Development of High Performance Breath Acetone Sensing Device
Thesis submitted in partial fulfillment
of the requirements for the degree of
Master of Science (by Research)
in
Electronics and Communications Engineering
by
Thati Anand
201250916
[email protected]
Center fo VLSI and Embedded Systems Technology
International Institute of Information Technology
Hyderabad - 500 032, INDIA
April 2017
c Thati Anand, 2017
Copyright All Rights Reserved
International Institute of Information Technology
Hyderabad, India
CERTIFICATE
It is certified that the work contained in this thesis, titled “Development of High Performance Breath
Acetone Sensing Device” by Thati Anand, has been carried out under my supervision and is not submitted elsewhere for a degree.
Date
Adviser: Dr. Shubhajit Roy Chowdhury
Co-adviser: Dr. Tapan Kumar Sau
To my Family & Friends
Acknowledgments
My journey in IIIT-Hyderabad has been a journey worth it. It could not have been so without the
support of many people. As I submit my MS thesis, I want to offer my gratitude to all those people who
helped me in successfully completing this journey. First of all, I want to thank my guide Dr. Shubhajit
Roy Chowdhury, for accepting me as a student and constantly guiding me. His guidance has helped me
improve not only as a researcher but also as a person. I can never thank him enough for providing me
support at my difficult times and helping me move forward. I would also like to thank my co-adviser
Dr. Tapan Kumar Sau for his continuous support and guidance during the course of my research. And
special thanks to Arunangshu for his contributions which aided my research work.
I am thankful to everyone in CVEST and CCNSB for providing such a positive work environment.
Many thanks to Harsh, Arpit, Vasu, Swathi , Ramakrishna and Harinath for all the help, discussions,
motivation and support. Thanks to Neeraj, Gopi, anwar, Vidya, Divya, Vigneswaran, Gugan, Bhuvanan
and Prateek for remaining by my side rock steady throughout this journey and for being my friends
through my ups and downs. Among my friends, Arpit deserves a special mention for being by my side
while I was going through some rough patches. Thanks to Raghuram, Narendra, ,Amitanshu, mohan,
parijat and karthik for their company, advice, help during courses, research and fun-filled moments
in IIIT. Thanks to Mr. Sathish Kumar, CVEST staff, and shireesha, CCNSB staff, for their help and
support.
I could not have accomplished it without the support and understanding of my parents. I wish to
thank my Mom, Dad, brothers Ravi and Nagesh for being my constant support and motivation. Last,
but not the least, thanks to IIIT community for giving me such a beautiful campus and environment to
grow.
v
Abstract
Researchers have demonstrated that breath acetone is an effective biomarker of Type 2 diabetes
which is the most common form of diabetes. Diabetes treatment requires intermittent monitoring of the
blood glucose levels in the patients. Conventional way of detecting glucose levels is through invasive
technique which involves pricking the finger and collecting blood sample. This is not only painful and
blood consuming but also time consuming and expensive. Therefore, there has been a great demand
for the non-invasive techniques of blood glucose determinations in the commercial market. Researchers
have been attempting to develop a number of non-invasive techniques where the glucose can be measured by different methods outside the body, without puncturing the skin or without taking the blood
sample. However, out of various non-invasive techniques developed, many of them are not suitable for
real-time, point-of-care, and routine uses because they involve high cost and require time-consuming
and complicated sample pretreatment, large space and infrastructure for operations. Detection of breath
acetone can be a rapid, noninvasive, and patient compliant viable alternative to the conventional methods
of blood glucose determination.
Acetone in the breath appears due to increased lipolysis. The breath acetone levels are found to be
less than 0.9 ppm in healthy people and more than 1.8 ppm in diabetes patients. This has led researchers
to develop alternative acetone sensors, especially semiconducting metal oxides (SMOs)-based chemoresistive sensors due to their several favorable attributes. However, achieving the target selectivity,
sensitivity, fast response and recovery, and low detection limit have often been very challenging for
the SMO-based sensors, especially for the sample like exhaled breath that consists of plethora of gases
including a large quantity of water vapor and carbon dioxide.
The thesis work involves designing an embedded system to measure blood glucose levels from breath
acetone. First, it involves testing with the existing SMO-based acetone sensor followed by reducing the
effects of other parameters by using Artificial neural network model. We observed that we still need
improvements in various fronts of sensor design. Sensors for breath acetone monitoring must show high
sensitivity and selectivity and very small response and recovery times, in addition to a good reversibility
and stability. Exhalation is a very fast process ( ≈ 3 s) and a small quantity of acetone (≈0.1 - 10 ppm)
is present in the exhaled breath. Therefore, detection of acetone below 10 ppm is of great significance
in breath acetone sensing. Towards this goal, we developed highly efficient SMO-based acetone sensors
made of Pd nanoparticle-loaded nanostructured SnO2 particles (Pd@SnO2 ). The sensors were prepared
by using simple chemical reduction and sol-gel synthesis methods with no requirements for dispersing
vi
vii
agents or complicated synthesis and fabrication steps. Our Pd@SnO2 sensors exhibited very high sensor
response with small response and recovery time at relatively low operating temperature. Further, the
sensors exhibit excellent reversibility, selectivity over ethanol and noninterference from water vapor
and CO2 , two major constituents of the exhaled breath. These parameters satisfy the requirements of a
real-time breath acetone sensor.
Contents
Chapter
Page
1
Introduction . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Importance of Breath Acetone Sensing . . . . . . . .
1.2 Origin of Breath Acetone . . . . . . . . . . . . . . .
1.3 History of Acetone monitoring . . . . . . . . . . . .
1.4 Glucose Monitoring Methods . . . . . . . . . . . . .
1.4.1 Traditional methods to diagnose Diabetes . .
1.4.2 Invasive detection of blood glucose levels: . .
1.4.3 Noninvasive detection of blood glucose levels
1.5 Problem statement . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
2
System Design and Testing with TGS822 Acetone sensor . . . . . . . . . . . . . . .
2.1 Breath Chamber: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1 TGS822 gas sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 DHT11 Digital Humidity and Temperature sensor . . . . . . . . . . . . .
2.1.3 BMP180 Digital barometric pressure sensor . . . . . . . . . . . . . . . .
2.2 Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Artificial Neural Network model . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Acetone characteristics of TGS822 Sensor . . . . . . . . . . . . . . . . . . . .
2.4.1 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . 8
.
8
.
9
.
11
.
11
.
11
.
12
.
13
.
15
.
17
3
High Performance Acetone Sensor Development . .
3.1 Introduction . . . . . . . . . . . . . . . . . . .
3.2 Sensor Fabrications and Characterizations . . .
3.3 Results and Discussion . . . . . . . . . . . . .
3.3.1 Sensitivity of the sensors . . . . . . . .
3.3.2 Humidity and CO2 gas testing: . . . . .
3.3.3 Response and Recovery time of sensors
3.4 Conclusions . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . 19
.
19
.
21
.
21
.
25
.
27
.
29
.
38
4
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
1
1
1
2
2
3
3
5
6
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
viii
List of Figures
Figure
Page
1.1
’Dextrostix’, Ames Reflectance Meter, used to test blood glucose levels based on color
change. courtesy: http://www.mendosa.com/, The pursuit of non invasive blood glucose
4
The ”Eyetone” blood glucose reflectance colorimeter. courtesy: http://www.mendosa.com/,
The pursuit of non invasive blood glucose . . . . . . . . . . . . . . . . . . . . . . . .
4
The improved version of kyoto Dai-ichi Dextrostix glucose monitoring meter with digital readout. courtesy: http://www.mendosa.com/, The pursuit of non invasive blood
glucose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
1.2
1.3
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
Block Diagram of the Noon-invasive blood glucose detection system through breath
acetone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
TGS822 sensor characteristics. Showing high sensitivity towards Acetone gas . . . . .
TGS822 gas sensor circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Artificial Neural Network Block Diagram . . . . . . . . . . . . . . . . . . . . . . . .
Artificial Neural Neuron Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Artificial Neural Network with 30 hidden neurons . . . . . . . . . . . . . . . . . . .
Acetone detection a). Under normal humidity b). Humidity=90% inside the chamber .
(a) Network with R=0.99662 (b) Trained network regression R=1 (c) Network under
validation R=1 (d) Network under test R=1 . . . . . . . . . . . . . . . . . . . . . . .
Variations in various parameters during monitoring of blood glucose levels over time:
(a) actual concentration of blood glucose levels of a patient over time during invasive
measurement; (b) voltage levels, (c) resistance, (d) pressure, (e) humidity and (f) temperature during non-invasive monitoring of blood glucose. . . . . . . . . . . . . . . .
Electric circuit diagram for the signal measurement of the sensor. . . . . . . . . . . .
XRD Images of as synthesized particles . . . . . . . . . . . . . . . . . . . . . . . . .
Typical SEM image showing the surface morphologies of the Pd@SnO2 particles . . .
A typical elemental composition mapping of the l-Pd@SnO2 sample determined by
EDS: Sn (in green), O (in red) and Pd (in yellow). . . . . . . . . . . . . . . . . . . .
Sensitivity plot for all sensors SnO2 , h-Pd@SnO2 and f-Pd@SnO2 under various acetone and ethanol concentrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A Typical plot showing the sensitivity of acetone sensing . . . . . . . . . . . . . . . .
Table showing the sensitivity (Slopes) of sensors prepared . . . . . . . . . . . . . . .
Typical response of the Half-coated Pd@SnO2 sensor exposed to CO2 gas. The sensor
showed no response to CO2 gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
9
10
10
12
13
14
15
16
18
22
23
23
24
26
27
27
28
x
LIST OF FIGURES
3.9
3.10
3.11
3.12
3.13
3.14
3.15
3.18
3.16
3.19
3.17
3.20
3.21
3.23
3.22
3.24
3.25
3.26
Typical response of the Half-coated Pd@SnO2 sensor exposed to Humidity. The plot
shows humidity sensor response and our sensor response. Humidity sensor showed good
response while our sensor showed almost zero response to humidity . . . . . . . . . .
Pd@SnO2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas
and air alternately. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
SnO2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas and
air alternately. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Response times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors
when exposed to Acetone gas under different concentrations . . . . . . . . . . . . . .
Recovery times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors
when exposed to Acetone gas under different concentrations . . . . . . . . . . . . . .
Response times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors
when exposed to Ethanol gas under different concentrations . . . . . . . . . . . . . .
Recovery times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors
when exposed to Ethanol gas under different concentrations . . . . . . . . . . . . . .
SnO2 sensor exposed to acetone gas concentrations ranging from 1ppm to 100ppm.
Shows increased sensitivity with respect to increase in concentration . . . . . . . . . .
Response reversibility of Pd@SnO2 sensor under 10ppm acetone gas . . . . . . . . .
SnO2 sensor exposed to ethanol gas concentrations ranging from 1ppm to 100ppm.
Shows increased sensitivity with respect to increase in concentration . . . . . . . . . .
Response reversibility of Pd@SnO2 sensor under 10ppm ethanol gas . . . . . . . . .
SnO2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to
100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Half-coated Pd@SnO2 sensor response acetone gas concentrations ranging from 1ppm
to 100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Half-coated Pd@SnO2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to 100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Half-coated Pd@SnO2 sensor response of ethanol gas concentrations ranging from 1ppm
to 100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Full-coated Pd@SnO2 sensor response of acetone gas concentrations ranging from 1ppm
to 100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Full-coated Pd@SnO2 sensor response of ethanol gas concentrations ranging from 1ppm
to 100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Full-coated Pd@SnO2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to 100ppm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
29
30
30
31
31
32
32
33
33
34
34
35
35
36
36
37
37
List of Tables
Table
Page
xi
Chapter 1
Introduction
1.1
Importance of Breath Acetone Sensing
There are many volatile organic compounds (VOCs) present in human breath. Over 1000 VOCs
have been detected to date in the ppmv (parts per million by volume) to pptv (parts per trillion by
volume) concentrations [1]. Acetone gas is one of many such VOCs and is present in the sub-ppm
range. Breath acetone shows a correlation with the blood glucose levels present in any human body.
