CONSIDERATIONS ON IMPROVING THE DESIGN OF CUFF

ELECTRODE FOR ENG RECORDING

Geometrical Approach, Dedicated IC, Sensitivity and Noise Rejection

Fabien Soulier, Lionel Gouyet, Guy Cath´ebras, Serge Bernard

David Guiraud and Yves Bertrand

LIRMM, Universit´e Montpellier II - CNRS - INRIA, 161 rue Ada, 34392 Montpellier, France

Keywords:

Multipolar cuff electrode, regular tessellation, electroneurogram, action potential, Laplacian, selectivity,

ASIC, multi-input differential ampliﬁer.

Abstract:

Cuff electrodes have several advantages for in situ recording of ENG signal. They are easy to implant and

not very invasive for the patient. Nevertheless, they are subject to parasitic background noise, especially the

EMG generated by the muscles. We show that the use of large numbers of poles can increase the sensitivity

of cuff electrodes as well as their selectivity with respect to a efﬁcient noise rejection. We investigate several

conﬁgurations and compare the performances of a tripolar cuff electrode versus a multipolar one in numerical

simulation.

One the other hand the use of cuff electrodes leads to the recording of the sum of the signals generated by all

the axons within the nerve. This puts in evidence the need of signal separation techniques that require a large

amount of information. Again, we show that multipolar electrodes can solve this problem since poles can be

switched one to another, provided that they are distributed along a regular tessellation.

Finally, we present the structure of an ASIC preampliﬁer aimed to obtain the Laplacian of the potential by

spatial ﬁltering and rejecting low-frequency noise..

1 INTRODUCTION

Functional Electrical Stimulation (FES) techniques

are used to restore motion or sensitive functions in

people with neural system pathologieslike spinal cord

injury. These techniques consist in generating artiﬁ-

cial contraction by electrical stimulation. In FES sys-

tem a direct opened loop control doesn’t allow efﬁ-

cient stimulation. In order to provide a loopback con-

trol we need sensory information (force, contact...)

(Djilas et al., 2006). An attractive solution consists

in using the natural sensors. The sensory information

is propagated by associated afferent ﬁbers and can be

recorded as an electric potential variation, the elec-

troneurogram (ENG).

Unfortunately, in peripheral nerves the complete

nerve activity due to the large number of axons makes

the extraction of the studied signal particularly hard.

Moreover the sensory signal seen through the nerve is

a very low amplitude signal compared with the ampli-

tude of parasitic signals. For instance, on a monopolar

recording, EMG created by muscle activity have am-

plitude about three orders of magnitude higher than

the ENG. In this context, the two main objectives to

be able to exploit natural sensors are:

• to ﬁnd a solution to separate the useful informa-

tion from the complete ENG signal;

• to reject the parasitic external signals.

The classical solution consists in using multipolar

electrodes, but from tripole (Ramachandran et al.,

2005) to nine pole electrode (Winter et al., 2000; Tay-

lor et al., 2004), the selectivity of the neural informa-

tion is not efﬁcient enough to be suitable in closed

loop FES system. To achieve both a better sensitivity

and efﬁcient background noise rejection we propose

a new conﬁguration of the cuff electrode with a large

number of poles regularly distributed onto the cuff. In

this conﬁguration, a group of poles can behave, with

suitable low level analog signal processing, like a kind

of a directive antenna. Moreover, the large number

of poles will allow enough channels in order to ap-

ply source separation signal processing on the ENG.

Of course, the directivity of the sensor relies on the

quality of the subsequent low-level analog signal pro-

cessing.

In this paper, we ﬁrst show how to generalize the

preprocessing operations on the recorded signal from

180

Soulier F., Gouyet L., Cathébras G., Bernard S., Guiraud D. and Bertrand Y. (2008).

CONSIDERATIONS ON IMPROVING THE DESIGN OF CUFF ELECTRODE FOR ENG RECORDING - Geometrical Approach, Dedicated IC, Sensitivity

and Noise Rejection.

In Proceedings of the First International Conference on Biomedical Electronics and Devices, pages 180-185

DOI: 10.5220/0001056301800185

 SciTePress

tripolar to multipolar conﬁguration using the Lapla-

cian formalism. Then we discuss on the optimal

pole placement around the nerve regarding tessella-

tion methods. Both the electrode conﬁguration and

the associated preprocessing circuit result from this

pole distribution and must be taken into account. We

particularly focus on the hexagonal seven-pole elec-

trode, presenting the associated seven input preampli-

ﬁer and preliminary simulation results.

