A

htc

tattcTwrP

K

1

oaSbtodb

Yf

0d

Journal of Neuroscience Methods 170 (2008) 25–34

A combined wireless neural stimulating and recordingsystem for study of pain processing

Thermpon Ativanichayaphong a,1, Ji Wei He b,1, Christopher E. Hagains b,1,Yuan B. Peng b,1, J.-C. Chiao a,∗,1

a Electrical Engineering Department, The University of Texas at Arlington, Arlington, TX 76019, USAb Psychology Department, The University of Texas at Arlington, Arlington, TX 76019, USA

Received 20 September 2007; received in revised form 11 December 2007; accepted 19 December 2007

bstract

Clinical studies have shown that spinal or cortical neurostimulation could significantly improve pain relief. The currently available stimulators,owever, are used only to generate specific electrical signals without the knowledge of physiologically responses caused from the stimulation. Wehus propose a new system that can adaptively generate the optimized stimulating signals base on the correlated neuron activities. This new methodould significantly improve the efficiency of neurostimulation for pain relief.

We have developed an integrated wireless recording and stimulating system to transmit the neuronal signals and to activate the stimulator overhe wireless link. A wearable prototype has been developed consisting of amplifiers, wireless modules and a microcontroller remotely controlled byLabview program in a computer to generate desired stimulating pulses. The components were assembled on a board with a size of 2.5 cm × 5 cm

o be carried by a rat. To validate our system, lumbar spinal cord dorsal horn neuron activities of anesthetized rats have been recorded in responseso various types of peripheral graded mechanical stimuli. The stimulation at the periaqueductal gray and anterior cingulate cortex with differentombinations of electrical parameters showed a comparable inhibition of spinal cord dorsal horns activities in response to the mechanical stimuli.

he Labview program was also used to monitor the neuronal activities and automatically activate the stimulator with designated pulses. Ourireless system has provided an opportunity for further study of pain processing with closed-loop stimulation paradigm in a potential new pain

elief method.ublished by Elsevier B.V.

ophys

aiiLsprB

eywords: Pain; Neurostimulation; Single neuron recording; Telemetry; Electr

. Introduction

Chronic pain is a debilitating health problem. The impactsf pain suffering are not only on individual’s life quality butlso on the family, society and national economics (Cramer andpilker, 1997; Phillips, 2003). Several major approaches haveeen used to ease chronic pain, including surgical implanta-ion of neurostimulators. Neurostimulation on the spinal cord

r primary motor cortex delivers low levels of electrical signalsirectly to nerve fibers or neurons to affect the neuronal mem-rane excitability, in turn to suppress pain signals by opening

∗ Corresponding author at: The University of Texas at Arlington, NH518, 416ates Street, Arlington, TX 76019, USA. Tel.: +1 817 272 1337;

ax: +1 817 272 2253.E-mail address: jcchiao@uta.edu (J.-C. Chiao).

1 These authors contribute equally to this project.

spudpbmg

165-0270/$ – see front matter. Published by Elsevier B.V.oi:10.1016/j.jneumeth.2007.12.014

iology; Dorsal horn

nd closing of ion channels, through activation of descendingnhibitory system. This form of therapy is attractive becauset is selective for pain and has few side effects (Fields andevine, 1984) compared to chemical approaches. Therapeutictudies have shown when used on carefully selected chronicain patients, neurostimulation can significantly improve painelief and reduce use of narcotic medications (North et al., 1991;urchiel et al., 1996; Cameron, 2004; Bittar et al., 2005).

The currently available stimulators, however, are open-loopystems in which the doctors can only obtain the results forain management from patients’ verbal feedback. The stim-lating signals are programmed during the installation of theevices and cannot be modified after the patients leave the hos-

ital. Over time, the same stimulating configuration might note effective to inhibit pain due to the possible resistance neuronsay develop or the changes in the electrode-tissue interface. Our

oal is to develop a closed-loop system that can automatically

2 f Neu

prvrfc(

detq2tafd

2

cmamucd2rbseowTv

Tt

rltrdrmcwoSubdwif

2

tfispcDs

2

Fbi

6 T. Ativanichayaphong et al. / Journal o

rovide optimal pain relief based on the physiological responsesecorded from the neurons associating to the specific pain. Toalidate animal models, a simple wireless device capable of bothecording and stimulating is therefore necessary for studies onreely moving animals in the near future. Ultimately, the systemould be implemented as a wireless implantable microsystemWise et al., 2004).

