of soleus (p80 = 0.05), and the 11-contact array recruited 72.8 ± 25.7 and 97.0 ± 4.3% of soleus (p40 = 0.07, p80 < 0.00001), respectively. The 7- and 11-contact arrays were more effective than the 5-contact array in recruiting small axons.

Slope of Recruitment Curve The effect of the electrode array on the recruitment curve slope was investigated by comparing slopes for all four electrode configurations (Table 3). The average slope was calculated between 20 and 80% recruitment levels. Slopes

Table 3. Average slope of LG and soleus (% activation per 10 mV stimulation) LG Simulations Tripolar Array-5 Array-7 Array-11 Experiments Tripolar Array-5 Array-7 Array-11

n

Soleus

62.1 2.3 3.5 5.8

± ± ± ±

1.8 0.6 1.1 0.8

60.1 2.4 4.9 11.1

± ± ± ±

3.1 1.1 1.4 3.0

10 10 10 10

18.1 2.7 2.1 2.5

± ± ± ±

8.8 1.2 0.2 1.2

25.6 2.4 4.7 6.3

± ± ± ±

10.6 0.8 4.1 3.4

8 3 3 4

Note. Slope for electrode array stimulations was at least four times lower than that of tripolar electrode in both computer simulations and experiments.

158

LERTMANORAT et al.

of all three electrode array configurations were lower than those of tripolar electrode. In computer simulations, the average slope of LG for the tripolar configuration was 62.1 ± 1.8% per 10 mV and at least 10 times higher than those of electrode arrays (p < 0.000001); 2.3 ± 0.6 (Array5), 3.5 ± 1.1 (Array-7), and 5.8 ± 0.8 (Array-11). The slope of soleus recruitment curve for the tripolar configuration was 60.1 ± 3.1 and also significantly higher than those of electrode arrays; 2.4 ± 1.1 (Array-5), 4.9 ± 1.4 (Array-7), and 11.1 ± 3.0 (Array-11). Similar results were observed in the experiments. The average experimental slope of LG for the tripolar configuration was 18.1 ± 8.8% per 10 mV and almost seven times higher than those of electrode arrays (p = 0.000004); 2.7 ± 1.2 (Array-5), 2.1 ± 0.2 (Array-7), 2.5 ± 1.2 (Array11). The slope of soleus for the tripolar configuration was 25.6 ± 10.6 and also significantly higher than those for electrode arrays; 2.4 ± 0.8 (Array-5), 4.7 ± 4.1 (Array-7), and 6.3 ± 3.4 (Array-11). All three electrode arrays decreased the recruitment curve slope when compared to tripolar electrode.

DISCUSSION AND CONCLUSION This study shows that selective activation of small axons can be achieved using an electrode array placed along the nerve. One advantage of this approach over other techniques is that short-duration stimulation pulses can be used. Currently proposed techniques require a long duration stimulation pulse (>500 µs)2,5,6,9,12,26,29,33,37 and large stimulation amplitude,2,6,12,26,29,33 which can cause electrode corrosion.27 In this study, electrode arrays of 5, 7, and 11 contacts were activated with 50 µs stimulation pulses in both computer simulation and experimentations. Computer simulations predicted that the 5-contact array would activate evenly soleus (8–12 µm) and LG (9–14 µm) and that the arrays of 7 and 11 contacts would activate soleus before LG (Fig. 5). These results were consistent with the results of a previous modeling study.14 That is, the 5-contact array with 0.65 mm separation activated axons having internodal distance close to or shorter than the contact separation (≤6.5 µm diameter in this case), and axons larger than 6.5 µm were equally less excitable compared to the smaller ones. The comparable excitability between soleus and LG for the 5-contact array can be explained by the fact that their axon diameter ranges are within the suppression range of the 5-contact array (>6.5 µm). The electrode arrays with 7 and 11 contacts were capable of suppressing selectively the excitability of axons having internodal distance close to the intercathodic distance: 13 ± 4 and 13 ± 2 µm for the arrays of 7 and 11 contacts, respectively.14 The 7- and 11contact arrays suppressed specifically the LG excitability (13 ± 2 µm) and therefore recruited soleus at lower stimulation amplitudes than LG.

