Biosensors and Bioelectronics 48 (2013) 188–196

Contents lists available at SciVerse ScienceDirect

Biosensors and Bioelectronics

journal homepage: www.elsevier.com/locate/bios

Toward embedded laboratory automation for smart lab-on-a-chip

embryo arrays
Kevin I-Kai Wang a, Zoran Salcic a, Johnny Yeh a, Jin Akagi b, Feng Zhu b, Chris J. Hall c,
Kathryn E. Crosier c, Philip S. Crosier c, Donald Wlodkowic b,d,n
a
Department of Electrical and Computer Engineering, University of Auckland, Auckland, New Zealand
b
The BioMEMS Research Group, School of Chemical Sciences, University of Auckland, Auckland, New Zealand
c
Department of Molecular Medicine and Pathology, University of Auckland, Auckland, New Zealand
d
The BioMEMS Research Group, School of Applied Sciences, RMIT University, Melbourne, Australia

art ic l e i nf o a b s t r a c t

Article history: Lab-on-a-Chip (LOC) biomicrofluidic technologies are rapidly emerging bioanalytical tools that can
Received 12 March 2013 miniaturize and revolutionize in situ research on embryos of small vertebrate model organisms such as
Received in revised form zebrafish (Danio rerio) and clawed African frog (Xenopus laevis). Despite considerable progress being
17 April 2013
made in fabrication techniques of chip-based devices, they usually still require excessive and manual
Accepted 17 April 2013
Available online 30 April 2013
actuation and data acquisition that significantly reduce throughput and introduce operator-related
analytical bias. This work describes the development of a proof-of-concept embedded platform that
Keywords: integrates an innovative LOC zebrafish embryo array technology with an electronic interface to provide
Microfluidics higher levels of laboratory automation for in situ biotests. The integrated platform was designed to
Lab-on-a-Chip
perform automatic immobilization, culture and treatment of developing zebrafish embryos during fish
Zebrafish
embryo toxicity (FET) biotests. The system was equipped with a stepper motor driven stage, solenoid-
Embryo
Bioassay actuated pinch valves, miniaturized peristaltic pumps as well as Peltier heating module. Furthermore, a
Automation Field Programmable Gate Array (FPGA) was used to implement an embedded hardware/software
Embedded systems solution and interface to enable real-time control over embryo loading and immobilization; accurate
FPGA microfluidic flow control; temperature stabilization and also automatic time-resolved image acquisition
Microcontroller of developing zebrafish embryos. This work presents evidence that integration of embedded electronic
CMOS interfaces with microfluidic chip-based technologies can bring the Lab-on-a-Chip a step closer to fully
Automated microscopy
automated analytical systems.
& 2013 Elsevier B.V. All rights reserved.

1. Introduction analysis in the physiological milieu of the intact, developing

The introduction of the 3 R (Reduction, Refinement and Replace-
ment) Animal Welfare legislation policies worldwide has facilitated
the accelerated development of alternative models for drug dis-
covery and environmental impact assessment (Lammer et al.,
2009a; Wlodkowic et al., 2011). In this regard toxicity biotests
performed on embryos of small multicellular model animals such as
zebrafish (Danio rerio) and clawed African frog (Xenopus laevis) offer
considerable bioanalytical advantages over costly and inherently
ethical controversial tests made on adult fish or rodents (Lammer
et al., 2009a; Parng, 2005; Rubinstein, 2006). Moreover, the embryo
biotests offer numerable advantages over bioassays performed on
cell lines and isolated cells/tissues as they support toxicological
n
Corresponding author at: The BioMEMS Research Group, School of Applied
Sciences, RMIT University, Melbourne, Australia. Tel.: +64 21 177 4369.
E-mail addresses: d.wlodkowic@auckland.ac.nz,
donald.wlodkowic@rmit.edu.au, d.wlodkowic@gmail.com (D. Wlodkowic).
vertebrate organism (Akagi et al., 2012a; Wheeler and Brandli,
2009). The optical transparency of fish embryos allows for con-
venient visualization of developing tissues in response to drug/toxin
treatment (Akagi et al., 2012a). The lack of efficient laboratory
automation that can perform analysis of statistically significant
cohorts of individual embryos in real-time hampers, however, the
large-scale industrial implementation of fish embryo toxicity (FET)
assays (Akagi et al., 2012a; Chang et al., 2012; Letamendia et al.,
2012; Wlodkowic et al., 2011).
Some progress has recently been made in robotic fish and frog
embryo handling, but the seamless integration of dispensing,
treatment and real-time analysis of sub-millimetre zebrafish
embryos is still not easily transferable to conventional, macro-
scale automation (Chang et al., 2012; Graf et al., 2011; Letamendia
et al., 2012; Wlodkowic et al., 2011). Therefore, there is a great
hope that advances in microfluidic Lab-on-a-Chip systems, where
most tasks are performed automatically without disturbing the
embryo, and without sudden changes to the embryo environment,