Although several wireless systems for animal tests have beeneveloped, they are not perfectly suitable in our applicationither because of the large size (Obeid et al., 2004), the shortransmission range (Peng et al., 2004), or the operating fre-uency that is not publicly allowed in United States (Nieder,000; Chien and Jaw, 2005; Schregardus et al., 2006). Never-heless, to our best knowledge, an integrated wireless recordingnd stimulating system that can provide an optimized signaleedback control of the electrical stimulation has yet beenemonstrated.

. Materials and methods

To avoid the complexity of wireless communications, wehose commercial wireless modules for our device. Wirelessodules have been deployed in a recording system (Hawley et

l., 2002) and a stimulating system (Xu et al., 2004) in freelyoving rats and achieved a communication range up to 300 m

sing industrial, scientific and medical (ISM) band. We haveombined both systems and modified to fit our application. Ourevice occupies a 2.5 cm × 5 cm × 2.7 cm volume and weighs0 g (without batteries) which is small enough to be carried by aat. The recording part transmits non-filtered signals to a receiverase station equipped with a 300-Hz to 10-kHz filter that extractsingle neuron action potentials (APs). The stimulating part deliv-rs bipolar pulses with voltage levels up to ±18 V. The numbers

f pulses, pulse durations, pulse intervals and voltage levels areirelessly adjustable from a computer using a Labview program.he wireless operation was conducted in anesthetized rats as aalidation of the system, for further use in freely moving animals.

Apt

ig. 1. A combined wireless neuronal stimulating and recording system. (a) The recoth communicate with the wireless module; (b) the block diagram of the telemetricts brain and spinal cord.

roscience Methods 170 (2008) 25–34

he device can be used with commercially available electrodeshat are suitable for specific recording and stimulating areas.

The recording section includes a receiver base station whichecords the amplified neuronal signals transmitted from the wire-ess transmitter on the rat (Fig. 1). The stimulating transmitter athe computer sends out commands to the receiver carried on theat. The microcontroller reads the command and generates theesired stimulating signals. The telemetric device carried on theat includes a wireless transmitter module, a wireless receiverodule and a microcontroller module. It also includes discrete

omponents for recording and stimulating circuitry. All partsere assembled on a PCB (printed circuit board) with a sizef 2.5 cm × 5 cm (Fig. 2). Two 6-V lithium batteries (2CR1/3N,anyo) were chosen due to their small size. Each battery wassed to operate each part separately to avoid interference. Theattery has a capacity of 160 mAh, allowing an experimentaluration for more than 6 h. The telemetric device can be putithin a jacket to be worn by a freely moving rat. However,

n this work, anesthetized rats were used to verify the deviceunctionalities.

.1. Wireless recorder

The wireless recording part consists of a transmitter board onhe rat and a receiver base station. The transmitter board ampli-es and transmits neuron signals to the receiver. The receivedignal is passed through a band-pass filter to extract the actionotentials. The signal is then input to a data acquisition unitonnected to a computer (CED 1401Plus, Cambridge Electronicesign). The data is recorded and analyzed using a commercial

oftware (Spike2, Cambridge Electronic Design).

.1.1. Transmitter board

The circuit diagram of the transmitter board is shown in Fig. 3.virtual ground was used to imitate positive and negative sup-

lies for the operational amplifiers (op-amp) to amplify the APshat occupy both positive and negative cycles.

eiver base station and stimulating transmitter are connected to a computer anddevice; (c) the telemetric device is worn on the back of a rat with electrodes to

T. Ativanichayaphong et al. / Journal of Neuroscience Methods 170 (2008) 25–34 27

CB w

2i(emeiftmtcaa1vt

ai

Ccr1

atMistarab

Fig. 2. (a) The telemetric device without batteries; (b) front side of the P

.1.1.1. Input stage. Several constraints were imposed at thenput stage due to the weak and noisy signals a neuron providesMartinoia et al., 1993). Three major parameters were consid-red. First, the amplitude of the extracellular AP is usually inicrovolt ranges. This low voltage is subjected to noises from

lectronic devices and nearby ac power lines. To eliminate thenterference signals, a differential amplifier is necessary. The dif-erential amplifier cancels the common noises presented at bothhe input and the reference electrodes. A high CMRR (common

ode rejection ratio) of the amplifier is preferred. Second, due tohe small sizes of neurons, the recording electrode is usually thinreating high impedance interface with the tissue. The amplifiert the input stage must have even higher input impedance tovoid attenuation of signals from the voltage divider (Geddes,972). Third, unlike an ideal voltage source, the neuron providesery limited currents to the amplifier. The input bias current of

he amplifier must be very low.