2 EMG NOISE REJECTION

Cuff electrodes have been the most used for ENG

recording in the last ten years (Haugland et al., 1994;

Jensen et al., 2002; Andreasen and Struijk, 2002).

They are relatively easy to implant, they are not in-

vasive for the nerve and implantation is very stable

and thus allows chronic experiments. ENG can be

a a

Figure 1: Tripolar electrode cuff model.

recorded as the potential difference created on the

electrodes by the charges associated to the action

potentials (AP) propagating along the nerve ﬁbers.

Fig. 1 shows a typical tripolar cuff electrode. When

recording with this kind of electrode, a classic method

to reject parasitic signals consists in calculating the

average of the potential differences between the cen-

tral pole and each of the outer poles (Struijk and

Thomsen, 1995; Pﬂaum et al., 1996):

rec

−V

) +(V

−V

)

= V

−

(1)

The last expression shows that this operation consists

in:

1. averaging the signal on the outer poles, i.e. apply-

ing a low-pass spatial ﬁlter.

2. subtracting the result to the signal of the central

pole, keeping only the high spatial frequencies.

Therefore, the recorded V

rec

signal can be considered

as spatial high-pass ﬁltered.

More precisely, this ﬁlter is a second-order one

considering that the expressions

−V

) and

−V

) (2)

evaluate the ﬁrst derivative

. Thus the difference



−V

)

−

−V

)





)

−V



(3)

denotes the second derivative

that is the one-

dimensional Laplacian of the potential. We can iden-

tify in the last expression the equation (1) without the

known constant factor −2a

−2

Laplacian ﬁlters can reject both homogeneous po-

tentials and linearly varyingones like those created by

far EMG sources. The purpose of this new design is to

build two-dimensional Laplacian using more poles to

obtain isotropic rejection.

Figure 2: Multipolar electrode cuff model.

3 POSITIONING THE POLES

A tripolar cuff electrode (Demosthenous and Triantis,

2005) provides only one recording which is the su-

perposition of all action potentials “seen” by the elec-

trode at a given moment. The use of several poles on

a cuff electrode (see Fig. 2) could allow us to record

more signals, thus increase the quantity of neural data

and facilitate the signal post-processing on the record-

ing system.

In order to achieve optimal placement of poles, we

must pay attention to three constraints:

1. The electrodes have to be placed all around the

nerve, thus the poles have to be distributed onto

the whole surface of the cuff.

2. The poles have to be equally spaced to simplify

electronics in charge of analog signal preprocess-

ing (weight coefﬁcients in Laplacian preampli-

ﬁer).

3. They have to be able to be substituted one to each

other, so we take beneﬁts of the maximum mea-

surement locations, allowing powerful signal pro-

cessing.

Since the cylindric shape of the cuff results from the

wrapping of an initially plane device, these condi-

tions imply to look for a regular tessellation of the

plane as the positions of the poles or, more precisely,

tessellations composed of regular polygons symmet-

rically tiling the plane. It is well known that there are

exactly three type of regular tessellations (Weisstein,

2002). They can be speciﬁed using the Schl¨aﬂi sym-

bols: {3,6}, {4,4} and {6,3}.

CONSIDERATIONS ON IMPROVING THE DESIGN OF CUFF ELECTRODE FOR ENG RECORDING - Geometrical

Approach, Dedicated IC, Sensitivity and Noise Rejection

181

{6,3} {4,4}

{3,6}

Figure 3: There are exactly three regular tessellations com-

posed of regular polygons symmetrically tiling the plane.

The ﬁrst symbol in the Schl¨aﬂi notation denotes the

shape of the patch (triangle, square or hexagon). On

the ﬁgure 3, each vertex corresponds to a pole. Each

of them being surrounded by a number of equidistant

poles given by the second Schl¨aﬂi symbol, respec-

tively 6, 4 and 3.

From the previous tessellations, one can build

three kinds of electrodes by selecting one central pole

and its closest neighbors. Namely, we can deﬁne a

mesh of:

• triangular 4-pole electrodes,

• squared 5-pole electrodes,

• hexagonal 7-pole electrodes.

These candidates can be seen on the ﬁgure 4 and the

resulting expressions for the Laplacian are:

rec

= V

−

∑

i=1

for {6,3} (4)

rec

= V

−

∑

i=1

for {4,4} (5)

rec

= V

−

∑

i=1

for {3,6} (6)

One can notice that the {4,4} conﬁguration corre-

spond to the 2D Laplacian ﬁlter used in image pro-

cessing (Gonzales and Woods, 1992).

