706 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 31, NO. 7, JULY 1990

A Nerve Cuff Technique for Selective Excitation of Peripheral Nerve Trunk Regions

JAMES D. SWEENEY, MEMBER, IEEE, DAVID A. KSIENSKI, MEMBER, IEEE, A N D

J . THOMAS MORTIMER

Abstract-Numerical modeling and experimental testing of a nerve cuff technique for selective stimulation of superficial peripheral nerve trunk regions is presented. Two basic electrode configurations (“snug” cuff monopolar and tripolar longitudinally aligned dots) have been considered. In addition, the feasibility of “steering” excitation into superficial nerve trunk regions using subthreshold levels of current flow from an electrode dot located on the opposite side of the nerve has been tested. Modeling objectives were to solve for the electric field that would be generated within a representative nerve trunk by each electrode configuration; and to use a simple nerve cable model to predict the effectiveness of each configuration in producing localized excitation. In three acute experiments on cat sciatic nerve the objective was to char- acterize the effectiveness of each electrode configuration in selectively activating only the medial gastrocnemius muscle.

Modeling and experimentation both suggest that longitudinally aligned tripolar dot electrodes on the surface of a nerve trunk, and bounded by a layer of insulation (such as a nerve cuff), will restrict excitation to superficial nerve trunk regions more successfully than will monopolar dot electrodes. Excitation “steering” will improve the spa- tial selectivity of both monopolar and tripolar electrode configurations.

INTRODUCTION UNCTIONAL neuromuscular Stimulation (FNS) is a F technique whereby electrical excitation of neural tis-

sue is used to elicit artificial control of paralyzed muscu- lature. Stimulation of a myelinated motor nerve can be achieved, in general, if an electric field is introduced in the region of the nerve that sufficiently depolarizes the excitable membrane at a node of Ranvier. With excitation of a motor nerve an action potential will be generated that propagates towards a muscle. In spinal cord injured peo- ple most motorneurons in the spinal cord will remain in- tact that are below the level of the injury. Electrical stim- ulation of the motorneuron axons can therefore be used to cause muscle contraction. For restoration of lost motor function (e.g., locomotion for paraplegics or hand-grasp for quadriplegics) coordinated control over graded con- traction of many muscles is necessary. In this paper we present the results of a study on a nerve cuff technique for selective excitation of peripheral nerve trunk regions and, therefore, multiple muscles.

Manuscript received January 11, 1989; revised December 7, 1989. J. D. Sweeney was with the Applied Neural Control Laboratory, De-

partment of Biomedical Engineering, Case Western Reserve University, Cleveland, OH 44106. He is now with the Department of Chemical, Bio and Materials Engineering, Arizona State University, Tempe, A 2 85287.

D. A. Ksienski is with the Department of Electrical Engineering and Applied Physics, Case Western Reserve University, Cleveland, OH 44106.

1. T. Mortimer is with the Applied Neural Control Laboratory, Depart- ment of Biomedical Engineering, Case Western Reserve University, Cleveland, OH 44106.

IEEE t o g Number 9035443. p

Electrically induced stimulation of motor axons can be achieved with a variety of existing electrode types. Stim- ulation using electrodes located on the skin surface is technically straightforward but relatively nonselective be- cause the electric field introduced into the body is neces- sarily widespread. Intramuscular electrodes [E], which can be percutaneously placed or fully implanted, and im- planted epimysia1 electrodes [ 121 enable selective acti- vation of individual muscles (or muscle regions). FNS systems that use intramuscular or epimysial electrodes are presently limited by the need to implant, maintain and use a large number of electrodes if many muscles are to be controlled (e.g., see [20]). It would be highly desirable if a stimulation system could be designed that would con- trol the activation of several muscles with only a single implanted electrode device.

Consideration of the intraneural topography of periph- eral nerves provides insight into how this goal may be achieved. The nerve fiber arrangement within the course of a nerve trunk has been studied by dissection and his- tology, correlating electrical stimulation with peripheral response, partial nerve injury with deficit study, and par- tial nerve lesioning with degeneration analysis (see [29] ). A general conclusion that can be drawn from this work is that, in the distal portion of nerve trunks, most axons will be arranged in fascicles (i.e., bundles) that innervate only a single muscle. In the proximal portion of a nerve trunk, intermixing of fascicles results in a more homogeneous arrangement of axons with respect to their end-organ in- nervation. Sunderland has published detailed investiga- tions of the intraneural topographies of the human sciatic nerve and its popliteal divisions [30] and the human me- dian and ulnar nerves [28]. These studies, and others [ 141- [16], [19] imply that potential peripheral nerve implant sites will exist where fascicles are not greatly intermixed. In these cases electrical stimulation of a single fascicle will result in selective activation of one muscle.

To realize selective activation of individual fascicles a “neural” electrode is needed that can create a highly fo- cal electric field within a nerve trunk. Several groups have reported on neural electrodes that are appropriate for gen- eral use [2] or for stimulation selective with respect to nerve fiber size [3], [9], [ 101. Epineural stimulation, i.e., surgical implantation of electrodes onto the epineurium of a nerve trunk, has been used by one research group [13]. Intraneural electrodes, which utilize implantation of coiled wire electrodes into the peripheral nerve itself, have been tested in animals [4]. Both epineural and intraneural elec-

00 18-9294/90/0700-0706$0 1 .OO O 1990 IEEE

SWEENEY et al.: NERVE CUFF TECHNIQUE 707

I I

trode systems are potentially suitable for selective acti- vation of nerve fascicles I321 and, because they are not introduced into or onto muscles, may be more reliable than intramuscular or epimysia1 electrodes. However, a nerve “cuff ’ electrode design that places electrodes adjacent to the nerve trunk without surgical connection to the trunk itself would be less invasive. McNeal and Bowman [18] tested in animals multiple-electrode cuffs intended to pro- vide selectivity of muscle activation. Electrode position and good electrode contact (i.e., close apposition of elec- trodes to the trunk) were found to be important factors in successfully achieving selectivity of stimulation with this design. Unfortunately, it has in general been recom- mended that nerve cuff electrodes be sized so that their internal diameters are 40 to 50% larger than the nerve trunk diameter in order to prevent tissue damage (see [23]). This criterion would prevent direct apposition of electrodes to the trunk surface. However, in preliminary animal experiments we have shown that a “spiral” nerve cuff design can be safely implanted that fits nerve trunks “snugly” (i.e., the cuff internal diameter is the same as the nerve trunk diameter) [22]. In this paper, we therefore present modeling and testing of a technique for selective excitation of peripheral nerve trunk regions that could be implemented using a snug spiral nerve cuff design.