It suggests that diabetes in which the blood glucose level increases can be diagnosed from the breath
acetone measurement. Breath acetone concentration increases in diabetic patients whereas concentration
levels are low for normal people [2] . Conventional diabetes monitoring is not only painful but also
needs blood sample and often the instruments are expensive. As the technology is moving forward in all
the directions it is the time to reinvent the diabetes monitoring instruments to make it more advanced,
pain-free, and user friendly. Monitoring acetone in the breath appears to be a great step in the diabetes
diagnosis.
1.2
Origin of Breath Acetone
Liver produces three ketone bodies: Acetone (C3 H6 O), acetoacetate (AcAc, C4 H6 O3 and 3-βhydroxybutyrate (3BH). Acetone is produced from these sources in the body: one is from the decarboxylation of acetoacetate and the other is from dehydrogenation of isopropanol [3]. The acetone conversion process through acetoacetate is:
CH3 COCH2 COO+H+ → CH3 COCH2 COOH → CH3 COCH3 +CO2
In fasting subject’s acetoacetate is observed to be ≈ 37% and in diabetic patients it is ≈ 52% [4, 5].
Acetone enters to lungs and it passes through exhaled breath, it also eliminated through urine from
lungs [6]. Reichard et al pointed out that breath and urinary excretion of acetone accounts for a 2 - 30%
of its endogenous production rate, and the in vivo metabolism accounts for the rest [4].
1
1.3
History of Acetone monitoring
Before semiconductor sensors came into picture acetone was measured by using conventional chemical experiments to a number of sophisticated techniques. For example, flame ionization selected ion
flow tube mass spectrometry (SIFT-MS) is the one such technique that was used for the accurate determination of the acetone concentration in breath. For this test it is required to collect breath sample in a
bag and send the bag to laboratory for analysis [7]. There are other methods also to detect acetone from
breath: gas chromatography [8],ion mobility spectrometry [9], cavity ringdown spectroscopy [10]. All
these methods requires a bag to collect the breath and send it to laboratory for analysis. Further, often
they involve high cost, time-consuming complicated sample preparations, etc. Therefore, they are not
suitable for real-time, point-of-care routine uses.
1.4
Glucose Monitoring Methods
Diabetes mellitus is a major health problem worldwide [11]. This health condition arises from many
complex metabolic disorders leading to abnormal glucose levels in a person [7]. Diabetes is a condition in which the body’s natural control of blood sugar has been lost. Whether it is termed type 1
(previously known as“juvenile-onset”), type 2 (“adult-onset”), or the gestational diabetes that is a complication of pregnancy, the end result is the same - glucose may be present in the blood in dangerously
low,”hypoglycemia”,or high,”hyperglycemia”,amounts. Without a means of measuring glucose, treatment is a dangerous guessing game of taking pills, injecting insulin, or deciding how much and what
kind of food to eat [12–14]. Insulin is the principal hormone that regulates the uptake of glucose from
the blood into most cells of the body, especially liver, muscle, and adipose tissue. Therefore, deficiency
of insulin or the insensitivity of its receptors plays a central role in all forms of diabetes mellitus [15].
The classical signal/symptoms of diabetes are weight loss, increased hunger (polyphegia), increased
urination (polyuria) and increased thirst (polydipsia) [16]. These symptoms may develop rapidly in
type-1 diabetes whereas in type-2 it is very slow. In addition it may cause kidney failure, cardiovascular
diseases, blurry vision, headache, fatigue and slow healing of wounds. Prolonged high glucose levels
can cause glucose absorption in the lens of eyes, which leads to blurry vision due to the change in lens
shape. The symptoms of hypoglycemia, low blood sugar, are sweating, trembling, feeling uneasy and
serious issues such as confusion, changes in behavior and (rarely) permanent brain damage or death in
severe cases. Mild to moderate cases are self treated by taking extra sugars. Severe cases can lead to
unconsciousness which should be treated with injecting glucagon.
If patients can measure their glucose levels, they can treat the diabetes easily. Diabetes patients can
lead a relatively normal life if they can maintain a proper diet, exercise, meditation and frequent glucose
measurements. In both - hyperglycemia and hypoglycemia - cases, it is required to measure patients’
2
blood glucose levels before taking medication in order to avoid further complications. This can be possible if painless, inexpensive, portable and reliable testing instruments are available to them.
1.4.1
Traditional methods to diagnose Diabetes
The disease has been known since ancient times. Since high blood glucose levels will cause the
kidneys to deposit glucose into urine, in early days Chinese started diagnosing diabetes by seeing if the
ants are attracted to the sugars in patient’s urine [17]. The process of diagnosing diabetes this way was
continued for a century and later chemical techniques were introduced. In 1941, the Ames Division of
Miles Laboratories, in Elkhart, Indiana, introduced ”Clinitest” tablets. The tablet based on a standard
test for certain sugars involving copper sulfate, called Benedict’s solution. These ”Clinitest” tablets can
be added to a few drops of urine. This will change the color from bright blue to orange and based
on the instruction sheet approximate glucose level can be estimated from urine sample [18]. However,
diagnosing diabetes from urine test has some serious issues. When a person first develops diabetes,
the level of glucose in urine is a reasonable indication of excessive amounts in the blood; however,
because both normal and low blood glucose levels result in no glucose in urine, it is never possible to
assess those blood levels by using urine tests. As the disease progresses over time, it becomes much
less reliable as a marker of high blood glucose. Even early on, its never an accurate measure, and
even though improved testing devices (dipsticks) have been developed over the years, its never been
more than a semi-quantitative test [19]. To get accurate values, its necessary to measure the amount of
glucose in the blood itself, and this has long been done in doctors offices and laboratories. However,
for people with diabetes to maintain healthy levels of glucose, there has always been a need for simple,
accurate tests they could perform at home [19].
1.4.2
Invasive detection of blood glucose levels:
In 1964, after developing many dipstick tests for urine, Ernest Adams of Ames developed a practical test strip for measuring glucose in blood named Dextrostix, after dextrose, another name for glucose [20,21]. Instead of using a chemical reaction to measure glucose, as Clinitest had done, Dextrostix
used a biochemical reaction with an enzyme called glucose oxidase, which reacted with glucose to
produce hydrogen peroxide. The hydrogen peroxide produced a color from another chemical called
o-tolidine, and the amount of color on the strip after exposing it to a drop of blood was a good measure
of the amount of glucose present [22]. At first, the strength of color developed was simply compared to
a series of printed colors on the label, and the glucose concentration was estimated by the color comparison. The procedure was not trivial but could be mastered by people with reasonable dexterity for home
use.
3
Figure 1.1: ’Dextrostix’, Ames Reflectance Meter, used to test blood glucose levels based on color
change. courtesy: http://www.mendosa.com/, The pursuit of non invasive blood glucose
The major limitation to this approach, aside from the timing and manipulation involved, is that visual
acuity and the ability to perceive color accurately decrease with age [23].The Dextrostix as shown in the
Figure 1.1is the first, developed at Ames by Anton Clemens, was called the Ames Reflectance Meter,
or A.R.M. According to interviews with Clements, he was ordered to drop the project several times
but somehow managed to bring it to the market, and the first electronic blood glucose device could be
purchased in about 1970 for about $400. Unfortunately, it had some reliability problems, mostly from
its rechargeable lead-acid batteries, and its use didnt become widespread [24].
Figure 1.2: The ”Eyetone” blood glucose reflectance colorimeter. courtesy: http://www.mendosa.com/,
The pursuit of non invasive blood glucose
4
The next electronic strip reader to appear was in about 1972, called the Eyetone,shown in the Figure
1.2, and was manufactured by a Japanese company, Kyoto Dai-ichi (which later changed the company
name to Ark-Ray). It also read Dextrostix, but used a plug-in AC adapter for power instead of batteries.
Albeit the increase in performance over Dextrostix, it has a higher standard deviation so that a baby with
hyperglycemia might be missed by Eyetone glucose measurement [25].
Figure 1.3 shows the improved version of Dextrostix meter.In about 1979, Kyoto Dai-ichi introduced an
improved Dextrostix meter with a digital readout, called the Dextrometer.
Figure 1.3: The improved version of kyoto Dai-ichi Dextrostix glucose monitoring meter with digital
readout. courtesy: http://www.mendosa.com/, The pursuit of non invasive blood glucose
After that there came many instruments to test although they failed at later stages. People came up with
the strips and digital meter to monitor the glucose levels. While glucose monitors and strips were once a
very profitable business, the market contraction since about 2005 has resulted in less research, reduced
sales forces and more intense price competition among the established companies. While it is possible
that this trend will also reduce the emphasis on a noninvasive monitoring solution, there is no indication
of a slowdown among inventors trying to provide novel approaches to solve the problem, even if the big
players appear to be less receptive to ideas presented by these inventors.
1.4.3
Noninvasive detection of blood glucose levels
As discussed in the invasive method, glucose level is determined from a small volume of blood sample collected by finger pricking. Though the test may not pose any risk to a healthy adult who goes
for the diabetes checkup in every 2 to 3 months, but it is very painful to the diabetic patients because
5
every time they have to prick the finger. The current invasive method is based on the enzymatic catalysis
principle where a thin needle is used to prick the finger of the patient to minimize the discomfort [26].
To avoid such painful diagnosis, extensive research has been devoted towards developing non-invasive
techniques that measure blood glucose levels without taking the blood sample [27–32].
Non-invasive diagnosis technique is becoming more prominent in diagnosing diseases due to their pain
free and simple monitoring methods. Non-invasive detection of blood hemoglobin was already reported
by our group in the earlier work [33]. Lieschnegg et al. have developed a sensor to detect failures and
material imperfections in total joint prosthesis based on acceleration measurement non-invasively [34].
Diabetes can also be detected using non-invasive methods. Frequent testing and accurate determination
of glucose levels is essential for diagnosis, effective management and treatment of diabetes mellitus.
Therefore, there have been constant efforts to develop efficient and sensitive techniques for the determination of blood glucose levels. As mentioned above, a number of invasive enzymatic and non-enzymatic
methods and systems have been reported for the detection of glucose [35–38].
Luaibi et al. used nuclear magnetic resonance technique to measure the blood glucose levels noninvasively [27]. Other non-invasive techniques used are electrical impedance, NIR spectroscopy, breath
analysis, ultrasound and thermal spectroscopy [36–38]. However, none of these methods seems to
achieve the desired accuracy due to varying environmental conditions and physical movements and
therefore none of them led to any accurate and safe commercial device. Further, compared to the breath
analyzer other techniques appear to be expensive due to the sensor components involved. This has led
researchers to develop alternative acetone sensors, especially a number of semiconducting metal oxides
(SMOs)-based sensors.
