LEARNING IN BIOLOGICAL NEUROPROCESSORS
USING A CENTER OF AREA METHOD
José M. Ferrández, Victor Lorente
Departamento de Electrónica, Tecnología de Computadores y Proyectos, Univ. Politécnica de Cartagena, Cartagena, Spain
Félix de la Paz, José Manuel Cuadra, José R. Álvarez-Sánchez
Departamento de Inteligencia Artificial, UNED, Madrid, Spain
Eduardo Fernández
Instituto de Bioingeniería, Univ. Miguel Hernández de Elche, CIBER-BBN, Elche, Spain
Keywords:
Cultured neural network, Induced plasticity, Multielectrode recordings, Robotic control.
Abstract:
Learning in a biological neuroprocessor is analyzed using human neuroblastoma cultures and a center of area
method in order to guide a robot to follow the light or the brightest area in a limited scenario. The main
setup consists in an inverted microscope where a multielectrode array is attached with the biological cultures.
This elements amplifies and send the weak neural signals to a D/A card where analyzing process is achieved,
computing the movement of the robot, that is remotely linked to this computer. The robot also sends the a
picture of the scenario to the computer in order to stimulate the culture with a center of area scheme. In this
paper, it is shown that learning is possible in this culture, and guiding the robot to a desired goal is a feasible
task.
1 INTRODUCTION
Mammalian nervous systems exhibit complex com-
putational functions including sensory functions, mo-
tor function and in humans, abstract thought. In par-
ticular, pattern recognition exhibited in our olfactory,
visual and auditory functions are of particular interest
to the electronic and computing communities. Mean-
while several approaches attempt to mimic/substitute
sensory or neural elements (missing by congenital
state or due to pathological processes) in order to en-
able/restore function by establishing neuro-electronic
interfaces.
Classical computational paradigms consist in se-
rial and supervised processing computations with
high-frequency clocks silicon processors, with mod-
erate power consumption, and fixed circuits structure.
In contrast, the brain uses millions of biological pro-
cessors, with dynamic structure, slow commutations
compared with silicon circuits, low power consump-
tion and unsupervised learning. There have been nu-
merous approaches to creating bioinspired parallel
processing. However, silicon provides a fundamen-
tally different technological platform to that of neu-
robiology. Neurons – the core technology component
has a huge number of interconnections compared to 3
in traditional transistors. This provides considerably
more computational power. Furthermore, this extraor-
dinary connectivity is coupled with natural unsuper-
vised learning based on varying connective efficiency.
Our learning experiments were performed in neu-
ral cultures containing 120.000 human neuroblastoma
SY-5Y, under the assumption that this kind of cells
are able to respond electrically to external stimuli and
modulate their neural firing by changing the stimu-
lation parameters. Such cultured neuroblastoma net-
works have shown dynamical configurations, being
able to grow and adapt functionally in response to
external stimuli over different configuration patterns.
We are especially interested in analyzing if popula-
tions of neuroblastoma cells are able to process and
store information, and if learning can be implemented
336
M. Ferrández J., Lorente V., de la Paz F., Manuel Cuadra J., R. Álvarez-Sánchez J. and Fernández E..
LEARNING IN BIOLOGICAL NEUROPROCESSORS USING A CENTER OF AREA METHOD.
DOI: 10.5220/0003084003360343
In Proceedings of the International Conference on Fuzzy Computation and 2nd International Conference on Neural Computation (ICNC-2010), pages
336-343
ISBN: 978-989-8425-32-4
Copyright
c
2010 SCITEPRESS (Science and Technology Publications, Lda.)
over this biological structure. The main objective of
this work will be to control a robot using this biolog-
ical neuroprocessor and a simple center of area learn-
ing scheme. The final system could be applied for
testing how chemicals affect the behavior of the robot
or to establish the basis for new hybrid optogenetic
neuroprostheses.
2 LEARNING IN HUMAN
NEUROBLASTOMA CULTURES
The physiological function of neural cells is modu-
lated by the underlying mechanisms of adaptation and
reconfiguration in response to neural activity. Heb-
bian learning describes a basic mechanism for synap-
tic plasticity wherein an increase in synaptic efficacy
arises from the presynaptic cell’s repeated and persis-
tent stimulation of the postsynaptic cell. The theory is
commonly evoked to explain some types of associa-
tive learning in which simultaneous activation of cells
leads to pronounced increases in synaptic strength.