ELECTRODE CONFIGURATIONS Coburn [6] and Rattay [26] have noted that, mathemat-

ically, it is the component of an electric field that lies parallel to the myelinated axons within a nerve trunk that interacts with the axons; a fact that was first studied in detail experimentally by Rushton [27]. Specifically, the “activating function” derived by Rattay [26] that links an extracellular field with transmembrane and intracellu- lar current flow possesses a term that is proportional to the longitudinal second spatial difference of the extracel- lular potentials at the nodes of Ranvier. This means that sharp transitions in current that flows longitudinally out- side of the nerve fibers within a trunk will elicit changes in axonal excitability dependent upon the transition po- larity (current driven into an axon by an anode will hy- perpolarize, current driven outward by a cathode will cause depolarization).

The insulating layer of a snugly fitting cuff electrode will have the desirable effect of restricting current flow near the nerve trunk surface into a longitudinal compo- nent only. Intuitively, discrete dot electrodes embedded in the inner surface of a cuff will localize transitions in longitudinal current flow to small regions better than cir- cumferential ring electrodes will. A monopolar dot elec- trode [Fig. l(a)] configuration, for example, will focus current well near the electrode. However, at the ends of the cuff far from the electrode, current will tend to enter the cuff uniformly. An obvious improvement over such a monopolar configuration is that of a longitudinally aligned tripole [Fig. l(b)]. With such a tripole, current will flow from two dot anodes to a (central) dot cathode. This con- figuration will suppress current from flowing deep into the nerve trunk; instead, most current should flow in a straight

I I

(b)

iii \I/

- - 71(1-- I l l

I I I +

INSULATION a= zero

+ INSULATION a= zero

z = zero Z = i k

(e) Fig. 1 . Cross-sectional views of four snug nerve cuff electrode configu-

rations and an idealized modeling representation: (a) Single monopolar dot, (b) tripolar dot, (c) monopolar dot with an excitation “steering” dot located 180” opposite, (d) tripolar dot with excitation steering, (e) equiv- alent model for electric field solutions.

708 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 37. NO. I, JULY 1990

longitudinal path from the anodes to the cathode. For fo- cusing excitation within a superficial nerve fascicle this is obviously superior.

A less obvious way to restrict excitation to small re- gions of a nerve trunk is to use transverse current flow to “steer” the field. As discussed above, it is the longitu- dinal component of current flow in a nerve trunk that can be used to excite the myelinated axons. Transverse cur- rent flow should, theoretically, have no excitatory effect. In practice, current flow between two dot electrodes lo- cated on opposite sides of a nerve trunk can be used to excite axons at higher thresholds (than for stimulation with longitudinally aligned electrodes) because current flow between two such dots is not completely transverse [25]. However, by passing (predominately) transverse current flow at a subthresold level from an oppositely located “steering” electrode it should be possible to shape the flow of longitudinal current in the nerve trunk. For ex- ample, with a dot anode located 180” opposite to a dot cathode (either a monopole (Fig. IC) or the cathode in a longitudinal tripole [Fig. l(d)] it should be possible to “repel” stimulation current away from the oppositely po- sitioned anode and into the region near the cathode. The net electric field distribution created is, of course, the su- perposition of i) the field created by the flow of current between the oppositely located “steering” anode and the cathode, and ii) the field created by current flow in the simple monopolar or tripolar dot configurations.

In the present study we have investigated, through modeling and acute animal experimentation, the feasibil- ity and effectiveness of four snug cuff electrode configu- rations (monopolar, tripolar, monopolar with excitation steering, and tripolar with excitation steering) for selec- tively activating myelinated motor axons located within one superficially located nerve trunk region. The model- ing objectives were to solve for the electric field that would be generated within a nerve trunk by each of the four electrode configurations; and to use a simple nerve cable model to predict the effectiveness of each configu- ration in producing localized excitation of myelinated ax- ons. In three acute experiments using the greater sciatic nerve of cats the objective was to characterize the perfor- mance of the four electrode configurations in selectively eliciting excitation of motor axons only within the medial gastrocnemius muscle fascicle.

METHODS Electric Field Model Formulation and Numerical Methods

A nerve trunk electric field model was constructed by solving Laplace’s equation in cylindrical coordinates. La- place’s equation may be obtained in an anisotropic me- dium such as a nerve trunk by considering the continuity equation :

where 5 is the current density and p is the charge density. The current density may be written in terms of the electric

v 5 = -ap/at = o (1)

field using cylindrical coordinates (r, 4, z) via the three scalar equations:

J , = a,E,

Jr = a,Er

J+ = U t E + (2) where the anisotropy has been explicitly denoted using a conductivity crZ in the longitudinal (“z”) direction and a conductivity U, in the transverse direction (both radially and azimuthally). In the quasistatic approximation [24] we may then denote the electric field E in terms of a scalar potential V as

E = -vv . ( 3 ) Equations (2) and (3) may be substitued into (1) which yields Laplace’s equation in cylindrical coordinates:

o = ( I / r ) ( a / a r ) ( r av/ar) + ( l / r 2 ) ( a ’ ~ / a 4 ~ ) + ( U , / U , > ( a2V/azZ). ( 4 )