1.5
Problem statement
The thesis work reports non-invasive systems for the determination of blood glucose levels from the
detection of the breath acetone. We have seen earlier that acetone is one of the volatile organic compounds (VOCs) present in the exhaled breath [39, 40]. The acetone present in the exhaled breath is a
metabolic product of the body-fat burning [39]. The breakdown of excess acetyl-CoA from fatty acid
metabolism in diabetic patients leads to increase in the levels of acetone in the blood and finally in the
exhaled breath. Therefore, the breath acetone levels could be a measure of the blood glucose levels of
a person [41]. The breath acetone concentration ranging from 1.7 ppm to 3.7 ppm can be detected in
diabetic patients [2, 42] , whereas it varies between 0.3 and 0.9 ppm for healthy humans [43]. Optical
fiber made of nanostructured films has been used for the detection of VOCs [44]. A number of metal
oxide nanomaterials have shown excellent gas sensing properties [45, 46] because large surface area
and nanoscale surface features of such nanomaterials result in increased sensitivity towards gases [47].
With the newly emerging semiconductor technologies it is possible to design and fabricate nanoparticle
6
gas sensors that can detect sub-ppm VOC concentration in the breath [48]. However, in addition to the
quantity of the gas, natures of both the gas and the metal oxide determine the degree of response of the
sensor [49]. Tin oxide (SnO2 ) based sensor have been widely used in the gas sensing applications due
to their unique gas detection properties [50]. Here we have used the SnO2 particle system as acetone
sensors. The resistance of this SMO sensor varies depending on the quantity of acetone present and can
be detected by potential divider circuit [51]. Several other parameters such as the temperature, humidity,
and pressure have been taken into consideration in the estimation of acetone because these parameters
affect the quantity of acetone sensed by the sensor. We have used the artificial neural network to calculate the glucose levels from breath acetone measurements. We have fabricated new Pd loaded SnO2
particle based systems as high performance acetone sensors.The thesis is organized as follows: First, we
discuss the use of the commercial TGS SnO2 particle system as the acetone sensor. Second, we explored
the acetone sensing performance of three SMO-based chemiresistive acetone sensor materials, namely
pristine SnO2 and two Pd nanoparticle-loaded nanostructured SnO2 particles (Pd@SnO2 ) prepared in
our laboratory. Finally, we compared the performances of the commercial and laboratory made sensors
and conclude the work.
7
Chapter 2
System Design and Testing with TGS822 Acetone sensor
Tin oxide (SnO2 ) based sensors have been widely used in the gas sensing applications due to their
unique gas detection properties. First, we discuss the use of the commercial TGS SnO2 particle system
as the acetone sensor. The resistance of this sensor varies depending on the quantity of acetone present
and can be detected by potential divider circuit. Several other parameters such as the breath chamber
temperature, humidity and pressure have been taken into consideration in the estimation of acetone
because these parameters affect the quantity of acetone sensed by the sensor. Further, the acetone
concentration in the breath chamber will not be the same for the same person every time due to different
flow rates, because all humans cant blow at the same rate into the mouth piece/breath chamber. This can
result in wrong diagnosis. Therefore, we specified the time duration for breathing into the mouth piece.
Finally, we have used the artificial neural network to calculate the glucose levels.
As mentioned earlier, non-invasive detection of blood glucose levels from breath involves monitoring
of several other parameters other than acetone itself. The working prototype of the system is designed
to monitor, record and compute the parameters. The Figure 2.1 represents the block diagram of the
Non-invasive glucose detection system through breath acetone.
The system is divided into three parts:
1. Breath Chamber
2. Controller
3. Artificial Neural Network
2.1
Breath Chamber:
A 215 cm3 mouth piece was designed and three sensors were placed inside the mouth piece for
analysis. Due to the presence of sensors inside the mouth piece the volume of this test chamber is taken
to be approximately 200 cm3 . For analysis of glucose levels from the breath, we considered a total of five
parameters from three different sensors: Voltage and resistance from acetone sensor (Figaro TGS822),
pressure from Digital barometric pressure sensor (BMP180) and temperature & humidity from DHT11
sensor.
8
Figure 2.1: Block Diagram of the Noon-invasive blood glucose detection system through breath acetone
a. TSG822 gas sensor
b. DHT11 Digital Humidity and Temperature sensor
c. BMP180 Digital barometric pressure sensor
2.1.1
TGS822 gas sensor
The TGS822 is gas sensor which has a high sensitivity towards acetone gas and minimally sensitive
to other gases like ethanol, Benzene, n-Hexane, Isobutene, CO and methane. The sensing element of the
figaro TGS822 sensor has very less conductivity at clean air and the conductivity increases in the presence of a detectable gas depending on the concentration. The acetone sensing application is used when
the ethanol is absent in the breath. The acetone sensing relies on the changes in electrical conductivity
due to the change in the sensor surface arising from the reactions between ionosorbed surface oxygen
and acetone gas. In the presence of a deoxidizing gas, the surface density of the negatively charged
oxygen decreases. This results in the decrease of the barrier height in the grain boundary and hence
decrease of sensor resistance.
Gas characteristics of the TGS822 sensor:
Features of Figaro TGS822 sensor:
1. High sensitive to organic solvent vapors such as acetone
2. High stability and reliability over a long period
3. Long life and simple circuitry
9
Figure 2.2: TGS822 sensor characteristics. Showing high sensitivity towards Acetone gas
From the Figure 2.2 it is clearly visible that the temperature is effecting the sensitivity of the sensor.
And also increasing humidity also effects the sensor’s sensitivity towards the gases.
Figure 2.3: TGS822 gas sensor circuit
VC - Sensor input voltage,
VH - Micro heater voltage,
RL - Load resistor.
VC = 5 V; VH = 5 V; RL = 100 K Ω
VRL =
Vc ∗ RL
RL + RS
10
(2.1)
VC
− 1) ∗ RL
(2.2)
VRL
Voltage VRL and RS are calculated from the above Figure 2 for analysis. The sensor accuracy,
especially at low analyte concentrations, shows nonlinear characteristics of the response and depends
on the temperature and relative humidity in addition to the nature of the gas [49] . As mentioned earlier,
the test cant be conducted when alcohol vapor is present in the breath. Lee et al [52] have studied the
fabrication and characteristics of SnO2 gas sensor array for many VOCs. A voltage detecting method
was used to calculate the sensitivity (S) of the sensor:
Rs = (
Rair − Rgas
∗ 100
Rair
and Rgas are the sensor resistances in normal air and under gas [32].
S=
where, Rair
2.1.2
(2.3)
DHT11 Digital Humidity and Temperature sensor
Flow rate and volume of breath blow into the mouth piece cant be controlled and are different for
each person. To compensate this, the effects of pressure, temperature and humidity levels have been
considered for each and every person apart from the actual parameters (voltage and resistance) detected
from the acetone sensor.
DHT11 is a Digital Temperature and Humidity sensor from micropik. It measures humidity, ranges
from 20-90% with a resolution of 1temperature from 0-50o C with a resolution of 1 at accuracy of 2o C.
The sensor is operated at 5VDC supply. DHT11 sensor sends 40bit data on a single data-line which
includes 16 bit Relative Humidity (8bit integer RH data + 8bit decimal RH data), 16 bit Temperature
(8bit integer temperature data + 8bit decimal temperature data) and 8bit checksum. The 40bit single
data-line is connected to the microcontroller board to read the Relative Humidity and Temperatures.
2.1.3
BMP180 Digital barometric pressure sensor
BMP180 is a Barometric pressure sensor from BOSCH Sensortec. It measures atmospheric pressure
ranges from 300-1100 hPa (hectopascal) with a resolution of 0.01 hPa under 0-65o C temperature conditions. The sensor is operated at 3.3V DC supply. The data is transferred via I2 C bus. Sensor is connected
to the microcontroller board via SCL and SDA lines to read data from I2 C bus. Microcontroller board
reads the data from all the three sensors and sends back to the artificial neural network model via serial
port.
2.2
Controller
All the three sensors data were given to the controller board (Arduino Uno R3). I2C communication
protocol is implemented in the controller to communicate with digital sensors (DHT11 and BMP180) to
read data. The read data is converted to standard values and transferred serially to the attached computer
11
where MATLAB program is running. MATLAB is used to plot the real-time values detected from all
the three sensors for mentioned duration of time. And it also saves all parameters for further analysis.
2.3
Artificial Neural Network model
Breath samples have been collected from 30 persons. Each person was advised to blow into the
mouth piece for 5 seconds continuously with the same flow rate. Sensor data was recorded on MATLAB
tool. In each of five parameters maximum and minimum values were taken in the specified time duration.
Invasive test was also performed with ACCUCHECK instrument immediately after taking the breath
test. To analyze and get the relation between the recorded parameters and glucose levels Neural Network
Tool in MATLAB has been used.
Figure 2.4 shows the modeled neural network block diagram. The network is selected to have 30
hidden layer neurons and 1 output layer neuron with 5 inputs and a single output. In hidden layer, the
neurons have the sigmoid transfer function at the output, and in the output layer neurons have a linear
function at the output.
Figure 2.4: Artificial Neural Network Block Diagram
A single neuron model is shown in Figure 2.5 with a sigmoid transfer function at the output.
Here the transfer function is
Gj = Σi xi ∗ Wij + Bias
(2.4)
Output of neuron is given to the sigmoid function
Out = Sigm(G) =
2
−1
1 + e−2G
(2.5)
For a hidden layer neuron, output = G which is the same as the sum of all the inputs to the neuron.
Figure 5 shows the complete neural network model with 5 inputs and 30 hidden layer neurons and one
output layer neuron. 30 samples of each 5 elements (V, R, P, H and T) have been given as the input
matrix to the network with a target output array of 30 Glucose samples, are given below
12
Figure 2.5: Artificial Neural Neuron Model

V1

 R1

Input = 
P 1

H1
T1
V2
R2
P2
H2
T2
V3
R3
P3
H3
T3
...
...
...
...
...

V 30

R30 

P 30 


H30
T 30
h
i
Input = G1 G2 G3 ... G30
2.4
Acetone characteristics of TGS822 Sensor
Acetone concentration detection is done in 200 cm3 test chamber. To create required concentration
inside the chamber we injected liquid acetone into the chamber. Acetone has a molecular weight of
MW=50.08 g/mol and density δ= 0.791 g/cm3 . The gas density of acetone is given by [33]
δ=
P ∗ MW
R∗T
Where,
δ = the density of the gas of acetone in g/L,
P = the standard Atmospheric Pressure (1 atm)
13
(2.6)
Figure 2.6: Artificial Neural Network with 30 hidden neurons
MW= Molecular Weight in g/mol
R= Universal gas constant in atm/mol.K (equal to 0.821 atm/mol.K),
T= temperature in Kelvin.
Thus, one gets 2.36g/L gas density for acetone.
100 ppm acetone stock solution was prepared for testing. For 1 ppm acetone gas concentration inside
the 200 cm3 test chamber required volume is
Vgas =
x
∗ 102
200
(2.7)
where x = 2 cm3
M ass = Vgas ∗ δ = Vliquid ∗ ρ
14
(2.8)
where Vgas is the volume occupied by the gas of acetone which is equal to 2 cm3 , δ is the density of
the gas of acetone as calculated before, ρ is the constant density of liquid acetone. Therefore,
Vliquid =
Vgas ∗ δ
ρ
(2.9)
From the eq. (9) one can calculate the liquid acetone required. Thus one gets Vliquid = 6 µ L for 2cm3
acetone gas that gives 1 ppm acetone concentration in the mouth piece. Therefore, for a concentration
of n ppm acetone inside the chamber n*6 µL liquid acetone is required.
2.4.1
Results and Discussion
Acetone concentrations have been studied in 1ppm - 10ppm levels with and without humidity effect.
First, Figaro TGS822 acetone sensor is tested under normal atmospheric conditions.
