Acta Physiologica Sinica, April 25, 2006, 58 (2): 183-188 183
A remote controlled multimode micro-stimulator for freely moving animals
SONG Wei-Guo1, CHAI Jie2, HAN Tai-Zhen2,*, YUAN Kui1,*
Hi-Tech Innovation Center, Institute of Automation, Chinese Academy of Sciences, Beijing 100080, China; 2Department of Physiology,
Xi’an Jiaotong University, Xi’an 710061, China
Abstract: This paper presents a remote controlled multimode micro-stimulator based on the chip nRF24E1, which consists mainly of
a micro-control unit (MCU) and a radio frequency (RF) transceiver. This micro-stimulator is very compact (18 mm×28 mm two layer
printed circuit board) and light (5 g without battery), and can be carried on the back of a small animal to generate electrical stimuli
according to the commands sent from a PC 10 meters away. The performance and effectiveness of the micro-stimulator were validated
by in vitro experiments on the sciatic nerve (SN) of the frog, where action potentials (APs) as well as artifacts were observed when the
SN was stimulated by the micro-stimulator. It was also shown by in vivo behavioral experiments on operant conditioned reflexes in rats
which can be trained to obey auditory instruction cues by turning right or left to receive electrical stimulation (‘virtual’ reward) of the
medial forebrain bundle (MFB) in a maze. The correct response for the rats to obey the instructions increased by three times and
reached 93.5% in an average of 5 d. This micro-stimulator can not only be used for training small animals to become an ‘animal robots’,
but also provide a new platform for behavioral and neurophysiological experiments.
Key words: electrical stimulation; electrophysiology; nRF24E1; bio-robotic; multimode micro-stimulator
宋卫国 1，柴 洁 2，韩太真 2,*，原 魁 1,*
中国科学院自动化研究所高技术创新中心，北京 100080；2 西安交通大学医学院生理教研室，西安 710061
参数的设置及结果分析均由 10 m 外的个人计算机进行。通过离体实验，即刺激蛙坐骨神经干时产生动作电位并可观察到伪
迹；及在体大鼠的训练操作性条件反射的行为学实验，即在 Y 臂迷宫中训练大鼠听到不同声音完成左右运动，如果完成正确
给予前脑内侧束以电刺激，我们观察到 5 d 内大鼠的正确率增加 3 倍，达到 93.5%。上述实验都验证了该刺激器的有效性与
稳定性。本系统具有体积小(18 mm×28 mm 双层线路板)、重量轻(不含电池 5 g)、简单、实用、可靠等特点，能有效地用
关键 词：电刺激；电生理；n R F 2 4 E 1 ；动物机器人；多模式微型刺激器
中图分类号：Q 4 2 4 ；R 3 3 -3 3
Electrical stimulation has been widely used in electrophysi- and rat[6,7], is also arousing refreshed interests from
ological research and for the treatment of a variety of neu- bioengineers and artificial intelligence (AI) researchers.
rologic and psychiatric disorders for a long time[1-5]. Re- These achievements not only provide insights into how
cently the combination of electrophysiology with behav- animals learn, which may eventually lead to better pros-
ioral science, such as the locomotion control of cockroach thetics[8-10], but also open new possibilities for AI study.
Received 2005-08-25 Accepted 2005-12-08
This work was supported by the National Natural Science Foundation of China (No. 60375026).
Corresponding author. HAN Tai-Zhen: Tel: +86-29-82655274; Fax: +86-29-82655274; E-mail: firstname.lastname@example.org; YUAN
Kui: Tel: +86-10-82614510; Fax: +86-10-62613695; E-mail: email@example.com
184 Acta Physiologica Sinica, April 25, 2006, 58 (2): 183-188
The commonly used electrical stimulators need high op- then transmitted to the micro-stimulator. The micro-simu-
erating voltage, isolated stimulating channels, or cable lator will generate a specified stimulus sequence to the brain
connections, which make them very complex and heavy, of the animal or drive the speaker to give a sound accord-
and thus limit their usage in the experiments. With the ad- ing the needs of the experiment. The behavior patterns of
vance of microelectronic technology and the validity of the animal can be tracked and recorded automatically by
direct brain stimulation by digital signal, different types of an image processing system developed in our lab at the
stimulators have emerged for different purposes. But most same time.