0956-5663/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.bios.2013.04.033
 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196
will produce significantly more reproducible data at a higher
throughput (Khoshmanesh et al., 2012; Wielhouwer et al., 2011;
Wlodkowic et al., 2011). The transfer of traditional in situ biotests
into a miniaturized format is still in its infancy but can prospec-
tively improve both quantitative and qualitative data acquisition.
In this regard, several innovative chip-based devices for analysis of
zebrafish embryos have recently been demonstrated. These
included use of a glass microwell flow-through system, micro-
droplet techniques, immobilization of zebrafish embryos using a
hydrodynamic fluidic traps and more recently also a 3D multilayer
plastic chip-based device capable of active trapping, immobiliza-
tion and recovery of zebrafish embryos in real-time (Akagi et al.,
2012b, 2012c; Son and Garrell, 2009; Wielhouwer et al., 2011).
Despite the considerable progress made in design and fabrica-
tion techniques, microfluidic devices usually still require excessive
and manual actuation and data acquisition that significantly
reduce throughput and introduce operator-related analytical bias
(Wang et al., 2012a, 2012b). Very limited progress has been made
in integration of LOC with the macroworld, design of user-friendly
off-chip components and also automation of many functional
mechanisms to provide truly turnkey and automated microfluidic
devices. Therefore key technological challenges that prevent wide-
spread adoption of LOC-based technologies still lie ahead. These
include automation of operation through the integration and
coordination of microfluidic, mechatronic and computer pro-
cesses, connection to the external world and advanced signal
processing and visualization techniques. This work describes the
development of a proof-of-concept embedded platform that inte-
grates a novel LOC zebrafish embryo array technology with
embedded electronic and microcontroller interface to provide
higher levels of laboratory automation for in situ biotests. The
integrated platform was designed to perform automatic immobi-
lization, culture and treatment of developing zebrafish embryos
during fish embryo toxicity (FET) biotests. We provide pioneering
evidence that by integrating state-of-the-art chip-based and soft-
ware technologies with embedded systems one can realize smart
Lab-on-a-Chip automation.
2. Materials and methods
2.1. Zebrafish husbandry and embryo culture
Wild type (AB line; Zebrafish International Resource Center,
USA) and transgenic Tg(fli1a:EGFP) adult zebrafish were used.
Embryos obtained from natural spawning were then collected in
E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM
MgSO4, 0.1% methylene blue) supplemented with penicillin/strep-
tomycin (Life Technologies Corp), kept at 28.5 70.5 1C and devel-
opmentally staged. Animal research was conducted with approval
from The University of Auckland Animal Ethics Committee
(approval ID R903). All control measurements are provided in
detail in the Figure legends, where appropriate, but in general
involved making direct comparisons between the chip-based
devices and static 24-well microtiter plates (Nalgene Nunc
Inc., USA).
2.2. Computational Fluid Dynamics (CFD) simulations
The chip design and optimization was guided by CFD simula-
tions as described before. Briefly, the 2D and 3D models of the
device were created with virtual embryos as spherical structures
inside the traps. The simulation was performed using Gambit
2.3 software (Fluent Inc., USA) to create the geometry and mesh
generation. Finite-volume based Fluent 6.3 software (Fluent Inc.,
USA.) was subsequently used to solve the associated differential
189
equations governing the balance of mass, momentum, chemical
species as described before (Khoshmanesh et al., 2012).
2.3. Multilayer 3D chip fabrication
The chip was designed and modeled using the CorelDraw X4
(Corel Corporation, Canada) and SolidWorks 2011 (Dassault Sys-
temes SolidWorks Corp, Concord, MA, USA) CAD packages. Physi-
cal prototyping of integrated 3D microfluidic devices was
performed in poly(methyl methacrylate) (PMMA/Acrylic; PSP
Plastics Ltd., Auckland, New Zealand) transparent thermoplastic
using a non-contact 30 W infrared laser micromachining system
with a 50 μm elliptical beam spot (Universal Laser Systems, USA).
PMMA layers were optically aligned and thermally bonded at
110 1C for up to 2 hours. Leak free connections were accomplished
by direct laser machining of fluidic interconnects to fit the 1/16 in.
PTFE tubing (Cole-Parmer, USA) with an internal diameter (ID)
1.5 mm. The tubing was reversibly plugged into the laser cut
connection ports.
2.4. FPGA-based embedded controller
The overall automation platform consisted of the microfluidic
actuation, temperature regulation and image acquisition units, and
required all units working in harmony to properly conduct auto-
matic experiments. One of the great advantages of LOC technology
was allowing massive parallelization and that was the main
motivation for designing a customized embedded controller using
FPGA-based technology. Other off-the-shelf microcontrollers,
while capable of providing similar control functionalities, were
significantly limited by their capabilities of handling parallel tasks
and interface I/Os. The prototyped automation platform used the
Altera Cyclone II chip that supports up to 622 I/O pins and ran at a
100 MHz system clock speed. The FPGA-based embedded con-
troller contained a soft-core NIOS-II microprocessor and three
customized hardware modules: Pulse Width Modulation (PWM)
module, bi-directional PWM module and stepper motor control
module, connected to the microprocessor through a standard bus
interface (Wang et al., 2012c).
2.5. Off-chip microfluidic actuation
The 3D chip required external microfluidic actuation for its
operation. The off-chip actuation consisted of three solenoid-
based pinch valves and two peristaltic pumps. Two pinch valves
were Biochem Fluidics 075P2NC12-02 S 2-way valves with 1/16”
silicone tubing and 12 VDC operating voltage. These two valves
controlled the embryo and chemical fluid entrance to the two
corresponding inlets of the chip. The other valve was a Biochem
Fluidics 075P3MP12-02 S 3-way valve with 1/16 in. silicone tubing
and 12 V operating voltage. This 3-way valve provided two
possible routes at the chip outlet for regulating chemical perfusion
or waste disposal operations. Kamoer DC and stepper motor-based
peristaltic pumps that both operated at 12 VDC were selected to
actuate the chip to provide liquid flow and embryo immobilize
suction at 1.05 mL/min flow rate. Each valve was controlled by an
individual digital I/O pin of the FPGA-based embedded controller,
via an H-bridge chip (SN754410, Texas Instruments) that accepted
3.3 V digital signal from the FPGA and powered the valve at 12 V.
The DC and stepper motor peristaltic pump were controlled by
customized bi-directional and PWM modules of the embedded
controller, via another H-bridge of the same type.
The embryos were to be cultivated at 28.5 70.5 1C regardless of
the ambient room temperature. The temperature sensing was
done using an negative temperature coefficient (NTC) type ther-
mistor that measures absolute temperature and works in the
190 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196
range of −40 1C to 125 1C with 0.5 1C tolerance. The analog