The N-methyl-D-aspartate (NMDA) receptor, a sub-
type of the glutamate receptor, has been implicated as
playing a key role in synaptic plasticity in the CNS
(Bading and Greenberg, 1991), where as dopamine
receptors are involved in the regulation of motor and
cognitive behaviors. For most synaptic ion channels,
activation (opening) requires only the binding of neu-
rotransmitters. However, activation of the NMDA
channel requires two events: binding of glutamate (a
neurotransmitter) and relief of Mg
++
block. NMDA
channels are located at the postsynaptic membrane.
When the membrane potential is at rest, the NMDA
channels are blocked by the Mg
++
ions. If the mem-
brane potential is depolarized due to excitation of the
postsynaptic neuron, the outward depolarizing field
may repel Mg
++
out of the channel pore. On the
other hand, binding of glutamate may open the gate
of NMDA channels (the gating mechanisms of most
ion channelsare not known). In the normal physiolog-
ical process, glutamate is released from the presynap-
tic terminal when the presynaptic neuron is excited.
Relief of Mg
++
block is due to excitation of the post-
synaptic neuron. Therefore, excitation of both presy-
naptic and postsynaptic neurons may open the NMDA
channels, this is closely related with Hebbian learn-
ing.
Another important feature of the NMDA channel
is that it conducts mainly the Ca
++
ion which may
activate various enzymes for synaptic modification,
even nitric oxide has been identified as a relevant el-
ement in synaptic regulation. The enhancement of
synaptic transmission is called the long-term poten-
Figure 1: Human neuroblastoma cells, showing different
growing stages and neuritic development.
tiation (LTP), which involves two parts: the induction
and the maintenance. The induction refers to the pro-
cess, which opens NMDA channels for the entry of
Ca
++
ions into the postsynaptic neuron. The subse-
quent synaptic modification by Ca
++
ions is referred
to as the maintenance of LTP.
A human neuroblastoma SY5Y cell line, that ex-
press clonal specific human dopamine receptors, and
also NMDA receptors, will be the biological platform
for studying learning in cultured cells.
Neuroblastoma SH-SY5Y cells are known to be
dopaminergic, acetylcholinergic, glutamatergic and
adenosinergic, so in this line they respond to differ-
ent neurotransmitters. The cells have very different
growth phases, as it can be seen in Figure 1. The
cells both propagate via mitosis and differentiate by
extending neurites to the surrounding area. The di-
viding cells can form clusters of cells which are re-
minders of their cancerous nature, but chemicals can
force the cells to dendrify and differentiate, in some
kind of neuritic growth.
As conclusion, neuroblastoma culture cells show
electrophysiological responses similar to standard
neurons, as potential actions generation sensible to
tetrodotoxin (TTX) and acetylcholyn. They have neu-
rotransmitters synthesis process and are able to neu-
ritic growth in culture medium.
3 EXPERIMENTAL SETUP
The neuro-physiology setup provides a complete so-
lution for stimulation, heating, recording, and data
acquisition from 64 channels. The MEA (microelec-
trode array) system, (Rolston et al., 2009; Hales et al.,
2010), is intended for extracellular electrophysiolog-
ical recordings in vitro of different applications that
include acute brain, heart, and retina slices; cultured
slices; and dissociated neuronal cell cultures, see Fig-
ure 3.
The basic components of the proposed system are
LEARNING IN BIOLOGICAL NEUROPROCESSORS USING A CENTER OF AREA METHOD
337
Cell Culture
Computer
Robot
Sensors
Stimulator
C
O
M
M
o
t
o
s
r
Figure 2: Experimental setup, see text for detailed explana-
tion.
Figure 3: Neuroblastoma cells over multielectrode array,
the picture shows a rectangle with sides about 1 mm. long.
shown in Figure 2. These components are:
1. A microelectrode array is an arrangement of 60
electrodes that allows the simultaneous targeting
of several sites for extracellular stimulation and
recording. Cell lines or tissue slices are placed
directly on the MEA, see figure , and can be cul-
tivated for up to several months. Almost all ex-
citable or spontaneously active cells and tissues
can be used.
2. Raw data from the MEA electrodes are amplified
by MCS filter amplifiers with custom bandwidth
and gain, which are built very small and compact
using SMD (Surface Mounted Devices) technol-
ogy. The small-sized amplifier combines the in-
terface to the MEA probe with the signal filtering
and the amplification of the signal. The compact
design reduces line pick up and keeps the noise
level down. The amplifiers are mounted over an
inverted microscopes.
