Available online at www.sciencedirect.

com

Progress in Natural Science 19 (2009) 1261–1269

www.elsevier.com/locate/pnsc

Microelectronics-embedded channel bridging and signal regeneration

of injured spinal cords
Zhigong Wang a,*, Xiaosong Gu b, Xiaoying Lü c, Zhenglin Jiang b, Wenyuan Li a,
Guangming Lü b, Yufeng Wang a, Xiaoyan Shen a, Xintai Zhao a, Huiling Wang c,
Zhenyu Zhang a, Hongmei Shen b, Yang Wu a, Weixing Shen b, Jingyang Zhang a,
Dong Chen c, Xiaoyi Mao a, Huaxiang Shen b
a
Institute of RF- & OE-ICs, Southeast University, Nanjing 210096, China
b
Key Laboratory of Neural Regeneration of Jiangsu Province, Nantong University, Nantong 226001, China
c
State Key Laboratory of Bioelectronics, Southeast University, Nanjing 210096, China

Received 19 January 2009; received in revised form 16 February 2009; accepted 27 February 2009

Abstract

Due to the difficulty in spinal cord regeneration with biological methods, the microelectronic neural bridge, a new concept based on
microelectronic technology, is presented. The microelectronic system has been realized in the forms of hybrid and integrated circuits.
The integrated circuits for neural signal detection, stimulation, and regeneration are realized in a CMOS process. In animal experiments with
100 toads, 48 rats, and 3 rabbits, nerve signals have been successfully detected from spinal cords and sciatic nerves, and functional electrical
stimulation has been carried out for spinal cords and sciatic nerves. When the microelectronic system is bridged between the controlling and
stimulated nerve, the relevant motion of legs and nerve signal waveforms, which are stimulated by the evoked or spontaneous nerve signal
through such a system, have been observed. Therefore, the feasibility of the presented method was demonstrated.
Ó 2009 National Natural Science Foundation of China and Chinese Academy of Sciences. Published by Elsevier Limited and Science in
China Press. All rights reserved.

Keywords: Nerve injury; Microelectrode arrays; Neural recordings; Neurostimulation; Functional recovery

1. Introduction or so and the total number of SCI patients is estimated to

be more than 2 million. All these patients suffer from loss
Spinal cord injury (SCI) is one of the most severe trau- of perception and motor functions.
mas for human beings. In accordance with the statistics of Substantial function regeneration of an injured spinal
the Fact Sheet published by the National Spinal Cord cord should include three aspects: (1) the proliferation of
Injury Statistical Center of the USA, approximately neurons to replace the lost or dead ones, (2) the regrowth
11,000 new cases of SCI occur annually in the USA, and of interrupted nerve fibres (dendrites and axons) of surviv-
the total number of SCI patients has been estimated to ing neurons to rebuild the original signal transmission
be about 253,000 as of June 2006. According to the statis- channels, and (3) the reconnection of the regrowing den-
tics of the Chinese State Administration of Work Safety, drites and axons with the destined neurons. On the first
the annual incidence of SCI in China is 130,000 new cases aspect, it was believed that the regeneration of the neurons
in the central nervous system of an adult mammalian is
*
impossible [1]. Therefore, further hope of neuronal regener-
Corresponding author. Tel.: +86 25 83793303 8808; fax: +86 25
83792882.
ation was pinned on the second and third aspects. One idea
E-mail address: [email protected] (Z. Wang). was to regenerate the axons of surviving neurons from the

1002-0071/$ - see front matter Ó 2009 National Natural Science Foundation of China and Chinese Academy of Sciences. Published by Elsevier Limited
and Science in China Press. All rights reserved.
doi:10.1016/j.pnsc.2009.02.005
1262 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269

