Available online at www.sciencedirect.com
ScienceDirect
IFAC PapersOnLine 53-2 (2020) 15970–15975
Development
of
a
Discrete
Spectrometric
Development
of
a
Discrete
Spectrometric
Development
of
a
Discrete
Spectrometric
NIR
Reflectance
Glucometer
Development
of a Discrete
Spectrometric
NIR
Reflectance
Glucometer
NIR Reflectance Glucometer ∗
NIR Reflectance
Glucometer∗
Jake D. Campbell ∗∗ Lui Holder-Pearson ∗∗
Jake
D.
Campbell
Lui Holder-Pearson
∗
∗
∗
Jake
D.
Campbell
Holder-Pearson
∗
∗
Christopher
Connor
∗ Lui
∗ Knopp ∗
Jake
D.
Campbell
Lui
Holder-Pearson
∗
∗ Jennifer
JakeG.
D.Pretty
Campbell
Lui Benton
Holder-Pearson
∗ Connor
∗
Christopher
G.
Pretty
Benton
Jennifer
Knopp
∗ ∗
Christopher
G.
Pretty
Connor
Benton
Jennifer
Knopp
∗
∗
∗
∗
J. Geoffrey
∗ Connor
∗ Jennifer Knopp ∗
∗
JakeG.
D.Pretty
Campbell
LuiChase
Holder-Pearson
Christopher
Benton
∗
Christopher
G.
Pretty
Connor
Benton
Jennifer
Knopp
∗
J.
Geoffrey
Chase
J.
Geoffrey
∗
∗
∗ ∗ Jennifer Knopp ∗
Christopher G. Pretty
ConnorChase
Benton
J.
Geoffrey
Chase
J.
Geoffrey
Chase
∗
∗
J. Geoffrey
Chase
Mechanical Engineering
Department,
University
of Canterbury,
∗
∗ Mechanical Engineering Department, University of Canterbury,
Department,
University
Canterbury,
∗ Mechanical Engineering
Christchurch,
New
Zealand
(e-mail:of
∗
Mechanical
Engineering
Department,
University
of
Mechanical
Engineering
Department,
University
of Canterbury,
Canterbury,
Christchurch,
New
Zealand
(e-mail:
Christchurch,
New
Zealand
(e-mail:
∗
jake.campbell@pg.canterbury.ac.nz).
Mechanical Engineering
Department,
University
of Canterbury,
Christchurch,
New
Zealand
(e-mail:
Christchurch,
New
Zealand
(e-mail:
jake.campbell@pg.canterbury.ac.nz).
jake.campbell@pg.canterbury.ac.nz).
Christchurch, New Zealand (e-mail:
jake.campbell@pg.canterbury.ac.nz).
jake.campbell@pg.canterbury.ac.nz).
jake.campbell@pg.canterbury.ac.nz).
Abstract:
Abstract:
Abstract:
Currently,
there are no continuous, non-invasive blood glucose monitors. With over 366
Abstract:
Abstract:
Currently,
there
are
no
continuous,
non-invasive
blood
glucose
monitors.
With
over
366
Currently,
there
are
no expected
continuous,
non-invasive
blood
glucose
monitors.
With
over
million
people
worldwide
to be
diagnosed as
diabetic
by 2030,
an alternative
to 366
the
Abstract:
Currently,
there
are
continuous,
non-invasive
blood
glucose
monitors.
With
366
Currently,
there
are no
no expected
continuous,
non-invasive
blood
glucose
monitors.
With over
over
366
million
people
worldwide
expected
to be
be
diagnosed as
diabetic
by
2030,
an
alternative
to
the
million
people
worldwide
to
diagnosed
as
diabetic
by
2030,
an
alternative
to
the
current
invasive
methods
is
critical.
This
paper
investigates
the
use
of
a
discrete
spectometric,
Currently,
there
are
no
continuous,
non-invasive
blood
glucose
monitors.
With
over
366
million people
people
worldwide
expected
to
be paper
diagnosed
as diabetic
diabetic
by
2030,
an
to
million
worldwide
expected
to
be
diagnosed
as
by
an alternative
alternative
to the
the
current
invasive
methods
is
critical.
This
investigates
the
use2030,
of aa discrete
discrete
spectometric,
current
invasive
methods
is
critical.
This
paper
investigates
the
use
of
spectometric,
NIR
reflectance
glucometer
to detect
a change
in glucose
concentration
in alternative
solution.
Attoeach
million
people
worldwide
expected
to
be
diagnosed
as diabetic
by
2030,
an
the
current
invasive
methods
is
critical.
This
paper
investigates
the
use
of
aa discrete
spectometric,
current
invasive
methods
is
critical.
This
paper
investigates
the
use
of
discrete
spectometric,
NIR
reflectance
glucometer
to
detect
a
change
in
glucose
concentration
in
solution.
At
each
NIR
reflectance
glucometer
to
detect
aa light,
change
in
glucose
concentration
in
solution.