Humidity effect has been calculated for acetone sensor under 85-90chamber. The test results are
shown in Figure 6. It is observed that there is an increase of 0.2V in high humidity sensing compared to
the low humidity sensing. Artificial neural network was used to compensate these effects. The network
0.5
0.45
Voltage variation in Volts
0.4
0.35
0.3
without humidity
with humidity
0.25
0.2
0.15
0.1
1
2
3
4
5
6
7
Acetone Concentration in PPM
8
9
10
Figure 2.7: Acetone detection a). Under normal humidity b). Humidity=90% inside the chamber
15
is trained such that the MSE (Mean Square Error) is very low and Regression reaches 1. The trained
network knowledge is stored in terms of weights and biases. The network is fixed with 180 I/O layer
weights and 31 neuron biases. The trained network MSE is 2.75 x e− 27 and the total regression ratio is
0.9962 as shown in Figure 7. The relation between the acetone concentration and sensor response with
humidity effect is already explained and is modeled in ANN tool [14].
From the collected samples it is observed that different values of voltage, resistance, pressure, humidity and temperature have been recorded for every person depending on their glucose levels and flow
rate.
(a)
(b)
(c)
(d)
Figure 2.8: (a) Network with R=0.99662 (b) Trained network regression R=1 (c) Network under validation R=1 (d) Network under test R=1
Figure 2.8 shows the network trained with regression coefficient R=1. Validation and Test results
also show regression coefficient R=1 as shown in Figures 2.8 (c) and 2.8 (d). After training the network
16
it has been tested with different breath inputs and the results compare closely with the actual glucose
levels.
From the results it is observed that the voltage and resistance are playing important roles in measuring the glucose level non-invasively whereas the other parameters affect the measurements very little.
The temperature effect is more compared to other two parameters, the relative humidity and pressure.
The training data is collected from non-diabetic patients with glucose levels in between 80 mg/dL and
140 mg/dL and a few pre-diabetic patients with glucose levels in between 140 mg/dL and 180 mg/dL.
Figure11 (a) shows the variation in concentration of actual blood glucose levels of a patient over time.
Figures 11b to 11f show the variation in different parameters during the non-invasive monitoring of
blood glucose over time.
From the Figures 2.9 (a) to (f) it can be observed that if the glucose levels are high at some point the
voltage will also be high and the resistance will go down and the other parameters are moderate. In some
cases, albeit the glucose levels are high, voltage levels are low because of considerably low pressure,
humidity and temperature levels, which indicate the person did not blow correctly into the mouth piece.
These effects are minimized with the help of neural network tool.
2.4.2
Conclusions
In this study, the applicability of the breath acetone sensing method to the determination of glucose
in human blood is demonstrated. We used acetone sensor for monitoring acetone levels in the exhaled
breath and compared with actual blood glucose levels. We also considered the effects of the pressure,
temperature and humidity parameters on the acetone sensing. This test involved studies of non-diabetic
and pre-diabetic persons. We have also used Artificial Neural Network model for analyzing the data.
The test results show that it is possible to measure the blood glucose levels via breath acetone sensing.
The accuracy of the system can be improved with a large set of data.
17
(a)
(b)
Figure 2.9: Variations in various parameters during monitoring of blood glucose levels over time: (a)
actual concentration of blood glucose levels of a patient over time during invasive measurement; (b)
voltage levels, (c) resistance, (d) pressure, (e) humidity and (f) temperature during non-invasive monitoring of blood glucose.
18
Chapter 3
High Performance Acetone Sensor Development
3.1
Introduction
Developing efficient sensors for the detection of low concentrations of various gases such as alcohol,
acetone, etc. has drawn a lot of attention due to their uses in the fields of biomedicines and public
safety. We have mentioned earlier that the detection and measurement of breath acetone can help in
the treatment and management of diabetes mellitus and drug-resistant epilepsy in children [3, 53, 54].
Breath acetone is an effective biomarker of Type 2 diabetes [55]. As mentioned earlier, diabetes is a
metabolic problem that leads to the rise in blood glucose (sugar) levels in a person. Blood glucose
levels are often measured by collecting blood via painful skin pricking. In order to avoid such painful
testing, a number of non-invasive methods for the determination of blood sugar level have been reported
[56–58]. Breath acetone sensing appears to be one of the promising non-invasive methods of blood sugar
level measurements [58, 59].Acetone in the breath appears due to increased lipolysis [3, 60].The breath
acetone levels are found to be less than 0.9 ppm in healthy people and more than 1.8 ppm in diabetes
patients [61]. A number of research demonstrated that the breath acetone concentration correlated with
the human blood sugar levels [3, 53, 58, 62]. Thus, detection of the breath acetone can be a rapid,
noninvasive, and viable alternative to the conventional methods of blood glucose determination. Further,
breath acetone had also been shown to be a measure of mild to moderate systemic ketosis [54, 63].
Ketogenic diet is prescribed for the treatment for the drug-resistant epilepsy. The treatment of both
the diabetes and the drug-resistant epilepsy require intermittent blood sampling which is not only time
consuming and expensive but also painful and blood consuming, especially for pediatric patients. Breath
acetone-based noninvasive measurements would allow frequent analyses without the requirement of any
blood sampling. This has paramount importance in the daily implementation of the diet, search for better
diets, etc.
A few sophisticated techniques like gas chromatography with flame ionization, selected ion flow tube
mass spectrometry, etc. has been used for the accurate determination of the breath acetone. However,
in addition to their high cost, this kind of instruments requires time-consuming, complicated sample
pretreatment and large space and infrastructure for operations. Therefore, they are not suitable for real19
time, point-of-care, and routine uses. This has led researchers to develop alternative acetone sensors,
especially a number of semiconducting metal oxides (SMOs)-based sensors [64–80]. Semiconducting
metal oxides (SMOs)-based sensors have been emerging as one of the most reliable, efcient, and costeffective sensors. A number of SMOs such as WO3 , SnO2 , ZnO, Fe2 O3 , Fe3 O4 , Co3 O4 , etc. have been
used for acetone sensing [68,73–82].We have chosen nanostructured SnO2 for our present studies. SnO2
is an inexpensive, nontoxic, and highly stable n-type SMO. It can be obtained by various fabrication
techniques. SnO2 has been extensively used to detect a number of gases and organic volatile compounds.
SnO2 has shown excellent sensitivity and response reversibility toward several target gases [83, 84]
Despite various favorable attributes of the SMO-based sensors, often additional treatments and improvements are required for target selectivity, enhanced response and recovery, and low detection limit.
SMOs often exhibit low sensitivity towards the target gases under ambient temperature conditions.
Therefore, SMOs require to be heated to high temperatures (≈300 - 500 o C) to show the desired response and recovery properties. The selectivity of the target gases poses another major problem due
to the similar redox behavior of many gases toward an SMO. For example, the exhaled breath consists
of plethora of gases including a large quantity of water vapor [67, 85]. Water vapor, like acetone, is a
reducing gas. The presence of water vapor poses severe challenges toward the breath acetone determination by using SMO-based sensors. SnO2 , an n-type semiconductor metal oxide, show a drop in the
resistance upon interacting with reducing gases like acetone, water vapor, etc.
Various efforts have been made to overcome such limitations and improve the gas sensing performance of the SMO-based sensors. For example, SMOs have been prepared in the nanostructural and
nanoparticulate forms for enhancing the efficiency of their gas sensing properties. Nanoparticles offer
larger surface areas, plenty of active sites, and higher reactivity for the gas adsorption [86]. The other
strategies have been doping and surface modification by various additives, like metals, metal oxides,
etc. [87–90, 90–93]. These additives enhance the SMO surface reactions of the target gases thereby
significantly enhancing the response of the SMOs. We would like to point out here that though we have
witnessed a great progress, we still need improvements in various fronts of sensor design. Often the
sensor preparations involve complicated synthesis and fabrication steps and the sensors require high operating temperatures for a good response. Further, sensors for breath acetone monitoring must show high
sensitivity and selectivity and very small response and recovery times, in addition to a good reversibility
and stability. Exhalation is a very fast process (≈3 s) and a small quantity of acetone (≈0.1 - 10 ppm)
is present in the exhaled breath. Therefore, detection of acetone below 10 ppm is of great significance
in breath acetone sensing. Here, we report a highly efficient SMO-based chemiresistive acetone sensor
made of Pd nanoparticle-loaded nanostructured SnO2 particles (Pd@SnO2 ). The sensors were prepared
by using simple chemical reduction and sol-gel synthesis methods with no requirements for dispersing
agents or complicated fabrications. Our Pd@SnO2 sensors exhibit a sensor response of 35 - 40 with a <
2s response and ≈8s recovery time at 2 ppm acetone (approx. onset concentration of diabetic zone) at
200o C operating temperature. These parameters satisfy the requirements of a real-time breath acetone
20
sensor. Further, the sensors exhibit excellent reversibility, selectivity over ethanol and noninterference
from water vapor and CO2, the two major constituents of the exhaled breath.
3.2
Sensor Fabrications and Characterizations
We have prepared three SMO-based chemiresistive acetone sensor materials, namely pristine SnO2
and two Pd nanoparticle-loaded nanostructured SnO2 particles (Pd@SnO2 ). The SnO2 particles were
prepared by facile sol-gel methods from stannous chloride (SnCl2 .2H2 O) precursor Pd particles were
generated by the sodium borohydride (NaBH4 ) reduction of potassium tetrachloropalladate (K2 PdCl4 )
salt. In order to see the effects of Pd contents on the acetone sensing behavior of our sensors elements,
we prepared Pd@SnO2 sensors with two different Pd concentrations, namely 0.5 mol% and 1.0 mol%
(l-Pd@SnO2 (half-Pd doped) and h-Pd@SnO2 (full-Pd doped)), respectively. Ma et al. reported that
the amount of Pd loading onto SnO2 particles reached a saturation level at ≈0.7 mol% precursor concentration for similar sized tin oxide particles [94, 95]. In fact, there was very little change in the Pd
loading after ≈0.5 mol% of precursor concentration. Therefore, we have chosen two Pd concentrations,
namely 0.5 mol% and 1.0 mol%, a little below and above the saturation points in order to verify the
effects of Pd amount on the sensor properties. The SnO2 and Pd@SnO2 particles were characterized
by using a scanning electron microscopy (SEM) (Carl Zeiss ultra 55) showed in Figure 3.3 equipped
with energy dispersive X-ray spectroscopy (EDS) showed in Figure 3.4, and X-ray diffraction (XRD)
showed in Figure 3.2.
The sensor was designed on a ceramic tube with preinstalled gold electrodes on it and platinum wires
attached to it. We had taken 20 mg of SnO2 or Pd@SnO2 particles as required and coated on the ceramic
tube. The sensor element construction was done according to ref [96]. The sensor element was kept
under room temperature for a while until it is dried. After that the sensor was placed in the circuit for
testing.
We used nichrome wire as a heating element that generates 200o C at 5V/0.32A. 200o C falls at the
relatively low temperature end of the optimal operating temperature range used for most of the acetone
and ethanol sensing experiments. This heating offers enhanced sensitivity and performance to the sensor
elements via inter-particle sintering and enhanced transducing and surface catalysis [97].
Sensor was tested with acetone vapor in the range of 1-100 ppm concentrations. Ethanol, moisture
(upto 90%), and CO2 testing were also performed to check the interference of the common gases. The
relative humidity (RH) was 40 - 50% while performing experiments.