To achieve these requirements, an integrated instrumentmplifier (AD620, Analog Device Inc.) was chosen because ofts small size, low supply currents and accurate gain. It has a

2soT

Fig. 3. Circuit diagram of t

ithout microcontroller and wireless modules; (c) back side of the PCB.

MRR of 80 dB, an input impedance of 1 G� and an input biasurrent of 2 nA. The gain of AD620 can be programmed by aesistor RG. A 5.1 k� for RG was chosen, resulting in a gain of0.7.

At the input, a high-pass filter C1 (0.1 �F) and R1 (1 M�) with1.6-Hz cut-off frequency was placed at the electrode to reduce

he dc potential from the animal that might create artifacts. The 1-� resistor also provides a current path to ground at the op-amp

nput. Without a resister, the charges accumulate and eventuallyaturate the op-amp (Kitchin and Counts, 2006). Depending onhe recording electrodes used, a low resistance value reduces themplifier input impedance and attenuates the signals while a highesistance value increases the op-amp dc offset and saturates themplifier (Wrobel et al., 2007). The value of 1 M� was achievedy experimental optimization.

.1.1.2. Gain stage and transmitter module. After the pre-amptage, the signal was amplified by a typical amplifier circuit. Inur design, an op-amp (TLV2264, Texas Instrument) was used.wo stages of non-inverting amplifiers were added with an equal

he transmitter board.

2 f Neuroscience Methods 170 (2008) 25–34

gaplmIs

2

riftfietpb

2

tdRcimmiPwAs

aTT(sIfF0rrvtmoclurpf

vi

2

rmrvwats

2

abvicNp

1tt

8 T. Ativanichayaphong et al. / Journal o

ain of 15. The total gain of the transmitter board was 2400. Themplified signal was connected to a high-pass filter (C4, R8) toull the signals from the virtual ground down to the 0–2.5-Vevel which is the analog input range of the wireless transmitter

odule (TX3A, Radiometrix). The module operates at 914 MHzSM band. It uses FM modulation and is capable of transmittingignals with a base band up to 35 kHz within a range of 300 m.

.1.2. Receiver base stationA wireless receiver module (RX3A, Radiometrix) was used to

eceive the FM modulated data at 914 MHz. The received signalncludes low-frequency fluctuation, mainly at 60 Hz couplingrom ac power lines, and high-frequency noises. To eliminatehe noises and extract the single unit APs, a unity gain band-passlter of 300 Hz to 10 kHz was used. The frequency range can beasily modified following a reference filter design (Carter, 2001)o record other types of neuron activities (i.e., compound actionotential at lower frequencies). Then the signal was amplifiedefore fed to the CED 1401Plus data acquisition unit.

.2. Stimulating part

Our neurostimulator follows the design by Xu et al. (2004)hat is capable of generating fixed ±5 V bipolar pulses. In ouresign, we utilize the 433-MHz wireless module pair (TX2/RX2,adiometrix) for communication. The transmitter TX2 wasonnected to an RS232 port in a computer, which sends out dig-tal commands generated by a Labview program. The receiver

odule RX2 on the rat received the commands and fed to aicrocontroller module (BS1-IC, Parallax Inc.) which has eight

nput/output (I/O) pins. The BS1-IC was programmed usingBASIC language provided by the manufacturer. One I/O pinas used to receive the commands from the wireless module.nother pin was connected to an LED to indicate the working

tatus during neurostimulation.In this work, we designed the stimulating system to gener-

te bipolar pulses up to ±18 V with adjustable voltage levels.he feature was achieved by using a charge pump (MAX202,exas Instrument) to increase the voltage level and a multiplexerCD4502B, Texas Instrument) to switch the voltage levels in fourteps. The MAX202 was operated by a 5-V supply from BS1-C, and the CD4502B was operated by a ±9-V supply generatedrom the MAX202. The simplified circuit diagram is shown inig. 4. Two of the I/O pins from BS1-IC were used to create–5 V stimulating pulses. The pulses were fed to the MAX202esulting in ±9 V signals. Each signal was fed to a series ofesistors (RX0-RX4 and RY0-RY4) to tap out four differentoltage levels, which can be arbitrarily adjusted by changinghe resistance values. The tapped out voltages were sent to the