of them are customized with special technology such as Micro-stimulator
application-specific integrated circuit (ASIC)[8,10-13], and the Experiments showed that the intensity needed for direct
stimulating parameters are preset or can only be set within neural stimulation is about 100 times smaller than what is
a few centimeters via radio frequency (RF) link. Instead needed for muscle stimulation, which means that the stimu-
of being used for small animal experiments, these stimula- lators can be more compact with the advances of system-
tors can only be used for the experiments with large ani- on-chip technology. The nRF24E1 chip, which has a mi-
mals or human beings. Since these stimulators, as well as cro-control unit (MCU) (8051) core and integrates a RF
the multi-array electrodes, usually need very complex sur- transceiver, is chosen as the main processor of the micro-
gery operation be implemented into animals to carry out an stimulator. The nRF24E1 has the properties of compact
experiment, the experiments are usually quite costly and package size, low power consumption, high transmission
inflexible. rate, etc. It also provides a pulse width modulation (PWM)
Apart from electrical stimuli, physical stimuli, such as analog output to drive the speaker. In order to minimize
sound stimuli, may also play an important role in behav- the weight and size of the micro-stimulator, all compo-
ioral research, which is lacking in the present stimulators. nents are chosen as surface mounted devices (SMD 0603
Therefore, the aim of our work is to develop a remote package), thus the total weight of the micro-stimulator is
controlled micro-stimulator, which is integrated with both less than 5 g. The outlook of the stimulator is shown in
audio and electrical stimuli, to explore the behavioral and Fig.2A and Fig.2B.
electrophysiological characteristics of small animals in nor- Stimulating pulses can be set either monopolar, as in the
mal living conditions. This work is meaningful for the re- traditional stimulator, or bipolar depending according to
search on animal’s locomotor behavior under direct brain the needs of the experiments and the characteristics of the
stimulation, and then for biomedical, AI and cognitive func- electrodes. Electrodes used for both recording and stimu-
tion researches[14,15]. It also has a potential possibility in lating have a sharp tip plated with gold, and bipolar stimu-
training an animal into “a remote controlled robot”, as well lating pulses are preferred, in order to minimize the po-
as in designing high-efficient brain-machine interfaces larization of the electrode as well as the permanent lesion
(BMI)[9,10,12,13]. to the tissue, which is caused by the charge accumulation
at the electrode. Bipolar pulses can be achieved by provid-
1 SYSTEM DESIGN ing another negative power supply, which means addi-
tional weight and size, so it is impractical for small animal
The system is composed of two parts: the remote con- use. By arranging P0 port (P0.3 P0.4), (P0.5 P0.6) of the
trolled multimode micro-stimulator, which is fixed on the nRF24E1 as two pairs and by setting high and low voltage
back of an animal, and the remote control station, which is alternatively and simultaneously for each pair, we can
a commonly used PC. The configuration of the system is achieve bipolar stimulating pulses effectively without the
shown in Fig.1. need of negative power.
The commanding parameters are set from the PC, and The stimulating electrodes used in this study are self-
made concentric needle electrodes, which have been proved
to be quite reliable under monopolar stimulation in our pre-
vious studies and monopolar pulses are chosen with ref-
erence to Stephen et al. The PWM port of the nRF24E1
is used to drive the speaker, and all the idle ports of the
nRF24E1 are set to low level in order to minimize the power
consumption and to minimize the electromagnetic
Fig. 1. The schematic representation of the system. interference. The remote controlled micro-stimulator de-
SONG Wei-Guo et al: Remote Controlled Multimode Micro-stimulator for Freely Moving Animals 185
Fig. 2. A rat with the stimulator (A), the main circuit board (B) and program flow chart of the stimulator (C).