3. The analog input signals are then acquired and
digitized by the MC-Card that is preinstalled on
the data acquisition computer, that supplies the
power for the amplifiers, and the pattern stimuli
to the stimulators.
4. The robot sends information about the environ-
ment to the computer using a bluetooth link. The
sensor consists in infrared sensors for detecting
obstacles.
It has been developed a system that provides a com-
plete robotic control platform over remote neural
cultures. The system includes free, open-source,
console-based programs written in C/C++ for real-
time robotic applications with embodied cultures. All
of this software has been developed for the Linux
Operating System and MCS hardware (MultiChan-
nel Systems, Reutlingen, Germany). Using this soft-
ware in conjunction with MEABENCH is specially
intended for close-loop experiments.
The software developed consists of the following
programs:
1. Cult2Robot: The main program. It has been de-
veloped as a MEABENCH module, so it can read
spikes information from MEABENCH spike de-
tector and compute a direction vector based on
MEA neural activity. The direction vector is cal-
culated based on the number of spikes per elec-
trode in t seconds and it can be weighted by the
height and width of the spikes. This direction vec-
tor can be sent to a robot to control its movement.
2. Stg_control: This program controls a general-
purpose two channel stimulus generator.
3. BT_server: Non-blocking Bluetooth server that
uses RFCOMM protocol to receive characters
from a specific MAC, process the information and
do some action, this protocol is used for remote
controlling the robot.
4. Remote environment: the computer that is com-
municating directly with the robot include TCP-
IP facilities for reading/sending information to the
bio-hybrid platform, wherever it is located.
Bluetooth client and server will be used with a hu-
manoid robot (Robonova, Hitec Robotics) and a two-
wheeled robot (e-puck, www.epuck.org) to send and
receive information about obstacles.
The system comprises:
1. A bio-hybrid robot control: It includes the devel-
opment of a whole bio-hybrid hardware/software
platform with for robotic guidance.
2. A remote communication system with the neural
culture: It has been developed a remote environ-
ment for communicating the robot with any bio-
hybrid control platform through TCP-IP links.
3. Creation of a bio-hybrid robotic control: Design
the neuromorphic processes, e.g. obstacle avoid-
ance tasks, implemented over the bio-hybrid sys-
tem.
4 METHODS
Human neuroblastoma cultures were produced us-
ing the commercial line SH/SY5Y. Neural cells were
ICFC 2010 - International Conference on Fuzzy Computation
338
then plated on Micro-Electrode Arrays -MEAs (Mul-
tiChannel Systems, Reutlingen, Germany). Initially
the nitrogen frozen cells, was immersed in a 37 de-
gree bath, and centrifuged at 1000 rpm during 5 min-
utes. When cells have grown in a uniform mono-
layer process, they are washed three time with buffer
Phosphate-buffered saline (PBS) for keeping the pH
approximately constant. 0,5 per cent trypsin was
added to the solution in order to re-suspend cells ad-
herent to the cell culture dish wall during the process
of harvesting cells. The cells were kept in the in-
cubator for 5 minutes and passed through a 40 µm.
cell strainer (Falcon, Bedford, MA) to remove large
debris. Finally the cells are transferred to a specific
medium in order to inactivate trypsin, and centrifuged
again during 5 minutes at 1000 rpm.
For seeding the plate cells are stained with try-
pan blue, (because cells that loose their permeabil-
ity get colored with this solution) and counted with a
Neubauer chamber. Finally, 80.000 or 120.000 total
neuroblastoma cells have been placed over the MEA
substrate.
Maintaining cells in culture is essential for study-
ing their physiological properties. Cell culturing is
dependent on the growth surfaces and cells must ad-
here to the electrode substrate in order to establish the
best connection with the electrodes material. For most
cultures coated tissue culture plates are prerequisite
for seeding. The most commonly used coatings are
positively charged polymers. In this work, the insu-
lation layer (silicon nitride) of some of the plates was
pre-treated with polyethyleneimine (PEI), showing no
advantages compared with no covered plates.
The neuroblastoma cultures are maintained in a 37
degree humidified incubator with 5 per cent CO
2
and
95 per cent O
2
with serum-free Neurobasal medium.