ultiplexer CD4502B into the X and Y channels. Three I/O pinsf BS1-IC were used to control both switches of CD4502B toonnect the outputs X and Y to any of the four tapped voltageevels. Three buffers (FDG6301, Fairchild Semiconductor) were

sed to translate the 5-V level from BS1-IC to the 9-V levelequired to control the CD4502B. The voltage of the bipolarulses between the outputs X and Y therefore can be selectedrom 0 to ±18 V. The stimulating pulse parameters including

tgia

Fig. 4. Simplified circuit diagram of the stimulating part.

oltage levels, numbers of pulses, pulse durations, and pulsentervals are controlled wirelessly from a Labview program.

.3. Feedback systems

To activate the wirelessly controlled stimulation with theecorded neuron activities, we used a data acquisition (DAQ)odule (USB-6008, National Instrument) to monitor the

eceived wireless signals from the rat’s spinal cord. The Lab-iew program was used to estimate the rate of APs (spikes/s)hich represented the pain level when mechanical stimuli were

pplied. The Labview program forms a feedback loop to activatehe stimulator on the rat’s brain when the rate is higher than apecific threshold corresponding to a pain threshold.

.3.1. Rate estimation of APsAlthough several commercial softwares such as Spike2 are

vailable to calculate the rates of APs accurately, they cannote easily adapted to integrate with our Labview codes to acti-ate the neurostimulator. To demonstrate the feedback loop, its more logical to use the same Labview codes for both tasks ofalculating AP rates and activation of neurostimulation pulses.evertheless, we also used Spike2 in parallel with our Labviewrogram to verify the accuracy.

The DAQ module has a maximum sampling rate of0 ksample/s which is sufficient enough to monitor the APs sincehe bandwidth of the pain action potential signal is mostly lesshan 5 kHz. Knowing the pulse shapes of action potentials and

he pulses have much higher amplitudes than those of back-round noises, we can count each AP when the measured signals higher than a certain threshold voltage. The threshold volt-ge, sampling rate and number of averaging data points can be

T. Ativanichayaphong et al. / Journal of Ne

aletsidpsgrt

lcsaorpm

2

2

evUg

i(pbecioebeth

2

ewatioptrtp

2

waabwbathe applied force (F) as a function of displacement (x) of theclamp was measured. Fig. 6 shows the measured results. Thegraph follows the Hooke’s law F = kx, where the slope k isthe spring constant of the clamp. The slopes are 31.17 N/m for

Fig. 5. Flow diagram of the feedback algorithm.

djusted for experiments. The numbers of APs were accumu-ated for a certain period of time called “pain time slot”. At thend of the pain time slot, if the number of APs was more thanhe pain threshold level, the stimulation would be activated. Theimplified operation of the feedback system is illustrated in Fig. 5n which x is the pain level threshold. The stimulation starts withose #1 (the lowest dose with a low voltage, fewer numbers ofulses, a short duration and a long interval). If the pain level istill higher than the threshold, more intense doses will be givenradually through this feedback mechanism. When the pain iseduced below the threshold, the stimulation will be stopped andhe loop begins again when the pain comes back.

More complex algorithms can be applied in the feedbackoop using the same hardware developed. Ultimately, a doctoran run a series of experiments to evaluate the efficiency of eachtimulating parameter associated to the pain levels of individu-ls. This database can be used in decision making providing notnly automatic stimulation but also in an efficient way of painelief. In this work, however, only a simple feedback loop waserformed to demonstrate the feasibility of automatic feedbackechanism.

.4. Experiments

.4.1. Animal preparationMale Sprague–Dawley rats (300–350 g) were used in the

xperiments. All surgical procedures were approved by the Uni-ersity of Texas at Arlington Institutional Animal Care andse Committee. The procedures were in accordance with theuidelines published by the Committee for Research and Eth-

Fm

uroscience Methods 170 (2008) 25–34 29

cal Issues of the International Association for Study of PainZimmermann, 1983). Animals were anesthetized using sodiumentobarbital (50 mg/kg, i.p.). The spinal cord was exposedy performing a 3–4 cm laminectomy over the lumbosacralnlargement. A cannula was inserted in the trachea for artifi-ial respiration if needed. The anesthesia was maintained byntravenous administration of sodium pentobarbital at a ratef 5 mg/(ml h). The pupil reflex was monitored periodically tonsure a proper depth of anesthesia. The spinal cord was immo-ilized in a stereotaxic frame and covered with mineral oil. Thend tidal CO2 was maintained at around 30 mmHg and the bodyemperature was maintained at 37 ◦C using a feedback controlledeating pad and a rectal thermal sensor probe.