codes commands from the RF data package and delivers a Table 1. Serial communication protocol
specified stimulating pulse sequence or drives the speaker Byte No. 1 2 3~7 This
to give sound. The flow chart of the main program is given
Data 0xFF 0xAA DATA CHECKSUM
in Fig.2C, and the program is written in C51 language and
stored in a memory chip (25AA320). The software run-
ning on the PC for commanding parameter setting and data which meets the experimental requirement. The protocol
analyzing is written in VC++ and, part of the interface is is shown in Table 1, where two bytes of synchronization
shown in Fig.3. head (0xFF, 0xAA) is followed by five bytes of command
package (DATA), in which each byte represents stimulat-
ing polarity (monopolar or bipolar), pulse duration (50 ms
~10 ms), pulse number (1~255), time interval (1 s~ 255
s), data frequency (0.1 Hz ~255 Hz), respectively. The
last byte (CHECKSUM) is the check sum of the whole
data package. An interruption mode is adopted for the
transmitter’s serial communication; receiving data from
the PC and transmitting them to the stimulator are accom-
plished in the interruption service routine, while the main
program will carry out the tasks of initialization and RF
2.2 Wireless communication
The active mode of nRF24E1 is operated at ShockBurst,
Fig. 3. The control and monitoring interface of the micro-stimulator. which can clock in data at a low data rate (depending on
the speed of the MCU) and transmit data at a high rate (up
to 1 Mbps), thus enabling extreme power reduction (60
2 COMMUNICATION MODES AND PROTO-
mW) and high sensitivity. While utilizing 250 kbps instead
of 1 Mbps would improve the receiver sensitivity by 10
2.1 Serial communication dB, we set a lower rate to achieve high sensitivity. As
The communication between the PC and the transmitter is the RF transceiver can only run in simplex mode, in
in serial mode and the baud rate is set to 19.6 kbps. Then, order to establish the bi-direction wireless data link, the
the time delay of the command package is less than 5 ms, nRF24E1s in both ends need to be configured properly.
186 Acta Physiologica Sinica, April 25, 2006, 58 (2): 183-188
The configuration words consist of 144 bits, which one end of the SN was placed on a recording electrode,
complete the ShockBurst configuration and the general while the other end was stimulated by the micro-stimulator,
configuration. The data package of the RF is listed in and a reference electrode was laid between them. Ringer
Table 2. solution was added to the SN in the experimental session
to keep a stable excitability of the sample. After being set
Table 2. Radio frequency communication protocol
from the PC, the command package was transmitted to
Byte No. 1 2~6 7~11 12~13 the micro-stimulator 10 meters away via RF. Then, the
Data PREAMBLE ADDRESS DATA CRC action potentials (APs) aroused from the nerve trunk were
sent to a data acquisition and analysis system (SMUP, Fudan
University) and to an oscilloscope simultaneously. Figure 4
The PREAMBLE is the package head, which is added
shows the recorded APs from the SN and the correspond-
and removed automatically at the transmitter and receiver
ing stimulating pulse (intensity: 3.3 V; monopolar; pulse
respectively. The ADDRESS is a receiving address, which
needs to be set at the transmitter and would be removed duration: 0.4 ms; frequency: 0.1 Hz; train number: 1 pulse).
automatically at the receiver. The DATA is the payload, the The Stimu and Spike in Fig.4 are corresponding to the
maximum width (bits) of which is 256 minus the address stimulating pulses and the APs, and the inside box is a
width (bits) and the check width (bits). The CRC repre- close-up of a single pair. It shows clearly that the stimulat-
sents an 8 or 16 bit checksum. The first byte of the receiv- ing pulses of the micro-stimulator could arouse APs from
ing address should not start with 0x55 or 0xAA, which the SN, and the waveforms of the APs and the shapes of
might be interpreted as part of preamble and cause address the stimulating artifact shows that it can be used for elec-
mismatch for the rest of the address. Also, the addresses trophysiology experiments effectively.
made by (5, 4, 3, or 2) equal bytes should be avoided
because it would make the packet-error-rate increase. In
general, more bits in the address and the checksum will
give less false detection, which in the end might give lower
data packet loss, but it would result in low transmission
In order to send commands (6 bytes) reliably and with
short delay, the application protocol is configured as single
receive channel, 40 bits receive address width, 16 bits
check width and 6 bytes of payload for the commands.
The nRF24E1 in the PC side is configured as transmitter
and in the stimulator side is configured as receiver, which
establishes the wireless link.
2.3 Power supply
The power supply of the whole system is composed of Fig. 4. The stimulating pulses (lower) and the action potentials (upper)
two separate parts: the power supply of the transmitter recorded from SN. The box shows an enlarged single stimulating
at the PC end is derived from the USB port of the host pulse and an action potential.