Under the aforementioned conditions we were able to
record stable electrophysiological signals over differ-
ent days in vitro (Div). The medium was replaced
one-half of the medium every 5 days.
5 RESULTS
The cultured neuroblastoma cells establish synap-
tic connections. It can be seen differentiated and
non-differentiated neuroblastoma cell bodies grow-
ing around the whole electrode population. The den-
dritic arborescence is more evident in the magnifica-
tion Figure 1 b) where differentiated neural cells sur-
round the four electrodes while the rest of the cells are
in their growing process. This Figure corresponds to
80.000 neuroblastoma cells seeded in a no-PEI MEA
at 2nd day in vitro (div).
40
20
0
-20
-40
40
20
0
-20
-40
40
20
0
-20
-40
40
20
0
-20
-40
40
20
0
-20
-40
40
20
0
-20
-40
40
20
0
-20
-40
40
20
0
-20
-40
21
22
23
24
25
26
27
28
31
32
33
34
35
36
37
38
41
42
43
44
45
46
47
48
51
52
53
54
55
56
57
58
61
62
63
64
65
66
67
68
71
72
73
74
75
76
77
78
82
83
84
85
86
87
12
13
14
15
16
17
µV
-1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1
sec
thresholds
-3 std. dev.
x
spikes
a)
population activity
200
150
100
60000
40000
20000
total spike number [%;spikes]
0
day 1 day 5 day 16
pre stimulus post stimulus
80000
b)
Figure 4: a) Spontaneous neural activity detected by the
multielectrode array. b) Induced neural activity by tetaniza-
tion stimuli.
The electrophysiological properties of the neu-
roblastoma cultures were analyzed by recording the
spontaneous activity of the network. Time course of
experiments was over 15 days; recordings were done
using two MCS-Meas with two neuroblastoma cell
cultures (but only in one the cells survivedtill day 15).
In vitro neuroblastoma networks show spontaneously
firing. This firing rates change during the culture de-
velopment with marked day differencesand the global
rate is closely related to the age of the network.
The physiological recordings correspond to neu-
roblastoma cultures in the range of 1-7 div. They
show bursting and spiking activity, with usually neg-
ative depolarizations. Figure 4 a) show the spiking
activity of the neural population with an automatic
detection level for each electrode. This is very con-
venient if one has multiple channels for extracting
spikes.
The standard deviation of each data trace is used
to estimate its spike threshold. A time interval of 500
ms is used to calculate the standard deviation. By fix-
ing the factor, by which the standard deviation is mul-
tiplied, the sign of the factor determines whether the
spike detection level is positive or negative, only val-
ues above this will be extracted as spiking activity.
A value between 1 and 4 is appropriate for most
applications the threshold was fixed at standard devi-
ation equal to 3 with respect to the electrode activity
LEARNING IN BIOLOGICAL NEUROPROCESSORS USING A CENTER OF AREA METHOD
339
in order to identify spikes embedded in the noisy sig-
nals.
During the neuroblastoma development, a wide
range of population bursting or synchronized activity
has been observed, according to some studies in neu-
ral cultures preparations (Wagenaar et al., 2006; Rol-
ston et al., 2007; Madhavan et al., 2007; Esposti et al.,
2009). The burst usually contains a large number of
spikes at many channels, with variable duration, from
milliseconds to seconds.
5.1 Tetanic Stimulation
Spontaneous activity was recorded for intervals of 3
minutes before stimulation (PRE-data), and the total
number of spikes extracted was counted. The biphasic
stimulus consists in a 10 trains of a 100 anodic-first
waveform with 1 Volt amplitude delivered to all 60
electrodes in order to propagate a tetanization stimu-
lus to the neuroblastoma culture.
In neurobiology, a tetanic stimulation consists of
a high-frequency sequence of individual stimulations
of a neuron. It is associated with long-term poten-
tiation, the objective of this work. High-frequency
stimulation causes an increase in transmitter release
called post-tetanic potentiation (Antonov et al., 2003).
This presynaptic event is caused by calcium influx.
Calcium-protein interactions then produce a change
in vesicle exocytosis. Some studies (Jimbo et al.,
1998) use repetitive stimulation for training neural
cultures, achieving activity potentiation or depres-
sion.