thermistor was powered with a high precision current source.
The measured voltage level was amplified and then converted by a
13-bit A/D converter (MPC3304, Microchip Technology Inc.) into a
digital signal, which was followed by a voltage translator
(MC14505B, Semiconductor Components Industries) to feed into
the FPGA-based embedded microcontroller at a matching voltage
level. The temperature regulation was done by software PI con-
troller commanding a thermoelectric Peltier device. The PI con-
troller regulated the temperature according to a predefined
temperature level by outputting the corresponding control signal
through a customized bi-directional PWM module to control the
Peltier device, via an H-bridge (SN754410, Texas Instruments) for
powering the Peltier device at 5 V.
2.6. Time-lapse image acquisition
The iDS-UI-1485LE CMOS camera (IDS Imaging Development
System GmbH) was selected to collect static image data at the
resolution of 2560  1920. This camera was connected to an
ordinary PC through the USB interface and provided a set of native
libraries in C++, which allowed programmable control and config-
uration during the experiments. The native libraries were wrapped
in a graphical user control interface that ran on the PC and
communicated with the FPGA-implemented embedded microcon-
troller via the UART (Universal Asynchronous Receiver/Transmit-
ter) interface. The camera was fixed directly above the chip, while
the chip was mounted on a slider stage. A stepper motor that was
controlled by the embedded microcontroller was used to actuate
the stage movement in a single dimension. The chip was designed
to move through a series of locations to allow camera to collect
real-time images of each embryo trapped in the chip in a time-
lapse fashion.
3. Results and discussion
3.1. Microfluidic chip-based device design and operation
The microfluidic 3D chip was fabricated in a biologically
compatible and optically transparent PMMA polymer using infra-
red laser micromachining. The microfluidic chip was designed as a
monolithic and fully integrated device with no moving parts and
encompassed four distinctive layers that were optically aligned
and thermally bonded (Fig. 1(A)). The chip design incorporated
three integrated modules: (i) a main channel (1.7 mm  1.5 mm 
55 mm) for zebrafish embryo loading, (ii) a linear array of 20
miniaturized traps for single embryo trapping and immobilization,
(iii) a plenum suction manifold (0.7 mm in depth) at the bottom
plane of the chip that created a drag force to immobilize embryos
(Fig. 1(B)). The chip design allowed for both single embryo
occupancy in the traps and unobstructed passage of other embryos
in the rectangular main channel following automated docking.
The trapping principles exploited the combined gravitational-
induced sedimentation of embryos and a low-pressure suction at
the bottom plane of the device to rapidly attract embryos into the
traps (Fig. 1(B, C)). Accordingly, in the presence of gravity the
embryos were deflected under the combined effect of suction flow
and gravity toward the aperture of the traps (Fig. 1(B, C)). The
embryos falling into traps were then immobilized by a continuous
suction at the bottom plane of the device (Fig. 1(C)). The process
was repeated until all traps were occupied and the loading was
then discontinued.
During experimental validation the system achieved 100%
trapping efficiency when actuated at a total flow rate of up to
0.6 mL/min as described before. Importantly, over 99% of trapped
embryos retained their position during the course of 72 h experi-
ments with no dislodgement observed when drug delivery mani-
fold was actuated. The hydrodynamic conditions inside the chip
with a maximum shear stress of 0.14 Pa at the lower surface of the
last four embryos were permissive for zebrafish embryo culture
over extended periods of time. Accordingly cumulative survival of
embryos cultured on a chip-based device for up to 72 h was over
9575% with the exception of chips kept at a static regimen or
perfused at flow rates lower than 0.1 mL/min. Importantly, the
embryo development and viability was uniform across all the traps
and did not result in any discernible developmental effects
irrespectively of the magnitude of flow rates applied. In this regard
our data were statistically comparable with static Petri dish
control experiments. Furthermore, the innovative chip design
allowed rapid recovery of developing embryos by simply reversing
the fluid flow and collecting the embryo at the input port. The
embryos after being released from the chip behaved the same as
controls and maintained cumulative survival of up to 95 75% with
no developmental abnormalities detected as compared to Petri
dish controls. The chip based- device allowed for rapid delivery of
drugs and exchange of media to immobilized embryos without
disturbing their position or introducing mechanical stress (Fig. 1(D)).
The average drug-exchange time was 60–90 s at the flow rate of
0.1 mL/min (Fig. 1(D)).
3.2. Off-chip automation design rationale
The embedded automation platform was designed by combin-
ing the state of the art FPGA-based embedded control, mecha-
tronics, and LOC technologies. The platform contained three
distinctive functional modules, such as: (1) microfluidic device
actuation, (2) chemical fluid temperature regulation, and
(3) automated image acquisition (Fig. 2). The microfluidic actuation
unit controlled the pinch valves and peristaltic pumps to regulate
the flow rate inside the chip-based device. The temperature
regulation unit monitored the fluid temperature and maintained
the fluid temperature at a predefined level. The image acquisition
unit acquired images in a time-lapse mode. The automation plat-
form used a common 12 VDC power adapter as the power source
for all the embedded devices, including pinch valves, peristaltic
pumps, thermoelectric devices and a stepper motor. In order to
examine the platform, each individual unit of the platform was
tested separately and together through real experiments with
different durations of up to 60 h.
The customized FPGA-based embedded microcontroller (Fig. 2(B))
was at the core of the automation platform, which performed
sensing of the status of individual elements and triggered and
controlled the overall operations of all the electrical–mechanical
devices. The embedded microcontroller contained a soft-core NIOS
II microprocessor customized with additional digital components,
including PWM modules, bi-directional PWM modules, stepper
motor control modules, SPI and UART interface modules. The final
design of the embedded microcontroller was accomplished on the
Terasic DE2-70 prototyping board with a Cyclone II chip and used
less than 20 K logic elements (LEs), which was 28% of the available
logic elements of the entire FPGA chip. In addition, the current
embedded microcontroller used 64 I/O pins, which were only 10% of
the 622 available I/O pins for Cyclone II chip, to interface with all the
embedded and microfluidic devices. However, 64 pins were already
beyond the I/O interface capacity of many off-the-shelf embedded
microcontrollers, for example, Atmel1280 that only has 54 I/O pins
(Wang et al., 2012a, 2012c). This indicated that the solution had high
flexibility, scalability and extensibility in supporting additional sen-
sing and actuating requirements if needed.
 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196 191