proximal stump to the distal one of the injured spinal cord. can regenerate both evoked and spontaneous neural
To demonstrate the possibility, a segment of sciatic nerve signals.
grafts was used to form a biological ‘‘bridge” over an In this paper, we report at first the system design and
injured section [2] or between the medulla oblongata and approaches of the microelectronics-embedded channel
the spinal cord [3]. Those experiments, however, demon- bridging and signal regeneration of the nervous system,
strated that there is a limitation in such cases that axons which are aimed at recovering the function of injured
regrow from the peripheral nervous system into the central spinal cords. Then, we show the results of the animal exper-
nervous system. This means that axon regrowth in a spinal iments and give discussions.
cord is difficult.
In recent years, pioneering neuroscientists introduced 2. System design and approaches
stem cell technology to establish the desired neural function
regeneration. Such research has shown that transplants of The proposed concept on microelectronics-embedded
olfactory ensheathing cells into the injured spinal cords channel bridging and signal regeneration for injured spinal
of laboratory rats had a remarkable capacity to repair cords is illustrated in Fig. 1(a), and the block diagram of
damaged tissue and lay a ‘bridge’ across a gap in the the microelectronic module is proposed as shown in
injured nerve fibres [4]. Anderson and his colleagues [5] Fig. 1(b).
have used adult human neural stem cells to regenerate The microelectronics system consists of two MEAs and
damaged spinal cord tissue and improve mobility in mice. two multi-channel neural signal regenerators (NSR).
These methods can be understood as the formation of bio- MEA1 and MEA2 are interfaced with the proximal and
logical ‘‘bridges” between proximal and distal stumps of distal stumps of an injured spinal cord, respectively. Part
injured spinal cord by means of newly introduced and dif- of the electrodes of MEA1 on the proximal stump is con-
ferentiable neurons, instead of the proliferated original nected to the first multi-channel neural signal regenerator,
ones. However, much more basic and preclinical research NSR1, and used to detect signals from the axons of the
must be completed before attempting human trials using neurons in the cortical centre. The output signals from
stem cell therapies to repair injured spinal cords. Therefore, NSR1 are supplied to part of the electrodes in MEA2 on
it is worth exploring alternative methods for repairing the distal stump and used to stimulate the desired motor
injured spinal cord. neurons. The downward channels in NSR1 are responsible
In more than 100 years, different electric or electronic for the regeneration of motor signals. Conversely, the rest
devices and systems have been introduced into the applica- of the electrodes of MEA2 on the distal stump are con-
tion of neural signal recoding and stimulation. With the nected to the second multi-channel neural signal regenera-
continuous rapid progress of microelectronics, integrated tor, NSR2, and used to detect signals on the axons from
circuits have reached the level of the so-called SOC (System the sensory neurons in the dorsal root ganglia. The output
on Chip). Moreover, by using micro-electro-mechanical- signals from NSR2 are supplied to the rest of the electrodes
system technology, micro-electrode-arrays (MEA) can be of MEA1 on the proximal stump and are used to stimulate
manufactured in different structures for implanted neural the desired axons that join the thalamus. The upward chan-
interfaces. Up to now, a lot of microelectrodes such as nels in NSR2 are responsible for the regeneration of sen-
microwire or cuff electrodes have been used to study the sory signals.
activity of the cerebral cortex and spinal cord [6–8]. Fur- The block diagram of the microelectronic module along
thermore, researchers have attempted to use such micro- with two MEAs is illustrated in Fig. 1(b). The core part is
electrodes incorporated with the SOC to restore the the NSR array, which consists of a line-up of downward
motor function of the upper limb of human beings [9]. and upward channels. Between the NSR array and the
Obviously, the chip implanted in an injured spinal cord is MEAs, two channel switching fabrics are inserted. Their
more pertinent, and therefore, more efficient for SCI function is to switch the downward and upward channels,
patients. so that the neural signals can be regenerated along the
Based upon SOC and MEA technology, we intro- intended channels even after implanting. The controlling
duced the concept of microelectronics-embedded channel unit carries out the switching function etc. The monitoring
bridging and signal regeneration of an injured spinal unit is used to monitor different parameters and functions
cord [10] in 2004 and have realized the prototypes of of the system.
the so-called microelectronic neural bridge in the forms The signal processing unit includes detecting and stimu-
of hybrid and integrated circuits [11]. The animal exper- lating modules, which detects and excites the neural signal
iments with 100 toads, 48 rats, and 3 rabbits demon- propagating along the given neuron in the spinal cord,
strated that the nerve signal from a sciatic nerve or a respectively [10]. The detecting circuit includes an RC net-
spinal cord can be regenerated to an interrupted sciatic work, a pre-amplifier, an active band-filtering stage, a
nerve in a toad or rat and an interrupted spinal cord notch network, and a shield-guarding circuit [12]. The stim-
can be connected bi-directionally in a rabbit, when the ulating module consists of a pre-amplifier and a post-
microelectronic system is bridged between the control- amplifier. The former amplifier is to amplify the input sig-
ling and the stimulated nerve. Both bridge formations nal to sufficient amplitude, and the latter one consisting of
 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269 1263