3.3
Results and Discussion
Metal oxide semiconductors have been intensively used as solid-state sensor materials for detecting
various hazardous, flammable, and pollutant gases. In addition to the chemical compositions and intrinsic structural features, materials properties such as size, shape, and morphological characteristics of
21
Figure 3.1: Electric circuit diagram for the signal measurement of the sensor.
the MOS play important roles in determining their gas sensing properties. Therefore, various synthesis
strategies such as sol-gel, solvothermal, hydrothermal, spray pyrolysis, etc. have been employed for
the preparation of the SMOs. We employed the facile sol-gel method to prepare nanostructured SnO2
particles. Materials in the nanoparticulate forms can be a smart choice as chemiresistive sensors due to
their large surface area to volume ratios, plenty of active sites, and high surface reactivity [98].
In the next step, Pd nanoparticles were directly deposited onto the nanostructured SnO2 particles by
the reduction of PdCl4 2− ions in the dispersion of SnO2 particles. PdCl4 2− complexes were supposed
to bind with surface hydroxyls of SnO2 particles in solution which were then reduced by the borohydride ions forming Pd nanoparticles on the nanostructured SnO2 particles. This protocol is suitable for
the large-scale, rapid preparation of nanoparticle-loaded SnO2 particles. There are a few metastable and
stable tin oxides formed during the chemical synthesis. We carried out XRD studies in order to know the
identity of the tin oxides formed by our chemical synthesis methods. The XRD pattern shown in Figure
3.2 indicates that the particles are mainly made of SnO2 particles along with some other tin oxides.
Measurements based on the full width at half maximum of the diffraction peaks in the Debye-Scherrer
equation gave an average crsytallite size of 10 nm which is in agreement with the SEM data given
below. Figure 3.3 shows the SEM images of pristine SnO2 and Pd@SnO2 particles. One should note
that Pd particles are hardly distinguishable from the background of nanostructured SnO2 . The lack in
contrast between Pd and SnO2 phases is expected because there is a very little difference in their atomic
numbers (46 Pd and 50 Sn) [99]. However, clear nanostructured surface roughness was visible in all the
samples. The roughness appears to result from the aggregation of several smaller nanoparticles created
by the aggregation/deposition of the nanoparticles. The particle size range from ≈10 nm to 25 nm.
Though the lack in phase contrast in SEM images could not confirm the presence of Pd particles on the
SnO2 background, the electron energy dispersive x-ray spectroscopy (EDS) studies clearly showed it.
Sn, O, and Pd elemental composition analysis and mapping of the samples were done by EDS studies.
22
Figure 3.2: XRD Images of as synthesized particles
Figure 3.3: Typical SEM image showing the surface morphologies of the Pd@SnO2 particles
23
Figure 3.4: A typical elemental composition mapping of the l-Pd@SnO2 sample determined by EDS:
Sn (in green), O (in red) and Pd (in yellow).
The elemental mapping Figure 3.4 shows that Pd nanoparticles were successfully and almost uniformly
distributed on the surfaces of the SnO2 particles. Rough surface nanostructures and evenly distributed
Pd nanoparticles multiply the accessible active sites and interactive surface area of the sensor materials
for the adsorption and diffusion of the target gas molecules; thereby enhancing the sensitivity and response rate of the sensors. It is interesting to compare the atom% compositions of the elements obtained
from the EDS analyses. For all the Pd@SnO2 samples, Pd mol% determined from EDS studies were
always higher than that expected from the calculations based on the precursor amounts. The analyses
give O:Sn atom% =2.56 for pristine SnO2 particles. Further, the O:Sn atom% ratio of the l-Pd@SnO2
>> h-Pd@SnO2 ≈ pristine SnO2 particles. The high oxygen quantity might arise due to adsorptions of
oxygen species from ambience.
In the following, we shall explore the main performance parameters, viz., the sensitivity, rate of response, selectivity, and signal-to-noise ratio limit of detection (LOD), of the sensors elements. The
sensitivity was determined by comparing the test signals of the sensor element before and after exposure to the target gas. SnO2 is a wide band-gap (3.6 eV) n-type semiconductor metal oxide (SMO).
Typically n-type semiconductor metal oxides show a drop in the resistance upon interacting with reducing gases like acetone, water vapor, etc. When an n-type SMO particle is exposed to air, atmospheric
oxygen adsorbs onto the surface of the SMO particles and captures electrons from the conduction band
of the SMO producing oxygen ions, O− (ads), O2 − (ads), etc. Such surface reactions lead to a decrease
in the carrier concentration (and hence increase in electrical resistance) due to the formation of depletion
layers on the surface of the SMO particles. The reducing target gases like acetone, ethanol, moisture,
etc., upon exposure to the SMO particles react with the surface-adsorbed oxygen species.
Acetone and ethanol are oxidized to CO2 and H2 O upon reaction with the surface-adsorbed oxygen
species. These reactions release the electrons captured by the oxygen species to the SMO thereby
decreasing the resistance. Since the value of the resistance decreased upon exposure to the reducing
gases, the sensitivity, S, was determined from the relation
S (%) = (Ra - Rg )/Rg x 100,
24
where Ra is the sensor resistance in air (original signal or base resistance) and Rg is the resistance in
the mixture of the target gas and air.
Figure 3.5 shows the dependence of the sensitivities of the pristine SnO2 and two Pd@SnO2 sensor
elements/ on the quantities of acetone and ethanol vapors in the range of 1 to 100 ppm. As discussed
above, ethanol can be a serious interfering agent to acetone sensing. Therefore, we have also examined
its sensor response along with acetone. It is clear from the figure that the sensitivity of all the sensor elements increased with the increase in the concentrations of both acetone and ethanol vapors. In the case
of pristine SnO2 , sensitivities of both acetone and ethanol sensing were almost similar. The sensitivity
increased dramatically upon loading of Pd nanoparticles onto the SnO2 particles. However, the increase
in the sensitivity was observed to be prominently higher in the case of acetone than that of ethanol. An
increase in the Pd nanoparticle loading showed an increase in the sensitivity.
3.3.1
Sensitivity of the sensors
Our sensor elements exhibited an almost linear, remarkably sharp increase in the sensitivity up to 10
ppm and then the increase was slow which appeared to be almost marginal in the case of ethanol. It is
worth noting that even at concentration as low as 1 ppm level of acetone, our sensor elements showed
a sensitivity of ≈25 - 30 magnitude which appears to be very promising from the application point of
view. Further, the sensitivity increases up to ≈65 when acetone concentration changes from 1 to 5 ppm
only as compared to ≈15 to 30 in the case of ethanol. 1 to 10 ppm is an important concentration range in
the acetone detection because a healthy human and a diabetic patient show respectively <0.9 ppm and
>1.8 ppm acetone in the exhaled breath. Further, it is reported that the concentration range of acetone in
exhaled breath lies in ≈0.1 - 10 ppm. Therefore, our Pd@SnO2 sensors can easily distinguish a healthy
person from a diabetic one.
We know that the resistance and hence the sensitivity change on exposure to the reducing gases is a
result of a number of surface-controlled phenomena such as the quantity of chemisorbed oxygen on the
surface of the SMO, the reactions between the chemisorbed oxygen and the target gases, the nature of
contacts among particles, etc. The receptor and transducer properties of the SMOs are determined by
these phenomena [100].
Researchers reported earlier that the loading SMOs with promoters like Pd, Ag, etc. change both the
receptor and transducer properties of the SMOs by changing their surface electronic structures [97,100].
For example, it is reported that Pd can activate the dissociation of molecular oxygen via the formation
of stable oxide, PdO, in air. P-type PdO causes a withdrawal of electrons from the n-type SnO2 particle
underneath [101]. Thus, Pd loading produces a wider electron-depleted layer on the SnO2 particle surface and hence a higher electrical resistance in air than that produced by the oxygen chemisorption only.
The space-charge layer disappears on exposure to reducing gas like acetone or alcohol, which reduces
PdO to Pd thereby releasing the electrons to the SnO2 surface. Thus, introduction of Pd nanoparticle
gives higher gas sensitivity than the pristine SnO2 . Pd nanoparticles being good oxidation catalysts can
increase the catalytic dissociation of the molecular oxygen and can decrease the activation energy barrier
25
of the reaction between the chemisorbed oxygen and the reducing gases. Further, the nanoparticulate
states of the Pd@SnO2 give enhanced adsorption and diffusion of the reacting species. Therefore, Pd
loading can increase not only the sensitivity but also the rate of response of the SMO sensors, which is
corroborated by our dynamic response experiments as given later.
Figure 3.5: Sensitivity plot for all sensors SnO2 , h-Pd@SnO2 and f-Pd@SnO2 under various acetone
and ethanol concentrations
The increase in the sensitivity with the increasing concentration of the target gas can be attributed to
the higher surface coverage of gas molecules. The slower increase in the sensitivity above certain gas
concentrations could be result of the availability of decreasing number of active sites with increasing gas
concentrations. Yamazoe reported that out of the possible three factors, viz., mean particle diameter, the
number of Pd particles per unit surface area of SnO2 , and the total surface area of loaded Pd particles, the
population density of Pd particles played a major role in their electronic and chemical sensitizations [97].
In addition to the sensitivity, selectivity is another important parameter of the sensor performance.
We find that pristine SnO2 hardly show any selectivity to acetone over ethanol. On the other hand,
Pd@SnO2 sensors exhibit certain selectivity for acetone over ethanol. It is not only the responses to
ethanol was less compared to acetone, a clear picture is also obtained from the plots of sensitivity vs.
26
log C, where C is the concentration of target gas as shown in Figure 3.6. A good linear relationship is
observed with clearly distinct slopes as shown in Table 3.7.
Figure 3.6: A Typical plot showing the sensitivity of acetone sensing
Figure 3.7: Table showing the sensitivity (Slopes) of sensors prepared
3.3.2
Humidity and CO2 gas testing:
Figures 3.8 and 3.9 show the effects of moisture and CO2 (major constituents present in the exhaled
breath) on the sensor elements. It is obvious that the conductivities of the sensors are hardly not affected
27
by the moisture and CO2 . This could be attributed to the presence of Pd on the SnO2 particles and a
different sensing mechanism.
Figure 3.8: Typical response of the Half-coated Pd@SnO2 sensor exposed to CO2 gas. The sensor
showed no response to CO2 gas
Figure 3.9: Typical response of the Half-coated Pd@SnO2 sensor exposed to Humidity. The plot shows
humidity sensor response and our sensor response. Humidity sensor showed good response while our
sensor showed almost zero response to humidity
Earlier researchers demonstrated that, in the presence of water vapor, pristine SnO2 and Pd@SnO2
particles have different major surface-adsorbed oxygen species, namely O and O2 , respectively [94].
The authors reported that Pd loading could suppress the water vapor adsorption onto the SnO2 parti28
cles due to the presence of O2 . The presence of different oxygen species could also be ascribed to the
change in sensor performance between the pristine SnO2 and Pd@SnO2 particles. However, the origin
of the difference in the acetone and ethanol gases toward a given sensor is not clear at present. It could
arise from the higher dipole moment of acetone than that of ethanol or some other favorable acetonePd@SnO2 interactions [81].
It requires further studies. However, this demonstrates that the Pd@SnO2 sensors are suitable for the
low concentration acetone detection in the presence of ethanol, CO2 , and moisture.
3.3.3
Response and Recovery time of sensors
Pd nanoparticle loading onto SnO2 particles showed spectacular performance in terms of the speed of
response. Figures 3.10 and 3.11 shows typical plots of the time-dependent signal changes of Pd@SnO2
and pristine SnO2 sensor elements exposed alternately to the target gas and air. The response and
recovery times were as usual taken as the time required for the signal to undergo a change of 90% in the
case of adsorption and desorption of the target gas, respectively. The test signal in the cases of Pd@SnO2
sensors shows almost instantaneous rise on exposure to the target gas. The signal remains almost stable
under constant concentration condition and drops abruptly as soon as the test gas is removed. It evident
Figure 3.10: Pd@SnO2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas and
air alternately.
that the range of the response and recovery times was very narrow, mostly in the single-digit numbers.