The ability of the electrode array to selectively activate small axons was validated successfully in animal experiments (Fig. 6). In every experiment, tripolar configuration activated LG before soleus. The arrays of 5 and 7 contacts shifted and reversed this recruitment order reproducibly and this effect was confirmed by switching among the 5-contact array, 7-contact array and tripolar stimulations in the same preparation by using different contact configurations without changing the electrode. The 11contact array was also capable of recruiting soleus before LG; however, the effect was variable within an experiment and not as highly reproducible as those of the 5and 7-contact arrays. The 11-contact array is 50% longer than the 7-contact array, making it difficult to implant the electrode while keeping the LG nerve in an ideal straight position due to the space constraint. The cause of the variability could also be attributable to the difficulty of maintaining a complex voltage profile along the nerve and the impedance imbalance among contacts. Since experiments with the 11-contact array stimulation were not performed in the same experiments as the 5- and 7contact arrays, their effectiveness could not be compared directly. Nonetheless, experimental results indicate that selective activation of small axons can be achieved using an electrode array with a number of contacts as low as five. All three electrode array configurations decreased the recruitment curve slope (Table 3). A low slope recruitment curve is an important characteristic for motor control.8,19 Although the electrode arrays required larger stimulation amplitudes than the tripolar electrode, the total charge injection of the array should be less than that by other stimulation techniques proposed to selectively activate small axons. Anodic-block and single-cathode techniques require supra-threshold stimulation amplitude (3–4 times) and long duration pulse (>500 µs),2,6,12,26,29,33 making the electrode array preferable for chronic implant in terms of electrode safety and power consumption. Although the recruitment curves from computer simulations and experiments were similar qualitatively, the two sets of data could not be compared directly. Computer simulations used numbers of recruited axons, whereas the experiments used muscle force, to determine recruitment order. The twitch peak is an indirect measurement of the recruitment order, and the relation of number of axons to generated force is not linear or well defined. However, computer simulations were essential for evaluating the feasibility of the electrode array to selectively activate small axons before conducting the experiments in this particular LG-soleus model where diameter distribution of small and large axons overlap. Although axon populations in the simulations included both efferent and afferent fibers, the numbers of recruited axon were normalized, thus eliminating the effect of relative percentage of efferent and afferent fibers on the recruitment order.


This study showed experimentally for the first time that selective activation of small axons could be achieved with narrow stimulation pulses (50 µs). Electrode array stimulations with long duration pulse (200 µs) were also performed in a limited number of experiments (data not shown) to evaluate the effect of stimulation pulse width on the electrode array. No difference in recruitment order was observed between the two pulse widths and the stimulation amplitudes for the 200 µs pulse were only 10% lower than those for the 50 µs pulse. The small amplitude drop could be due to the characteristic of voltage source stimulation. When a rectangular voltage pulse is applied, the actual voltage shape delivered to the tissue is not rectangular, but drops as a function of time due to the capacitive component of the stimulating contact,39 implying that only the narrow section at the beginning of voltage pulse contributes to the neural excitation. In spite of the capacitive component of the stimulation contact, voltage sources were used rather than current sources in this study. Voltage sources can control the voltage profile inside the insulating electrode more accurately than the current sources since voltage sources eliminate the effect of the variation of impedance of the volume conductor within the electrode. However, this method of stimulation requires that all stimulating contacts have similar contact impedances. For the Ag/AgCl electrodes used in this study, the impedance of individual contacts was adjusted by controlling the thickness of AgCl layer. Moreover, the AgCl layer also reduces the capacitance of the contacts. Ag/AgCl is not suitable for human applications and an alternative electrode to control extracellular voltage profile is required. Only one contact separation (0.65 mm) of the arrays was tested in this study. The 0.65 mm contact separation was chosen in order to suppress the excitability of axons with diameters around 11–14 µm (LG). With contact separations between 0.5–1.5 mm, it is likely that the 5-contact array would induce a similar recruitment order between LG and soleus since the 5-contact array is not as effective at selectively activating small axons when the diameter distributions of small and large axons are not clearly separated based on a previous study.14 On the other hand, the arrays of 7 and 11 contacts with smaller contact separation (0.4–0.5 mm) are likely to suppress the excitability of soleus, whereas the arrays with larger contact separation (>0.8 mm) are not likely to suppress any axons. In either case, LG would be more excitable than soleus and the recruitment order would be similar to that obtained with the tripolar electrode. The effective axon-contact range for the array stimulation needs to be investigated. A previous study suggested that the electrode array would be more effective if the axon-contact distance was minimized.13 The electrode array with only one opening height was tested in this study (0.5 mm). The electrode slightly reshaped the nerve to a flat geometry to minimize the axon-contact distance. The