The electric fields produced by the four snug nerve cuff electrode configurations of Figs. l(a)-(d) can be solved using the idealized geometry depicted in Fig. l(e). In this problem we assume that over a length I the nerve trunk is bounded by an insulator (such as a “snug”cuff or air). At the two ends of the problem ( z = 0, 1 ) the nerve trunk is assumed to lie in a conductive bath that effectively clamps the potential of the trunk at zero potential (“ground”). These two boundary conditions may be stated as

VI ( 2 = 0 , 1 ) = O

Given the geometry of Fig. l(e), (4) may be solved using separation of variables and (6) incorporated to give

w m

v = C C (A,, cos (n+) + B,, sin (n4)) n = O m = l

I , ( ( u , / u , ) ’ / ~ ( rna r l l ) ) sin(rnaz/l) (7)

where the Fourier series coefficients A,, and B,, are de- termined from the boundary conditions and I , is a modi- fied Bessel function [ 11. Equation ( 5 ) is subject to the ex- ception that current can and does flow radially out of the nerve trunk when an electrode is placed on the surface. The nature of such current flow on an electrode will in general not be a simple function. For example, current densities will in general be much higher along electrode edges. The accurate modeling of such variations in the current density would be very difficult using the continu- ous functions which are summed in (7). Therefore the current distribution has been modelled as uniform over the surface of all electrodes. With this assumption an ef- ficient evaluation of the coefficients A,, and B,, is pos- sible given the orthogonal nature of the trigonometric functions of 4 and z (see below).

SWEENEY et al. : NERVE CUFF TECHNIQUE 709

1- * . . . . Vi.n vi‘n+t

. . . .

Fig. 2. A simple resistive ladder network has been used to model the ef- fects of imposed extracellular fields on myelinated axons within the nerve trunk. The parameters of the cable have been based on those of idealized 15 pm diameter mammalian myelinated nerve fibers. Axons were as- sumed to produce an action potential if the transmembrane potential at any node of Ranvier in the cable became depolarized by 25 mV or more.

Since all of the terms in (7) inherently satisfy (6) all that remains is to impose the boundary condition of ( 5 ) and boundary conditions due to any electrodes. The elec- trode boundary conditions can be translated into a con- straint on the derivative of V with respect to r at r = ro by writing

Jr = Z/S = a,Er = U, ( - av/ar) ( 8 )

where Z is the electrode current and S is the surface area of the electrode in contact with the nerve. Thus, the boundary conditions over the surface of the nerve ( r = ro) are only in terms of dV/dr. That is

on the electrode

not on the electrode. (9)

Since the variations in both 4 and z are represented in (7) by trigonometric functions, (9) may be satisfied by using a two-dimensional Fourier series to determine the coeffi- cients A,, and B,,.

The summations for n and m were truncated at n = 50 and m = 150, which provided adequate accuracy (i.e., use of additional terms did not significantly affect the re- sults). Each calculation involved 7500 summations for each point at which the potential was required. To permit straightforward experimentation with the program, all necessary field potentials were obtained for a given elec- trode (e.g., dot cathode, dot anode, etc.) and then stored. This permitted varying the relative strengths of electrodes (in a configuration) without recomputation. This process paralleled the experimental procedure where, for exam- ple, a dot anode (located 180” opposite to a stimulating cathode dot) might be used to “steer” the excitation while the current strength of the dot cathode was varied to achieve different recruitment levels.

Modeling Effects of Electric Fields on Nerve Fibers A simple nerve fiber compartmental cable model was

used to determine whether, for a given location in the nerve trunk ( r and 4 specified), excitation of a myelinated axon would result due to an imposed electric field. This cable model, seen in Fig. 2, was similar in construction to that of McNeal [17], with the following differences:

i) The parameters of the cable were based on mam- malian myelinated nerve properties (see [31]) rather than those of frog.

ii) The cable was purely resistive in nature. No active conductances or nodal capacitances were included. Ex- citation was assumed to result for a given axon if any point in the cable became depolarized by 25 mV or greater. A simple constant voltage threshold model of this kind pre- dicts stimulation thresholds similar to those at the rheo- base current (i.e., pulse width at least several times grea- tee than chronaxie) for an active model with nodal capacitance such as that of Sweeney et al. [31].

iii) The resistances in the cable were calculated to re- flect the properties of “intermediate” size 15 pm diame- ter myelinated motor fibers (rather then the large 20 pm diameter fibers of McNeal [ 171 ).

Formulation of a nerve cable model in this straightfor- ward manner permits rapid exploration of the effects of imposed fields on fiber excitability. At a node “n” Kir- choffs current law is used to derive (using the notation of McNeal)

0 = Ga { ( ~ n - 1 - 2 ~ n + ~ n + l )

where Ga is the axoplasmic conductance between nodes of Ranvier, G, is the transmembrane conductance at a node, the V, terms reflect the nodal transmembrane volt- ages (i.e., V,,, - Ve,, - V,,,) and the nodal extracellular potential terms ( Ve,J comprise the component of Rattay’s [26] ‘’activating function” by which an extracellular field interacts with a nerve fiber. The nerve fiber cable end con- dition was that of a zero current boundary (i.e., the cable terminated in an infinite impedance compartment). Given the parameters used (see below) 18 to 20 internodal lengths separated the cable ends from the electrode posi- tion(s). Matrix formulation of such equations for an N compartment cable results in an N x N sized matrix con- taining terms along the primary diagonal and the two nearest diagonals with all other elements being zero (see [7]). This is a tridiagonal matrix, and since the matrix is also positive definite, an extremely rapid solution is ob- tainable.