The deposition of Pd nanoparticles onto the SnO2 particle surface decreased the response and recovery
times significantly in the case of acetone. The faster response and recovery processes can be ascribed to
the Pd nanoparticle catalyzed faster surface oxygen dissociation and reactions as well as faster diffusion
and transport of the gas molecules and their reaction products.
29
Figure 3.11: SnO2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas and air
alternately.
We achieved a response times as small as 1 s and 4 s and recovery times 7 s and 4 s for 2 ppm
and 100 ppm acetones, respectively, for h-Pd@SnO2 . It demonstrates that our Pd@SnO2 sensors have
extremely fast response and recovery rate for acetone. Such fast response and recovery to a target gas are
very useful for practical applications, especially for a reliable determination of acetone concentration in
the exhaled air during the fast exhalation process ( 3 s). However, it was observed that the Pd loading
could not improve the recovery process in the case of ethanol, though the response speed was enhanced
considerably shown in Figures 3.12 to 3.15.
Figure 3.12: Response times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors when
exposed to Acetone gas under different concentrations
30
Figure 3.13: Recovery times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors when
exposed to Acetone gas under different concentrations
Figure 3.14: Response times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors when
exposed to Ethanol gas under different concentrations
We have tested the response reversibility of the sensors for all concentrations of both acetone and
ethanol. Pd@SnO2 sensors showed good response reversibility. Figures 3.16 and 3.17 shows typical
response and recovery cycles of the sensor elements toward 10ppm acetone gas and 10ppm ethanol
gas. It is clear from the figures that there was neither sign of any decay of response amplitude nor
aging performance over the repeating test cycles. When the sensor was exposed to air, the resistance
returned nearly to the baseline level. On the other hand, the response and recovery curves for the pristine
SnO2 particles were irregular in shape which can be attributed to the restriction of the gas reaction and
diffusion.
31
Figure 3.15: Recovery times of SnO2 , Half-coated Pd@SnO2 and Full-coated Pd@SnO2 sensors when
exposed to Ethanol gas under different concentrations
Figure 3.18 shows the sensitivity of the SnO2 sensor under different gas concentrations ranging
from 1ppm to 100ppm. From the plot it is clearly observed that the sensitivity is increasing linearly
with increase in acetone gas concentration. At low concentrations the sensor showed low response
and recovery times as compared to the high concentrations. The sensitivity difference between 1ppm
acetone and 2ppm acetone concentrations is about 2% which can be identified easily. After exposing to
each gas concentration it was exposed to air to come back to recovery state.
Figure 3.18: SnO2 sensor exposed to acetone gas concentrations ranging from 1ppm to 100ppm. Shows
increased sensitivity with respect to increase in concentration
32
Figure 3.16: Response reversibility of Pd@SnO2 sensor under 10ppm acetone gas
Ethanol sensing was also carried on the same sensor and plotted the response in Figure 3.19. It is
observed that ethanol resonse is almost similar to the acetone gas as shown in Figure 3.20
Figure 3.19: SnO2 sensor exposed to ethanol gas concentrations ranging from 1ppm to 100ppm. Shows
increased sensitivity with respect to increase in concentration
33
Figure 3.17: Response reversibility of Pd@SnO2 sensor under 10ppm ethanol gas
Figure 3.20: SnO2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to
100ppm.
To improve the sensitivity and response time we have use palladium. It is coated on top of the
SnO2 nanoparticles as explained in the synthesis secetion. The half Pd coated SnO2 sensor was tested
with acetone and ethanol gases and plotted the results. Figure 3.21 shows the increased sensitivity of
half-Pd@SnO2 sensor when exposed to acetone gas. In the above figure we can see the response and
recovery time are reduced as compared to SnO2 sensor response. Ethanol gas concentrations were also
exposed to the half-Pd@SnO2 sensor and plotted the results as shown in Figure 3.22 We have compared
34
Figure 3.21: Half-coated Pd@SnO2 sensor response acetone gas concentrations ranging from 1ppm to
100ppm.
acetone and ethanol responses and plotted them in Figure 3.23. From the figure it is clearly observed
that the acetone response is increased more than the ethanol gas. We can also observe that the response
and recovery times are decreased to a considerable amount in acetone than in ethanol.
Figure 3.23: Half-coated Pd@SnO2 sensor response of ethanol and acetone gas concentrations ranging
from 1ppm to 100ppm.
35
Figure 3.22: Half-coated Pd@SnO2 sensor response of ethanol gas concentrations ranging from 1ppm
to 100ppm.
The amount of palladium is further increased (doubled) to increase the sensitivity of acetone gas.
This showed an increased response in acetone and ethanol. Figure 3.24 shows the sensitivity of the
Pd@SnO2 sensor exposed to different concentrations of acetone gas.
Figure 3.24: Full-coated Pd@SnO2 sensor response of acetone gas concentrations ranging from 1ppm
to 100ppm.
The sensor was also exposed to ethanol gas and plotted the results as shown in Figure 3.25.
36
Figure 3.25: Full-coated Pd@SnO2 sensor response of ethanol gas concentrations ranging from 1ppm
to 100ppm.
Comparing both acetone and ethanol gas responses it is clearly observed that acetone sensitivity
higher than the ethanol gas. It also shows the increased response and recovery time for both acetone and
ethanol gases.
Figure 3.26: Full-coated Pd@SnO2 sensor response of ethanol and acetone gas concentrations ranging
from 1ppm to 100ppm.
37
3.4
Conclusions
Our nanostructured Pd@SnO2 particles were prepared by facile chemical synthesis methods. The
particles were characterized by FE-SEM-EDX and XRD. Nanostructured Pd@SnO2 particles showed
excellent acetone sensing performances in terms of high sensitivity, fast response and recovery speed,
and high signal-to-noise ratio. The fabricated sensors offered certain selectivity over ethanol and noninterference from two major constituents of the exhaled breath, water vapor and CO2 gas. Further,
Pd@SnO2 sensors showed excellent response reversibility over the repeating test cycles at all the concentrations of both acetone and ethanol. These parameters satisfy the requirements of a real-time breath
acetone sensor and therefore, our Pd@SnO2 sensors could be used for the fabrication of point-of-care
breath acetone sensors. Our Pd@SnO2 sensors appeared to be better than many acetone sensors and
comparable to the best performing SMO-based sensors reported so far.
38
Chapter 4
Conclusion
Tin oxides are inexpensive, stable, semiconductive metal oxides. They have been used for the sensing of reducing gases such as acetone, alcohols, etc. We have chosen tin oxides for the sensing of
breath acetone. In the first experiment, the applicability of the breath acetone sensing method to the
determination of glucose in human blood is demonstrated. We have used commercial tin oxide (Firgaro TGS822) for monitoring acetone levels in the exhaled breath and compared with the actual blood
glucose levels. We have considered the effects of the pressure, temperature and humidity parameters
on the acetone sensing. This test involved studies of over 30 non-diabetic and pre-diabetic persons.
We have also used Artificial Neural Network model for analyzing the data. The observed correlation
between the acetone concentration and the actual blood glucose level show that it is possible to measure
the blood glucose levels by the breath acetone sensing. The accuracy of the system is maintain with
an error ±7.5mg/dL. From this experiment we have observed that humidity and pressure changes the
actual results and gives an erroneous output. Further, one needs to improve on the fronts of sensitivity,
selectivity, rate of response, and so on.
Towards this goal, we have attempted to design our own acetone sensors. We have fabricated three
different sensors (SnO2 , half-Pd@SnO2 and full-Pd@SnO2 ) and have explored the performances of the
sensors towards acetone sensing. All the three sensors have been tested with acetone, ethanol, moisture,
and CO2 gases with different concentrations ranging from 1ppm to 100ppm. From the experiments we
have observed that Pd coated SnO2 sensors show far better acetone sensing response than the pristine
SnO2 . Increasing the palladium content increases the sensitivity and also reduces the response and
recovery time. The fabricated sensors exhibit certain selectivity over ethanol and do not show any
interference from humidity and CO2 gas. The sensor performance is observed to be better than the most
of the similar sensors reported so far.
39
Related Publications
[1]. Thati A, Biswas A, Chowdhury SR, Sau TK. BREATH ACETONE-BASED NON-INVASIVE
DETECTION OF BLOOD GLUCOSE LEVELS. International Journal on Smart Sensing Intelligent
Systems. 2015 Jun 1;8(2).
[2]. Biswas A,Thati A, Sau TK. Polypyrrole gold nanoparticle composites: catalytic and sensing
properties. CompFlu 2016, IIIT Hyderabad from 12th - 14th Dec 2016.
[3]. Thati A, Biswas A, Mukharjee D, Chowdhury SR, Sau TK. High Performance acetone sensing
by nanostructured Pd@SnOX , ACS Sensors ( Journal under review).
40
Bibliography
[1] D. Hill and R. Binions, “Breath analysis for medical diagnosis,” International Journal on Smart
Sensing and Intelligent Systems, vol. 5, no. 2, pp. 401–440, 2012.
[2] S. Chakraborty, D. Banerjee, I. Ray, and A. Sen, “Detection of biomarker in breath: A step
towards noninvasive diabetes monitoring,” Current science, vol. 94, no. 2, pp. 237–242, 2008.
[3] Z. Wang and C. Wang, “Is breath acetone a biomarker of diabetes? a historical review on breath
acetone measurements,” Journal of breath research, vol. 7, no. 3, p. 037109, 2013.
[4] G. Reichard Jr, A. Haff, C. Skutches, P. Paul, C. Holroyde, and O. Owen, “Plasma acetone
metabolism in the fasting human,” Journal of Clinical Investigation, vol. 63, no. 4, p. 619, 1979.
[5] O. Owen, V. Trapp, C. Skutches, M. Mozzoli, R. Hoeldtke, G. Boden, and G. Reichard, “Acetone
metabolism during diabetic ketoacidosis,” Diabetes, vol. 31, no. 3, pp. 242–248, 1982.
[6] E. M. P. Widmark, “Studies in the acetone concentration in blood, urine, and alveolar air. ii: The
passage of acetone and aceto-acetic acid into the urine,” Biochemical Journal, vol. 14, no. 3-4, p.
364, 1920.
[7] C. Turner, C. Walton, S. Hoashi, and M. Evans, “Breath acetone concentration decreases with
blood glucose concentration in type i diabetes mellitus patients during hypoglycaemic clamps,”
Journal of breath research, vol. 3, no. 4, p. 046004, 2009.
[8] M. Phillips, “Method for the collection and assay of volatile organic compounds in breath,” Analytical biochemistry, vol. 247, no. 2, pp. 272–278, 1997.
[9] V. Ruzsanyi, J. I. Baumbach, S. Sielemann, P. Litterst, M. Westhoff, and L. Freitag, “Detection of human metabolites using multi-capillary columns coupled to ion mobility spectrometers,”
Journal of Chromatography A, vol. 1084, no. 1, pp. 145–151, 2005.
[10] C. Wang and A. Mbi, “A new acetone detection device using cavity ringdown spectroscopy at
266 nm: evaluation of the instrument performance using acetone sample solutions,” Measurement
Science and Technology, vol. 18, no. 8, p. 2731, 2007.
41
[11] J. Gavin, “The importance of monitoring blood glucose,” US Endocrine Disease, vol. 1, no. 1,
pp. 1–3, 2007.
[12] “About diabetes,” World Health Organization, Retrieved 4 April 2014.
