Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

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

Microfluidic optoelectronic sensor for salivary diagnostics of stomach

cancer
Yael Zilberman, Sameer R. Sonkusale n
NanoLab, Electrical and Computer Engineering Department, Tufts University, Medford, MA 02155, USA

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

Article history: We present a microfluidic optoelectronic sensor for saliva diagnostics with a potential application for
Received 6 June 2014 non-invasive early diagnosis of stomach cancer. Stomach cancer is the second most common cause of
Received in revised form cancer-related deaths in the world. The primary identified cause is infection by a gram-negative
23 August 2014
bacterium Helicobacter pylori. These bacteria secrete the enzyme urease that converts urea into carbon
Accepted 2 September 2014
dioxide (CO2) and ammonia (NH3), leading to their elevated levels in breath and body fluids. The
proposed optoelectronic sensor will detect clinically relevant levels of CO2 and NH3 in saliva that can
Keywords: potentially be used for early diagnosis of stomach cancer. The sensor is composed of the embedded in a
Optoelectronic sensor microfluidic device array of microwells filled with ion-exchange polymer microbeads doped with various
Microfluidics
organic dyes. The optical response of this unique highly diverse sensor is monitored over a broad
Stomach cancer
spectrum, which provides a platform for cross-reactive sensitivity and allows detection of CO2 and NH3 in
Salivary diagnostics
saliva at ppm levels.
& 2014 Elsevier B.V. All rights reserved.

1. Introduction diagnostic accuracy of 95–99% (Dooley et al., 1984). However, the

Gastric cancer is one of the most common causes of death from
cancer worldwide. In the United States, the estimated new cases
from gastric cancer in 2013 are 21,600, including 10,990 deaths
(NCI, 2014). Because of unspecific clinical symptoms and purely
defined risk factors, the diagnosis is often delayed, leading to low
survival rate. In Europe, the 5-year survival is below 25%, and the
situation is even worse in the United States. Earlier diagnosis
substantially improves the rate of survival: 95% of patients with
cancerous growth, which is confined to the inner stomach lining,
will survive longer than 5 years (Crew and Neugut, 2006). The
primary marker for early identification of stomach cancer is
infection with gram-negative bacterium Helicobacter pylori (Chui
et al., 2005; Ma et al., 1998; Piazuelo et al., 2010; Uemura et al.,
2011). It has been shown that when the bacterium is eliminated in
treated patients at early stages of gastric cancer, the risk of
developing a second gastric cancer decreases by two-thirds (NCI,
2014). To survive in the harsh, acidic environment of the stomach,
H. pylori secretes the enzyme urease that converts urea into carbon
dioxide and ammonia (neutralizing the acidic environment
locally).
Gastric cancer diagnosis is normally done by endoscopy with
biopsy and histopathological evaluation which provides a high
n
Corresponding author.
E-mail address: sameer@ece.tufts.edu (S.R. Sonkusale).
invasive and relentless character of this procedure makes it less
suitable for fast screening (Chen et al., 2009). In addition, this
method is relatively expensive and requires highly skilled medical
personnel. Moreover, some early gastric cancers provide very little
optical contrast with the surrounding tissues and could be missed
during endoscopic examination. Endoscopy with biopsy can be
also used for the identification of H. pylori infection, when
combined with a rapid urease test and microbial culture. Urease
can be also detected in serum, using enzyme-linked immunosor-
bent assay (ELISA), requiring however collecting blood (Parsonnet
et al., 1993). Another approach is breath analysis: isotopically-
labeled urea is orally ingested by the patient and the breath
sample is subsequently analyzed by mass spectrometry to test
for (isotopically labeled) nitrogen and carbon. This approach is
noninvasive, but requires expensive equipment.
Recently, analyzing saliva has been proposed as an alternative
method for early screening of gastric cancer. Salivary diagnostics is
much more convenient than biopsy or blood test: sample collec-
tion is simple, non-invasive and, therefore, not painful (Martin
et al., 2012; Soini et al., 2010; Streckfus and Bigler, 2002). With
respect to the gastric cancer screening, saliva is commonly
analyzed for the presence of abnormal genetic material, or of
specific biomarkers, which are detected by analyte-specific inter-
actions using immunoassay arrays (Ivnitski et al., 2004; Jokerst
et al., 2009; Tan et al., 2008). While these techniques are effective
in early detection of gastric cancer, they are less appropriate for
fast and high throughput screening, particularly outside of medical

http://dx.doi.org/10.1016/j.bios.2014.09.006
0956-5663/& 2014 Elsevier B.V. All rights reserved.

Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
2 Y. Zilberman, S.R. Sonkusale / Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎

facilities. In addition, immunoassays are analyte-specific and,

therefore, limited to measurement of just a few, among many,
biomarkers.
It has been shown recently by means of gas chromatography/
mass spectrometry (GC–MS) that, just as in the case of exhaled
breath, there are a variety of volatile compounds in saliva (Cao and
Duan, 2006; Soini et al., 2010). Some volatile compounds found in
exhaled breath, such as ammonia (NH3) and carbon dioxide (CO2),
are related to gastric cancer (Lechner et al., 2005; Timmer et al.,
2005; Xu et al., 2013). Since volatile compounds exhaled from the
oral cavity will be also present in saliva and at much higher
concentrations, salivary analysis of NH3 and CO2 has a potential for
non-invasive early screening of gastric cancer. However, GC–MS is
time consuming, not portable and requires complex and expensive
equipment, which makes it less suitable for fast and portable
screening.
An alternative approach is implementation of gas-phase cross-
reactive sensor arrays known as ″electronic noses″, using chemir-
esistors or field-effect transistors that provide a collective elec-
trical response (Leja et al., 2013; MacNaughton et al., 2011; Xu
et al., 2013; Zilberman et al., 2011, 2009, 2010). The research in this
field is quite advanced and this concept has been already im-
plemented for detection of numerous volatiles related to various
types of cancer, including gastric cancer (Xu et al., 2013). Another Fig. 1. The concept of early detection of stomach cancer by salivary diagnostics.
approach is based on the use of optical arrays, composed of Helicobacter pylori infection is detected by analyzing saliva for the presence of
elevated levels of dissolved NH3 and CO2, which are products of the H. pylori
various organic dyes which are sensitive to a particular group (or
metabolism (decomposition of urea by the enzyme urease).
groups) of volatiles, providing an ensemble optical response which
can be used for analyte classification (Rakow and Suslick, 2000;
Suslick, 2004; Suslick et al., 2004). We have recently demonstrated aqueous environments) were used. Optical sensors typically rely
that such optical arrays based on halochromic organic dyes on a response at a single wavelength or a narrow range. Our cross-
embedded in microfluidic devices can be effectively used for reactive sensor reads the entire visible spectrum and, therefore,
detection of ppm levels of CO2 and NH3 in liquid (aqueous) phase could extract unique multi-dimensional signatures. Such readouts
(Zilberman et al., 2014a, 2014b). As compared to other, more capture multiple interactions between the dissolved gases (NH3
conventional methods (e.g. electrochemical detection) that can and CO2) and the composite sensing material which contributes to
typically provide good sensitivity, the use of microfluidic compo- both selectivity and sensitivity. An important advantage of the use
site optoelectronic sensors provides a number of important of multiple halochromic materials with different sensing mechan-
advantages, such as small sample volume, improved selectivity isms is that a single device can be used for detection of multiple
etc. (Zilberman et al., 2014a, 2014b). A brief overview of detection dissolved compounds in the same sample. Here, we demonstrate
methods of dissolved CO2 and NH3 can be found in our previous the excellent performance of our miniature optical sensor for
publications (Zilberman et al., 2014a, 2014b). analyzing ppm-level concentrations of NH3 and CO2 in spiked
Though several novel methods have been developed for detect- human saliva samples.
ing dissolved CO2 and NH3, these methods have not been im-
plemented yet for simultaneous detection of NH3 and CO2 in
saliva, whereas there are some important advantages in analyzing 2. Materials and methods
liquid samples of saliva. First, collected samples can be easily
stored, since only tiny volumes are required. Second, it is simpler 2.1. Materials
to prevent contamination from the environment. Finally, salivary
concentrations are much higher than those in breath samples, Two halochromic materials were used: the pH-sensitive dye
where the exhaled compounds are largely diluted by air. For Cresol Red ion-paired to the quaternary ammonium and the
instance, typical NH3 concentrations in gastric juice vary from metalloporphyrin dye Zinc tetraphenylporphyrin (further denoted