710 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 31, NO. 7, JULY 1990

Model Parameter Selection and Excitation Selectivity Analysis

Solutions were found for fields within an idealized cat sciatic nerve. The nerve trunk diameter (2r0) modeled was 4 mm. The total length of nerve trunk (1 = 6 cm) was intentionally chosen to be relatively large to insure that stimulation would not be dependent on insulation (i.e., cuff or air) length. The transverse neural conductivity (a,) was chosen to be equal to 0.001 S/cm and the longitu- dinal conductivity (a,) to be 0.01 S/cm [ I l l . Dot elec- trodes were modelled as ring sections on the nerve trunk surface (1 mm long in z direction and subtending 25” in 4). Field potentials within the nerve trunk were calculated for 3549 positions in all (for 39 nodal locations of 91 nerve fibers; the 91 nerve fibers were positioned at radial increments of 0.2r0 and azimuthal increments of 20”; given 1 of 6 cm and the relationship that internodal length divided by nerve fiber diameter (LID) = 100, a potential needed to be calculated external to 39 nodes of Ranvier for each fiber). The time required to do this calculation at 39 points on each of 91 nerve fiber radial locations was approximately 30 min on a Macintosh 11. This speed was achieved by precalculating all of the coefficients used in each summation and exploiting the various orders of the Bessel functions and their derivatives [ 11. The portion of the program that determined the effect of imposed fields on nerve fiber excitability through matrix solution using (10) was not timed; but in a sample run the matrix prob- lem for each nerve fiber position was solved, each fiber was checked to see if threshold was exceeded, and a plot of the results was completed in under three seconds.

The outer 35% of nerve trunk area (i.e., r > 0.8r0) was considered to represent nonexcitable tissue such as fat or connective tissue [29]. For each of the four electrode con- figurations the longitudinal current levels for 20, 50, and 80% recruitment of excitable tissue area were found (using linear interpolation to estimate the boundary of the exci- tation region between model fibers). These values were chosen to reflect three gross regions of excitation: adja- cent to the cathode, at an intermediate distance into the nerve trunk, and well removed from the cathode.

Experimental Procedure

In each experiment a cat was preanesthetized with IM ketamine hydrochloride and subsequently anesthetized with IV doses of sodium pentobarbital as necessary. The greater sciatic nerve was surgically isolated in the poplit- eal fossa of one hindlimb and lifted up out of a saline pool so that the air around the nerve acted as an efficient in- sulator. Thus, the approximate geometry of a long and snugly fitting nerve cuff electrode (about 3 to 4 cm in length) was created. The lateral gastrocnemius muscle was surgically divided from the medial gastrocnemius and its tendon transected. The medial gastrocnemius (MG), so- leus (SO), and tibialis anterior (TA) muscle tendons were isolated, transected and attached to force transducers. A mm by 1 mm square platinum dot electrode (to be used

as a cathode) was placed on the sciatic nerve several cen- timeters proximal to its branch point; using visual inspec- tion to place the electrode at the approximate location where the MG nerve fascicle should pass. Radiant heat was used to maintain the nerve-muscle preparation at body temperature.

For monopolar stimulation the saline pool was used as an anode and a regulated-current stimulator output stage was used to pass single 10 ps rectangular stimuli. This pulse width was chosen to enhance the selectivity of stim- ulation as related to nerve fiber diameter [2 11. For mon- opolar stimulation with excitation steering, one output stage was still used in this manner; but a second stage was used to pass a just-subthreshold 10 ps stimulus current from a dot anode located 180” around from the cathode on the nerve trunk surface. For tripolar stimulation two dot electrodes were placed longitudinally k 2 mm (edge to edge) on either side of the cathode and used as anodes (instead of the saline pool).

The lengths of the MG, SO, and TA muscles were ad- justed so that isometric contractions produced maximal forces in response to single supermaximal stimuli. In test- ing the effectiveness of the various electrode configura- tions the three muscle force signals were recorded on strip- chart paper. The amplitudes of each response were later measured, normalized to maximal obtainable during the run, and plotted versus longitudinal stimulus current am- plitude (i.e., the stimulus current from the monopole bath anode, or tripole dot anodes, to the cathode).

At the conclusion of each experiment the animal was given an overdose of sodium pentobarbital. A piece of 6-0 polypropylene suture material inserted through the epineurium was used to mark the location on the nerve trunk surface where the dot cathode had been placed. The sciatic nerve was dissected at the level of the cathode into its major divisions (common peroneal, soleus-lateral gas- trocnemius, medial gastrocnemius and tibial) and a gross cross-sectional fascicular ‘‘map” constructed.

MODELING RESULTS

In Tables I-A and B are listed summaries of the major modeling results. In Table I-A the longitudinal current levels needed for each configuration to elicit stimulation of 20,50, and 80% of the excitable area of the nerve trunk are presented. In Table I-B these data have been placed into ratio form to illustrate the effects of changes in elec- trode configurations.

Steering Current Excitation Threshold

The just subthreshold level for excitation of the nerve fiber just under the cathode due to current flow from a dot anode (located 180” around from the cathode) was 0.070 mA. This value was used for excitation steering of both monopolar and tripolar stimulation. Note that this value is about an order of magnitude less than the values typi- cally found in the animal experiments (see below and dis- cussion).

SWEENEY et al.: NERVE CUFF TECHNIQUE 711

TABLE I-A MODEL STIMULATION CURRENT LEVELS (IN mA)

Area Monopolar + Tripolar + Excited Monopolar Steering Tripolar Steering

20 % 0.41 1 0.370 0.800 0.710 50 % 0.813 0.813 3.100 3.100 80 % 1.100 1.220 6.900 7.600

TABLE I-B MODEL STIMULATION CURRENT RATIOS

Mono + Monopolar = Tripolar 3 Steering =

Area Mono + Trip + Monopolar - Trip + Excited Steering Steering Tripolar Steering

20 % 0.90 0.89 1.95 1.92 50 % 1 .oo 1 .oo 3.81 3.81 80 % 1.11 1.10 6.27 6.23

t"KNYam 20%

EXCITATION

McFxxoLAR 50%

EXCITATION

4 = zero degrees

McFxxoLAR 80%

EXCITATION

trunk surlace

0.411 mA 0813 mA 1.100 mA

(a) (b) (C) Fig. 3. Model excitation maps for simple monopolar stimulation of mye-

linated axons in three peripheral nerve trunk regions. A cathode "dot" electrode was modelled as a 1 mm length ring section that subtended f 12.5' of arc at 4 = 0". The outer 35% of each nerve trunk was assumed to consist of nonexcitable tissue (fat, connective tissue). At longitudinal current levels of 0.411, 0.813, and 1.100 mA regions of stimulation (dark) occurred in (a) 20% (b) 50%, and (c) 80% of the excitable portion of the nerve trunk, respectively. Complete results of such excitation mapping for all four electrode configurations modeled are contained in Tables I-A and B.