Fig. 1. Concept of the channel bridging and signal regeneration of an injured spinal cord by using two NSRs (a) and the block diagram of the
microelectronic module along with two MEAs (b).

two amplifiers can generate two signals with 180° phase dif- system, respectively. The self-made needle electrode arrays
ference, which can be used to amplify the signal further so were made of acupuncture needles whose body was covered
as to obtain an effective stimulating voltage to drive the fol- with Parylene except for the tip (Cookson Electronics Spe-
lowing microelectrodes [13]. For simplicity and low-power cialty Coating Branch, Shanghai). Fig. 2(a) shows one
application, all signals are processed in an analogue four-needle electrode array. The aluminum sheath was
domain. The active band-filtering stage realizes only a fil- used to wrap the needles for supporting and shielding. A
tering function to remove noises and interferences of the three-arm stereotactic positioner (Shanghai Alcott Biotech
50 Hz power supply. CO. Ltd.) was used to hold and position the three-needle
For our study, three types of electrodes, self-developed electrode arrays as shown in Fig. 2(b). The three regions
microelectronic systems, and a series of commercial instru- to be positioned were the spinal cord, the left sciatic nerve,
ments have been used. and the right sciatic one.
The hook-type electrode (Quanshui Experimental The microelectronic systems for in vitro experiments
Devices Co., Nanjing, China) was made of two parallel were realized by using the printed circuit board (PCB) tech-
metal wires pressed into the surface layer of a hook-formed nique. Besides simple discrete devices such as resistors and
plastic plate. The cuff electrode arrays of different sizes capacitors, precision operational amplifiers OP27 or
were obtained from the Fraunhofer Institute for Biomedi- MAX4168 (Maxim, USA) were chosen for all gain stages
cal Engineering, Germany. Its substrate was made of poly- [15,16]. In the system, there are eight groups of low-noise
imide (Pyralin PI 2611, Du Pont), the contact metal high-gain amplifiers, analogue signal processing units,
platinum/iridium, the conductive strips gold, and the bond and FES units. The gain of each amplifier can be switched
pads solderable layer [14]. Two cuff-type MEAs with differ- among 10 values: 1, 10, 50, 100, 200, 500, 1000, 2000, 5000,
ent contact configurations, which are 3  4 and 3  6, were and 10,000. One amplifier plus one signal processing unit
used to form four and six channels of the triple-electrode can form a neural signal detector or monitor. An FES unit
1264 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269

Fig. 2. Four-needle electrode array made by the Institute of RF- and OE-ICs, Southeast University, China (a) and photograph of the three-arm
stereotactic positioner for holding and positioning three-needle electrode arrays (b).