50 ppm for healthy individuals to
200 ppm for those infected as CRIP and ZnTPP). The Cresol Red-based ion pair is preferably
with H. pylori (Yang et al., 1995), while breath NH3 concentrations sensitive to dissolved CO2 due to specific interaction between
are only 100 ppb–2 ppm, making NH3 detection in the breath quite dissolved CO2 and the quaternary ammonium ion and the re-
challenging (Kearney et al., 2002). Salivary NH3 concentrations are sponse of the pH-sensitive dye to changes in Brønsted acidity. The
just slightly lower than those in the gastric juice ranging from metalloporphyrin Zinc tetraphenylporphyrin responds to analytes

20 ppm (Huizenga and Gips, 1982). capable of Lewis acidity, while NH3 can act as Lewis base in
In this contribution, we present a portable optoelectronic aqueous environments.
microfluidic sensor for direct detection of NH3 and CO2 in saliva The CRIP preparation procedure is based on our previously
with potential for early screening of stomach cancer ( Figs. 1 and published work (Zilberman et al., 2014a). Briefly, the pH-sensitive
2). The device is based on a microfluidic platform and the sensing dye Cresol Red (Aldrich) and tetraoctylammonium hydroxide
material is composed of organic dyes encapsulated in ion ex- (TOAOH) as an ion-pairing agent were used. The dye containing
change resin microbeads, providing high sensitivity and cross- aqueous solution was prepared by dissolving Cresol Red in 0.1 M
reactive selectivity. The pH-sensitive dye Cresol Red paired to a NaOH and TOAOH solution in organic phase (toluene) was pre-
quaternary ammonium ion (selectively sensitive to dissolved CO2) pared from tetraoctylammonium bromide (TOABr) (Aldrich) and
and Zinc tetraphenylporphyrin that responds to analytes capable silver oxide (Aldrich). These solutions were mixed and stirred and
of Lewis acid/base interactions (NH3 can act as Lewis base in the organic phase was subsequently removed and washed with

Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
 Y. Zilberman, S.R. Sonkusale / Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3

Fig. 2. Sensing device and experimental setup. The microfluidic device embeds an array of microwells filled with the sensing material, ion-exchange polymer microbeads
doped with the Cresol Red-quaternary ammonium ion pair and Zinc tetraphenylporphyrin (chemical structures of the dyes are shown). The device has inlets and outlets used
for sensing material deposition and exposure experiments. The setup includes a flow system used for microbeads deposition and sensing experiments and the sensor itself,
which is composed of the sensing device, a light source (LED), an optical fiber and the USB spectrometer.

0.1 M NaOH. The purified organic phase contained the ion pair of
[Cresol Red-/TOA þ ] dissolved in toluene. The ZnTPP dye solution
was prepared by dissolving 22.5 mg of 5,10,15,20-Tetraphenyl-
21H,23H-porphine zinc dye (Sigma-Aldrich) in 1 ml of ethanol
(4 99.5%, Sigma-Aldrich), followed by addition of 4 ml of deio-
nized water (Zilberman et al., 2014b).
CRIP was encapsulated within ion exchange polymer by mixing
560 mg anion exchange resin microbeads (Dowexs chloride form,
200–400 mesh, Sigma-Aldrich), 1 ml of the CRIP solution in
toluene and 2 ml of toluene and stirring for 1 h at room tempera-
ture. The resulted suspension was dried in an oven at 50 °C
overnight in order to evaporate the toluene. The dry powder was
redispersed in 5 ml of ethanol and 5 ml of deionized water was
subsequently added. ZnTPP was encapsulated by mixing 560 mg
cation exchange resin microbeads (Dowexs 50WX4 hydrogen
form, 200–400 mesh, Sigma-Aldrich), 5 ml of the previously
prepared ZnTPP solution and 5 ml of deionized water and stirring
for 5 h at room temperature.
In order to prepare saliva for analysis, a healthy volunteer was
instructed to expectorate every 30 s into a 50 ml glass beaker after
being asked to stay 30 min without food and liquid. A total of
10 ml of whole saliva was collected, which was diluted 10-fold by
adding deionized water under continuous stirring using a Teflon
magnetic stirrer bar. Diluted samples of saliva were further used
without any filtration for preparation of solutions mimicking
elevated levels of CO2 and NH3, immediately after the saliva
sample collection.
2.2. Microfluidic device fabrication
holes were punched in the upper part of the device using a 16-
gauge needle. The upper and bottom parts of the device (Fig. 2)
were then placed into the March CS-1701F Reactive Ion Etcher
(50 mT) and oxygen was flowed for 30 s at 50 W RF power.
Immediately after removing from the RIE chamber, the PDMS
molds were pressed gently, ensuring that a strong bond is formed.
Ion exchange resin microbeads (
40–80 mm) doped with
either CRIP or ZnTPP were deposited into the array of microwells
(500 mm diameter and depth) of the microfluidic device by
circulating microbeads mixtures through the device using peri-
staltic pumps (Fig. 2). The middle channel was used for exposure
experiments, while the two side channels of the device were used
for deposition of two different types of doped microbeads. The
reason is that cation and anion exchange microbeads used for
ZnTPP and CRIP encapsulation have opposite charge and coagulate
severely if mixed in the same channel.
2.3. Experimental setup
The experimental setup included a flow system for microbeads
deposition and exposure experiments in a range of concentrations
of dissolved CO2 and NH3 and an optoelectronic setup, Fig. 2.
White LED was used as a light source. The light transmitted
through the sensor array was concentrated using a collimating
lens (74-UV UV/vis, 200–2000 nm, Ocean Optics) and guided by an
optical fiber (QP400-2-UV–vis 400 mm premium fiber, UV/vis) to a
portable spectrometer (USB650, Ocean Optics) that recorded
transmittance in the visible range.