Once the tetanization stimulus was applied to the
whole population 5 minutes after the stimulation a
3 minutes interval was recorded (POST-data). Only
neuronal signals which had at least a 2:1 signal:noise
ration were valued as "spikes". Again, the total num-
ber of spikes extracted was counted. This process was
made for cultures at 1 day in vitro (div), 5 div and 16
div. Figure 4 b) represents the counted spikes with
bar charts for the different recordings. The conclu-
sion from this Figure is:
1. While the neuroblastoma culture is growing new
connections are created, and the number of spikes
increases as the culture expands over the MEA.
2. After a tetanic stimulation the cells continue with
their increased spiking rate, providing a persistent
change in the culture behavior. When this change
in the network response lasts, these changes can
be called learning.
In all the experimentation performed, tetanic stimula-
tion was applied as training method, and the electro-
physiological properties of the neuroblastoma culture
a)
b)
?
c)
d)
e)
f)
g)
h)
Figure 5: Obstacles avoidance using area center. These pic-
tures form a sequence showing how a robot overcomes an
obstacle using the center of area method. The current ad-
vance sector is the dark blue area, discarded sectors are in
light blue. a) Area center (magenta/dark circle) is accessi-
ble: follow it (using full advance sector). b) Area center
is still accessible: follow it (using full advance sector). c)
Area center became inaccessible (red/dark circle): shrink
and split the advance sector, choose one side and turn to its
area center (yellow/lighter or brown/darker circles), remem-
ber (local coordinates) the split point (marked with arrow).
d) Left side has been chosen: as robot turns the sector is
expanded to the right to cover the split point and increase
restricted advance sector radius as function of sector ampli-
tude increment, follow sector area center (yellow/light cir-
cle). e) Robot is going past the split point: continue expand-
ing the sector, continue following sector area center (using
restricted advance sector). f) Split point has been gone past,
the advance sector has been expanded to its initial radius
and amplitude: forget split point, continue following sector
area center (magenta/dark), the full advance sector has been
recovered. g) Area center is accessible: follow it (using full
advance sector). h) Robot has overcome the obstacle, area
center is accessible: follow it.
change, getting a potentiation effect on the sponta-
neous firing, modulating in this way the culture neural
activity.
5.2 Robotic Control
For controlling the direction of the robot we propose
to compute the winner neurons (that is the ones that
increase more its firing characteristics) resulting from
neural activity recorded in the human neuroblastoma
culture stimulated using a center of area method (Ál-
varez Sánchez et al., 2010; Álvarez Sánchez et al.,
2009).
This method is a new brand of navigation meth-
ICFC 2010 - International Conference on Fuzzy Computation
340
Figure 6: Image acquired by a robot over MEA.
ods, at this time a reactive version has been devel-
oped. The method computes the center of area of a
frontal sector, the advance sector, of the robot free
perceived area and use its position for robot driving.
The robot approximately follows the center of area
path, see Figure 5. The idea is that the center of area
is normally a safe place to go, but when an obstacle
lies near robot path, the center of area could became
inaccessible inside the obstacle, so there is no way
to follow it. In this case the advance sector is split
in two shorter sectors using a point of the obstacle
as a reference. If the centers of area of both sectors
are accessible one of them is chosen, randomly or by
some external preference. If only one center is acces-
sible it is selected. If both centers are inaccessible, no
way, then the method determines the center of area of
the robot shrunk rear area and robot turns to it as an
escape maneuver. Successive splits may be done, if
needed, so robot can drive through complex configu-
ration of obstacles in a safe way, even passing through
narrow places and following smooth paths.
A modification of the method makes it capable of
goal reaching avoiding obstacles, still being, at least
at these preliminary stages of its development, a reac-
tive method.
Note that this center of area concept is a construc-
tion that emerges from the visual or ranging percep-
tion of the robot. In Figure 6, it can be seen a simu-
lation of a robot walking through a cave. This image
will be digitized in three grey levels, black, white and
grey, in order to provide three differentstimulations to
the neural culture, no stimulation, high tetanization,
and medium stimulation respectively.
The resulting stimulation configuration is shown
in Figure 7 a). White boxes correspond to
no-stimulation, red boxes correspond to medium
tetanization, while blue electrodes will deliver high
tetanization according with the acquired cave im-
age. Medium tetanization will consist in five trains
of a hundred anodic first pulses with 1 V amplitude,
while high tetanization will provide 1,5 V anodic first
1 2
1 3
1 4
1 5
1 6
1 7
82
83
84
85
86
2 2
2 1
2 3
2 4
2 5
2 6
31
32
33
34
35
38
41
42
43
44
51
52
53
54
55
58
61
62
63
64
71
72
73
74
75
76
78
... 100 pulses
... 100 pulses
high
medium
mV.