can be used alone or combined with a neural signal detec- All animal experiments were completed in the labora-
tor to build up an NSR-channel. That means, a maximal of tory of the Southeast University and Jiangsu Key Labora-
eight NSR-channels can be built up. One applicable possi- tory of Neuroregeneration of Nantong University and
bility is that four channels are used for downward signal performed in compliance with the Medicine Institutional
regeneration and the others for the upward one. The func- Animal Care and Use Committee of Nantong University
tion transforming and gain adjusting were carried out by and national regulations and policies.
the switches and potentiometers on the front board of
the box. 3. Experimental results
The integrated circuits (ICs) for implantable systems
were designed in the so-called fabless mode [12,13]. A Using all three types of electrodes and the discrete or
0.6 lm CMOS 2P(oly)2 M(etal) technology (CSMC, integrated microelectronic neural regenerating systems
Wuxi, China) was chosen as the foundry for the chip mentioned above, we have carried out, in the last 4 years,
fabrication. The computer-aided design programs Smart- 15 animal experiments, and a total of 100 toads, 48 rats,
Spice (Silvaco, Silicon Valley, USA) and Zeni (Huada, and 3 rabbits were used as models in the experiments.
Beijing, China) were used for the circuit simulation The significant results obtained are as follows.
and the layout design, respectively. The chips were fab-
ricated through the Multi Project Wafer (MPW) service
3.1. In vitro nerve signal regeneration from a toads’ left leg to
of ICC, Shanghai, China. The sizes of the die and the
the right one
packaged chip were 2.82  2 and 14.2  14.2 mm,
respectively.
First, the spinal cord of a toad was destroyed and then,
The measuring instruments of both the PCB- and IC-
two hind limbs were separated and the relevant sciatic
type microelectronic systems mainly include an arbitrary
nerves were exposed. In this experiment, two five-needle
pulse generator (Agilent 33220A, USA) and an oscillo-
electrode arrays were used to pierce into the toad’s left
graph (Tektronix TDS5104, USA). The same pulse gener-
and right sciatic nerves, respectively. The schema of the
ator and oscillograph were also used in the animal
channel bridging and signal regeneration are shown in
experiments as the pulse signal source of the FES driver
Fig. 3(a), and the experimental results are shown in
and for signal monitoring and recording, respectively.
Fig. 3(b). Here, an FES signal was stimulated by a pulse
 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269 1265

reaction parts of the body, were tested by means of the

cuff-type MEAs and needle-type electrode arrays. Up to
now, we have determined more than 30 spinal cord points,
at which the distinguishable reaction of limbs could be
watched. For data reading, the upper border of a defined
vertebra, the posterior median sulcus, and the surface of
the spinal cord were used as the origins of the x-, y-, and
z-coordinates, respectively. Adding voltage to a needle-type
electrode array, which was inserted into the spinal cord
with a distance of 0.6 mm left from the posterior median
sulcus and at a depth of 1.8 mm from the surface of the
spinal cord at vertebra L1, for instance, we found the
actions of the toe of the left hind foot. It was also found
that with a needle-type electrode array, most threshold val-
ues of the stimulating voltage could be lower than 1 V,
while with a cuff-type MEA, the threshold values were
mostly higher than 3 V.

3.3. Nerve signal regeneration from the left leg to the right
one

The schema of our experiment of the channel bridging

and signal regeneration on the rats’ sciatic nerves is shown
in Fig. 4(a), and the experimental results are shown in
Fig. 4(b). An FES signal generated by a pulse generator
was supplied to the hooked electrode A, which was inter-
faced on the ventral spinal cord of the rat where the corti-
Fig. 3. Schema of neural channel bridging between sciatic nerves in toads’ cospinal tract runs downwards. The first cuff MEA, B, was
legs (a) and the stimulating signal on point A in the left sciatic nerves and encircled on the intact left sciatic nerve to detect the neural
the regenerated nerve signal on point C2 in the right ones (in top–down signal for regeneration, while the others were on the distal
direction) (b). stump of the transected right one. The first channel joined
to C1 was used to evoke a new neural signal, and the sec-
generator with a needle electrode on point A. Another nee- ond one to C2 for monitoring the regenerated signal.
dle electrode was employed to detect the nerve signal acti- Under the condition of narcotism and external stimula-
vated by the foregoing one from A on point B. Grounded tion, the signal waveforms shown in Fig. 4(b) were
electrodes were arranged on point G between the stimulat- obtained from four electrodes, respectively. Then, the pulse
ing and detecting electrodes to realize the artefact suppres- generator was switched off and the waveforms were cap-
sion. Then, a nerve signal regeneration chip was used to tured from electrodes A, B, C1, and C2 (Fig. 4(c)). In this
transmit the nerve signal from the left sciatic nerve to the case, the spontaneous neural signal was regenerated from
right one. After that the regenerated signal was added on the left sciatic nerve to the right counterpart and detected
the right sciatic nerve through a needle electrode on C1. successfully.
Finally, a nerve signal was detected from another needle A pattern recognition raster diagram for signal C2 was
electrode on C2 by an oscillograph, as shown in Fig. 3(b). created (Fig. 4(d)), by adopting the coding method used
After the aforementioned in vitro experiments, in vivo by Mazurek et al. [17] and Jones et al. [18]. It is well known
ones were carried out. Sprague–Dawley (SD) rats weighing that each pulse corresponds to a neural spike. Therefore,
about 350 g were used in the following experiments. We the spike numbers of the five bursts from B, C1, or C2
anesthetized them with chloral hydrate (300 mg/kg), are 10, 9, 6, 3, and 1.
exposed the spinal cord in L10 (approximately 5 mm wide
and 15 mm long) and removed dura mater over the 3.4. Nerve signal regeneration between the spinal cord and
exposed area. transacted sciatic nerve