.4.2. Recording sitesA tungsten microelectrode (10–12 M�, FHC) was used for

lectrophysiological recordings in the spinal cord dorsal horn,here nociceptive primary afferent fibers terminate (Kandel et

l., 2000). The L5 and L6 regions were chosen which responds tohe rat’s hind paw that the pain would be later applied by mechan-cal stimulation. The electrode was connected to the amplifiern the wireless device. Single unit extracellular recordings wereerformed from the neurons that gave noticeable responses tohe mechanical stimulation of the receptive fields in the plantaregion of the hind paw. The data were recorded wirelessly usinghe CED 1401Plus and Spike2 software to extract the actionotential signals.

.4.3. Mechanical stimulationGraded mechanical (brush, pressure and pinch) stimulations

ere applied to the receptive fields in the hind paw. Brush waspplied by a camel hair brush moving over the receptive fields inrhythmic fashion which was innocuous. Pressure was appliedy a venous bulldog clamp (6 cm long, straight, serrated jaws)hich was between innocuous and noxious. Pinch was appliedy an arterial bulldog clamp (3 cm long, straight, serrated jaws)s a noxious stimulus. To analyze the pain level quantitatively,

ig. 6. Measured force as a function of displacement for the clips used forechanical stimuli.

30 T. Ativanichayaphong et al. / Journal of Neuroscience Methods 170 (2008) 25–34

F ponse

tpcfr

amsw

2

b((Ttbecfe0Wws

3

3

f

bwq5wsapaaw(t

3

amitEips

stTt

ig. 7. (a) Recorded signals with an amplitude of 500 �Vp-p; (b) frequency res

he pressure and 573 N/m for the pinch stimuli. As the clampressed onto the rat’s feet within a 25-mm2 area (A), and thelamp opened by 3 mm, the mechanical pressures were 0.54 psior the pressure stimulus and 9.35 psi for the pinch stimulus,espectively, calculated by P = F/A = kx/A.

Each mechanical stimulus was applied once for 10 s withn inter-stimulus interval of 20 s. The pain response to eachechanical stimulus was measured as the number of APs per

econd. Wide dynamic range (WDR) spinal dorsal horn neuronsere selected for these studies (Chung et al., 1986).

.4.4. Stimulating sitesInhibition of spinal cord dorsal horn neuron activity has

een demonstrated by stimulating midbrain periaqueductal grayPAG) (Peng et al., 1996), as well as anterior cingulate cortexACC) (Senapati et al., 2005) using conventional wired systems.hese results suggest that the PAG and ACC could be the poten-

ial stimulating areas for pain relief. We thus focused on the samerain area in our studies. After craniotomy, a bipolar stimulatinglectrode (Science Products) was placed in the PAG, 7.04 mmaudal to bregma, 0.5 mm lateral to the midline and 4.2 mm deeprom the brain surface (Paxinos and Watson, 1998). Anotherlectrode was placed in the ACC, 0.26 mm caudal to bregma,.5 mm lateral to the midline and 2.00 mm deep (Paxinos andatson, 1998). The electrode was selectively connected to theireless device for bipolar stimulation depending on the desired

timulation area.

. Results

.1. Device performance

The device was first tested using synthesized sinusoidal wave-orms as inputs. Fig. 7 shows the output signals from the receiver

nmAp

of the system; (c) recorded action potentials; (d) neurostimulation pulses.

ase station. At 1 kHz, the 500 �Vp-p signals were recordedith clear shapes (Fig. 7(a)) without visible distortion. The fre-uency response was measured from 100 Hz to 10 kHz with the00 �Vp-p input signals (Fig. 7(b)). The system gain at 1 kHzas around 15,000 or 84 dB. The 3-dB bandwidth of the systems

panned from 300 Hz to 4 kHz. The device then was used withn anesthetized rat and Fig. 7(c) shows typical recorded actionotentials, with a single action potential waveform expandeds indicated by the arrow. Four of bipolar pulses with a ±1 Vmplitude, a 1-ms duration and a 5-ms interval (or 200 Hz)ere generated from the device and recorded by an oscilloscope

Fig. 7(d)). The results showed that the wireless system achievedhe desired electrical performance.