1000
500
0
-500
-1000
1500
-1500
1000
500
0
-500
-1000
a)
b)
Figure 7: Selective electrode tetanization corresponding to
data acquired from image shown in Figure 6. a) MEA
representation showing three groups of electrodes, white
ones are not stimulated, light blue ones are stimulated with
high tetanization and red (darker) ones are stimulated with
medium tetanization. b) High and medium tetanization
pulses train, anodic first waveform.
pulses, Figure 7 b). From this example, it is expected
that electrodes that cover the center of area of the grey
and white image, that is the electrodes centered at the
bottom of the image, electrodes 48 and 58, will in-
crease more their activity. This winner neuron coordi-
nates will be provided to the robot in order to guide its
movement. In the new robot position the camera will
send a new image, and the information will be passed
to the computer in order to induce a selective tetaniza-
tion of the biological neural network for changing the
resulting direction vector.
Again, human neuroblastoma cultures were pro-
duced using the commercial line SH/SY5Y. Cell cul-
ture of SH SY5Y was grown in DMEM (Gibco) com-
pleted with 10% of fetal bovine serum at 37
o
C in 5%
CO
2
and humidify atmosphere.
The electrophysiological properties of the neu-
roblastoma cultures were analyzed by recording the
spontaneous activity of the network. In vitro neurob-
lastoma networks showed spontaneously firing. This
firing rates changed during the culture development
with marked day differences and the global rate was
closely related to the age of the network.
Recordings of Neuroblastoma SH-SY5Y has the
disadvantage of having a very low signal to noise ra-
tio. As we have shown in previous papers, the elec-
trophysiological properties of the culture change with
the age of the culture, getting a potentiation effect
in the spontaneous firing. A young neuroblastoma
culture (1-5 DIV) has a low spontaneous firing ac-
tivity, with a signal to noise ratio barely higher than
1:1. A mature neuroblastoma culture (1-15 DIV)
have a higher spontaneous firing activity and its snr
may be higher than 2:1, but still is lower than snr
of other cells. The physiological recordings corre-
spond to neuroblastoma cultures in the range of 1-7
div. They showed bursting and spiking activity, with
usually negative depolarizations. It was used also an
automatic detection level for each electrode. This is
LEARNING IN BIOLOGICAL NEUROPROCESSORS USING A CENTER OF AREA METHOD
341
50
40
30
20
10
0
1
2 3
4 5
6 7 8 9
E58
E19
E22
E26
E29
E30
It. E58 E19 E26 E29 E30
1 7 11 4 7 17 4
2 7 7 3 10 12 9
3 4 8 5 9 11 5
4 6 12 5 9 11 5
5 15 9 10 6 13 4
6 8 6 4 7 8 1
7 8 2 3 14 15 1
8 12 2 10 9 12 4
9 9 6 6 8 10 4
E22
2360 25 28 31 34 36 61
20 21 24 29 35 38 39
18 19 27 32 37 40 41
15 16 17 26 33 42 43 44
14 13 12 3 56 47 46 45
11 10 7 2 57 52 49 78
9 8 5 0 59 54 51 50
62 6 4 1 58 55 53 63
Figure 8: Spontaneous neural activity detected previous to
tetanization. Spikes number (y-axis) from six electrodes
were recorded 9 times (x-axis). Left table shows data in nu-
meric format, right table shows a MEA representation with
selected electrodes marked out. Spikes number is always
below 20.
very convenient if one has multiple channels for ex-
tracting spikes.
The standard deviation of each data trace was
used to estimate its spike threshold and computing the
spikes per channel. A time interval of 500 ms was
used to calculate the standard deviation. The thresh-
old was fixed at standard deviation equal to 4 with
respect to the electrode activity in order to identify
spikes embedded in the noisy signals.
Spontaneous activity was recorded for intervals of
5 seconds before stimulation, and the total number of
spikes extracted was counted for each channel. Figure
8 shows the total number of spikes per channel, and
a graphic visualization of this data for six different
electrodes located at different positions at the neural
culture.