[13] K. G. M. M. Alberti and P. f. Zimmet, “Definition, diagnosis and classification of diabetes mellitus and its complications. part 1: diagnosis and classification of diabetes mellitus. provisional
report of a who consultation,” Diabetic medicine, vol. 15, no. 7, pp. 539–553, 1998.
[14] B. Tripathy, H. B. Chandalia, and A. K. Das, RSSDI textbook of diabetes mellitus.
Ltd, 2012.
JP Medical
[15] “Insulin basics,” American Diabetes Association, 24th April-2014.
[16] D. W. Cooke and L. Plotnick, “Type 1 diabetes mellitus in pediatrics,” Pediatr Rev, vol. 29, no. 11,
pp. 374–84, 2008.
[17] D. S. Mull, N. Nguyen, and J. D. Mull, “Vietnamese diabetic patients and their physicians: what
ethnography can teach us,” Western journal of medicine, vol. 175, no. 5, p. 307, 2001.
[18] M. M. Belmonte, E. Sarkozy, and E. R. Harpur, “Urine sugar determination by the two-drop
clinitest method,” Diabetes, vol. 16, no. 8, pp. 557–559, 1967.
R a qualitative dipstick test for micro[19] P. Spooren, J. Lekkerkerker, and I. Vermes, “Micral-test:
albuminuria,” Diabetes research and clinical practice, vol. 18, no. 2, pp. 83–87, 1992.
[20] R. Jarrett, H. Keen, and C. Hardwick, “instant blood sugar measurement using dextrostix and a
reflectance meter,” Diabetes, vol. 19, no. 10, pp. 724–726, 1970.
[21] B. Schersten, C. Kuhl, A. Hollender, and R. Ekman, “Blood glucose measurement with dextrostix
and new reflectance meter,” Br Med J, vol. 3, no. 5927, pp. 384–387, 1974.
[22] E. Mazzaferri, R. Lanese, T. Skillman, and M. Keller, “Use of test strips with colour meter to
measure blood-glucose,” The Lancet, vol. 295, no. 7642, pp. 331–333, 1970.
[23] K. Junker and J. Ditzel, “Inaccuracy of test strips with reflectance meter in determination of high
blood-sugars,” The Lancet, vol. 299, no. 7755, pp. 815–817, 1972.
[24] S. Clarke and J. Foster, “A history of blood glucose meters and their role in self-monitoring of
diabetes mellitus,” British journal of biomedical science, vol. 69, no. 2, p. 83, 2012.
[25] W. Hay and I. Osberg, “The” eyetone” blood glucose reflectance colorimeter evaluated for in
vitro and in vivo accuracy and clinical efficacy.” Clinical chemistry, vol. 29, no. 3, pp. 558–560,
1983.
[26] S. Updike and G. Hicks, “The enzyme electrode,” Nature, vol. 214, pp. 986–988, 1967.
42
[27] A. Y. Luaibi, A. J. Al-Ghusain, A. Rahman, M. H. Al-Sayah, and H. A. Al-Nashash, “Noninvasive
blood glucose level measurement using nuclear magnetic resonance,” in GCC Conference and
Exhibition (GCCCE), 2015 IEEE 8th. IEEE, 2015, pp. 1–4.
[28] S. Ramasahayam, S. H. Koppuravuri, L. Arora, and S. R. Chowdhury, “Noninvasive blood glucose sensing using near infra-red spectroscopy and artificial neural networks based on inverse
delayed function model of neuron,” Journal of medical systems, vol. 39, no. 1, pp. 1–15, 2015.
[29] U. Müller, B. Mertes, C. Fischbacher, K. Jageman, and K. Danzer, “Non-invasive blood glucose
monitoring by means of near infrared spectroscopy: methods for improving the reliability of the
calibration models.” The International journal of artificial organs, vol. 20, no. 5, pp. 285–290,
1997.
[30] D. Guo, D. Zhang, L. Zhang, and G. Lu, “Non-invasive blood glucose monitoring for diabetics
by means of breath signal analysis,” Sensors and Actuators B: Chemical, vol. 173, pp. 106–113,
2012.
[31] S. K. Vashist, “Non-invasive glucose monitoring technology in diabetes management: A review,”
Analytica chimica acta, vol. 750, pp. 16–27, 2012.
[32] M. Goodarzi, S. Sharma, H. Ramon, and W. Saeys, “Multivariate calibration of nir spectroscopic
sensors for continuous glucose monitoring,” TrAC Trends in Analytical Chemistry, vol. 67, pp.
147–158, 2015.
[33] C. Sharma, S. Kumar, A. Bhargava, and S. R. Chowdhury, “Field programmable gate array based
embedded system for non-invasive estimation of hemoglobin in blood using photoplethysmography,” Int. J. Smart Sens. And Intel. Sys, vol. 6, pp. 1268–1282, 2013.
[34] M. Lieschnegg, M. Zacherl, B. Lechner, C. Weger, and A. Fuchs, “Non-invasive characterization
of total hip arthroplasty by means of passive acceleration measurement,” International Journal
on Smart Sensing and Intelligent Systems, vol. 3, no. 1, pp. 75–87, 2010.
[35] M. M. Rahman, A. Ahammad, J.-H. Jin, S. J. Ahn, and J.-J. Lee, “A comprehensive review of
glucose biosensors based on nanostructured metal-oxides,” Sensors, vol. 10, no. 5, pp. 4855–
4886, 2010.
[36] T. Wang, Y. Yu, H. Tian, and J. Hu, “A novel non-enzymatic glucose sensor based on cobalt
nanoparticles implantation-modified indium tin oxide electrode,” Electroanalysis, vol. 26, no. 12,
pp. 2693–2700, 2014.
[37] A. A. Saei, J. E. N. Dolatabadi, P. Najafi-Marandi, A. Abhari, and M. de la Guardia, “Electrochemical biosensors for glucose based on metal nanoparticles,” TrAC Trends in Analytical
Chemistry, vol. 42, pp. 216–227, 2013.
43
[38] D. Zhai, B. Liu, Y. Shi, L. Pan, Y. Wang, W. Li, R. Zhang, and G. Yu, “Highly sensitive glucose
sensor based on pt nanoparticle/polyaniline hydrogel heterostructures,” ACS nano, vol. 7, no. 4,
pp. 3540–3546, 2013.
[39] M. Righettoni, A. Amann, and S. E. Pratsinis, “Breath analysis by nanostructured metal oxides
as chemo-resistive gas sensors,” Materials Today, vol. 18, no. 3, pp. 163–171, 2015.
[40] P. Mochalski, K. Unterkofler, G. Teschl, and A. Amann, “Potential of volatile organic compounds
as markers of entrapped humans for use in urban search-and-rescue operations,” TrAC Trends in
Analytical Chemistry, vol. 68, pp. 88–106, 2015.
[41] Q. Zhang, P. Wang, J. Li, and X. Gao, “Diagnosis of diabetes by image detection of breath using
gas-sensitive laps,” Biosensors and Bioelectronics, vol. 15, no. 5, pp. 249–256, 2000.
[42] C. Deng, J. Zhang, X. Yu, W. Zhang, and X. Zhang, “Determination of acetone in human breath
by gas chromatography–mass spectrometry and solid-phase microextraction with on-fiber derivatization,” Journal of Chromatography B, vol. 810, no. 2, pp. 269–275, 2004.
[43] T. I. Nasution, I. Nainggolan, S. D. Hutagalung, K. R. Ahmad, and Z. A. Ahmad, “The sensing
mechanism and detection of low concentration acetone using chitosan-based sensors,” Sensors
and Actuators B: Chemical, vol. 177, pp. 522–528, 2013.
[44] C. Elosua, I. R. Matias, I. Bariain, and F. J. Arregui, “Detection of volatile organic compounds
based on optical fibre using nanostructured films,” 2008.
[45] Y.-F. Sun, S.-B. Liu, F.-L. Meng, J.-Y. Liu, Z. Jin, L.-T. Kong, and J.-H. Liu, “Metal oxide
nanostructures and their gas sensing properties: a review,” Sensors, vol. 12, no. 3, pp. 2610–
2631, 2012.
[46] V. Khanna, “Nanoparticle-based sensors,” Defence Science Journal, vol. 58, no. 5, p. 608, 2008.
[47] C.-C. Wang, Y.-C. Weng, and T.-C. Chou, “Acetone sensor using lead foil as working electrode,”
Sensors and Actuators B: Chemical, vol. 122, no. 2, pp. 591–595, 2007.
[48] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, and M. Meyyappan, “Carbon nanotube sensors for gas and
organic vapor detection,” Nano letters, vol. 3, no. 7, pp. 929–933, 2003.
[49] B. Hakim and D. Zohir, “Enhancement of the neural network modeling accuracy using a submodeling decomposition-based technique, application in gas sensor,” Neural Computing and Applications, vol. 21, no. 8, pp. 1981–1986, 2012.
[50] S. Dmitriev, “Nanosensors engineering: I. structurally modulated sno2 nanowires,” International
Journal on Smart Sensing and Intelligent Systems, vol. 3, no. 4.
44
[51] A. Zvyagin, A. Shaposhnik, S. Ryabtsev, D. Shaposhnik, A. Vasilev, and I. Nazarenko, “Determination of acetone and ethanol vapors using semiconductor sensors,” Journal of Analytical
Chemistry, vol. 65, no. 1, pp. 94–98, 2010.
[52] N. K. Suryadevara, S. C. Mukhopadhyay, and L. Barrack, “Towards a smart non-invasive fluid
loss measurement system,” Journal of medical systems, vol. 39, no. 4, pp. 1–10, 2015.
[53] K. Musa-Veloso, S. S. Likhodii, E. Rarama, S. Benoit, Y.-m. C. Liu, D. Chartrand, R. Curtis,
L. Carmant, A. Lortie, F. J. Comeau, et al., “Breath acetone predicts plasma ketone bodies in
children with epilepsy on a ketogenic diet,” Nutrition, vol. 22, no. 1, pp. 1–8, 2006.
[54] S. S. Likhodii, K. Musa, and S. C. Cunnane, “Breath acetone as a measure of systemic ketosis
assessed in a rat model of the ketogenic diet,” Clinical chemistry, vol. 48, no. 1, pp. 115–120,
2002.
[55] N. Makisimovich, V. Vorotyntsev, N. Nikitina, O. Kaskevich, P. Karabun, and F. Martynenko,
“Adsorption semiconductor sensor for diabetic ketoacidosis diagnosis,” Sensors and Actuators
B: Chemical, vol. 36, no. 1, pp. 419–421, 1996.
[56] D. C. Klonoff, “Noninvasive blood glucose monitoring,” Diabetes Care, vol. 20, no. 3, pp. 433–
437, 1997.
[57] I. Harman-Boehm, A. Gal, A. M. Raykhman, J. D. Zahn, E. Naidis, and Y. Mayzel, “Noninvasive
glucose monitoring: a novel approach,” Journal of diabetes science and technology, vol. 3, no. 2,
pp. 253–260, 2009.
[58] T. D. C. Minh, D. R. Blake, and P. R. Galassetti, “The clinical potential of exhaled breath analysis
for diabetes mellitus,” Diabetes research and clinical practice, vol. 97, no. 2, pp. 195–205, 2012.
[59] A. Thati, A. Biswas, S. R. Chowdhury, and T. K. Sau, “Breath acetone-based non-invasive detection of blood glucose levels,” International Journal on Smart Sensing & Intelligent Systems,
vol. 8, no. 2, 2015.