159

safety of nerve reshaping has been established in chronic animals.17,36 This study demonstrates that the electrode array can alter the recruitment order of peripheral nerve stimulation. The gastrocnemius and soleus models are appropriate and have been used1 to examine the physiological recruitment order of the motor control system.10 However, the overlapping LG and soleus fiber diameter distributions did not permit the analysis of the fiber diameter resolution of this approach. The electrode array should be most effective for applications that attempt to reverse the recruitment order of distinctly separate fiber populations such as sacral ventral root stimulation for bladder control.25 In conclusion, this study shows that the axonal excitability and the recruitment order can be manipulated by reshaping the extracellular voltage profile along the nerve with the 5-, 7-, and 11-contact arrays. The 5- and 7-contact arrays are preferable to the 11-contact array due to the simplicity. The 7-contact array is more effective than the 5-contact array when the fiber diameter distributions of small and large axons are not clearly separated. Diameter selective activation can be achieved with narrow stimulation pulse. The electrode array will be beneficial to neural prosthesis applications that require diameter selective stimulation such as motor control, bladder control. The electrode array also decreases the recruitment curve slope, providing a distinct advantage for neural control over other currently available electrodes. REFERENCES 1

Baratta, R., M. Ichie, S. K. Hwang, and M. Solomonow. Orderly stimulation of skeletal muscle motor units with tripolar nerve cuff electrode. IEEE Trans. Biomed. Eng. 36(8):836–843, 1989. 2 Bhadra, N., V. Grunewald, G. Creasey, and J. T. Mortimer. Selective suppression of sphincter activation during sacral anterior nerve root stimulation. Neurourol. Urodyn. 21(1):55–64, 2002. 3 Burke, R. E., and P. Tsairis. Anatomy and innervation ratios in motor units of cat gastrocnemius. J. Physiol. 234(3):749–765, 1973. 4 Choi, A. Q., J. K. Cavanaugh, and D. M. Durand. Selectivity of multiple-contact nerve cuff electrodes: A simulation analysis. IEEE Trans Biomed. Eng. 48(2):165–172, 2001. 5 Deurloo, K. E., J. Holsheimer, and P. Bergveld. The effect of subthreshold prepulses on the recruitment order in a nerve trunk analyzed in a simple and a realistic volume conductor model. Biol. Cybern. 85(4):281–291, 2001. 6 Fang, Z. P., and J. T. Mortimer. A method to effect physiological recruitment order in electrically activated muscle. IEEE Trans. Biomed. Eng. 38(2):175–179, 1991. 7 Geddes, L. A., and L. E. Baker. The specific resistance of biological material—A compendium of data for the biomedical engineer and physiologist. Med. Biol. Eng. 5(3):271–293, 1967. 8 Gorman, P. H., and J. T. Mortimer. The effect of stimulus parameters on the recruitment characteristics of direct nerve stimulation. IEEE Trans. Biomed. Eng. 30(7):407–414, 1983. 9 Grill, W. M., and J. T. Mortimer. Stimulus waveforms for selective neural stimulation. IEEE Eng. Med. Biol. Mag. 14(4):375– 385, 1995.

160 10

LERTMANORAT et al.

Henneman, E., G. Somjen, and D. O. Carpenter. Functional significance of cell size in spinal motoneurons. J. Neurophysiol. 28:560–580, 1965. 11 Hursh, J. M. Conduction velocity and diameter of nerve. Am. J. Physiol. 127:131, 1939. 12 Koldewijn, E. L., N. J. Rijkhoff, E. V. van Kerrebroeck, F. M. Debruyne, and H. Wijkstra. Selective sacral root stimulation for bladder control: Acute experiments in an animal model. J. Urol. 151(6):1674–1679, 1994. 13 Lertmanorat, Z., and D. M. Durand. A novel electrode array for diameter-dependent control of axonal excitability: A simulation study. IEEE Trans. Biomed. Eng. 51(7):1242–1250, 2004. 14 Lertmanorat, Z., and D. M. Durand. Extracellular voltage profile for reversing the recruitment order; of peripheral nerve stimulation: A simulation study. J. Neural Eng. 1:202–211, 2004. 15 Lertmanorat, Z., and D. M. Durand. Reversing the recruitment order with electrode array stimulation. In Proceedings—25th International Conference—IEEE/EMBS. 2003. Cancun, Mexico. 16 Lertmanorat, Z., K. J. Gustafson, and D. M. Durand. Orderly recruitment order of peripheral nerve stimulation with electrode array. In Proceedings—2th International Conference— IEEE/EMBS on Neural Engineering. 2005. VA, USA. 17 Leventhal, D. K., and D. M. Durand. Chronic measurement of the stimulation selectivity of the flat interface nerve electrode. IEEE Trans. Biomed. Eng. 51(9):1649–1658, 2004. 18 McNeal, D. R. Analysis of a model for excitation of myelinated nerve. IEEE Trans. Biomed. Eng. 23(4):329–337, 1976. 19 McNeal, D. R., L. L. Baker, and J. T. Symons. Recruitment data for nerve cuff electrodes: Implications for design of implantable stimulators. IEEE Trans. Biomed. Eng. 36(3):301–308, 1989. 20 McPhedran, A. M., R. B. Wuerker, and E. Henneman. Properties of Motor Units in a Homogeneous Red Muscle (Soleus) of the Cat. J. Neurophysiol. 28:71–84, 1965. 21 Mortimer, J. T. Motor prostheses. In: Handbook of Physiology: The Nervous System, edited by V. B. Brooks. Bethesda, MD: American Physiological Society, 1981, pp. 155–187. 22 Peckham, P. H., and G. H. Creasey, Neural prostheses: clinical applications of functional electrical stimulation in spinal cord injury. Paraplegia, 1992, 30(2): pp. 96–101. 23 Ranck, J. B., and S. L. Be Ment. Specific impedance of the dorsal columns of cat; An anisotropic medium. Exp. Neurol. 11:451–463, 1965. 24 Richardson, A. G., C. C. McIntyre, and W. M. Grill. Modelling the effects of electric fields on nerve fibres: Influence of the myelin sheath. Med. Biol. Eng. Comput. 38(4):438–446, 2000.