The schema of the channel bridging and signal regener-

3.2. Mapping relation between spinal cord stimulation and ation from the spinal cord to a sciatic nerve of a rat is
the motion of muscles shown in Fig. 5(a), and the experimental waveforms are
shown in Fig. 5(b). Here, the spontaneous neural signal
A series of important ‘‘mapping”, that is, the relation was detected by A from the intact spinal cord and sent into
between the stimulating points in the spinal cord and the the neural signal regenerator. After being amplified and
1266 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269

Fig. 5. Schema of the channel bridging and signal regeneration from the
spinal cord to a sciatic nerve bundle of a rat (a) and three signal
waveforms recorded from A, B1, and B2 (in top–down direction) while the
left sciatic nerve of the rat was flickering (b).

3.5. Bi-directional nerve signal regeneration of the transacted

spinal cord

Based on the experiments above, we carried out the bidi-

rectional signal channel bridging and regeneration experi-
ment on the transacted spinal cord of a rabbit, as
schematically shown in Fig. 6(a), and the signal waveform
shown in Fig. 6(b). Two cuff MEAs of large dimension
(U  10 mm) were used. Electrode A2, neural signal regen-
Fig. 4. (a) Schema of neural channel bridging between sciatic nerves with erating circuit NSR1, and electrode B1 formed the neural
evoked potential spikes from the spinal cord. (b) Four waveforms signal regeneration channel in the downward direction
recorded from electrodes A, B, C1, and C2 (in top–down direction) with while B3, NSR2, and A4 formed the neural signal regener-
stimulation on A. (c) Four waveforms recorded from the same electrodes
ation channel in the upward direction.
without stimulation on A. (d) The raster diagram of signal B in (c)
obtained by means of pattern recognition. First, the rabbit was narcotized, and an external stimu-
lating signal was applied to evoke the neural spikes. In
order to avoid self-oscillation, we adjusted the gain, G, of
processed, a related FES signal was generated and applied the neural signal regenerators NSR1 and NSR2 from
onto B1, which was joined to the distal stump of the trans- >500 to <500 and obtained the signal waveforms recorded
acted left sciatic nerve. The neural spikes evoked by B1 from A1, B2, B3, and B4 (in top–down direction) as shown
were detected by B2, which was interfaced to the site near in Fig. 6(b). Then, under the condition that there was no
B1, and sent to the oscillograph. external stimulation and the inner side of the right thigh
An interesting phenomenon was observed. After the sys- of the rabbit was knocked, the signal waveforms recorded
tem was switched on, the muscle near the distal stump of from B4, A4, and A3 (in top–down direction) were cap-
the transacted left sciatic nerve bundle twitched irregularly. tured, as shown in Fig. 6(c). The result demonstrated that
At the same time, we monitored the waveforms from A, B1, the motorial and sensory signals were regenerated in the
and B2 as shown in Fig. 5(b). bidirectional system.
 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269 1267