Monopolar Excitation Threshold for excitation with simple monopolar exci-

tation was 0.065 mA. As seen in Fig. 3(a), at a current level of 0.41 1 mA, 20% of the excitable tissue in the nerve trunk was excited. At current levels of 0.813 and 1.100 mA, the excitation level increased to 50 and 80%, re- spectively [Figs. 3(b) and (c)]. With the addition of the 0.070 mA steering current, the selectivity of monopolar stimulation was improved. The current necessary to ex- cite 20% of the nerve dropped to 0.370 mA while the current needed to recruit 80 % of the nerve fibers increased to 1.220 mA. Note that the steering current flow has no effect on the longitudinal current needed to excite 50% of

the excitable tissue in the nerve trunk. This is due to the fact that, by symmetry, the field created by the steering current can not possess a longitudinal component at the center of the nerve trunk (in the + = 90" and 270" plane). In Table I-B these data are listed in ratio form. Thus, the monopolar longitudinal currents necessary to elicit 20, 50, and 80% activation using excitation steering were 90, 100, and 1 11 % , respectively, of the longitudinal currents needed without steering.

Tripolar Excitation The threshold for excitation using two dot anodes lo-

cated * 2 mm (edge to edge) longitudinally from the

712 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING. VOL. 37. NO. 7. JULY 1990

cathode was 0.060 mA. The simple tripolar current levels needed to achieve 20, 50, and 80% excitation with and without excitation steering are listed in Table I-A. In Ta- ble I-B are listed the longitudinal current ratios for tripolar to monopolar stimulation, tripolar with steering versus without steering, and tripolar with steering versus mono- polar with steering. Note, first, that simple tripolar exci- tation restricts current flow to the superficial region be- neath the electrodes. This has the effect of progressively increasing the ratios for tripolar versus monopolar stim- ulation (both with and without field steering) as a greater percentage of the nerve trunk is activated. Second, note that in terms of ratios, addition of steering current has a similar effect in improving the selectivity of both mono- polar and tripolar excitation (lowering the current needed for 20% excitation and raising the current for 80% exci- tation).

EXPERIMENTAL RESULTS In Tables 11-A and B are listed summaries of monopolar

and tripolar excitation results for all three animal experi- ments. To construct these tables the simple stimulation current at the 50% force level was found (using linear interpolation between the two closest data points) for each recruitment curve. Table 11-A lists the average values of such current levels (with standard deviations) for the three different muscles and four electrode configurations. Stu- dent’s t-test ( P = 0.05) was used to test, for each config- uration, if soleus and tibialis anterior recruitment differed significantly from medial gastrocnemius recruitment (as denoted by an *); and if tibialis anterior recruitment dif- fered significantly from that of soleus (T). To construct Table 11-B these same data were placed in ratio form to characterize the effects of electrode configuration changes. Significant shifts in recruitment of force in each muscle due to these configuration changes are noted (*) based on Student’s t-test ( P = 0.05).

Steering Current Excitation Threshold

The just subthreshold current levels (for the “steering” current pair) averaged 1.01 mA (SD = 0.18 mA). In each experiment the just subthreshold value was determined and subsequently used in both the monopolar and tripolar excitation steering configurations.

Monopolar Excitation

Example monopolar excitation muscle force recruit- ment curves are seen in Fig. 4(a). Simple location of the cathode adjacent to the medial gastrocnemius nerve fas- cicle always resulted in recruitment of force in the medial gastrocnemius muscle at lower longitudinal current levels (threshold mean = 0.47 mA; SD = 0.09) than for soleus or tibialis anterior. The degree of selectivity was never sufficient, however, to allow recruitment of medial gas- trocnemius force between 0 and 100% without any acti- vation of the other two muscles. Addition of excitation steering current increased the selectivity. The recruitment

TABLE 11-A EXPERIMENTAL STlMULhTlON CURRENT LEVELS ( IN mA) TO ACHIEVE 50%

FORCE RECRUITMENT

Muscle Monopolar + Tripolar + Excited Monopolar Steering Steering Tripolar

MG 0.51 (.12) 0.15 ( .03) 0.92 ( . 2 3 ) 0.22 (.07) SO 0.77 ( . I O ) * 0.69 ( . I O ) * 2.54 (.52)* 2.15 (.91)* TA 0.79 (.19) 0.84 (.18)* 6.26 (.27)*t 7.42 (.25)*t

TABLE 11-B EXPERIMENTAL STIMULATIOV CURRENT RATIOS AT 50% FORCE

RECRUITMFNT

Mono + Monopolar 5 Tripolar =I Steering =

Muscle Mono + Trip + Monopolar = Trip + Excited Steering Steering Tripolar Steering

MG 0.28 (0.02)* 0.24 (0.04)* 1.83 (0.53)* 1.53 (0.54) SO 0.91 (0.20) 0.82 (0.21) 3.36 (1.01)* 3.01 (0.97)* TA 1.06 (0.03) 1.18 (0.03)* 8.26 (2.01)* 9.18 (2.19)*

Mean values ( N = 3) with standard deviations in parentheses.

of medial gastrocnemius muscle force was significantly shifted to lower current levels, soleus recruitment was rel- atively unaffected, and tibialis anterior recruitment was in all three experiments raised to higher current levels. While this shift in tibialis anterior recruitment was not in itself significant (Table 11-B), tibialis anterior recruitment with steering was significantly different from medial gastroc- nemius recruitment; and not significantly different with- out steering (Table 11-A).