25

Rijkhoff, N. J., E. L. Kldewijn, W. d’Hollosy, F. M. Dbruyne, and H. Wijkstra. Morphometric data of canine sacral nerve roots with reference to electrical sacral root stimulation. Neurourol. Urodyn. 15(3):235–248, 1996. 26 Rijkhoff, N. J., L. B. Hendrikx, P. E. van Kerrebroeck, F. M. Debruyne, and H. Wijkstra. Selective detrusor activation by electrical stimulation of the human sacral nerve roots. Artif. Organs 21(3):223–226, 1997. 27 Robblee, L. S., and T. L. Rose. Electrochemical guidelines for selection of protocols and elctrode materials for neural stimulation. In: Neural Prostheses: Fundamental Studies, edited by W. F. M. Agnew, D. B. McCreery. Englewood Cliffs, NJ: Prentice Hall, 1990, pp. 25–66. 28 Romero, E., O. Cuisenaire, J. F. Denef, J. Delbeke, B. Macq, and C. Veraart. Automatic morphometry of nerve histological sections. J. Neurosci. Methods 97(2):111–122,2000. 29 Rozman, J., B. Sovinec, M. Trlep, and B. Zorko. Multielectrode spiral cuff for ordered and reversed activation of nerve fibres. J. Biomed. Eng. 15(2):113–120, 1993. 30 Sassen, M., and M. Zimmermann. Differential blocking of myelinated nerve fibres by transient depolarization. Pflugers Arch. 341(3):179–195, 1973. 31 Schwarz, J. R., G. Reid, and H. Bostock. Action potentials and membrane currents in the human node of Ranvier. Pflugers Arch. 430(2):283–292, 1995. 32 Solomonow, M. External control of the neuromuscular system. IEEE Trans. Biomed. Eng. 31(12):752–763, 1984. 33 Tai, C., and D. Jiang. Selective stimulation of smaller fibers in a compound nerve trunk with single cathode by rectangular current pulses. IEEE Trans. Biomed. Eng. 41(3):286–291, 1994. 34 Tanner, J. A. Reversible blocking of nerve conduction by alternating-current excitation. Nature 195:712–713, 1962. 35 Tasaki, I. Nerveous Transmission. Springfield, IL: Charless C Thomas, 1953. 36 Tyler, D. J., and D. M. Durand. Chronic response of the rat sciatic nerve to the flat interface nerve electrode. Ann. Biomed. Eng. 31(6):633–642, 2003. 37 Van Bolhuis, A. I., J. Holsheimer, and H. H. Savelberg. A nerve stimulation method to selectively recruit smaller motor-units in rat skeletal muscle. J. Neurosci. Methods 107(1–2):87–92, 2001. 38 Weerasuriya, A., R. A. Spangler, S. I. Rapoport, and R. E. Taylor. AC impedance of the perineurium of the frog sciatic nerve. Biophys J. 46(2):167–174, 1984. 39 Weinman, J., and J. Mahler. An Analysis of Electrical Properties of Metal Electrodes. Med. Electron. Biol. Eng. 33:299–310, 1964.