are interrupted and neural signals cannot be transmitted

through the injured section. Our idea for the channel
bridge and the signal regeneration of an injured spinal
cord is to build up such a system in which the transmis-
sion channels of the neural signals are bridged by means
of an embedded microelectronic device or module instead
of a nervous graft as made by Richardson et al. [2] and
David et al. [3].
The possibility that a microelectronic system can be
applied to connect the injured spinal cord is based on the
fact that a neural signal travelling in a nerve fibre is similar
to an electromagnetic wave travelling along a transmission
line or cable. Therefore, it is possible to detect or monitor
the neural signal with proper electrode or electrode array.
The challenge in detecting signals from a spinal cord is that
an array of electrodes should be arranged properly such
that the unnecessary signals from the nearby nerve fibres
can be shielded.
We choose cuff-type MEAs at first. As indicated by its
name, the cuff MEA becomes columnar when it is inserted
into the spinal cord and automatically encircles the cord.
Therefore, the cuff MEA has the advantage that for inter-
face with the spinal cord or nerve, it is non-invasive and
self-grasping. Another advantage is that all signals in the
spinal cord or nerve surrounded by the cuff MEA are
detectable in principle. For the primary experiments,
four-channel MEAs with 12 contact dots were selected
for all of the sciatic nerves and spinal cords of rats, and
for the experiments with the spinal cords of rabbits, six-
channel MEAs with 18 contact dots were used. For clinic
patients in the future, MEAs with channel numbers of
16, 32, 64, or more should be feasible. One disadvantage
of today’s cuff MEAs is that their contact dots are inter-
faced with the perineurium of the spinal cord and not
directly with the nerve fibres to be bridged. Thus, its thresh-
old value for an evident reaction is high and the selectivity
is poor. Therefore, we have developed a needle-type elec-
trode array, with which the threshold valve could be
reduced to <1 V. However, such electrode arrays are diffi-
cult to fix on a spinal cord. In order to overcome the disad-
vantages of both the cuff- and needle-type electrode arrays,
Fig. 6. (a) Schema of the signal channel bridging and bidirectional signal a new type of MEA by incorporating the cuff structure with
regeneration experiment on the spinal cord of a rabbit. (b) Signal needle-type electrode arrays has been developed and will be
waveforms recorded from A1, B2, B4, and A3 (in top–down direction)
when we adjusted the gain, G, of the neural signal regenerators NSR1 and
reported.
NSR2 from G > 500 to G < 500. (c) Signal waveforms recorded from B4, The channel bridging and signal regeneration is signifi-
A4, and A3 (in top–down direction) under the condition that the inner side cant for the following three cases:
of the right thigh of the rabbit was knocked.
(1) An injured spinal cord. This is the main goal of this
study.
4. Discussion (2) An injured peripheral nerve. The inosculation of the
injured peripheral nerves needs a long time, some-
A nervous system is, in fact, a complex communica- times many months, while the microelectronic-
tion network. Its function is mainly to realize the gener- embedded channel bridging can be made directly at
ation, processing, transmission, and reception of neural the accident site to maintain the activity of the distal
signals. Like an interrupted physical (sonic, optical, or peripheral nerves.
electrical) communication system, an injured spinal cord (3) An artificial channel from the spinal cord to a periph-
means that a part or all of the transmission channels eral nerve.
1268 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269