Fig. 1. Microfluidic chip-based device for long-term culture of zebrafish embryos. (A) 3D device with integrated manifolds that allow for suction-assisted embryo trapping
and immobilization in linear array of up to 20 miniaturized embryo traps; (B) 3D streamlines of fluid flow through the embryo traps colored by velocity (m/s), obtained by
CFD simulations. Black spheres denote simulated embryos. Arrows denote fluid flow direction; (C) Cross-sectional view of the device depicting the four layers of PMMA and
integrated modules such as main channels, trapping array and suction manifold. Note clearly visible immobilized embryos inside the trapping array. Embryos are deflected
toward the traps under the combined effects of suction flow and gravity. Blue arrows denote the direction of the fluid flow; (D) Cross-sectional view of the device depicting
time-lapse imaging of drug delivery to embryos immobilized inside the microfluidic embryo array. Chip was infused with a 0.04% Trypan Blue dye at a volumetric flow rate of
0.2 mL/min.

3.3. Fluidic system performance
One of the key requirements of automating the chip was to
precisely actuate the fluid flow rate and to control the correct flow
path during the course of the experiments. Three microfluidic
devices, solenoid-actuated pinch valves (Fig. 2(C)), a stepper motor
peristaltic pump and a DC motor peristaltic pump, were used to
regulate the flow paths, to immobilize embryos in the array and to
provide precise flow rate respectively, as shown in Fig. 2(D) (Wang
et al., 2012a, 2012c).
The two-way pinch valves controlled the two inlets of the chip
to feed the embryos and the chemical fluid. The three-way pinch
valve regulated the outlet to perform either chemical perfusion or
waste disposal. The pinch valves required 12 VDC input voltage to
192 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196