When the tetanization configuration shown in Fig-
ure 7 was applied, the spiking characteristics of the
neuroblastoma culture changed. The computed spikes
per channel is shown in Figure 9 during the tetaniza-
tion process. It can be seen that the most significant
increment registered is at electrode 58, that matches
the center of area of the provided image, guiding in
this way the robot to the light. When the image of
the cave was presented once again, that is the same
selective stimulation was provided, the registered ac-
tivity was again modified. In Figure 10 it can be seen
a clearly potentiation effect in electrode 58 about 4
times, while the rest of the electrodes did not show
any significant increase. In this way, selecting the ori-
entation of the robot as the neuron or group of neu-
rons that increase in a more quantitative aspect its fir-
ing characteristics, it is possible to guide the robot to
the light or the brightest area of the discretized scene.
This changes last for seconds, in this case a biped
robot will be the perfect candidate due to its limited
50
40
30
20
10
0
1
2 3
4 5
6
E58
E19
E22
E26
E29
E30
It. E58 E19 E22 E26 E29 E30
1 5 7 6 16 6 4
2 5 1 6 12 18 4
3 3 9 6 10 6 3
4 46 13 7 12 6 4
5 9 7 3 4 14 3
6 7 4 7 8 7 4
2360 25 28 31 34 36 61
20 21 24 29 35 38 39
18 19 27 32 37 40 41
15 16 17 26 33 42 43 44
14 13 12 3 56 47 46 45
11 10 7 2 57 52 49 78
9 8 5 0 59 54 51 50
62 6 4 1 58 55 53 63
Figure 9: Spontaneous neural activity detected during
tetanization. It shows a considerable increase in activation
at electrode 58. See figure 8 caption for more explanations.
250
200
150
100
50
0
1
2 3
4 5
6
E58
E19
E22
E26
E29
E30
It . E58 E19 E22 E26 E29 E30
1 23 10 8 8 5 5
2 148 10 11 11 20 8
3 198 15 7 15 15 3
4 33 15 5 9 16 11
5 24 9 4 8 14 6
6 3 14 9 8 13 4
2360 25 28 31 34 36 61
20 21 24 29 35 38 39
18 19 27 32 37 40 41
15 16 17 26 33 42 43 44
14 13 12 3 56 47 46 45
11 10 7 2 57 52 49 78
9 8 5 0 59 54 51 50
62 6 4 1 58 55 53 63
Figure 10: Spontaneous neural activity detected after se-
lective tetanization. Note the different y-axis scale in this
figure respect to figures 8 and 9. Activation at electrode 58
is bigger than before.
movement. The robotic control will be refined in fu-
ture works, while in this paper the plasticity of the
center of area stimulation is presented.
6 CONCLUSIONS
Learning in cultured neuroblastoma networks by a
stimulation process requires identifying the correct
stimuli to provide to the neurons in culture. These
neuroblastoma networks form a large culture covering
the whole electrode array and generating a rich den-
dritic configuration. The connectivity can be modu-
lated by external stimulation as has been described in
many studies, (Bakkum et al., 2007; Bakkum et al.,
2008b; Bakkum et al., 2008a; Chao et al., 2008), but
also the activity of the network can be modulated with
the appropriate stimulation scheme. Tetanization con-
sists in high-frequency stimulation to the culture, in
order to cause an increase in transmitter release called
post-tetanic potentiation. The results illustrate the ex-
istence of qualitatively different responses to stimu-
ICFC 2010 - International Conference on Fuzzy Computation
342
lation. Our results indicate the existence of a clear
facilitation mechanism in response to the tetaniza-
tion stimuli at different stages of cell development.
By selective tetanizing some parts of the culture, the
network changes its firing characteristics in seconds,
modifying in this way its electrical behavior, and it
has been shown that if the brightest area of the sce-
nario, induces more stimulation in its corresponding
part of the culture, then the increase in the firing prop-
erties of the neurons that represent the area where
light is detected is observed. Future work consists in
determining the optimal stimulation to apply for in-
ducing permanent firing changes in the culture. These
aspects will then constitute the basis for analyzing the
behavior change by adding chemicals to the culture,
and for designing new optogenetic hybrid learning
schemes. A more detailed robotic control will be also
studied analyzing the culture time responses.
