Portable Memristive Biosensing System as Effective

Point-of-Care Device for Cancer Diagnostics

Ioulia Tzouvadaki1 and Abuduwaili Tuoheti2, Giovanni De Micheli1, Danilo Demarchi2, Sandro Carrara1
1. Integrated System Laboratory, EPFL, Lausanne, Switzerland
2. Politecnico di Torino, Torino, Italy
Email: ioulia.tzouvadaki@alumni.epfl.ch

Abstract— Memristive biosensors have been proved as

excellent candidates for ultrasensitive biosensing. In this work, a
novel portable bio-detection system based on memristive
biosensors is designed, developed and tested for providing a
significantly fast, automatic and simultaneous sensing output of
multiple memristive biosensors on a single chip. The suggested
compact and independent bio-sensing prototype is achieved
through the realization of an electronic board designed for
addressing the specifications of the memristive biosensor signal
acquisition. Memristive bio-sensing-chips are specially designed
and fabricated as well. The system was tested for successfully
sensing of Prostate Specific Antigen, one of the main biomarkers
of prostate cancer, at fM concentrations. Overall, this novel
scheme resembles to a memristive-biosensing-kit approach,
paving the way for fast and ultrasensitive PoC (point-of-care) Fig.1. Electrical characteristics of a memristive biosensor acquired with the
devices. probe station and Keithley configuration. The pinched loop characterizing
the memristive devices is lost upon the bio-functionalization. Hysteretic
I. INTRODUCTION electrical properties (a) and the voltage gap (b) are the main features of
interest of the memristive biosensors.
Semiconductor nanowires are considered as promising
building blocks for miniaturized bioassays providing low-cost
applications minimizing the time needed for diagnostic
microchips for applications in the medical practice. In some
procedures is a highly crucial aspect, there is the need to develop
cases, silicon nanowire-arrays exhibiting memristive electrical
new acquisition platforms to provide a fast sensing output. As a
properties [1-3] are bio-functionalized with receptor molecules,
first effort towards this direction, a measurement interface for
such as antibodies or DNA aptamers, giving rise to the so-called
memristive biosensors with a manually performed drain
memristive biosensors. These memristive nano-bio-sensors,
selection capability, by means of a mechanical switch, was
have been already successfully implemented in both the
previously realized [8]. Nevertheless, despite successfully
diagnostics and the therapeutics field, demonstrating their
reducing the measurement time, this configuration only replaced
immense potential for precise biosensing. Most specifically,
the probe station, while connection with a Source Measure Unit
ultrasensitive detection is achieved for cancer biomarkers,
(SMU) was still required. This aspect determinately constrains
pushing the sensing performance at atto-molar concentrations
the implementation of the ultrasensitive memristive nano-bio-
[4] as well as effective screening of therapeutic compounds
sensors in the environment of a hospital or medical clinic.
along with the possibility for continuous drug monitoring [5].
This ultrasensitive bio-detection paradigm is based on the In this work, the design and realization of a completely new,
hysteresis modification appearing in the electrical characteristics portable memristive biosensing platform is developed and tested
upon treatment of the nanodevices’ surface with charged successfully, providing a fast, fully-automatized and
biological species [6, 7]. More specifically, the pinched simultaneous sensing output of multiple individual memristive
hysteresis is lost upon the bio-functionalization of the biosensors. The realized system is independent from SMU
nanodevice and a voltage gap is introduced in the semi- configurations allowing a flexible and unrestricted sensing
logarithmic current-to-voltage characteristics as a further procedure. Moreover, single chips consisting of nanofabricated
memory effect to the voltage scan across the nanowire (Fig. 1). silicon wires exhibiting memristive electrical response,
So far, the electrical monitoring of memristive biosensors conjugated with metallic extension electrodes are also
was performed by sweeping a drain-to-source voltage forward fabricated, allowing an integrated measurement procedure. Bio-
and backward using a Cascade Microtech Probe Station and a functionalization with receptor molecules, finally provides
Hewlett-Packard 4165A Precision Semiconductor Parameter bioassay sensing-chips based on series of memristive
Analyzer while implementing tungsten needles to directly biosensors. The full system was successfully validated for
contact the NiSi pads of the nanostructures. This experimental prostate cancer biomarkers detection, namely PSA at fM
setup resulted to an overall unpractical measuring procedure concentrations (critical level of PSA 4 ng/mL ca. 133 pM),
considering that the sensing could be performed only in the therefore demonstrating high potential for the implementation as
specific laboratory and was likewise significantly time- a novel ultrasensitive medical tool for efficient bio-sensing at
consuming. However, considering the fact that in medical early stages of the disease.