Fig. 2. Off-chip FPGA-based embedded automation platform. (A) Conceptual automation platform consisted of a microfluidic chip-based device, microfluidic valves and
pumps, temperature sensor and thermoelectric devices, and a CMOS camera and a stepper motor actuated 1D slider stage, integrated and orchestrated by a FPGA-based
embedded microcontroller; (B) FPGA-based embedded microcontroller contained a soft-core NIOS II microprocessor with customized digital components, including PWM
and bi-directional PWM modules, a stepper motor control module SPI and UART communication modules; (>C) Two-way pinch valve for regulating microfluidic flow path;
(D) Stepper motor-based peristaltic pump for precisely suction manifold actuation to trap and immobilize zebrafish embryos at a flow rate of 1.05 mL/min; (E) Close-loop
temperature regulation unit with an analog thermistor-based temperature sensor which interfaces with the embedded microcontroller through an A/D converter via a
customized SPI module. A software PI controller running on the embedded microcontroller actuates the Peltier device based on the collected temperature data; (F) High
precision current source that produces exactly 45μA flowing through the thermistor; (G) Two cascaded Peltier devices thermally pasted with aluminum heat sink for
maintaining reference temperature as ambient room temperature; (H) Chip-based device integrated with stepper motor controlled 1D slider stage; (I) CMOS camera with
objective lens which provides magnification of up to 5 times.
 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196
Fig. 3. The calibration curve for the selected negative temperature coefficient
(NTC-type) thermistor. With an external 13-bit A/D converter (resolution of
1.22 mV), the temperature sensor can achieve a theoretical resolution of approxi-
mately 0.01 1C. The resolution will vary at different temperature due to the
nonlinear characteristic of thermistors. The temperature readings were converted
into the linear temperature scale based on this curve.
operate properly. In order to drive the pinch valve while achieving
digital control at the same time, the digital output (3.3 V) from the
FPGA-based embedded controller was used to control an H-bridge,
which in turn powered and controlled the pinch valve at 12 VDC.
The purpose of the stepper-motor based peristaltic pump was
to provide a suction force through the suction manifold in order to
load and to immobilize the embryo in each trap. The pump was
equipped with its own driving circuit, where the rotational speed
of the pump and hence the flow rate was controlled according to
the frequency of an input PWM signal. A customized PWM module
was designed as part of the FPGA-based embedded controller to
provide end users with easy programmability to modify both the
signal frequency and duty cycle by changing corresponding regis-
ter values. The output of the PWM module was fed as the input to
the driving circuit of the stepper motor peristaltic pump. The
stepper motor peristaltic pump operated at a flow rate of 1.05 mL/min
and was able to load the chip with good trapping efficiency.
The DC motor peristaltic pump operated at a higher rotational
speed to maintain the flow rate at 1.5 mL/min for chemical liquid
perfusion. The pump was a two-input terminal device and could
change its rotational direction (or flow direction) by reversing the
input voltage polarities at the two input terminals. A customized
bi-directional PWM module was designed in the embedded
controller. This module generates two signal outputs which con-
trolled and drove an external H-bridge (SN754410, Texas Instru-
ments). One output produced PWM signal with adjustable
frequency and duty cycle, which controlled the power delivered
to the actual device through the H-bridge. The other output
maintained at constant ground voltage. The role of these two
outputs could be reversed by the software through changing a
register value in the embedded microcontroller. This change
caused the voltage of opposite polarity to be delivered to the
end device. The DC motor peristaltic pump, driven by the external
H-bridge and the bi-directional PWM module could rotate in both
clockwise and anti-clockwise directions, depending on the voltage
polarity provided at its input terminals. This allowed the DC motor
pump to perform either chemical perfusion or unloading the
embryos for disposal at different experiment stages.
Throughout the trial experiments, the embedded controller
was capable of real-time controlling each individual and pre-
programmed experiment stages such as (1) automatic embryo
loading into the chip-base device; (2) immobilization of embryos
in individual traps; (3) perfusion of embryos with E3 medium at a
desired flow rate and (4) release and unloading of immobilized
and viable embryos at the end of the experiment.
193
3.4. Temperature regulation
The microenvironment had a great influence on the embryonic
development and results of the experiments. Zebrafish, as a
tropical fish, needed to be cultivated at a stable temperature range
of 28.5 70.5 1C and the chemical fluid that perfused through the
chip must be within this range in order to hatch the embryos
(Akagi et al., 2012a, 2012b; Lammer et al., 2009b). Fig. 2(E) depicts
a close loop temperature regulation unit, which consisted of a
temperature sensing and actuation devices. This module was
developed to monitor and control the temperature of fluid being
delivered to the embryos in real-time. In order to measure the
absolute temperature with high precision, thermistor was selected
over other types of temperature sensors (Wang et al., 2012c). The
thermistor used in the current platform is an NTC (negative
temperature coefficient) type. In order to properly operate the
thermistor, a high precision current source was necessary to
power up the thermistor, as shown in Fig. 2(F). The current source
provided precisely 45 μA flowing through the thermistor, which
allowed 2.9 V to be measured across the thermistor at 25 1C, after
amplification. With the provided current source, the voltage varied
at 0.1 V/1C and was fed into a 13-bit A/D converter with a
resolution of 1.22 mV. This design allowed temperature sensor to
achieve a theoretical resolution of approximately 0.01 1C. Although
the sensor could achieve a theoretical resolution of approximately
0.01 1C, the resistance–temperature relationship was calibrated
using a water bath with a measuring resolution of 0.1 1C and the
resistance–temperature curve was presented in Fig. 3. The resolu-
tion varied at different temperature due to the nonlinear char-
acteristic. The temperature readings were converted into the linear
temperature scale using the polynomial equation in Fig. 3. The A/D
converter used in the current platform provided common SPI bus
for communication and data acquisition. The FPGA-based
embedded controller collected real-time temperature information
from the A/D converter through a customized SPI unit.
In addition to temperature sensing, a thermoelectric (Peltier)
device was employed as the actuator of the temperature regula-
tion unit (Fig. 2(G)). It has the advantage of being able to actively
heat or cool its surface, relative to a reference temperature. In
order to achieve absolute temperature control, one side was fixed
at the ambient room temperature by attaching a stable heat sink
module with thermal paste, which maintained the reference side
at the room temperature. The common 12 VDC power supply of
the automation platform was stepped down to 5 VDC using a
voltage regulator. The same H-bridge IC (SN754410, Texas Instru-
ments) was used to drive the Peltier device. The customized
bi-directional PWM module designed for the DC peristaltic pump,
which allowed voltage across the device to be applied in both
directions, was replicated to control the H-bridge. By reversing the
voltage polarity, active heating or cooling of the delivered fluid
could be achieved.
Software PI (proportional integral) controller was designed to
run on the embedded microcontroller to regulate the H-bridge
output and to maintain the temperature at the pre-specified level
(Fig. 4(C)). With this configuration, temperatures between 15 1C
and 35 1C can be obtained on the controlled side of the Peltier
device, while the reference side was maintained at the room
temperature of 25 1C.
During the embryo development experiments, the thermistor
was attached to the tubing surface close to the inlet of the chip
instead of measuring the direct fluid temperature. It was found
that the actual fluid temperature was 3 1C less than the tempera-
ture of the heated tubing surface and therefore the desired
temperature was set to 32 1C during the experiments. The soft-
ware PI controller designed to run on the embedded microcon-
troller was able to maintain the temperature at 32 1C with 70.5 1C
194 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196