4 ENG AMPLIFIER

Because of the very low level of processed signals we

propose to perform the maximum of signal processing

as close as possible to the nerve. The more complex

operations to be considered are those with the hexag-

onal electrode.

{6,3}

{4,4}

{3,6}

Figure 4: Three possible conﬁgurations of electrodes.

Figure 5: Seven input preampliﬁer.

For this purpose, we have designed a seven channels

ASIC. Each channel compute a weighted difference

between the measurement point and the six closest

surrounding points. This is done in the analog do-

main using the preampliﬁer shown on ﬁgure 5. This

preampliﬁer is build around a differential pair whose

negative input transistor was split into six transistors

(six times smaller, of course). It has a voltage gain

that is about 100 and it is followed by an instrumen-

tation ampliﬁer whose gain is conﬁgurable between

6dB and 80dB. Each channel is composed of one

preampliﬁer followed by an instrumentation ampli-

ﬁer.

This circuit was designed to give an input-referred

noise below 1µV

rms

, a CMMR above 60 dB and a sufﬁ-

cient gain, i.e greater than 60 dB ; all these parameters

in the bandwidth of interest (1Hz ≤ f ≤ 3kHz). The

performances expected for this ampliﬁer are given

in Table 1 (the noise is measured in the band 1Hz-

3kHz).

A microphotography of the fabricated circuit is

presented Fig. 6. This circuit was designed in CMOS

AMS 0.35-µm technology.

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

182

Table 1: Ampliﬁer characteristics (simulation).

Active area (7 channels) 1.16mm

Supply voltage 3.3 V

DC Current (Preamp) 20µA

Voltage gain (Preamp) 100 (40 dB)

CMRR (Preamp) 80dB (10 kHz)

Voltage gain (Inst amp) 2 ≤ G ≤ 10000

CMRR (Full amp) 80dB (10kHz)

Input-ref. noise (Preamp) 0.672µV RMS

Input-ref. noise (Full amp) 0.677µV RMS

Bandwidth (Full amp) 76kHz

Figure 6: Microphotograph of the seven-channel prototype.

5 NUMERICAL RESULTS

5.1 Action Potential Modeling

In order to evaluate the performances of multipolar

electrodes, we need a model for the extracellular elec-

tric ﬁeld created by an action potential. Let us con-

sider a 10µm diameter myelinated axon. Its Ranvier

nodes are 1µm long, while their diameter is 6 µm and

their spacing is 1 mm. Let us call Ω the center of the

Ranvier node. When the AP is present at this node, we

can model it as a 6 µm diameter circle, perpendicular

to the axon axis, with a positive charge +q at its cen-

ter (Ω) and a negative charge −q spread on the circle.

The potential created at a point M of the space by this

AP can be approximated by:

V(M) =

8πε



1−

sin



(7)

In this expression, a is the radius of the Ranvier node

(3µm), r is the distance between Ω and M, while ψ

is the angle between the axe of the axon and

−−→

ΩM.

This approximation, valid for r ≫ a, is in good ac-

cordance with measurements. In particular, we can

see thatV(M) is negative for ψ = π/2 (Stein, 1980, p.

81). Last, q can be easily estimated from the charac-

teristics of the Ranvier node. For this study, we took

q ≃ 20fC and ε

≃ 80.

The model given by the equation 7 was used to

evaluate the sensitivity of the electrodes to action po-

tentials occurring inside the nerve. For the evaluation

of the rejection of parasitic signals, we must ﬁrst re-

call that EMG are also action potentials, creating the

same kind of electric ﬁeld. But, in this case, we can-

not make any assumption on the value of ψ. So, to

evaluate the external sensitivity of electrodes, we only

used a 1/r

model, unable to give voltages, but sufﬁ-

cient to compare the sensitivities of the electrodes.

We have limited the numerical study to the com-

parison of the classical tripolar cuff with the heptapo-

lar (hexagonal shape {3,6}) electrode. We have also

studied the effect of the wrapping to the performance

of the heptapolar electrode: we consider a plane elec-

trode and then a cylinder-wrapped one.

Given the position of a single AP we can easily

calculate the induced potential on each pole of the

cuff, since they are very small. For the tripolar cuff,

we need to average the potential on each ring. This

lead to an elliptic integral we have solved using nu-

merical methods.