[60] K. Roberts, A. Jaffe, C. Verge, and P. S. Thomas, “Noninvasive monitoring of glucose levels:
is exhaled breath the answer?” Journal of diabetes science and technology, vol. 6, no. 3, pp.
659–664, 2012.
[61] N. Teshima, J. Li, K. Toda, and P. K. Dasgupta, “Determination of acetone in breath,” Analytica
Chimica Acta, vol. 535, no. 1, pp. 189–199, 2005.
[62] Z. Gong, M. Sun, C. Jian, Z. Wang, M. Kang, Y. Li, and C. Wang, “A ringdown breath acetone
analyzer: performance and validation using gas chromatography-mass spectrometry,” Journal of
Analytical & Bioanalytical Techniques, vol. 2015, 2014.
45
[63] B. E. Landini and S. T. Bravard, “Breath acetone concentration measured using a palm-size enzymatic sensor system,” IEEE Sensors Journal, vol. 9, no. 12, pp. 1802–1807, 2009.
[64] L. Wang, K. Kalyanasundaram, M. Stanacevic, and P. Gouma, “Nanosensor device for breath
acetone detection,” Sensor Letters, vol. 8, no. 5, pp. 709–712, 2010.
[65] G.-F. Meng, Q. Xiang, Q.-Y. Pan, and J.-Q. Xu, “The selective acetone detection based on fe3o4
doped wo3 nanorods,” Sensor Letters, vol. 9, no. 1, pp. 128–131, 2011.
[66] R. C. Biswal, “Pure and pt-loaded gamma iron oxide as sensor for detection of sub ppm level of
acetone,” Sensors and Actuators B: Chemical, vol. 157, no. 1, pp. 183–188, 2011.
[67] M. Righettoni, A. Tricoli, and S. E. Pratsinis, “Si: Wo3 sensors for highly selective detection of
acetone for easy diagnosis of diabetes by breath analysis,” Analytical chemistry, vol. 82, no. 9,
pp. 3581–3587, 2010.
[68] Q. Qi, T. Zhang, L. Liu, X. Zheng, Q. Yu, Y. Zeng, and H. Yang, “Selective acetone sensor based
on dumbbell-like zno with rapid response and recovery,” Sensors and Actuators B: Chemical,
vol. 134, no. 1, pp. 166–170, 2008.
[69] R. Rella, J. Spadavecchia, M. Manera, S. Capone, A. Taurino, M. Martino, A. Caricato, and
T. Tunno, “Acetone and ethanol solid-state gas sensors based on tio 2 nanoparticles thin film
deposited by matrix assisted pulsed laser evaporation,” Sensors and Actuators B: Chemical, vol.
127, no. 2, pp. 426–431, 2007.
[70] A. Teleki, S. Pratsinis, K. Kalyanasundaram, and P. Gouma, “Sensing of organic vapors by flamemade tio 2 nanoparticles,” Sensors and Actuators B: Chemical, vol. 119, no. 2, pp. 683–690,
2006.
[71] H.-W. Zan, C.-H. Li, C.-C. Yeh, M.-Z. Dai, H.-F. Meng, C.-C. Tsai, et al., “Room-temperatureoperated sensitive hybrid gas sensor based on amorphous indium gallium zinc oxide thin-film
transistors,” Applied Physics Letters, vol. 98, no. 25, 2011.
[72] L. Torsi, A. Dodabalapur, L. Sabbatini, and P. Zambonin, “Multi-parameter gas sensors based on
organic thin-film-transistors,” Sensors and Actuators B: Chemical, vol. 67, no. 3, pp. 312–316,
2000.
[73] L. L. Deng, C. X. Zhao, Y. Ma, S. S. Chen, and G. Xu, “Low cost acetone sensors with selectivity
over water vapor based on screen printed tio 2 nanoparticles,” Analytical Methods, vol. 5, no. 15,
pp. 3709–3713, 2013.
[74] D. Chen, X. Hou, T. Li, L. Yin, B. Fan, H. Wang, X. Li, H. Xu, H. Lu, R. Zhang, et al., “Effects
of morphologies on acetone-sensing properties of tungsten trioxide nanocrystals,” Sensors and
Actuators B: Chemical, vol. 153, no. 2, pp. 373–381, 2011.
46
[75] S. F. Bamsaoud, S. Rane, R. Karekar, and R. Aiyer, “Nano particulate sno 2 based resistive films
as a hydrogen and acetone vapour sensor,” Sensors and Actuators B: Chemical, vol. 153, no. 2,
pp. 382–391, 2011.
[76] S. Wang, L. Wang, T. Yang, X. Liu, J. Zhang, B. Zhu, S. Zhang, W. Huang, and S. Wu, “Porous
α-fe 2 o 3 hollow microspheres and their application for acetone sensor,” Journal of Solid State
Chemistry, vol. 183, no. 12, pp. 2869–2876, 2010.
[77] R. Khadayate, J. Sali, and P. Patil, “Acetone vapor sensing properties of screen printed wo 3 thick
films,” Talanta, vol. 72, no. 3, pp. 1077–1081, 2007.
[78] N. Rezlescu, N. Iftimie, E. Rezlescu, C. Doroftei, and P. Popa, “Semiconducting gas sensor for
acetone based on the fine grained nickel ferrite,” Sensors and Actuators B: Chemical, vol. 114,
no. 1, pp. 427–432, 2006.
[79] P. Sahay, “Zinc oxide thin film gas sensor for detection of acetone,” Journal of Materials Science,
vol. 40, no. 16, pp. 4383–4385, 2005.
[80] X.-L. Li, T.-J. Lou, X.-M. Sun, and Y.-D. Li, “Highly sensitive wo3 hollow-sphere gas sensors,”
Inorganic Chemistry, vol. 43, no. 17, pp. 5442–5449, 2004.
[81] S. Liu, F. Zhang, H. Li, T. Chen, and Y. Wang, “Acetone detection properties of single crystalline
tungsten oxide plates synthesized by hydrothermal method using cetyltrimethyl ammonium bromide supermolecular template,” Sensors and Actuators B: Chemical, vol. 162, no. 1, pp. 259–268,
2012.
[82] J. Shi, G. Hu, Y. Sun, M. Geng, J. Wu, Y. Liu, M. Ge, J. Tao, M. Cao, and N. Dai, “Wo 3
nanocrystals: synthesis and application in highly sensitive detection of acetone,” Sensors and
Actuators B: Chemical, vol. 156, no. 2, pp. 820–824, 2011.
[83] M. Parthibavarman, B. Renganathan, and D. Sastikumar, “Development of high sensitivity
ethanol gas sensor based on co-doped sno 2 nanoparticles by microwave irradiation technique,”
Current Applied Physics, vol. 13, no. 7, pp. 1537–1544, 2013.
[84] S.-W. Tsai and J.-C. Chiou, “Improved crystalline structure and h 2 s sensing performance of
cuo–au–sno 2 thin film using sio 2 additive concentration,” Sensors and Actuators B: Chemical,
vol. 152, no. 2, pp. 176–182, 2011.
[85] L. Ferrus, H. Guenard, G. Vardon, and P. Varene, “Respiratory water loss,” Respiration physiology, vol. 39, no. 3, pp. 367–381, 1980.
[86] X. Zhao, B. Cai, Q. Tang, Y. Tong, and Y. Liu, “One-dimensional nanostructure field-effect
sensors for gas detection,” Sensors, vol. 14, no. 8, pp. 13 999–14 020, 2014.
47
[87] N. Gogurla, A. K. Sinha, S. Santra, S. Manna, and S. K. Ray, “Multifunctional au-zno plasmonic nanostructures for enhanced uv photodetector and room temperature no sensing devices,”
Scientific reports, vol. 4, p. 6483, 2014.
[88] N. Murata, T. Suzuki, M. Kobayashi, F. Togoh, and K. Asakura, “Characterization of pt-doped
sno 2 catalyst for a high-performance micro gas sensor,” Physical chemistry chemical physics,
vol. 15, no. 41, pp. 17 938–17 946, 2013.
[89] F. Gyger, A. Sackmann, M. Hübner, P. Bockstaller, D. Gerthsen, H. Lichtenberg, J.-D. Grunwaldt,
N. Barsan, U. Weimar, and C. Feldmann, “Pd@ sno2 and sno2@ pd core@ shell nanocomposite
sensors,” Particle & Particle Systems Characterization, vol. 31, no. 5, pp. 591–596, 2014.
[90] D. Zhang, A. Liu, H. Chang, and B. Xia, “Room-temperature high-performance acetone gas
sensor based on hydrothermal synthesized sno 2-reduced graphene oxide hybrid composite,” RSC
Advances, vol. 5, no. 4, pp. 3016–3022, 2015.
[91] L. Wang, Z. Lou, R. Zhang, T. Zhou, J. Deng, and T. Zhang, “Hybrid co3o4/sno2 core–shell
nanospheres as real-time rapid-response sensors for ammonia gas,” ACS applied materials &
interfaces, vol. 8, no. 10, pp. 6539–6545, 2016.
[92] M. Yuasa, T. Kida, and K. Shimanoe, “Preparation of a stable sol suspension of pd-loaded sno2
nanocrystals by a photochemical deposition method for highly sensitive semiconductor gas sensors,” ACS applied materials & interfaces, vol. 4, no. 8, pp. 4231–4236, 2012.
[93] L. Bagal, J. Patil, I. Mulla, and S. Suryavanshi, “Influence of pd-loading on gas sensing characteristics of sno 2 thick films,” Ceramics International, vol. 38, no. 6, pp. 4835–4844, 2012.
[94] N. Ma, K. Suematsu, M. Yuasa, T. Kida, and K. Shimanoe, “Effect of water vapor on pd-loaded
sno2 nanoparticles gas sensor,” ACS applied materials & interfaces, vol. 7, no. 10, pp. 5863–
5869, 2015.
[95] N. Ma, K. Suematsu, M. Yuasa, and K. Shimanoe, “Pd size effect on the gas sensing properties
of pd-loaded sno2 in humid atmosphere,” ACS applied materials & interfaces, vol. 7, no. 28, pp.
15 618–15 625, 2015.
[96] C. Wang, X. Cui, J. Liu, X. Zhou, X. Cheng, P. Sun, X. Hu, X. Li, J. Zheng, and G. Lu, “Design of
superior ethanol gas sensor based on al-doped nio nanorod-flowers,” ACS Sensors, vol. 1, no. 2,
pp. 131–136, 2015.
[97] N. Yamazoe, “New approaches for improving semiconductor gas sensors,” Sensors and Actuators
B: Chemical, vol. 5, no. 1-4, pp. 7–19, 1991.
[98] T. K. Sau and A. L. Rogach, “Nonspherical noble metal nanoparticles: colloid-chemical synthesis
and morphology control,” Advanced Materials, vol. 22, no. 16, pp. 1781–1804, 2010.
48
[99] A. V. Marikutsa, M. N. Rumyantseva, L. V. Yashina, and A. M. Gaskov, “Role of surface hydroxyl
groups in promoting room temperature co sensing by pd-modified nanocrystalline sno 2,” Journal
of Solid State Chemistry, vol. 183, no. 10, pp. 2389–2399, 2010.
[100] H.-S. Woo, C. W. Na, and J.-H. Lee, “Design of highly selective gas sensors via physicochemical
modification of oxide nanowires: Overview,” Sensors, vol. 16, no. 9, p. 1531, 2016.
[101] S. Matsushima, Y. Teraoka, N. Miura, and N. Yamazoe, “Electronic interaction between metal
additives and tin dioxide in tin dioxide-based gas sensors,” Japanese journal of applied physics,
vol. 27, no. 10R, p. 1798, 1988.
49