Fig. 4. Graphical user control interface. (A) The first window allowed users to configure and control each individual experiment stage, and also to manually control each
peristaltic pump and the 1D slider stage. Real-time video can be viewed in this window as well; (B) The second window provided the capabilities to adjust camera settings
and to manually trigger camera operations and control bottom lighting levels; (C) The third window showed the collected temperature information and allowed users to
adjust the desirable temperature setting to actuate the thermoelectric Peltier devices.

fluctuation for up to 60 h. The results of the experiments showed
that 100% of embryos that were viable 12 h after loading into the
chip developed fully into the larval forms, which supported the
notion of appropriate levels of temperature stabilization.
3.5. Graphical user control interface
color gain, and lighting source levels. The effects on the image
could be observed in the first service window in real-time. The
third service window displayed real-time monitored temperature
information retrieved from the embedded microcontroller
through UART interface and allowed users to adjust the desired
temperature level for the experiments.

In addition to the FPGA-based embedded microcontroller and
control software, a user control interface was designed to run on a
PC to bridge the operations between the camera and the
embedded controller, and to provide synchronized real-time
monitoring and configuration capabilities to the end users. The
control interface contained three service windows for performing:
(1) real-time configuration and monitoring of the experiment,
(2) camera configuration, and (3) temperature monitoring and
regulation, as shown in Fig. 4. The first service window provided
users the ability to manually configure various embedded devices,
including the peristaltic pumps and the chip positioning stepper
motor, for examining their functionalities before commencing the
experiments. It also allowed users to manually perform and
examine each experiment stage, including purging the chip,
loading the embryos, performing chemical perfusion, and unload-
ing the embryos. Different time intervals for conducting auto-
mated experiments could be configured for each experiment stage
through this window, as well. The second service window allowed
users to adjust camera settings, including image/video modes,
3.6. Automatic monitoring and image acquisition
Time-lapse image acquisition of developing embryos cultured
on the LOC devices was a crucial part of system design. The ability
to perform real-time monitoring and analysis would greatly
improve the throughput and turnaround time of the experiments
(Akagi et al., 2012a). In the current automation platform, the real-
time monitoring consisted of two parts, the LOC device positioning
and the image acquisition, as shown in Fig. 2(H) and (I)
respectively.
For the bioassay experiments, automatic image acquisition
should be performed in a time-lapse mode every 30 min. The
iDS-UI-1485LE CMOS camera provided native libraries that were
wrapped in a customized user control interface (Fig. 4) to allow
easy monitoring, controlling and configuration by the end users.
Although the camera was controlled by the customized software
running on a PC, the image acquisition must be performed in
harmony with the embedded controller, following strict and
precise timing and positioning of the chip.
 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196 195