5.2 Tripolar and Heptapolar Electrodes

Models

In the following, we compare a tripolar cuff electrode,

whose diameter is 2R = 3mm and ring spacing is

a = 4R, with one patch of the hexagonal cuff. To get

comparable results, this hexagonal cuff has the same

diameter (2R) and the spacing between poles is d = R.

Since this patch is partially wrapped around the nerve,

we considered also another patch perfectly ﬂat.

For all the calculations, the coordinates were ﬁxed

as follow: the origin O is at the center of the cuff elec-

trode. The Ox axis is the axis of the nerve (and, obvi-

ously, of the cuff). The Oy axis passes by the center

of the considered patch (which is perpendicular to this

axe). Last the Oz axe is placed to form a direct trihe-

dron with Ox and Oy.

5.3 Internal Sensitivity

Figure 7 shows the radial sensitivities of the three

electrodes (tripolar cuff, planar hexagonal patch and

wrapped hexagonal patch) that we compare. The ver-

tical axis is the value of V

rec

(in dBµV) calculated for

an AP placed on the Oy axis, at abscissa yR. The graph

shows clearly that while the sensitivity of the tripolar

cuff is quasi constant on the section of the nerve, the

sensitivity of the hexagonal patch is far higher (up to

30 dB) when considering an AP located between the

center of the patch and the center of the cuff.

CONSIDERATIONS ON IMPROVING THE DESIGN OF CUFF ELECTRODE FOR ENG RECORDING - Geometrical

Approach, Dedicated IC, Sensitivity and Noise Rejection

183

Figure 7: Radial sensitivities of (A) a tripolar cuff elec-

trode, (B) a planar hexagonal patch and (C) a bent hexago-

nal patch. The vertical axis is in dB µV and the unit for the

horizontal axis is the radius R of the electrode.

Figure 8: Longitudinal sensitivities on the axe of (A) a

tripolar cuff electrode, (B) a planar hexagonal patch and (C)

a bent hexagonal patch. The vertical axis is in dB µV and

the unit for the horizontal axis is the radius R of the elec-

trode.

Figure 9: Longitudinal sensitivities on an off-center (80 %

of R) axis of (A) a tripolar cuff electrode, (B) a planar

hexagonal patch and (C) a bent hexagonal patch. The ver-

tical axis is in dBµV and the unit for the horizontal axis is

the radius R of the electrode.

Figure 10: External relative sensitivity along Ox and Oz

axes for (A) a tripolar cuff electrode, (B) a planar hexagonal

patch and (C) a bent hexagonal patch. The vertical axis is

in dB and the unit for the horizontal axis is the radius R of

the electrode.

Figures 8 and 9 show the longitudinal sensitivities of

the three considered electrodes. On ﬁgure 8, the sen-

sitivity is computed for an AP placed on the Ox axe,

while, on ﬁgure 9, the AP is placed on a line, parallel

to Ox, cutting Oy at abscissa 0.8R. On this later ﬁg-

ure, we can see an increase of sensitivity of the tripo-

lar cuff in the vicinity of the rings, but this remains

far lower than the sensitivity of any of the hexagonal

patches.

5.4 External sensitivity

The ﬁgure 10 show the external sensitivities of our

three electrodes for an AP placed on the Ox or on the

Oz axis of the electrode. As stated above, the quan-

tity plotted is not a voltage, but is homogeneous to

the reciprocal of the cube of a distance. Nevertheless,

we can see on these two graphs that the hexagonal

patches exhibit a better rejection of parasitic signals

than the tripolar cuff. This improvement is of 32 dB

for Ox and 20 dB for Oz.

Figure 11: External relative sensitivity along the two halves

of the Oy axis for (A) a tripolar cuff electrode, (B) a planar

hexagonal patch and (C) a bent hexagonal patch. The ver-

tical axis is in dB and the unit for the horizontal axis is the

radius R of the electrode.

The same study conducted along the Oy axis (ﬁg-

ure 11) shows that, while the planar patch continues

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

184

to have the better rejection of parasitic signals, the

wrapped hexagonal patch has a sensitivity decreasing

slowly along this Oy axis. In fact, the bent hexago-

nal patch only begins to have larger sensitivity than

the tripolar cuff for action potentials placed at more

than ﬁfty times the radius of the cuff, corresponding

to approximately 7cm. At this distance, the parasitic

signal could be neglected in comparison to ENG sig-

nal.