At
each
wavelength,
an
LED
is
used
to
emit
and
a
reverse-biased
LED
detects
light
using
current
invasive
methods
is
critical.
This
paper
investigates
the
use
of
a
discrete
spectometric,
NIR
reflectance
glucometer
to
detect
change
in
glucose
concentration
in
solution.
At
each
NIR
reflectance
glucometer
to to
detect
changeand
in glucose
concentration
solution.
At using
each
wavelength,
an LED
LED
is used
used
to
emita light,
light,
and
reverse-biased
LED indetects
detects
light
using
wavelength,
an
is
emit
aaa reverse-biased
LED
light
wavelengths
660 LED
nm, 850
nm,to940
nm,
nm,
1550
nm, 1650
nm. The
discharge
time
of a
NIR
reflectance
glucometer
detect
change
in
glucose
concentration
solution.
At using
each
wavelength,
an
is
to
emit
light,
and
LED
detects
light
wavelength,
an
LED
is used
used
to
emita1450
light,
and
a reverse-biased
reverse-biased
LED in
detects
light
using
wavelengths
660
nm,
850
nm,
940
nm,
1450
nm,
1550
nm,
1650
nm.
The
discharge
time
of
a
wavelengths
660
nm,
850
nm,
940
nm,
1450
nm,
1550
nm,
1650
nm.
The
discharge
time
of
reverse-biased
LED
is
proportional
to
the
temporal
integral
of
the
detected
light
intensity.
wavelength,
an
LED
is
used
to
emit
light,
and
a
reverse-biased
LED
detects
light
using
wavelengths 660
660
nm, is850
850
nm, 940
940 nm,
nm,
1450
nm,
1550 nm,
nm,
1650ofnm.
nm.
The
discharge
time
of aaa
wavelengths
nm,
nm,
1450
nm,
1550
1650
The
discharge
time
of
reverse-biased
LED
proportional
to
the
temporal
integral
the
detected
light
intensity.
reverse-biased
LED
is
proportional
to
the
temporal
integral
of
the
detected
light
intensity.
The
sensor’s 660
response
toproportional
changing
glucose
was
tested
both
water
and
porcine
wavelengths
nm, is
nm, 940 nm,
nm,
1550 nm,
1650
The
discharge
time
of a
reverse-biased
LED
to
the
temporal
integral
of
the
detected
light
intensity.
reverse-biased
LED
is850
proportional
to 1450
theconcentration
temporal
integral
ofnm.
thein
detected
light
intensity.
The
sensor’s
response
to
changing
glucose
concentration
was
in
both
water
and
porcine
−1 tested
The
sensor’s
response
to
changing
glucose
concentration
was
tested
in
both
water
and
porcine
blood.
Glucose
concentration
was
increased
by
0.5
mmol
l
and
compared
with
a
finger
stick
reverse-biased
LED
is
proportional
to
the
temporal
integral
of
the
detected
light
intensity.
The
sensor’s
response
to
changing
glucose
concentration
was
tested
in
both
water
and
porcine
−1
The
sensor’s
response
to changing
glucose concentration
was
tested
in both with
wateraa and
porcine
blood.
Glucose
concentration
was
increased
by
0.5
ll−1
and
compared
finger
stick
blood.
Glucose
concentration
was
increased
by
0.5 mmol
mmol
compared
with
finger
stick
glucometer.
Each
wavelength
exhibited
an expected
change
in and
adsorption
given
only
an increase
−1
The
sensor’s
response
to changing
glucose
concentration
was
tested
in both
water
porcine
blood.
Glucose
concentration
was
by
ll−1
and
compared
with
aa and
finger
stick
blood.
Glucose
concentration
was increased
increased
by 0.5
0.5 mmol
mmol
and
compared
with
finger
stick
glucometer.
Each
wavelength
exhibited
an
expected
change
in
adsorption
given
only
an
increase
glucometer.
Each
wavelength
exhibited
an
expected
change
in
adsorption
given
only
an
increase
−1
in
glucose
concentration.
The
inverted
exponential
increase
in
absorption
is
explained
by
Beer
blood.
Glucose
concentration
was
increased
by
0.5
mmol
l
and
compared
with
a
finger
stick
glucometer.
Each wavelength
wavelength
exhibited
an expected
expected increase
change in
ininadsorption
adsorption
given
only an
an increase
increase
glucometer.
Each
exhibited
an
change
given
only
in
glucose
concentration.
The
inverted
exponential
absorption
is
explained
by
Beer
in
glucose
concentration.
The
inverted
exponential
increase
in
absorption
is
explained
by
Beer
Lambert’s
law.
Wavelengths
660
nm,
850
nm
and
1450
nm
showed
minimal
change
to
absorption,
glucometer.
Each
wavelength
exhibited
an
expected
change
in
adsorption
given
only
an
increase
in
glucose
concentration.
The
inverted
exponential
increase
in
absorption
is
explained
by
Beer
in
glucose
concentration.
The
inverted
exponential
increase
in
absorption
is
explained
by
Beer
Lambert’s
law.