In Fig. 4(a), we choose both the left and the right sciatic (5) The waveform from C2 is especially interesting. At
nerves because the cuff MEAs are 12 mm long and it is dif- first, it is similar and synchronous to the ones from
ficult to wrap two cuff MEAs around one sciatic nerve. B and C1. With this argument, we can say that the
Another argument for this schema is that the action of spike from C2 is a regenerated version of the one
the left leg can be regarded as a reference to the right one. from B. Secondly, the narrower the spike from C2 ,
The essential characteristic of the experimental systems the stronger the ringing becomes, and the period of
shown in Figs. 3–6(a) is that they are of the hybrid elec- the ringing is shorter than the counterparts from B
tro-nervous system because, in each system, there is at least and C1. All these characteristics reveal that the wave-
one electronic system, one nervous system, one electro-to- form is more independent of the evoking spike from
nervous interface, and one nervous-to-electric interface. C 1.
The most important is that the electronic system can be
embedded into the nervous system. All the signals, from B, C1, and C2, shown in Fig. 4(c),
Our goal of the animal experiments is to demonstrate are typical spontaneous neural spikes. Comparable results
whether the hybrid electro-nervous system functioned as are the waveforms obtained by Vetter et al. [22].
desired. In comparison with the experiments on the left and right
In the in vitro toad’s experiment, when the stimulating sciatic nerves, the signal regeneration from the spinal cord
signal was added to the point A and the nerve signal regen- to a sciatic nerve bundle, as shown in Fig. 5(a), is more
eration chip did not work, we could just observe the twitch interesting, because it is an artificial transmission system
of the left hind limb of the toad. Then, after the chip was completely. The observed twitch of the muscle near the sci-
switched on, the right hind limb began to twitch synchro- atic nerve bundle in the left leg of the rat and the wave-
nously with the left one. At the same time, we monitored forms shown in Fig. 5(b) indicate that the distal stump of
the waveforms from C2, which is shown in Fig. 3(b). The the interrupted left sciatic nerve has been bridged to the
upper was a stimulating signal added to the left sciatic spinal cord.
nerve from signal regeneration, while the nether detected In the bidirectional experimental system in Fig. 6(a), we
a signal from the right sciatic nerve. As can be seen, after observed an oscillation when the gain of the neural signal
a direct electrical coupling signal, two pulses appeared in regenerators was greater than 500. This is not surprising
the nether. Both maxima of the widths of the two pulses since the bidirectional neural signal regenerating system
were all about 1 ms, which is well known as the typical along with the nervous network may form a closed feed-
value of one of the nerve signals. Therefore, we can con- back loop under special conditions. According to the
clude that a nerve signal has been regenerated by the micro- cybernetics, a closed feedback loop will oscillate by itself
electronic chip system successfully. when the loop gain is P1 and the loop phase is equal to
2np (n = 0, 1, 2, . . .). When, in our case, the system does
From Fig. 4(b), we can see the following characteristics: oscillate, it gives us an important indication: the channel
bridging system is established. Then, the next task is to
(1) The evoked potential signal, from B, is differentiable adjust the selectivity and the parameters of the electronic
from the FES square waveform applied to electrode system such that the self-oscillation is suppressed and the
A. This might be due to the nonlinearity of the ner- system functions correctly.
vous network. Through the preceding experiments, we are confident
(2) The waveform of the evoked potential signal, from B, that a microelectronics-embedded signal channel bridge
is similar to those obtained by Watkins et al. [19], can be established between the proximal and distal stumps
which means that the detected waveform is a neural of an interrupted nerve bundle. Moreover, both the evoked
signal. and spontaneous neural signals can be regenerated by such
(3) The positive potential spike from B, which is evoked an artificial electronic system.
by the rising flank of the square wave from A, corre-
sponds to the starting point of depolarization of neu- Acknowledgements
rons. Furthermore, the regenerated signals from C1
and C2 include a main spike with a period of about This work was supported by the National Natural Sci-
several milliseconds. This result is comparable with ence Foundation of China (Grant Nos. 69825101,
the one given by Patrick et al. [20] and Linderman 30270391, 30070255, and 90377013). We are grateful to
et al. [21], which means that it is the neural spikes, Dr. Klaus Peter Koch of the Fraunhofer Institute for Bio-
which are evoked by the rising flank of the square medical Engineering, 66386 St. Ingbert, Germany, for the
wave from A, that are regenerated from the left sci- support with MEAs and useful discussions.
atic nerve to the right one.
(4) In the waveforms from B, C1, and C2, there is a ring- References
ing procedure with decaying amplitude. This phe-
nomenon can only be explained again by the [1] Cajal YR. Degeneration and regeneration of the nervous system. New
nonlinearity of neural networks. York: Oxford Univ Press; 1991.
 Z. Wang et al. / Progress in Natural Science 19 (2009) 1261–1269 1269