6 CONCLUSIONS AND

PERSPECTIVES

We have presented here a method to build multipo-

lar cuff electrodes and how to extract useful informa-

tions from the multiple channels. Although numeri-

cal investigations are still necessary to an exhaustive

comparison of multipolar structures, the comparison

between the classical tripolar cuff electrode for ENG

recording and a multipolar electrode has shown that

this new type of the design is very promising. In ev-

ery simulation, multipolar electrodes prove to be more

sensitive to sources located inside the nerve, and in al-

most every case they show better far source rejection.

We hope the improvement of the recorded signal

given by this new design will allow the use of sig-

nal processing techniques such as source separation.

Then, multipolar conﬁgurations could make it possi-

ble to estimate more precise parameters like the speed

and the direction of propagation of the AP (Taylor

et al., 2004; Rieger et al., 2006).

Future works include the complete test of the

seven-channel ampliﬁer. It requires special attention

to electromagnetic interferences due to the high gain

it can reach. Then, we will be able to process the in

vivo qualiﬁcation of the whole system.

REFERENCES

Andreasen, L. N. S. and Struijk, J. J. (2002). Signal strength

versus cuff length in nerve cuff electrode recordings.

49(9):1045–1050.

Demosthenous, A. and Triantis, I. F. (2005). An adaptive

ENG ampliﬁer for tripolar cuff electrodes. 40(2):412–

421.

Djilas, M., Azevedo Coste, C., Yoshida, K., and Cath´ebras,

G. (2006). Interpretation of ENG signal for FES

closed-loop control. In IFESS’06: 11th Annual

Conference of the International Functional Electrical

Stimulation Society, pages 85–87, Miyagi-Zao, Japan.

Gonzales, R. C. and Woods, R. E. (1992). Digital Image

Processing. Addison-Wesley.

Haugland, M. K., Hoffer, J. A., and Sinkjaer, T. (1994).

Skin contact force information in sensory nerve sig-

nals recorded by implanted cuff electrodes. IEEE

Transactions on Rehabilitation Engineering, 2(1):18–

28.

Jensen, W., Sinkjaer, T., and Sepulveda, F. (2002). Improv-

ing signal reliability for on-line joint angle estimation

from nerve cuff recordings of muscle afferents. IEEE

Transactions on Neural Systems and Rehabilitation

Engineering, 10(3):133–139.

Pﬂaum, C., Riso, R. R., and Wiesspeiner, G. (1996). Per-

formance of alternative ampliﬁer conﬁgurations for

tripolar nerve cuff recorded ENG. In Engineering in

Medicine and Biology Society, 1996. Bridging Disci-

plines for Biomedicine. Proceedings of the 18th An-

nual International Conference of the IEEE, volume 1,

pages 375–376, Amsterdam.

Ramachandran, A., Sacristan, J., Lago, N., D¨orge, T., Os`es,

M., Koch, K., and Hoffmann, K. (2005). Multipolar

cuff electrodes with integrated pre-ampliﬁer & ﬁlter

to interface peripheral nerves for FES application. In

10th Annual Conference of the International FES So-

ciety, Montreal, Canada.

Rieger, R., Schuettler, M., Pal, D., Clarke, C., Langlois,

P., Taylor, J., and Donaldson, N. (2006). Very low-

noise ENG ampliﬁer system using CMOS technology.

IEEE Transactions on Neural Systems and Rehabilita-

tion Engineering, 14(4):427–437.

Stein, R. B. (1980). Nerve and Muscle. Plenum Press.

Struijk, J. J. and Thomsen, M. (1995). Tripolar nerve cuff

recording: stimulus artifact, EMG and the recorded

nerve signal. In Engineering in Medicine and Biol-

ogy Society, 1995. IEEE 17th Annual Conference, vol-

ume 2, pages 1105–1106, Montreal, Que.

Taylor, J., Donaldson, N., and Winter, J. (2004). Multiple-

electrode nerve cuffs for low-velocity and velocity-

selective neural recording. Medical and Biological

Engineering and Computing, 42(5):634–643.

Weisstein, E. W. (2002). Tessellation. From MathWorld–A

Wolfram Web Resource. http://mathworld.wolfram.

com/Tessellation.html.

Winter, J., Rahal, M., Taylor, N., Donaldson, N., and Stru-

ijk, J. (2000). Improved spatial ﬁltering of ENG sig-

nals using a multielectrode nerve cuff. In 5th Annual

Conference of the International Functional Electrical

Stimulation Society, Aalborg, Denmark.

CONSIDERATIONS ON IMPROVING THE DESIGN OF CUFF ELECTRODE FOR ENG RECORDING - Geometrical

Approach, Dedicated IC, Sensitivity and Noise Rejection

185