In the current design iteration platform, a stepper motor was
employed to precisely move the chip through a series of pre-
defined positions underneath the camera to allow proper image
acquisition of each individual embryo. A customized stepper
motor control module was designed in the embedded microcon-
troller to provide four synchronized PWM signals that drive the
Fig. 5. Control flow and operation synchronization between the control software
on embedded microcontroller and graphical user control interface on a PC. Process
synchronization is achieved between image acquisition and chip-based device
location.
stepper motor. This module provided two modes of driving the
stepper motor. The number of rotation steps, rotation speed and
rotation direction could be modified by simply changing register
values through software running on the embedded microcontrol-
ler. The first mode controlled the motor to rotate through a
discrete number of steps, which allows user to precisely configure
the linear moving distances in a resolution of 0.01 mm between
each step. The second mode controlled the motor to rotate
continuously to move the LOC device back to its initial starting
location.
In order to properly collect images of each embryo periodically,
synchronization was achieved between the camera operation and
the chip locations controlled by the embedded controller (Fig. 5).
For this purpose, the embedded controller was configured by the
user control interface to operate in a time-lapse mode to move the
chip through a series of fixed locations in a regular time interval
and eventually returned to the starting position. At the end of each
movement, an interrupt request signal (IRQ) was emitted, signal-
ing the completion of operation. The embedded microcontroller
passed this completion status to the control interface via a UART
message to trigger the camera operation. Once the images were
acquired another UART message was sent from the control inter-
face to the embedded controller, signaling the chip could be
moved to its next location.
In the course of the experiments, color time-lapse images were
collected every hour and the chip positioning operation was
synchronized with the image acquisition. Images of developing
embryos were collected using the embedded CMOS camera and a
reference stereomicroscope (Leica MZ7.5 equipped with a trans-
parent base stage) as shown in Fig. 6(A and B). Both imaging
systems were operated at a resolution of 2048  1536 pixels (3.1
Megapixels). Comparison between resolution, light sensitivity and
illumination sources achieved by embedded iDS-UI-1485LE CMOS