978-1-5386-4881-0/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: University of Gent. Downloaded on September 16,2022 at 13:37:36 UTC from IEEE Xplore. Restrictions apply.
 II. SYSTEM ARCHITECTURE
A. Memristive biosensing board prototype
The proposed system, shown in Fig. 2, consists of the sensor
module, analog-to-digital converter (ADC) and multiplexer
(MUX) parts, the computing block, a data acquisition and
storage scheme and a power supply. The schematic of the board
is realized using Altium Designer, while the PCB layout is
designed using OrCAD PCB Designer. For the sensor part,
twelve independent sensing outputs, and one common excitation
input are considered. This configuration corresponds to the
twelve individual memristive biosensors, fabricated as discussed
in section III. The realized portable (175 mm x 72 mm) printed
circuit board (PCB) prototype for memristive biosensors is Fig.3. The memristive biosensing board prototype. The realized electronic
presented in Fig. 3. The sensor part is placed separately on one board mainly consists of the sensor module, the ADC and MUX part, the
side of the board in order to avoid any electromagnetic microcontroller, a data storage scheme and a power supply.
interference from other components. For the MUX
(ADG428BRZ) and the ADC (AD7785BRUZ) part the accuracy some tens of µA, they are very prone to electrical noise.
of the measurement is very high thanks to the 20-bit ADC. Therefore, the noise reduction is the key step in obtaining
Furthermore, with the MUX each channel can be measured reliable measurement results. For this reason, suitable input bias
programmatically through the microcontroller. Therefore, the current operational amplifies are used and a very low tolerance
board can individually address and read signals from the surface mount resistors and capacitors are chosen. In addition, a
individual sensors of the biosensing chip. For the computing part board-level shielding box (RFI SHIELD CAN 50X25X5MM,
the high speed STM32 microcontroller (STM32F446RE) is 952-2642-ND) is introduced to minimize the external noise,
used. It has a clock speed that can reach 180 MHz and 16 substituting the Faraday cage included in the measurement
independent counters that can allow very accurate generation of configuration involving the probe station and the Keithley. The
Pulse Width Modulation (PWM) signal, with the possibility noise shielding enclosure is incorporated on board with the
reaching a pulse with 1ms period. For the data acquisition and possibility to be stabilized during the measurements through the
elaboration an on-board serial port is incorporated and can be use of on-board shield clips (952-1475-1-ND) and to be easily
used through wire-connection with a computer or a Raspberry Pi removed for loading/unloading of the sensing module. The
(RPi), allowing the data processing. In addition, a uSD memory humidity and temperature values can be continuously monitored
card socket (DM3AT-SF-PEJM5, 2427719, series DM3, Micro through a sensor (HDC1050DMBT Texas Instruments)
SD) is introduced on the board, combined with a microSDXC incorporated on the board. Finally, for the power supply a
card (64 GB, SDCA3, speed class 10) with a fast reading and voltage regulator that can accept input voltage in the range up to
writing speed of 90 MB/s and 80 MB/s, that is necessary for 25 V is considered and can also be powered by 5V batteries.
direct and quick signal acquisition. The main purpose of the B. Sensing Module
memory card is the recording of the full current-to-voltage The sensing part in Fig. 4 is the core of the novel board
curves enabling the transferring of this information to a platform and consists of twelve identical nanosensors,
computer, for further analysis or as back-up files useful for demonstrating same range and accuracy. Each nanosensor
future analysis. In addition, a flash 32 MB memory card is results to an independent signal that is considered through each
included for storing measurement parameters, and the possibility individual drain. All nanosensors are subjected to the same
for a future introduction of a Wi-Fi chip that will enable wireless excitation through the common source terminal (Fig.5). The
data transmission is also taken into consideration. Moreover, nanosensors require input signal excitation in a form of a double
considering the very low currents obtained by these nanosensors source-to-drain voltage sweep following the experimental
that are in the range from some tens of pA (current minima) till protocol established by the measurements with the Keithley. The
negative biasing for the nanosensors is obtained through the
control of digital-to-analog converter (DAC) (2112662,
AD5686RBRUZ) with the supervision of the microcontroller.
The microcontroller and DAC module are communicating
through Serial Peripheral Interface (SPI) Bus. Each signal
generator is realized by the microcontroller that sends an
increasing or decreasing digital value to the specific DAC which
converts it to an analog voltage. Overall, positive and negative
potential values are applied to the nanodevices during the
voltage sweep. During the voltage sweep, the measured current
signals from these nanosensors are fed into the twelve-
independent analog readout circuitry each consisting of two
Fig.2. Overall block diagram of the portable memristive biosensors stages, current to voltage conversion stage and offset
platform. The nanosensor socket with the multichannel system block cancellation stage. The DC and low frequency gain of the 1st
interfaces with the micro-controller that drives the voltage generation stage is equal to –Rf that is -20 MΩ, while the gain of 2nd stage
circuits and receives ouput values from the readout circuits. is equal to the ratio of the output voltage to the input voltage at