ACKNOWLEDGEMENTS
This work was supported by the Spanish Govern-
ment through grants TIN2008-06893-C03,TEC2006-
14186-C02-02 and SAF2008-03694, Cátedra Bidons
Egara, Fundación Séneca 08788/PI/08, CIBER-BBN
and by the European Comission through the project
"‘NEUROPROBES"’ IST-027017.
REFERENCES
Álvarez Sánchez, J. R., de la Paz López, F., Cuadra
Troncoso, J. M., and de Santos Sierra, D.
(2010). Reactive Navigation in Real Environ-
ments Using Partial Center of Area Method.
Robotics and Autonomous Systems. In press,
http://dx.doi.org/10.1016/j.robot.2010.05.009.
Álvarez Sánchez, J. R., de la Paz López, F., Cuadra Tron-
coso, J. M., and Rosado Sánchez, J. I. (2009). Par-
tial Center of Area Method Used for Reactive Au-
tonomous Robot Navigation. In Mira, J., Ferrández,
J. M., Álvarez, J. R., de la Paz, F., and Toledo, F. J.,
editors, Bioinspired Applications in Artificial and Nat-
ural Computation, volume 5602 of LNCS, pages 408
418. Springer Verlag.
Antonov, I., Antonova, I., and Kandel, E. (2003). Activity-
dependent presynaptic facilitation and hebbian LTP
are both required and interact during classical condi-
tioning in aplysia. Neuron, 37(1):135–147.
Bading, H. and Greenberg, M. (1991). Stimulation of pro-
tein tyrosine phosphorylation by NMDA receptor ac-
tivation. Science, 253(5022):912–914.
Bakkum, D. J., Chao, Z. C., and Potter, S. M. (2008a).
Long-term activity-dependent plasticity of action po-
tential propagation delay and amplitude in cortical
networks. PLoS One, 3(5):e2088. Online Open-
Access paper.
Bakkum, D. J., Chao, Z. C., and Potter, S. M. (2008b).
Spatio-temporal electrical stimuli shape behavior of
an embodied cortical network in a goal-directed learn-
ing task. Journal of Neural Engineering, 5:310–323.
Bakkum, D. J., Gamblen, P. M., Ben-Ary, G., Chao, Z. C.,
and Potter, S. M. (2007). MEART: the semi-living
artist. Frontiers in NeuroRobotics, 1(5):1–10. Online
Open-Access paper.
Chao, Z. C., Bakkum, D. J., and Potter, S. M. (2008).
Shaping embodied neural networks for adaptive goal-
directed behavior. PLoS Computational Biology,
4(3):e1000042. Online Open-Access paper, supple-
ment, and movie.
Esposti, F., Signorini, M. G., Potter, S. M., and Cerutti,
S. (2009). Statistical long-term correlations in dis-
sociated cortical neuron recordings. IEEE Transac-
tions on Neural Systems & Rehabilitation Engineer-
ing, 17(4):364–9.
Hales, C. M., Rolston, J. D., and Potter, S. M. (2010).
How to culture, record and stimulate neuronal net-
works on micro-electrode arrays (MEAs). JoVE,
39. doi: 10.3791/2056. Online video tutorial:
http://www.jove.com/index/Details.stp?ID=2056.
Jimbo, Y., Robinson, H., and Kawana, A. (1998). Strength-
ening of synchronized activity by tetanic stimulation
in cortical cultures: application of planar electrode ar-
rays. IEEE transactions on Biomedical Engineering,
45(11):1297–1304.
Madhavan, R., Chao, Z. C., and Potter, S. M. (2007). Plas-
ticity of recurring spatiotemporal activity patterns in
cortical networks. Physical Biology, pages 181–193.
Rolston, J. D., Gross, R. E., and Potter, S. M. (2009). A
low-cost multielectrode system for data acquisition
and real-time processing with rapid recovery from
stimulation artifacts. Frontiers in Neuroengineering,
2(12):1–17. Online Open-Access paper.
Rolston, J. D., Wagenaar, D. A., and Potter, S. M. (2007).
Precisely timed spatiotemporal patterns of neural ac-
tivity in dissociated cortical cultures. Neuroscience,
148:294–303.
Wagenaar, D. A., Pine, J., and Potter, S. M. (2006). An ex-
tremely rich repertoire of bursting patterns during the
development of cortical cultures. BMC Neuro-science,
7:11.
LEARNING IN BIOLOGICAL NEUROPROCESSORS USING A CENTER OF AREA METHOD
343