Fig. 6. Comparison between resolution, light sensitivity and illumination sources achieved by embedded iDS-UI-1485LE CMOS camera and conventional stereomicroscope
(Leica MZ7.5 with transparent base). (A) Time-lapse imaging of Tg(fli1a:EGFP) transgenic zebrafish embryos cultured for up to 35 hours post-fertilization (hpf) on a chip-
based device. During the time-lapse experiments, the E3 medium was supplemented with 0.003% 1-phenyl-2-thiourea (PTU, Life Technologies Corp.) to maintain the optical
transparency of the zebrafish embryos by inhibiting melanocyte formation; (B) Wild-type zebrafish embryos cultured for up to 60 hours post-fertilization (hpf) on a chip-
based device without melanocyte formation inhibitor PTU. Note that embedded CMOS sensor and dedicated optics, despite shorter focal depth, provided sufficient resolution
and sensitivity to detect characteristic patterns of tissue formation during embryogenesis such as eye formation, melanocyte formation, tail detachment, etc. Custom-built
LED array combined with CMOS camera provided superior degree of illumination to acquire high-definition images of developing zebrafish embryos.
196 K.I.-K. Wang et al. / Biosensors and Bioelectronics 48 (2013) 188–196
camera and conventional stereomicroscope (Leica MZ7.5 with
transparent base) were favorable for the embedded CMOS camera.
Despite shorter focal depth of the CMOS optics the system was
capable of acquiring images with sufficient resolution and sensi-
tivity to detect characteristic patterns of tissue formation during
embryogenesis such as eye formation, melanocyte formation, tail
detachment (Fig. 6A and B)). Most notably the embedded CMOS
camera achieved much better image quality when applied to wild-
type embryos that featured higher contrast and thus more
optically pronounced tissue patterns due to culture without
0.003% of melanocyte formation inhibitor 1-phenyl-2-thiourea
(PTU, Life Technologies Corp.) (Fig. 6(B)). When applied to time-
resolved imaging of Tg(fli1a:EGFP) transgenic embryos cultured in
the presence of PTU both CMOS and stereomicroscope acquired
only satisfactory images mostly due to high-level of embryo
transparency (Fig. 6(A)). Most notably the custom-build LED array
that was embedded into the custom imaging system provided
superior degree of illumination to acquire high-definition images
of developing zebrafish embryos as compared with a stock
transparent-base illumination employed by the stereomicroscope
(Fig. 6(B)). Our work provides proof-of-concept that highly custo-
mizable system based solely on off-the-shelf CMOS camera and
optics can be built at a fraction of cost and power usage of
conventional microscopes while also providing substantial space
savings and user-friendly operation. At the same time, when
proper illumination sources are applied such system will provide
resolution and sensitivity matching that of the conventional
microscopes and thus opening new avenues for low-cost automa-
tion of fish embryo assays.
4. Conclusions
The presented Lab-on-a-Chip automation prototype outlines
the future avenue for development of fully integrated LOCs
targeted at in situ analysis of small vertebrate model organisms
such as zebrafish. This proof-of-concept technology provides a
new rationale for rapid and automated biotests performed on
developing zebrafish embryos. The conceptual system was made
as reduced functionality prototype that integrated the innovative
microfluidic LOC technology, a customized FPGA-based embedded
microcontroller, off-the-shelf components, and de-centralized
control over several modules. We anticipate that the next stage
of prototyping will involve a completely integrated off-chip digital
and electronic automation platform, which combines all critical
mechanical, electrical, and software components with a graphical-
user interface (GUI). Moreover, development of customized auto-
matic image acquisition algorithms and real-time image analysis
are also underway, which will improve the throughput providing
instantaneous quantification of acquired results. Our work pro-
vides a new rationale to realize smart and low-cost Lab-on-a-Chip
automation in biotechnology and predictive toxicology.
Acknowledgments
Grant sponsor: Faculty Research Development Fund, University
of Auckland (KW, ZS, JY, JA, DW); Ministry of Science & Innovation
(MSI) New Zealand (JA, CJH, KEC, PSC, DW); Early Career Research
Excellence Award, University of Auckland (DW); Australian
Research Council (DW); Vice-Chancellor's Senior Research Fellow-
ship, RMIT University (DW).
The authors thank Mr. Alhad Mahagaonkar for his expert
management of the zebrafish facility and Dr. Khashayar Khoshma-
nesh for his help in CFD simulations.
References
Akagi, J., Khoshmanesh, K., Evans, B., Hall, C.J., Crosier, K.E., Cooper, J.M., Crosier, P.S.,
Wlodkowic, D., 2012a. PLoS One 7 (5), e36630.
Akagi, J., Khoshmanesh, K., Hall, C.J., Cooper, J.M., Crosier, K.E., Crosier, P.S.,
Wlodkowic, D., 2012b. Sensors and Actuators, B: Chemical.
Akagi, J., Takeda, K., Fujimura, Y., Matuszek, A., Khoshmanesh, K., Wlodkowic, D.,
2012c. Sensors and Actuators, B: Chemical.
Chang, T.Y., Pardo-Martin, C., Allalou, A., Wahlby, C., Yanik, M.F., 2012. Lab on a Chip
12 (4), 711–716.
Graf, S.F., Hotzel, S., Liebel, U., Stemmer, A., Knapp, H.F., 2011. Journal of the
Association for Laboratory Automation 16 (2), 105–111.
Khoshmanesh, K., Akagi, J., Hall, C.J., Crosier, K.E., Crosier, P.S., Cooper, J.M.,
Wlodkowic, D., 2012. Biomicrofluidics 6 (2), 24102–2410214.
Lammer, E., Carr, G.J., Wendler, K., Rawlings, J.M., Belanger, S.E., Braunbeck, T.,
2009a. Comparative Biochemistry and Physiology Part C: Toxicology and
Pharmacology 149 (2), 196–209.
Lammer, E., Kamp, H.G., Hisgen, V., Koch, M., Reinhard, D., Salinas, E.R., Wendler, K.,
Zok, S., Braunbeck, T., 2009b. Toxicology In Vitro 23 (7), 1436–1442.
Letamendia, A., Quevedo, C., Ibarbia, I., Virto, J.M., Holgado, O., Diez, M., Izpisua
Belmonte, J.C., Callol-Massot, C., 2012. PLoS One 7 (5), e36690.
Parng, C., 2005. Current Opinion in Drug Discovery and Development 8 (1),
100–106.
Rubinstein, A.L., 2006. Expert Opinion on Drug Metabolism and Toxicology 2 (2),
231–240.
Son, S.U., Garrell, R.L., 2009. Lab on a Chip 9 (16), 2398–2401.
Wang, K.I.K., Salcic, Z., Yeh, J., Akagi, J., Wlodkowic, D., 2012a. 7th IEEE International
Symposium on Industrial Embedded Systems, Karlsruhe pp. 327–330.
Wang, K.I.-K., Bonnetat, A., Andrews, M., Salcic, Z., Akagi, J., Wlodkowic, D., 2012b.
19th International Conference on Mechatronics and Machine Vision in Practice,
Auckland.
Wang, K.I.-K., Yeh, J., Salcic, Z., Akagi, J., Wlodkowic, D., 2012c. 19th International
Conference on Mechatronics and Machine Vision in Practice19th IEEE M2VIP,
Auckland.
Wheeler, G.N., Brandli, A.W., 2009. Developmental Dynamics 238 (6), 1287–1308.
Wielhouwer, E.M., Ali, S., Al-Afandi, A., Blom, M.T., Riekerink, M.B., Poelma, C.,
Westerweel, J., Oonk, J., Vrouwe, E.X., Buesink, W., vanMil, H.G., Chicken, J., van’t
Oever, R., Richardson, M.K., 2011. Lab on a Chip 11 (10), 1815–1824.
Wlodkowic, D., Khoshmanesh, K., Akagi, J., Williams, D.E., Cooper, J.M., 2011.
Cytometry A 79 (10), 799–813.