A detailed procedure for the microfluidic device fabrication can
be found in our previously published work (Zilberman et al.,
2014a). Briefly, a mask with desired master features was pre-
viously fabricated and used to make a master. PDMS molds were
prepared by pouring the PDMS mixture on top of the master,
followed by vacuum degassing and curing. The PDMS molds were
peeled up from the master wafer and cut. Fluid inlet and outlet
3. Results and discussion
Fig. 3 shows transmission spectra recorded under the sensor
exposure to selected concentrations of NH3, CO2 and their mixture
dissolved in diluted saliva. Two solutions, one with 150 ppm of
CO2 and another one with 20 ppm of NH3 were prepared first
using saliva diluted 10-fold (as described in Section 2.1) by

Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
4 Y. Zilberman, S.R. Sonkusale / Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 3. Transmittance spectra recorded under the sensor exposure to 20 ppm NH3 (a), 150 ppm CO2 (b) and mixtures of 135 ppm CO2 and 2 ppm NH3 (c) and 75 ppm CO2 and
10 ppm NH3 (d) dissolved in diluted saliva (saliva sample was diluted 10-fold by deionized water as described in Section 2.1).

deionized water containing CO2 (bubbled until saturation and
diluted accordingly with pH monitoring) or a standard NH3
solution prepared using deionized water. Saliva diluted 10-fold
by deionized water (without addition of CO2 and NH3) was further
used for dilution to obtain various concentrations. Further denoted
concentrations of CO2 and NH3 do not account for CO2 and NH3
which were already present in the saliva sample. Therefore, the
concentrations denoted correspond to elevated levels of CO2 and
NH3 in saliva. Much lower concentration was used for NH3, since
the sensor was found to be very sensitive to dissolved NH3. All
spectra were first normalized to the maximal intensity and the
reference spectrum obtained with deionized water was subse-
quently subtracted, i.e. ΔT¼ 0 means no response, while higher
values of ΔT indicate higher responses.
There are clear differences in the response magnitude and in
the shape of the obtained spectra depending on the analyte and its
concentration. Spectral shifts in different ranges can be attributed
to specific responses to NH3 (positive responses at 430–470 nm
and 600–680 nm, Fig. 3a) and CO2 (negative shift at 430–500 nm
and positive response at 610–670 nm). The observed spectral
shifts correspond to the changes in the dye color: the colors of
ZnTPP and CRIP change from green to purple and from purple to
yellow under the exposure to dissolved NH3 and CO2, respectively.
Importantly, while these shifts can be still recognized in the sensor
response to the mixture of CO2 and NH3, the responses to mixtures
have some unique characteristics which cannot be attributed to
purely additive effects (Fig. 3c and d). In principle, such unique ″
signatures″ can be used to distinguish dissolved CO2 and NH3 from
other analytes that present in human saliva, as well as in other
complex environments.
Fig. 4 shows representative examples of the sensor response to
different concentrations of NH3, CO2 and their mixtures dissolved
in diluted saliva (prepared as described in Section 2.1) at selected
wavelengths. In all experiments, the exposure time was 30 s (of
pumping the solution containing different concentrations of NH3,
CO2 or both through the device), followed by the recovery with
deionized water (flow rate of 500 μL/min was used in all experi-
ments). The response is shown in terms of the measured transmis-
sion intensity (I) normalized to the baseline intensity (Ib) recorded
when the device is recovered by deionized water. The sensor is
sensitive to ppm-level concentrations of CO2 and NH3 in the
salivary background.
Importantly, the response shape, magnitude and sign vary
significantly depending on the wavelength. This demonstrates
the potential of the sensor for selective detection in complex
environments (such as saliva), since such variability can be used to
obtain specific signatures. It should be emphasized that the sensor
was composed of only two dyes and its cross-reactive variability
can be further increased by incorporating additional dyes which
would response in different ways resulting in a multi-dimensional

Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
 Y. Zilberman, S.R. Sonkusale / Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Fig. 4. Sensor response to different concentrations of CO2 (a), NH3 (b) and their
mixture (c) dissolved in diluted saliva (Section 2.1). Denoted concentrations do not
account for CO2 and NH3 which were already present in the collected saliva sample.
highly-diverse response. It also should be noted that the diverse
character of the sensor response is a direct consequence of the
analysis over the entire visible spectrum, instead of using a
particular wavelength or a short range. With respect to the sensor
stability and reproducibility, it can be seen that the response to
CO2 and NH3 does not change significantly with time, if comparing
three consecutive exposures over a time course of
1 h (Fig. 4a
and b), though some decrease in the response magnitude was
detected for the mixture (Fig. 4c).
Fig. 5 summarizes the obtained results in terms of the peak
response magnitude in the range of concentrations of CO2 (Fig. 5a)
and NH3 (Fig. 5b) in (diluted) saliva for selected wavelengths and
in the wavelength vs. concentration plane (Fig. 5c–g). Relatively
small error bars in Fig. 5a and b (showing standard deviation
between three measurements) indicate that the sensor response is
stable and repeatable over the time course. Evidently, the sensor is
sensitive to different concentrations of CO2 and NH3 (Fig. 5c–e) in
salivary background and provides distinctive signatures which can
be very useful for detection in complex environments. The sensor
is also sensitive to diluted saliva (Fig. 5f), which is expected since
saliva of a healthy person always contains ppm-level concentra-
tions of CO2 and NH3, which are products of respiration and
normal metabolism. On the other hand, addition of extra levels of
CO2 and NH3 (mimicking H. pylori stomach infection in the
stomach) leads to a variety of responses (Fig. 5c–e). It is interesting
that the sensor response to CO2 and NH3 dissolved in (deionized)
water resembles that to diluted saliva (compare Fig. 5f and g). This
indicates that the sensor is preferably sensitive to CO2 and NH3
and, therefore, can be used for their detection in complex
Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
5
environments, e.g. with varying pH. This enhances sensitivity can
be related to the special sensing mechanisms of the halochromic
materials used, the pH indicator-quaternary ammonium ion pair
(CRIP) selectively sensitive to CO2 and the metalloporphyrin
(ZnTPP) sensitive to Lewis acidity (Zilberman et al., 2014a, 2014b).
Finally, we tested the discriminative power of our sensor by
applying a pattern recognition technique, the principle component
analysis (PCA), which is a statistical method that finds new
orthogonal axes (PCs) as a linear combination of the input
variables, computing these factors to maximize the variance
between the classes. This transformation is defined in such a
way that the first principal component has the largest possible
variance (that is, accounts for as much of the variability in the data
as possible) and each succeeding component in turn has the
highest variance possible under the constraint that it is orthogonal
to (i.e., uncorrelated with) the preceding components. The PCA
reduces the dimension of the input variables by linear combina-
tion and provides a new “map” with two or three dimensions
which can be used for discrimination. Fig. 6 shows the PCA of the
sensor response to NH3 and CO2 dissolved in water and diluted
saliva and to diluted saliva. The three types of analytes produced
distinctive signal response patterns by forming well separated
clusters in principle component space. Importantly, there is very
good separation between the response to saliva without addition
of NH3 and CO2 (“healthy” saliva) and saliva with spiked elevated
levels of NH3 and CO2 (mimicking H. pylori infection). Thus, the
sensor provides a potential for early detection of gastric cancer.
4. Conclusions
We have developed a novel optoelectronic microfluidic sensor
for direct detection of dissolved CO2 and NH3 in human saliva,
with a potential for early stomach cancer diagnostics. The sensor is
based on a microfluidic platform and contains a cross-reactive
halochromic array composed of the metalloporphyrin dye and the
pH-sensitive dye/quaternary ammonium ion pair ionically bonded
to ion-exchange polymer microbeads. The use of the composite
sensing material and of wide spectrum detection offers important
advantages for sensitive and selective detection of dissolved CO2
and NH3 in complex environments, directly in liquid samples and
without the use of any separation membrane. Strong ionic inter-
actions between the ion-exchange microbeads and the dyes
prevent the wash out of the sensing material.
The procedure for composite microbeads deposition into the
sensing device is simple and robust and allows easy incorporation
of virtually any combination of organic dyes, providing a route for
cross-reactive sensing. Due to the flexible optical setup, the sensor
can be easily modified for a variety of applications, where portable
detection of dissolved volatiles in small liquid samples is required.
The overall platform is low cost and portable. The sensor is
sensitive to ppm levels of dissolved CO2 and NH3 in human saliva.
Our results are encouraging in the sense that the sensor can
distinguish between the saliva of a healthy person and the saliva
with elevated concentrations of CO2 and NH3, which is typically
associated with H. pylori infection. This provides a potential for a
variety of medical diagnostic applications and for early gastric
cancer diagnosis in particular. From a more general perspective,
the use of composite optical arrays containing various halochromic
materials with different sensing mechanisms can allow detection
of multiple dissolved compounds or specific (e.g. healthy and
unhealthy) backgrounds using a single device, in the same sample.
6 Y. Zilberman, S.R. Sonkusale / Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 5. Sensor response (peak values) in the range of concentrations of CO2 (a) and NH3 (b) in (diluted) saliva for selected wavelengths (error bars show standard deviation
between three measurements) and “fingerprints” (c–g), i.e. sensor peak responses in the wavelength vs. concentration plane, for CO2 and NH3 in saliva (c, d), mixtures of CO2
and NH3 in saliva (e), diluted saliva with no addition of either CO2 or NH3 (f) and aqueous mixtures of CO2 and NH3 (dissolved in deionized water).

Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
 Y. Zilberman, S.R. Sonkusale / Biosensors and Bioelectronics ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Fig. 6. Principle component analysis (PCA) showing discrimination between sensor
responses to diluted saliva, aqueous solution of CO2 and NH3 and CO2 and NH3
dissolved in saliva.
Acknowledgments
Device was fabricated in Tufts Micro and Nanofabrication
Facility. We thank Arturo Parrales Salinas (Tufts University) for
help with implementation of PCA. Project was founded by National
Science Foundation, United States under the Emerging Frontiers in
Research and Innovation (EFRI) program through Grant number
EFRI-1240443.
References
Cao, W., Duan, Y., 2006. Clin. Chem. 52 (5), 800–811.
Chen, C.Y., Kuo, Y.T., Lee, C.H., Hsieh, T.J., Jan, C.M., Jaw, T.S., Huang, W.T., Yu, F.J.,
2009. Radiology 252, 410–417.
Chui, S.Y., Clay, T.M., Lyerly, K.H., Morse, M.A., 2005. Cancer Epidemiol. Biomark.
Prev. 14 (8), 1883–1889.
Please cite this article as: Zilberman, Y., Sonkusale, S.R., Biosensors and Bioelectronics (2014), http://dx.doi.org/10.1016/j.
bios.2014.09.006i
7
Crew, K.D., Neugut, A.I., 2006. World J. Gastroenterol. 12 (3), 354–362.
Dooley, C.P., Larson, A.W., Stace, N.H., Renner, I.G., Valenzuela, J.E., Eliasoph, J.,
Colletti, P.M., Halls, J.M., Weiner, J.M., 1984. Ann. Intern. Med. 101 (4), 538–545.
Huizenga, J.R., Gips, C.H., 1982. J. Clin. Chem. Clin. Biochem. 20, 571–574.
Ivnitski, D., Sitdykov, R., Ivnitski, N., 2004. Anal. Chim. Acta 504, 265–269.
Jokerst, J.V., Raamanathan, A., Christodoulides, N., Floriano, P.N., Pollard, A.A.,
Simmons, G., Wong, J., Gage, C., Furmaga, W.B., Redding, S.W., McDevitt, J.T.,
2009. Biosens. Bioelectron. 24, 3622–3629.
Kearney, D.J., Hubbard, T., Putnam, D., 2002. Dig. Dis. Sci. 47 (11), 2523–2530.
Lechner, M., Karlseder, A., Niederseer, D., Lirk, P., Neher, A., Rieder, J., Tilg, H., 2005.
Helicobacter 10, 385–390.
Leja, M., Liu, H., Haick, H., 2013. Expert Rev. Gastroenterol. Hepatol. 7 (5), 389–391.
Ma, J.-l., You, W.-c., Gail, M.H., Zhang, L., Blot, W.J., Chang, Y.-s., Jiang, J., Liu, W.-d.,