Tripoiar Excitation

Tripolar excitation muscle force recruitment curves typical of those obtained experimentally are depicted in Fig. 4(b) (threshold mean = 0.80 mA; SD = 0.16). As in the model the effect of the tripolar electrode configu- ration was to restrict most current flow to the superficial region of the nerve trunk beneath the electrodes (in this case the medial gastrocnemius nerve fascicle). This re- sulted in a significant increase in the longitudinal current levels needed to achieve 50% force recruitment in all three muscles (Table 11-B); however, this increase was dispro- portionate so that selectivity of stimulation was increased. The use of an excitation steering current resulted in a sig- nificant decrease in the longitudinal current level for me- dial gastrocnemius recruitment, a nonsignificant decrease in soleus recruitment, and a significant increase in tibialis anterior recruitment current (in comparison to simple tri- polar stimulation).

Fascicular Maps

The sciatic nerve fascicular map from the experiment that yielded the data in Fig. 4(a) and (b) is depicted in Fig. 5 . The location of the cathode dot electrode is also shown. As can be seen, in this experiment, the fascicle

T

SWEENEY et a l . : NERVE CUFF TECHNIQUE

M O N O P O L A R S T I M U L A T I O N

W

U

II

I t- H 3

W

U

3

0 N

A

I: 0

H

1

0 . 6

0 . 6

0 . 4

0 . 2

0

0 0 . 5 1 1 . 5

L O N G I T U D I N A L S T I M U L A T I O N C U R R E N T (mA)

(a) T R I P O L A R S T I M U L A T I O N

w * U

II

I U o.e H 3

W

CO

I O

H A

I

0 . 6

; 0 . 4

4 0 . 2

0

0 2 4 6 e 1 0

L O N G I T U D I N A L S T I M U L A T I O N C U R R E N T (mA1

(b) Fig. 4. Normalized force recruitment curves versus monophasic rectan-

gular stimulation current (pulse width equal to I O ps) for medial gas- trocnemius (o ) , soleus (*), and tibialis anterior (+) muscles. Arrows reflect shift in recruitment curves with addition of 1.20 mA excitation steering current. (a) monopolar configurations, (b) tripolar configura- tions.

1 rnrn

Fig. 5 . Fascicular map of the sciatic nerve trunk yielding the data of Fig- ure 4. The medial gastrocnemius (MG), soleus-lateral gastrocnemius (S), common peroneal (CP), and tibial ( T ) nerve fascicles are shown at the level of the stimulating cathode (top).

713

innervating the medial gastrocnemius muscle lay under one end of the cathode; with the soleus-lateral gastroc- nemius nerve fascicle lying deeper in the trunk. In the other two experiments visual placement of the cathode was also successful in positioning the electrode over the me- dial gastrocnemius fascicle. In one of these experiments the soleus-lateral gastrocnemius fascicle again lay be- neath the medial gastrocnemius; while in the other nerve the soleus-lateral gastrocnemius branch was positioned both lateral and deep to the medial gastrocnemius fasci- cle. The common peroneal division (containing the inner- vation of the tibialis anterior [ 161 ) comprised roughly one half of the sciatic nerve in all three experiments with the same orientation as in Fig. 5.

7 I4 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 31, NO. 7. JULY 1990

DISCUSSION Numerical modeling and animal testing of a technique

for selective excitation of peripheral nerve trunk regions have been presented. Modeling and experimentation both suggest that tripolar dot electrode configurations bounded by a layer of insulation (such as a “snug” nerve cuff) will restrict excitation to superficial nerve trunk regions more successfully than will monopolar electrodes bounded by a cuff. This result is most clearly illustrated by the ratios in Tables I-B and 11-B on changing from monopolar to tripolar configurations. Modeling of the effects of neural electric fields on myelinated nerve fibers predicted, and experimentation confirmed, that excitation “steering” (using current flow from a dot anode located 180” oppo- site to the stimulating cathode) will improve the selectiv- ity of both monopolar and tripolar configurations. Note that in modeling and testing the effect of addition of steer- ing current on excitation selectivity was very similar for both monopolar and tripolar stimulation. This effect was somewhat less pronounced in the modeling results due to the relatively small steering current that could be used without exceeding threshold (0.070 mA in the model, 1.01 mA on average in experiments). This disparity in steering current levels (and a smaller disparity in simple stimula- tion current thresholds between the models and experi- mentation) is undoubtedly due to the idealized nature of the nerve trunk electric field model and the simplicity of modeling field effects on nerve fibers through a resistive cable network. ’ In many respects, however, the field and cable modeling results (Tables I-A and B) correlated re- markably well with the outcome of the animal experimen- tation (Tables 11-A and B) given the arrangement of the sciatic nerve fascicular divisions (e.g., Fig. 5). The me- dial gastrocnemius experimental results are, not surpris- ingly, similar to the modeling results for 20% excitation. Given that the nerve fascicle containing the innervation of the soleus muscle was found to sit deep and/or adjacent to that of medial gastrocnemius, we would expect that the experimental results for soleus activation would lie inter- mediate between the modeling 20 and 50% results. The innervation of tibialis anterior passes through the common peroneal nerve division of the sciatic nerve before sepa- rating into several individual fascicles more distally [ 161. Experimentally the effect of steering current on tibialis anterior recruitment was either a non-significant or sig- nificant increase in the stimulation current needed to pro- duce 50% of maximal force (Table 11-A and B). We would therefore hypothesize that the tibialis anterior experimen- tal results might best be compared with the 50 and 80% modeling results.