.2. Inhibitory effects on action potentials

To find out the relationship between the neuronal responsesnd the given stimulating pulses, the Labview program was firstanually controlled to activate the stimulator on the rat dur-

ng the recording. Series of pulses were given to the rat fourimes during the 10-s periods of the mechanical stimulation.ach stimulation lasted for 1 s. Various stimulating parameters

ncluding voltage levels, number of pulses, pulse durations andulse intervals were used to observe the inhibitory effects of thetimulation.

The single neuron recording examples, during the pressuretimuli on the rat’s paw, with wirelessly controlled stimulation inhe PAG and ACC areas are shown in Figs. 8 and 9, respectively.he lower trace (a) shows the recorded signals and the middle

race (b) indicates wireless command pulses that activate the

eurostimulation which lasted for 1 s after the end of the com-and pulses. The upper trace (c) shows the rate histogram ofPs (spikes/s) identified from the trace in (a). The stimulationarameters are 100 pulses, ±1.0 V, 100 Hz and 0.5-ms duration

T. Ativanichayaphong et al. / Journal of Neuroscience Methods 170 (2008) 25–34 31

F o. 6-4a

fd

rTpirtdrlsowa

1tsfiAthbwWme

Fr

ig. 8. Inhibition effects with wireless PAG stimulation (rat no. RT071207, cell nction potentials.

or PAG stimulation, and 50 pulses, ±16 V, 50 Hz and 0.5-msuration for ACC stimulation.

In Fig. 8, when we started to apply pressure stimuli to theat, the rate of APs increased to 28 spikes/s (at 1273–1274 s).he rate of APs reduced to 4 spikes/s when the first stimulatingulses were applied (at 1274–1275 s). This corresponds to annhibition percentage of 86%. After the stimulation ended, theate of APs rose back to 11 spikes/s (at 1275–1276 s). Whenhe second stimulating pulses were applied, the rate of APsecreased again achieving an inhibition of 55%. The similarlyepeating cycles continued with the third and fourth stimu-ations, with inhibition of 80 and 100%, until the pressure

timuli were released after 10 s. The same phenomenon was alsobserved for ACC stimulation. The AP rate reduced when theireless stimulating pulses were applied. In Figs. 8 and 9, the

rrows indicate the inhibition percentages compared with the

wfp

ig. 9. Inhibition effects with wireless ACC stimulation (rat no. RT062707, cell noecorded action potentials.

). (a) Rate of action potentials; (b) command pulses of stimulation; (c) recorded

-s period earlier when no neurostimulation is applied. Fromhese results, with specific stimulating parameters, it seems pos-ible to achieve inhibition of near 100% as the neuron stopsring during the stimulation. Another observation is that theCC stimulation requires a much higher voltage than that in

he PAG case to achieve similar results. In our experiments, thisigh voltage introduced stimulation artifacts coupling throughoth air medium and the rat skin from brain to spinal cordhich was also observed in literatures (McLean et al., 1996).e reduced the interference by electrical grounding with alu-inum foil wrapped around cables and the rat skin between two

lectrodes.

Several experiments with various stimulating parameters

ere conducted under brush, pressure and pinch stimuli. Weound that the inhibition highly depends on the stimulationarameters. The results varied with rats and recorded neurons

. 2-2). (a) Rate of action potentials; (b) command pulses of stimulation; (c)

32 T. Ativanichayaphong et al. / Journal of Neuroscience Methods 170 (2008) 25–34

F 507-

alr

3

tSuT

dcsrhti

Fa

ig. 10. Automatic pain reduction using fixed stimulating pulses (rat no. RT100

s well. With this wireless device, combinations of stimu-ating parameters can be given in order to achieve optimalesults.

.3. Stimulation with a feedback loop

More recordings were performed for preliminary studies of

he feedback loop mechanisms using the method explained inection 2.3. The results of automatic activation of neurostim-lation with fixed stimulating pulses are shown in Fig. 10.he stimulating parameters were ±2.6 V, 100 pulses, 1.0-ms

outo

ig. 11. Automatic pain reduction using various stimulating doses (rat no. RT10120ction potentials.