978-1-5386-4881-0/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: University of Gent. Downloaded on September 16,2022 at 13:37:36 UTC from IEEE Xplore. Restrictions apply.
 Fig. 6. SEM micrographs illustrating the nanofabricated wires anchored
between the NiSi pads (a) and detail of an individual nanowire-array
configuration (b). Optical microscopy image depicting the system
nanowire-pads integrated with the extension Pt electrodes (c).

are considered for individual measurement of 12 integrated

Fig.4. Sensing module detailed architecture for the nanosensor socket and
sensors that allow possibility for statistical analysis with average
low bias Transimpedance amplifiers. value and standard deviation. This statistical analysis is required
for obtaining reliable results and overcome the variability of the
sensors originating from the etching process. In addition, these
multiple sensors allow the possibility for simultaneous sensing
of different biomarkers. First, a photolithography process is
implemented in order to define the pattern of the lines. A resist
bi-layer (a lift-off resist followed by a positive photoresist) is
spin-coated on the surface and the exposure of the lines design
is performed through maskless direct laser writer (MLA150,
Heidelberg Instruments). A 20 nm Titanium (Ti) and a 100 nm
Fig.5. The realized chip-holder board including a sensing chip that consists Platinum (Pt) layer are introduced through physical vapor
of twelve sensors with independent outputs and common excitation. This deposition (PVD) followed by a liftoff process for the creation
scheme offers disposable sensing modules that can be easily connected to of the metal electrodes. The wafer is diced into single chips of 1
the main electronic board apparatus for the sensing procedures.
cm × 1 cm. Then the chips consisting of memristive nanowires
conjugated with the metal lines are wire-bonded to a specially
the inverting input, that is -233.5. Both stages are in an inverting
designed chip-holder (as shown in Fig.5) with connections
configuration, so the gain is negative in each stage.
corresponding to the main PCB and integrated through board-to-
C. Data-acquisition programming board connectors. The memristive silicon nanowires are first
The high-precision ADC has 3 differential analog inputs, so subjected to a soft oxygen plasma treatment for 15 min for the
to measure all the 12 output signals of the readout circuitry, two generation of hydroxyl terminating groups on the surface, then
multiplexers are adopted. At each applied potential, the functionalized by exposure of the surface to antibody against
multiplexers select the output signals sequentially so that all of Prostate Specific Antigen (PSA), i.e. anti-PSA antibody
them can be measured and the result is stored in a 12-element (Abcam-ab10185), solution in Phosphate Buffered Saline (PBS)
array. After all output signals are measured, the array is sent (pH 7.4 Sigma-Aldrich) for 4 h at room temperature, thoroughly
through serial port of micro-controller to RPi for data rinsed with the same buffer and gently dried with N2 flow. The
processing. antigen uptake is performed through successive 45 min
incubations of the sensing chip in solutions of PSA (Millipore
III. MATERIALS AND METHODS Angebot R-1939458.1; 539834 purchased from Merck) in PBS.
A. Nano-Engineering of Memristive Biosensors
Two-terminal, Schottky-barrier silicon nanofabricated wires B. Measurement procedure with the realized board
arrays demonstrating memristive electrical properties are To carry out the voltage sweep, an applied voltage in the
acquired through a top-down nanofabrication process using range of -0.4 V to +0.4 V is considered as a common excitation
Silicon-on-Insulator (SOI) wafers. The wires are suspended and input. The generated current signals from the memristive bio-
anchored between Nickel Silicide (NiSi) pads for the electrical sensors are in range of pA, so a pre-processing is necessary
characterization. Those pads are defined through e-beam before carrying out the measurements. This pre-processing stage
lithography and realized with Nickel (Ni) evaporation followed includes: (a) current to voltage conversion and filtering (b)
by lift-off and annealing procedures. voltage signal amplification and voltage level correction. The
Finally, the nanowires are defined by e-beam lithography memristive sensors react to the applied voltage immediately
and etched through Deep Reactive Ion Etching (DRIE) cycles of with a current peak and then the current becomes stable after 40-
the Si. Metal lines are integrated on top of the already fabricated 60 ms. Therefore, the measurement for each sensor are applied
nanostructures and serve as extension electrodes to the NiSi pads
at least 40 ms later after applying the excitation voltage. The
of the devices (Fig. 6). A common source and separated drains
duration of the measurement for each sensor is 60 ms to go