‘In particular, our use of a relatively “high” neural tissue anisotropy ratio of 10:l may have accentuated the tendency for steering current to exceed stimulation threshold. Due to the strength-duration characteristics of nerve fibers we might have expected that the modeling predictions for simple stimulation currents (which should reflect rheobase values) would have been significantly lower than the actual experimental values obtained (using a brief pulse width of ps) . The high model anisotropy ratio used may also account for the lack of this result.

In future studies we will investigate the effectiveness of snug nerve cuffs which incorporate a large array of dot electrodes in selectively exciting not just one superficially placed nerve fascicle, but any of a number of desired “wedges” of a nerve trunk. More quantitatively accurate nerve trunk field and nerve fiber cable models will be de- veloped to aid in the design and use of these nerve cuffs. More detailed modeling of nerve trunk structure through explicit representation of connective tissues, fat layers, and fascicles [32], [33] should yield a more quantitatively accurate electric field solution. An important aspect of se- lective stimulation not yet taken into account by our mod- eling is selectivity dependent upon nerve fiber size rather than fiber position. The well-known “inverse recruit- ment” tendency of simple electrical stimulation (exciting larger myelinated axons in a nerve trunk before smaller) [see, e.g. , 171 should be observable if a realistic nerve fiber diameter histogram is incorporated [33]. This effect undoubtedly helped separate, for example, medial gas- trocnemius force recruitment from that of soleus in our present studies; given that in the cat the mixed fiber type medial gastrocnemius tends to be innervated by larger di- ameter axons than the predominately slow fiber type so- leus [5]. Future prototype electrode designs will be tested in both acute and chronic animal experiments to deter- mine their ability to control selectively recruitment of force in multiple muscles.

REFERENCES [ I ] M. Abramowitz and I. A. Stegun, Eds.. Handbook of Mathemarical

Functions. New York: Dover, 1970. 12) W. F . Agnew, D. B. McCreery. T. G. H. Yuen. and L. A. Bullara.

“Histologic and physiologic evaluation of electrically stimulated pc- ripheral nerve: Considerations for the selection of parameters,” Ann. Biunied. E n g . , vol. 17, pp. 39-60, 1989.

131 R. Baratta, M. Ichie. and M. Solomonow, “Properties of orderly re- cruited motor units with tripolar cuff electrode,” in Proc. /Orb Ann. Con$. IEEE-EMBS, 1988, p. 1681.

[4] B. R . Bowman and R. C. Erickson, “Acute and chronic implantation of coiled wire intraneural electrodes during cyclical electrical stimu- lation,” Ann. Biomed. Eng. , vol. 13, pp. 75-93, 1985.

[SI 1. A. Boyd and M. R. Davey, Composition of Peripheral Nerves. London: Livingstone, 1968.

[6] B. Coburn, “Electrical stimulation of the spinal cord: Two-dimen- sional finite element analysis with particular reference to epidural electrodes.” Med. B i d . Eng. Cornput., vol. 18, pp. 573-584, 1980.

[7] -. “A theoretical study of epidural electrical stimulation of the spinal cord-Part 11: Effects on long myelinated fibers,” IEEE Trans. Biomed. Eng. , vol. BME-32. pp. 978-986, 1985.

[8] P. E. Crago, P. H. Peckham, and G. B. Thrope, “Modulation of muscle force by recruitment during intramuscular stimulation,’’ lEEE Trans. Biomed. Eng . , vol. BME-27, pp. 679-684. 1980.

191 Z. Fang and J . T . Mortimer, “Selective activation of small motor axon5 by quasitrapezoidal current pulses,” IEEE Trans. Biomed. Eng. , to be published.

1101 -, “A method to effect physiological recruitment order in electri- cally activated muscle.” IEEE Trans. Biomed. E n g . , to be published.

I1 1 1 L. A. Geddes and L. E. Baker, “The specific resistance of biological material-A compendium of data for the biomedical engineer and physiologist,” Med. B i d . Eng. Compur., vol. 5, pp. 271-293, 1967.

I121 P. A. Grandjean and J . T. Mortimer, “Recruitment properties of monopolar and bipolar epimysia1 electrodes, ’’ Ann. Biomed. Eng . ,

1131 J . Holle, M. Frey, H. Gruber, H. Kern, H. Stohr, and H . Thoma, “Functional electrostimulation of paraplegics: Experimental investi- gations and first clinical experience with an implantable stimulation device,” Orrhop., vol. 7, pp. 1145-1155, 1984.

vol. 1 4 , ~ ~ . 53-66, 1986.

SWEENEY et al.: NERVE CUFF TECHNIQUE 715

[I41 W. M. Kraus and S. D. Ingham, “Electrical stimulation of peripheral nerves exposed at operation,” J . Amer. Med. Assoc., vol. 74, pp. 586-589, 1920a.

[15] -, “Peripheral nerve topography, 77 observations of normal and diseased peripheral nerves,’’ Arch. Neurol. Psychiat., Chicago, vol.

[I61 J . N. Langley and M. Hashimoto, “On the suture of separate nerve bundles in a nerve trunk and on internal nerve plexuses,” J . Physiol.,

[17] D. R. McNeal, “Analysis of a model for excitation of myelinated nerve,” IEEE Trans. Biomed. Eng., vol. BME-23, pp. 329-337, 1976.

[I81 D. R. McNeal and B. R. Bowman, “Selective activation of muscles using peripheral nerve electrodes,” Med. Biol. Eng. Comput. , vol.

[19] P. Marie, H. Meige, and A. Gosset, “Les localisations motrices dans les nerfs peripheriques,” Bull. Acad. Med., vol. 74, pp. 798-811, 1915.

[20] E. B. Marsolais and R. Kobetic, “Functional electrical stimulation for walking in paraplegia,” J . Bone Joint Surg., vol. 69-A, pp. 728- 733, 1987.