PC after voltage regulation (LM2576) from 12 V to 3.3
V, which is very convenient and practical especially in 3.2 In vivo
the open field. The power supply for the micro-stimula- Three adult male Sprague-Dawley rats, weighing 290 ~
tor was a lithium cell (3.6 V, 650 mAh, 9 g), which can 320 g, were used (provided by Xi'an Jiaotong University
sustain for about 20 h when the micro-stimulator works Animal Center). The surgery operation was performed
continuously. under anesthesia, induced by administration of 2% sodium
pentobarbital (50 mg/kg). Supplemental anesthesia was
3 METHODS AND RESULTS given when necessary. The animals were stereotaxically
3.1 In vitro planted with several self-made microelectrodes (a teflon
At first, a piece of sciatic nerve (SN) trunk was separated coated stainless steel with diameter of 50 mm was put in a
from a frog and put in a custom-made shield box. Then, cannula with a diameter of 80 mm) in the medical fore-
SONG Wei-Guo et al: Remote Controlled Multimode Micro-stimulator for Freely Moving Animals 187
brain bundle (MFB) (L: –7.5 mm; A: 2.6 mm; P: 2 mm), about 100 kΩ at 100 Hz, thus the effective stimulating cur-
and four stainless steel screws were driven in the skull rent is around 30 μA. The parameters of the conditional
around the cannula evenly. After the surface of the skull stimulation sound, used as instruction cues, were as follows:
was cleaned of all fascias and thoroughly dried, it was one 1-minute of 1 kHz for left turn and three 0.5-minute of
treated with erythromycin ointment. The screws and the 1 kHz for right turn. At the end of the behavioral tests,
cannula were fixed to the skull firmly with dental cement. each rat was deeply anaesthetized with sodium pentobar-
And then, the animals were injected with antibiotics for bital and the stimulating site was marked by the equipment
recovery in the following three days. Training was per- (SS-202J, Nihon Kohden, Japan). Then brain sections (40
formed one week after surgery, when the animals had fully mm in thickness) were prepared for identifying the elec-
recovered and were behaving normally. A Y-maze para- trode tip in the MFB (Fig.5). Statistical results showed that
digm was used to start the following task: for five days the
in an average of five days rats reached 93.5% correct cri-
rats were trained to run down the maze and respond to
terion for acquisition and the averaged correct turns in-
auditory instruction cues by turning right or left to receive
creased by three times (Fig.6). Behavioral experiments
an electrically stimulating reward (MFB). It has been found
showed the preference of the animal for the MFB stimula-
that stimulating the MFB in the hypothalamus is a good
tion as well as the effectiveness of the micro-stimulator
enough reward so that the animal will work to obtain the
and the method for ‘intelligent-rat’ controlling.
stimulation just as it will for food or water[20,21]. Therefore,
this method can be used as a ‘virtual’ reward for training.
A daily training session consisted of two 5-minute ses-
sions with 10 min interval in between (each tone presented
randomly). Whenever the maze presented the rats a choice
of turning left or right, the micro-stimulator generated a
specific sound. When the rat made the corresponding turn,
it was immediately given reward stimulation in the MFB
and recorded as ‘correct’. If the rat made an incorrect
choice, it was deprived of reward stimulation and recorded
as a ‘fault’. The animal’s behavior was recorded by a CCD
video tracking system.
Parameters used for MFB stimulating were as follows:
intensity 3.3 V, monopolar pulse width 0.5 ms, pulse inter- Fig. 5. Representative histological section showing an electrode track
val 10 ms, pulse number 10, train number 1. The mea- (red line), marking lesion (black arrow) at the bottom of the track and
sured resistance of the brain tissue and the electrodes, by the reference marks. A and B are the third ventricle and the
using an oscilloscope (VC-10, Nihon Kohden, Japan), is periamygdaloid cortex, respectively. Scale bar, 100 μm.
Fig. 6. The training process for three rats. In an average of five days, the rats reached 93.5% correct criterion for acquisition (A) and the averaged
correct turns increased by three times (B).
188 Acta Physiologica Sinica, April 25, 2006, 58 (2): 183-188
4 DISCUSSION the ventral tegmental area in rats. Behav Brain Res 1998; 93: 119-
This paper presents a remote controlled multimode micro- 6 Holzer R, Shimoyama I. Locomotion control of a bio-robotic
stimulator, which can generate physical (audio) and elec- system via electric stimulation. Proc IEEE/RSJ Inter Conf 1997;
trical stimulations. The electrical stimulating pulses can be 3: 1514-1519.
set to monopolar or bipolar via each pair of isolated channels, 7 Talwar SK, Xu SH, Hawley ES, Weiss SA, Chapin JK, Moxon
and the audio stimuli can generate different tones. The elec- KA. Behavioural neuroscience: Rat navigation guided by remote
trical stimulating intensity is 3.3 V, and the maximum stimu- control. Nature 2002; 417: 37-38.