3). (a) Recorded signals and stimulating pulses; (b) rate of action potentials.

urations and 10-ms intervals applied to the PAG area. Thelosed-loop pain relief mechanism can be explained in fourteps. In the 1st period, the mechanical stimuli were appliedesulting in high rates of APs representing severe pain. Theigh rates exceeded the pain threshold and the neurostimula-ion was activated in the 2nd period. In this period, pain wasnhibited by stimulation and the rate of APs reduced. A sec-

nd stimulating pulse train was activated in the 2nd periodntil the rate was lower than the threshold. Thus the stimula-ion stopped in the 3rd period. Without stimulation, the ratesf APs gradually increased meaning pain gradually came back.

7-3). (a) Recorded action potential signals; (b) stimulating pulses; (c) rate of

l of Ne

Oumi

ie1cwWhIwa3±tt±tscp

4

sarbliteoIiHtaptbpmm

vpmmoTai

A

daff

R

B

B

C

C

C

C

C

F

G

H

K

K

M

M

N

N

O

P

P

P

P

T. Ativanichayaphong et al. / Journa

nce the rate was again higher than the threshold, the stim-lation was activated again, as shown in the 4th period. Theechanism continued until the pain signal was completely inhib-

ted.Another experiment showed the stimulation with various

ntensities. In this experiment, two sets of stimulation param-ters were used for PAG stimulation. The first set was ±1.3 V,00 pulses, 0.5-ms durations and 10-ms intervals, which wasonsidered as a light dose. The second set was a stronger doseith ±2.6 V, 200 pulses, 1.0-ms durations and 5-ms intervals.hen the stimulation was activated by the action potential rate

igher than the threshold, the lower intensity was applied first.f the feedback indicated ineffectiveness, the stronger intensityas then used for the rest of the stimulation periods. The results

re shown in Fig. 11. When the high pain occurred (at the26th s), the stimulation was activated giving the pulses with1.3 V amplitude. However the stimulation did not decrease

he pain level effectively as the rate of APs remained higherhan the threshold. The 2nd and the 3rd stimulations with

2.6 V amplitude were thus activated. After the 3rd stimula-ion, the rate of APs decreased lower than the threshold so thetimulation stopped. When the pain came back, a new cycleontinued again with the 4th stimulation starting from ±1.3 Vulses.

. Discussion

We have developed a wireless system that can record andtimulate at the same time with features of a feedback loopnd decision making to activate the stimulation according to theecorded action potential signals. Our device is small enough toe carried by a rat. The device is capable of generating stimu-ating pulses with voltages up to ±18 V. The preliminary resultsn anesthetized rats show the feasibility of using our systemo study neuronal activities for pain management. From thexperiments and results, we found that there would be certainptimal stimulating parameters that give the highest inhibition.n general, more pulses, longer pulse durations, shorter pulsesntervals and higher voltage levels give better inhibition effects.owever, muscle contraction on the rats was observed when

he stimulating doses were too high. Unnecessary stimulationlso consumes extra battery power making practical implantower-inefficient and so inconvenient for patient. This implieshat the stimulation intensity should be kept as low as possi-le in practical uses without sacrificing pain reduction. Ourroposed feedback mechanism, integrated with wireless com-unication, can potentially reach optimal pain reduction withinimal stimulation.Given that responses from different neurons and rats may

ary, finding optimal parameters to inhibit pain at differentotential brain areas then requires further systematic experi-ents. Our next step is to conduct the experiment in freelyoving animals. This wireless system can be used with vari-

us electrode configurations that are suitable for specific areas.his reported device provides a new tool for studying neuronalctivities and potentially enables a new era of chronic pain reliefn humans.

S

S

uroscience Methods 170 (2008) 25–34 33

cknowledgements

This project was supported by the National Science Foun-ation, ECS Division, IHCS Program grant ECS-0601229. Theuthors would like to thank Shaohua Xu and Emerson Hawleyor original Labview stimulation program and PBASIC codeor BS1-IC.

eferences

ittar R, Kar-Purkayastha I, Owen S, Bear R, Green A, Wang S, et al. Deep brainstimulation for pain relief: a meta-analysis. J Clin Neurosci 2005;12:515–9.

urchiel K, Anderson V, Brown F. Prospective, multicenter study of spinalcord stimulation for relief of chronic back and extremity pain. Spine1996;21:2786–94.