978-1-5386-4881-0/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: University of Gent. Downloaded on September 16,2022 at 13:37:36 UTC from IEEE Xplore. Restrictions apply.
through the full voltage range with a step of 10 mV. The
measurement data from ADC for each sensor channel is stored
inside the micro-controller and after the measurement is done for
all sensors, the results are sent to the RPi through a serial port
and the data analysis is carried out. Namely, the connection with
the serial port transmits the data in real time and is elaborated
inside the RPi. The Tera-term software (Tera Term Project,
Japan) is used at the beginning for receiving continuously the
values for the serial port. Then a user-friendly graphical interface
(GUI) is developed using C++ in Qt inside RPi and it contains a
‘Save’ Button that carries out the elaboration of data and finally Fig.8. Sensing output for PSA as acquired with the proposed memristive
saves the processed data in an .xlsx file. biosensing platform.

IV. SYSTEM TESTING accordingly adjusting the sampling parameters. More

specifically, the excitation signal waveform requires four
A. Analytical Performance of the realized board parameters from the graphical user interface: (a) the initial
The memristive biosensing board prototype is tested to voltage (b) the final voltage, (c) the voltage step and (d) the
verify the successful signal acquisition of memristive biosensors sweep time. The waveform starts from the initial voltage and
and for preserving their particular electrical characteristics. The keeps the voltage level for a duration equal to the sweep time,
realized board allows successful simultaneous acquisition of the then increases by the voltage step value and again keeps the
signal from all the individual sensors included in the chip, and it voltage level for the duration of the sweep time and then
is confirmed that the obtained electrical characteristics are in increases till it reaches the final voltage level. Following that, it
very good agreement with the measurements for memristive starts to decrease by voltage step, keeps the duration of sweep
biosensors performed using the probe station as shown in Fig. 1. time and repeats the steps till it reaches the initial voltage. At this
Indicative raw data from current-to-voltage measurements stage, one complete measurement period is accomplished.
acquired is presented in Fig. 7. The crucial aspect of the
B. Sensing application of prostate cancer biomarkers
hysteresis is fulfilled, and the electrical response express the
characteristic hysteresis expected for a memristive biosensor, Antigen coupling and sensing events are detected by
proving that the board successfully records the particular measuring the current-to-voltage characteristics following
electrical signal of the nanosensors (Fig. 7a). Most importantly, incubations of the sensing chip in increasing PSA concentrations
the voltage gap that consists the main bio-detection parameter is in the range of fM. Increasing concentration of antigens results
also successfully obtained with the board, and presented in to a decreasing trend of the voltage gap [7]. This trend is also
figure (Fig. 7b). It is worth highlighting that a very important successfully acquired with the electronic board prototype.
benefit is the significant decrease of the output acquisition time. Indicative results for the uptake of a low and a high PSA
The same measurement procedure that with the probe station concentration on the same sensor are demonstrated in Fig. 8 as
configuration requires approximately 40 min now with the a-proof-of concept. Overall the obtained findings further
implementation of the realized board it only requires some validate the realized platform for detection of extremely small