lating current can reach 15 mA for monopolar or bipolar 8 Heiduschka P, Thanos S. Implantable bioelectronic interfaces
square pulses. The effectiveness and performance of this for lost nerve functions. Prog Neurobiol 1998; 55: 433-461.
system were validated from both in vitro and in vivo 9 Mussa-lvaldi FA, Miller LE. Brain-machine interfaces: compu-
experiments. The results showed that the system can not tational demands and clinical needs meet basic neuroscience. TINS
only used as behavioral and neuroscience research plat- 2003; 26(6): 329-334.
form for freely moving small animals, it can also be used 10 Nicolelis MAL. Actions from thoughts. Nature 2001; 409: 403-
as a tool for small animal’s locomotion training based on 407.
electrical stimulation, although more improvements are still 11 Arabi K, Sawan MA. Electronic design of a multichannel pro-
needed. grammable implant for neuromuscular electrical stimulation. IEEE
As an ongoing work, we have chronically implemented Trans Rehabil Eng 1999; 7(2): 204-214.
12 deCharms RC, Blake DT, Merzenich MM. A multielectrode
electrodes in the brain of the rats, and observed some be-
implant device for the cerebral cortex. J Neurosci Methods 1999;
havioral effects (including learning and memory) by using
the system to stimulate different nuclei of the brain. At the
13 Chapin JK, Moxon KA, Markowitz RS, Nicolelis MA. Real-
same time, we are improving our system by increasing its
time control of a robot arm using simultaneously recorded neurons
ability to record neuronal activity of the brain and making it
in the motor cortex. Nature Neurosci 1999; 2(7): 664-
a remote controlled stimulator and telemetric device for
freely behaving small animals. We hope to explore the rela-
14 Kohler E, Keysers C, Umilta MA, Fogassi L, Gallese V, Rizzolatti
tionships among electrical stimulation, spiking character- G. Hearing sounds, understanding actions: action representation
istics and behavioral responses[14,15]. Still further, we hope in mirror neurons. Science 2002; 297: 846-848.
it can be used to form a closed loop between the brain and 15 Romo R, Hernandez A, Zainos A, Lemus L, Brody CD. Neu-
machine as a BMI [10,12,13]. ronal correlates of decision-making in secondary somatosensory
cortex. Nature Neurosci 2002; 5(11): 1217-1225.
16 Tehovnik EJ. Electrical stimulation of neural tissue to evoke
1 Zhou BH, Baratta RV, Solomonow M, Matsushita N, D'Ambrosia behavior responses. J Neurosci Meth 1996; 65: 1-17.
RD. Open-loop tracking performance of a limb joint controlled 17 Yang DW, Pan B, Han TZ, Xie W. Sexual dimorphism in the
by random, periodic, and abrupt electrical stimulation inputs to induction of LTP: Critical role of tetanizing stimulation. Life Sci
the antagonist muscle pair. IEEE Trans Bio-Med Eng 1998; 45 (4): 2004; 75:119-127.
511-519. 18 Stephen DM, Susan LA, William ACJ. Early developmental
2 Jiang F, Racine R, Turnbull J. Electrical stimulation of the septal exposure to methylphenidate reduces cocaine-induced potentia-
region of aged rats improves performance in an open-field maze. tion of brain stimulation rewarding rats. Biol Psychiat 2005; 57:
Physiol Behav 1997; 62(6): 1279-1282. 120-125.
3 Mogilner AY, Benabid AL, Rezai AR. Brain stimulation: current 19 Bao XM, Shu SY. The Stereotaxic Atlas of the Rat Brain. Beijing:
applications and future prospects. Thalamus Relat Syst 2001; 1: People’s Sanitation Press, 1991.
255-267. 20 OLDS J, MILNER P. Positive reinforcement produced by elec-
4 Abbas JJ, Gillette JC. Using electrical stimulation to control trical stimulation of septal area and other regions of rat brain. J
standing posture. IEEE Contr Syst Mag 2001; 8: 80-90. Comp Physiol Psychol 1954; 47(6): 419-427.
5 Watanabe T, Morimoto K, Nakamura M, Suwaki H. Modifica- 21 Olds ME, Fobes JL. The central basis of motivation: Intracranial
tion of behavioral responses induced by electrical stimulation of self-stimulation studies. Ann Rev Psychol 1981; 32: 523-574.