ameron T. Safety and efficacy of spinal cord stimulation for the treat-ment of chronic pain: a 20-year literature review. J Neurosurg: (Spine 3)2004;100:254–67.

arter B. Filter design in thirty seconds, application report. SLOA093: TexasInstrument Inc; 2001.

hien C-N, Jaw F-S. Miniature telemetry system for the recording of action andfield potentials. J Neurosci Methods 2005;147:68–73.

hung JM, Surmeier DJ, Lee KH, Sorkin LS, Honda CN, Tsong Y, et al. Classifi-cation of primate spinothalamic and somatosensory thalamic neurons basedon cluster analysis. J Neurophysiol 1986;56:308–27.

ramer J, Spilker B. Quality of life & pharmacoeconomics: an introduction.Lippincott Williams & Wilkins; 1997.

ields H, Levine J. Pain-mechanisms and management. West J Med1984;141:347–57.

eddes LA. Electrodes and the measurement of bioelectric events. New York:Wiley-Interscience; 1972. pp. 39–43.

awley ES, Hargreaves EL, Kubie JL, Rivard B, Muller RU. Telemetry sys-tem for reliable recording of action potentials from freely moving rats.Hippocampus 2002;12:505–13.

andel E, Schwartz J, Jessell T. Principles of neural science. New York:McGraw-Hill; 2000. p. 475.

itchin C, Counts L. A designer’s guide to instrumentation amplifiers. 3rd ed.Analog Device Inc.; 2006.

artinoia S, Bove M, Carlini G, Ciccarelli C, Grattarola M, Storment C, et al.A general-purpose system for long-term recording from a microelectrodearray coupled to excitable cells. J Neurosci Methods 1993;48:115–21.

cLean L, Scott R, Parker P. Stimulus artifact reduction in evoked potentialmeasurements. Arch Phys Med Rehabil 1996;77(12):1286–92.

ieder A. Miniature stereo radio transmitter for simultaneous recording ofmultiple single-neuron signals from behaving owls. J Neurosci Methods2000;101:157–64.

orth R, Ewend M, Lawton M, Kidd D, Piantadosi S. Failed back surgerysyndrome: 5-year follow-up after spinal cord stimulator implantation. JNeurosurg 1991;28:692–9.

beid I, Nicolelis M, Wolf P. A multichannel telemetry system for single unitneural recordings. J Neurosci Methods 2004;133:33–8.

axinos G, Watson C. The rat brain in stereotaxic coordinates. San Diego:Academic Press; 1998.

eng YB, Lin Q, Willis WD. The role of 5-HT3 receptors in periaqueductal gray-induced inhibition of nociceptive dorsal horn neurons in rats. J PharmacolExp Ther 1996;276:116–24.

eng C-W, Chen J-J, Lin C-C, Poon P, Liang C-K, Lin K-P. High frequencyblock of selected axons using an implantable microstimulator. J NeurosciMethods 2004;134:81–90.

hillips CJ. Pain management: health economics and quality of life considera-tions. Drugs 2003;63:47–50.

chregardus S, Pieneman W, Maat A, Jansen R, Brouwer T, Gahr M. Alightweight telemetry system for recording neuronal activity in freely behav-ing small animals. J Neurosci Methods 2006;155:62–71.

enapati AK, LaGraize SC, Huntington PJ, Wilson HD, Fuchs PN, Peng YB.Electrical stimulation of the anterior cingulate cortex reduces responses of

3 f Neu

W

W

Xu S, Talwar SK, Hawley ES, Li L, Chapin JK. A multi-channel telemetrysystem for brain microstimulation in freely roaming animals. J NeurosciMethods 2004;133:57–63.

4 T. Ativanichayaphong et al. / Journal o

rat dorsal horn neurons to mechanical stimuli. J Neurophysiol 2005;94:845–51.

ise K, Anderson D, Hetke J, Kipke D, Najafi K. Wireless implantable microsys-

tems: high-density electronic interfaces to the nervous system. Proc IEEE2004;92(1):76–97.

robel G, Zhang Y, Krause HJ, Wolters N, Sommerhage F, Offenhausser A.Influence of the first amplifier stage in MEA systems on extracellular signalshapes. Biosens Bioelectron 2007;22:1092–6.

Z

roscience Methods 170 (2008) 25–34

immermann M. Ethical guidelines for investigations of experimental pain inconscious animals. Pain 1983;16:109–10.