Hu, Y.-r., Brown, L.M., Xu, G.-w., Fraumeni, J.F., 1998. Int. J. Epidemiol. 27,
570–573.
MacNaughton, S., Surwade, S., Ammu, S., Manohar, S., Sonkusale, S., 2011. IEEE Sens.
J., 125–128.
Martin, H.J., Riazanskaia, S., Thomas, P.C.L., 2012. Analyst 137, 3627–3634.
NCI, 2014. 〈www.cancer.gov〉.
Parsonnet, J., Samloff, M.I., Nelson, L.M., Orentreich, N., Vogelman, J.H., Friedman, G.
D., 1993. Cancer Epidemiol. Biomark. Prev. 2, 461–466.
Piazuelo, B.M., Epplein, e., Correa, P., 2010. Infect. Dis. Clin. N. Am. 24, 853–869.
Rakow, N.A., Suslick, K.S., 2000. Nature 406, 710–714.
Soini, H.A., Klouckova, I., Wiesler, D., Oberzaucher, E., Grammer, K., Dixon, S.J., Xu,
Y., Brereton, R.G., Penn, D.J., Novotny, M.V., 2010. J. Chem. Ecol. 36, 1035–1042.
Streckfus, C.F., Bigler, L.R., 2002. Oral Dis. 8 (3), 69–76.
Suslick, K.S., 2004. MRS Bull., 720–725.
Suslick, K.S., Rakow, N.A., Sen, A., 2004. Tetrahedron 60, 11133–11138.
Tan, W., Sabet, L., Li, Y., Yu, T., Klokkevold, P.R., Wong, T., Ho, C.-M., D., 2008. Biosens.
Bioelectron. 24, 266–271.
Timmer, B., Olthuis, W., Berg, A..v.d., 2005. Sens. Actuators B 107, 666–677.
Uemura, N., Okamoto, S., Yamamoto, S., Matsumura, N., Yamguchi, S., Yamakido, M.,
Taniyama, K., Sasaki, N., Schlemper, R.J., 2011. N. Engl. J. Med. 345 (11),
784–789.
Xu, Z.-q., Broza, Y.Y., Ionsecu, R., Tisch, U., Ding, L., Liu, H., Song, Q., Pan, Y.-y., Xiong,
F.-x., Gu, K.-s., Sun, G.-p., Chen, Z.-d., Leja, M., Haick, H., 2013. Br. J. Cancer 108,
941–950.
Yang, D.H., Bom, H.S., Joo, Y.E., Choi, S.K., Rew, J.S., Yoon, C.M., 1995. Dig. Dis. Sci. 40
(5), 1083–1086.
Zilberman, Y., Ameri, S.K., Sonkusale, S.R., 2014a. Sens. Actuators B 194, 404–409.
Zilberman, Y., Chen, Y., Sonkusale, S.R., 2014b. Sens. Actuators B 202, 976–983.
Zilberman, Y., Ionescu, R., Feng, X., Mullen, K., Haick, H., 2011. ACS Nano 5 (8),
6743–6753.
Zilberman, Y., Tisch, U., Pisula, W., Feng, X., Mullen, K., Haick, H., 2009. Langmuir 25
(9), 5411–5416.
Zilberman, Y., Tisch, U., Shuster, G., Pisula, W., Feng, X., Müllen, K., Haick, H., 2010.
Adv. Mater. 22, 4317–4320.