seconds and the sensing result can be available immediately. traces of cancer markers along with the high potential for
Namely, the whole measurement procedure takes up to 80 efficient future sensing applications with memristive biosensors.
seconds to complete and the sensing parameter i.e. voltage gap
is calculated automatically for each sensor finally resulting to a V. CONCLUSIONS
calibration curve that indicates the relationship between voltage In this work, the design, realization and validation of a novel
gap and concentration of cancer molecules. It is important to and portable, ultrasensitive sensing system based on memristive
highlight that even-though the measurement procedure is fast, effect is presented. The suggested bio-sensing platform achieved
the memristive hysteresis is not affected. The exact a significantly fast, automatized and simultaneous sensing of
measurement time depends on the chosen measurement multiple memristive nano-bio-sensors on a single chip. In
parameters, and the hysteresis maintenance is ensured by addition, the realized system paves the way for the actual
implementation of such sensors in medical environment and as
point-of-care devices, thanks to the platform compactness and
portability along with a user-friendly interface. The system
successfully achieved detection of PSA, the main biomarker of
prostate cancer, at concentrations below the critical level of the
biomarker qualifying its suitability for early stage diagnostics.
The suggested paradigm may further be implemented for bio-
detection involving even more complex schemes (i.e. cells,
tumor extracts) holding great promise for fast and highly
sensitive sensing in clinical diagnostics and therapeutics.

Fig.7. Indicative electrical characteristics aquired with the memristive ACKNOWLEDGMENT

biosensing board prototype demonsrating the characteristic hysteresis (a) Financial support received from the multidisciplinary FNS
and the voltage gap (b) that consits the main bio-detection parameter.
Grant CR32I3 156915.

978-1-5386-4881-0/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: University of Gent. Downloaded on September 16,2022 at 13:37:36 UTC from IEEE Xplore. Restrictions apply.
 REFERENCES [6] S. Carrara, D. Sacchetto, M.A. Doucey, C. Baj-Rossi, G. De Micheli and
Y. Leblebici, “Memristive-biosensors: A new detection method by using
[1] L. Chua, “Memristor-The missing circuit element” IEEE Trans. Circuit nanofabricated memristors”, Sensor Actuat. B-Chem. 171-172:449-457,
Theory, 18(5):507 – 519, 1971. 2012.
[2] L. Chua, S. Kang, “Memristive Devices and systems”, Proc. IEEE, [7] I. Tzouvadaki, F. Puppo, M.A. Doucey, G. De Micheli, S Carrara,
64(2):209–223, 1976. “Computational Study on the Electrical Behavior of Silicon Nanowire
[3] L. Chua, “Resistance switching memories are memristors” Appl. Phys. A, Memristive Biosensors”, IEEE Sensors Journal, 15(11), 6208-6217, 2015.
102(4):765-783, 2011. [8] S. Naus and I. Tzouvadaki, P-E. Gaillardon, A. Biscontini, G. De Micheli,
[4] I. Tzouvadaki and P. Jolly, X. Lu, S. Ingebrandt, G. de Micheli, P. Estrela S. Carrara, “An Efficient Electronic Measurement Interface for
and S. Carrara, “Label-Free Ultrasensitive Memristive Aptasensor” Nano Memristive Biosensors” IEEE International Symposium on Circuits and
Lett. 16: 4472−4476, 2016. Systems ISCAS 2017.
[5] I. Tzouvadaki and N. Aliakbarinodehi, G. De Micheli and S. Carrara,
“The memristive effect as a novelty in drug monitoring” Nanoscale, 9,
9676 - 9684, 2017.

978-1-5386-4881-0/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: University of Gent. Downloaded on September 16,2022 at 13:37:36 UTC from IEEE Xplore. Restrictions apply.