[2] Richardson PM, McGuinness UM, Aguayo AJ. Axonal elongation [14] Stieglitz T, Meyer JU. Implantable microsystems: polyimide-based
into peripheral nervous system ‘‘bridge” after central nervous system neuro-prostheses for interfacing nerves. Med Device Technol
injury in adult rats. Nature 1980;284:264–5. 1999;10:28–30.
[3] David S, Aguayo AJ. Axons from CNS neurons regenerate into PNS [15] Wang HL, Lü XY, Wang ZG, et al. Circuit design for neural signal
grafts. Science 1981;214:931–3. recording with cuff electrode. In: Proceedings of the world engineers’
[4] Raisman G. Olfactory ensheathing cells and repair of brain and convention. Shanghai, China; 2004. p. 157–61.
spinal cord injuries. Cloning Stem Cells 2004;6:364–8. [16] Li WY, Zhang ZY, Wang ZG. Function electrical stimulation
[5] Cummings BJ, Uchida N, Tamaki SJ, et al. Human neural stem cells signals generator circuits for the central nerve and the sciatic nerve.
differentiate and promote locomotor recovery in spinal cord-injured In: Proceedings of the 2005 IEEE engineering in medicine
mice. PNAS 2005;102:14069–74. and biology 27th annual conference. Shanghai, China; 2005. p.
[6] Moritz CT, Lucas TH, Steve I, et al. Forelimb movements and muscle 183–6.
responses evoked by microstimulation of cervical spinal cord in [17] Mazurek ME, Shadlen MN. Limits to the temporal fidelity of cortical
sedated monkeys. J Neurophysiol 2007;97:110–20. spike rate signals. Nat Neurosci 2002;5:463–71.
[7] Jackson A, Fetz EE. Compact movable microwire array for long- [18] Jones LM, Depireux DA, Simons DJ, et al. Robust temporal coding
term chronic unit recording in cerebral cortex of primates. J in the trigeminal system. Science 2004;304:1986–9.
Neurophysiol 2007;98:3109–18. [19] Watkins PT, Kier RJ, Lovejoy RO, et al. Signal amplification,
[8] Jackson A, Mavoori J, Fetz EE. Long-term motor cortex plasticity detection and transmission in a wireless 100-electrode neural
induced by an electronic neural implant. Nature 2006;444(7115): 56–60. recording system. Circuits and systems. ISCAS 2006;4:2193–6.
[9] Jackson A, Moritz CT, Mavoori J, et al. The neurochip BCI: towards [20] Patrick E, Ordonez M, Alba N, et al. Design & fabrication of a
a neural prosthesis for upper limb function. IEEE Trans Neural Syst flexible substrate microelectrode array for brain machine interfaces.
Rehabil Eng 2006;14(2):187–90. In: The 28th international conference of the IEEE engineering in
[10] Wang ZG, Lu XY, Li WY, et al. Study of microelectronics for medicine and biology society, August 30-Sept. 3. New York City,
detecting and stimulating of central neural signals. In: IEEE USA; 2006. p. 2966–9.
proceedings of international conference on neural interface and [21] Linderman MD, Gilja V, Santhanam G, et al. An autonomous,
control. Wuhan, China; 2005. p. 192–5. broadband, multi-channel neural recording system for freely behav-
[11] Wang ZG, Gu XS, Lu XY. China Patent ZL 2005 1 0135541.6; 2008. ing primates. In: The 28th international conference of the IEEE
p. 12–24. engineering in medicine and biology society, August 30-Sept. 3. New
[12] Wang YF, Wang ZG, Lü XY, et al. A multichannel neural signal York City, USA; 2006. p. 1212–5.
detecting module: its design and test in animal experiments. Prog Nat [22] Vetter RJ, Miriani RM, Casey BE, et al. Development of a
Sci 2007;17(6):675–80. Microscale Implantable Neural Interface (MINI) probe system. In:
[13] Li WY, Wang ZG, Lü XY, et al. Multi-channel neural signal The 27th international conference of the IEEE engineering in
stimulating module and in-vivo experiments. J Southeast Univ medicine and biology society, October 19-October 22. Shanghai,
2007;23(1):26–30. China; 2005. p. 7341–4.