[21] J . T. Mortimer, “Motor prostheses,” in Handbook of Physiology, Section I : The nervous system, Vol. 11, Motor control, Part I , J . M.

4 , pp. 259-296, 1920b.

vol. 51, pp. 318-346, 1917.

23, pp. 249-253, 1985.

Brookhart, V. B. Mountcastle, V. B. Brooks, and S. R. Geiger, Eds. Bethesda, MD: Amer. Physiol. Soc., 1981, pp. 155-187. G. G. Naples, J . T. Mortimer, A. Scheiner, and J . D. Sweeney, “A spiral nerve cuff electrode for peripheral nerve stimulation,’’ IEEE Trans. Biomed. Eng., vol. BME-35, pp. 905-916, 1988. G. G. Naples, J . T . Mortimer, and T. G. H. Yuen, “Overview of peripheral nerve electrode design and implantation,” in Neural Prostheses, W. F. Agnew and D. B. McCreery, Eds. Englewood Cliffs, NJ: Prentice-Hall, 1989. R. Plonsey and D. Heppner, “Considerations of quasi-stationarity in electrophysiological systems,” Bull Math. Biophys., vol. 29, pp. 657- 664. 1967. , - - - - ..

[25] J. B. Ranck, “Which elements are excited in electrical stimulation of mammalian central nervous system: A review,” Brain Res., vol. 98,

[26] F. Rattay, “Analysis of models for external stimulation of axons,” IEEE Trans. Biomed. Eng., vol. BME-33, pp. 974-977, 1986.

[27] W. A. H. Rushton, “The effect upon the threshold for nervous ex- citation of the length of nerve exposed, and the angle between current and nerve,” J. Physiol., vol. 63, pp. 357-377, 1927.

[28] S. Sunderland, “The intraneural topography of the radial, median, and ulnar nerves,” Brain, vol. 68, pp. 243-298, 1945.

[29] -, Nerves and Nerve Injuries. New York: Churchill Livingstone, 1978.

[30] S. Sunderland and L. J . Ray, “The intraneural topography of the sciatic nerve and its popliteal divisions in man,” Brain, vol. 71, pp.

[31] J . D. Sweeney, J . T. Mortimer, and D. Durand, “Modeling of mam- malian myelinated nerve for neuromuscular stimulation,’’ in Proc. 9th Ann. Conf. IEEE-EMBS, 1987, pp. 1577-1578.

[32] P. H. Veltink, J . A. van Alste, and H. B. K. Boom, “Simulation of intrafascicular and extraneural nerve stimulation,” IEEE Trans. Biomed. Eng., vol. BME-35, pp. 69-75, 1988.

[33] P. H. Veltink, J . A. van Alste, and H. B. K. Boom, “Influences of stimulation conditions on recruitment of myelinated nerve fibers: A model study,” IEEE Trans. Biomed. Eng., vol. BME-35, pp. 917- 924, 1988.

pp. 417-440, 1975.

242-273, 1948.

James D. Sweeney (S’85-M’89) was born in Pittsburgh, PA, in 1957. He received the Sc.B. degree in biomedical engineering from Brown University, Providence, RI, in 1979; and the M.S. and Ph.D. degrees in biomedical engineering from Case Western Reserve University, Cleveland, OH, in 1983, and 1988, respectively.

During part of his training he was a Predoctoral Fellow under the sponsorship of the Paralyzed Veterans of America-Spinal Cord Research Foundation. From 1988 to 1989 he was a faculty

Research Associate in the Department of Biomedical Engineering at Case Western Reserve University. He is currently an Assistant Professor of Bioengineering in the Department of Chemical, Bio and Materials Engi- neering at Arizona State University, Tempe. His research interests in gen- eral are in biomedical engineering, rehabilitation engineering, and neuro- physiology. Specific areas of interest and current research include: applied neural control; neural prosthesis design; neural modeling; implantable electrode system design; the effects of electric fields on nerve structure and function; and bioelectric phenomena.

Dr. Sweeney is a member of the Society for Neuroscience and the Amer- ican Association for the Advancement of Science.

David A. Ksienski (S’78-M’85) was born in Los Angeles, CA, in 1958. He received the B.S. de- gree from the Ohio State University, Columbus, in 1980, and the M.S. and Ph.D. degrees from the University of Michigan, Ann Arbor, in 1981 and 1984.

Currently he is with the Case Western Reserve University, Cleveland, OH, as an Assistant Pro- fessor of Electrical Engineering and Applied Physics. In addition to research in numerical mod- eling of bioelectrical phenomena, his interests in-

clude electromagnetic theory and inverse scattering and he is a member of Commission B of the International Union of Radio Science.

Dr. Ksienski is a member of the Eta Kappa Nu, Tau Beta Pi, and Phi Kappa Phi honor societies.

J. Thomas Mortimer received the B.S.E.E. de- gree from Texas Technological College, Lub- bock, in 1964, and the Ph.D. degree from Case Western Reserve University, Cleveland, OH, in 1968.

From 1968 to 1969 he was a Research Asso- ciate at Chalmers Institute of Technology, Goth- enberg, Sweden. In 1969 he joined the Biomedi- cal Engineering Faculty, Case Western Reserve University. From 1977 to 1978 he was a Visiting Professor at the Institute for Biocybernetics, Uni-

versity of Karlsruhe, Karlsruhe, Germany. Currently, he is Professor of Biomedical Engineering, and Director of the Applied Neural Control Lab- oratory, Case Western Reserve University. His past and present research, in areas involving electrical stimulation, includes pain suppression, resto- ration of hand function in quadriplegic patients, scoliosis correction, elec- trophrenic respiration and neural block.

Dr. Mortimer was the recipient of the Humboldt Preis (Senior U.S. Sci- entist Award), Alexander von Humboldt Foundation, Federal Republic of Germany. He is a member of Tau Beta Pi, Eta Kappa Nu, and Sigma Xi.