Wavelengths
660
nm,
850
nm
and
1450
nm
showed
minimal
change
to
absorption,
Lambert’s
law.
Wavelengths
660
nm,
850
nm
and
1450
nm
showed
minimal
change
to
absorption,
while
940
nm,
1550
nm
and
1650
nm
showed
considerable
change
in
absorption.
The
1550
nm
in
glucose
concentration.
The
inverted
exponential
increase
in
absorption
is
explained
by
Beer
Lambert’s
law.
Wavelengths
660
nm,
850
nm
and
1450
nm
showed
minimal
change
to
absorption,
Lambert’s
law. 1550
Wavelengths
660
nm,
850
nm and
1450 nm showed
minimal
change
to
absorption,
while
940
nm,
nm
and
1650
nm
showed
considerable
change
in
absorption.
The
1550
nm
−1
−1
while
940
nm,
1550
nm
and
1650
nm
showed
considerable
change
in
absorption.
The
1550
LED
gave
the
greatest
increase
in
absorption
with
a 7%
rise
over minimal
4.3in
mmol
l−1
to 20.6
mmol
lnm
.
Lambert’s
law.
Wavelengths
660
nm,
850
nm
and
1450
nm
showed
change
to
absorption,
while
940
nm,
1550
nm
and
1650
nm
showed
considerable
change
absorption.
The
1550
−1
while
940
nm,
1550
nm
and
1650
nm
showed
considerable
change
in
absorption.
The
1550
nm
−1
−1 .
LED
gave
the
greatest
increase
in
absorption
with
aa 7%
rise
over
4.3
mmol
ll−1 to
20.6
mmol
llnm
LED
gave
the
greatest
increase
in
absorption
with
7%
rise
over
4.3
mmol
to
20.6
mmol
−1
Ratios
of
absorption
responses
(R
,
R
and
R
)
each
proportional
−1gave
−1 ..
while
940
nm,
1550
nm
and
1650
nm
showed
considerable
change
in
absorption.
The
1550
nm
LED
gave
the
greatest
increase
in
absorption
with
a
7%
rise
over
4.3
mmol
l
to
20.6
mmol
l
1550/1650
1550/1450
940/850
LED gave
the
greatest responses
increase in(R
absorption
with
a 7% rise
over
4.3 mmol
l−1gave
to 20.6
mmol l−1 .
Ratios
of
absorption
,, R
and
R
)) each
proportional
1550/1650
1550/1450
940/850
Ratios
of
absorption
and
R
proportional
1550/1650
1550/1450
940/850
increases
inthe
absorption
with increasing
glucose
concentrations.
LED gave
greatest responses
increase
in(R
absorption
with
a 7% rise
over
4.3 mmol
l gave
to 20.6
mmol l .
Ratios
of
absorption
responses
(R
,, R
R
and
R
)) each
each
gave
proportional
1550/1650
1550/1450
940/850
Ratios
of
absorption
responses
(R
R
and
R
each
gave
proportional
increases
in
absorption
with
increasing
glucose
concentrations.
1550/1650
1550/1450
940/850
increases
in
absorption
with
increasing
glucose
concentrations.
Ratios
of
absorption
responses
(R
,
R
and
R
)
each
gave
proportional
increases
in
absorption
with
increasing
glucose
concentrations.
1550/1650
1550/1450
940/850
license
BY-NC-ND
CC
the
under
article
access
open
an
is
This
Authors.
The
2020
©
increases
in
absorption
with
increasing
glucose
concentrations.
Copyright
increases inDiscrete
absorption
increasing non-invasive,
glucose concentrations.
licenses/by-nc-nd/4.0)
Keywords:
NIRwith
spectrometry,
glucose sensor, LED-LED detection, Beer
(http://creativecommons.org/
Keywords:
Discrete
NIR
spectrometry,
non-invasive,
glucose
sensor,
LED-LED
detection,
Beer
Keywords:
Discrete
NIR
spectrometry,
non-invasive,
glucose
sensor,
LED-LED
detection,
Beer
Lambert’s
Law
Keywords:
Discrete
NIR
spectrometry,
non-invasive,
glucose sensor,
sensor, LED-LED
LED-LED detection,
detection, Beer
Beer
Keywords:
Discrete
NIR
spectrometry,
non-invasive,
glucose
Lambert’s
Law
Lambert’s
Law
Keywords:
Discrete NIR spectrometry, non-invasive, glucose sensor, LED-LED detection, Beer
Lambert’s Law
Law
Lambert’s
Lambert’s
Law
1. INTRODUCTION
Continuous non-invasive glucose sensing has been a topic
1. INTRODUCTION
INTRODUCTION
Continuous
non-invasive
glucose
sensing has
been
a
topic
1.
Continuous
non-invasive
glucose
has
been
a
topic
of
research for
over 20 years
(Liu sensing
et al., 2005).
Techniques
1. INTRODUCTION
INTRODUCTION
Continuous
non-invasive
glucose
sensing
has
a
1.
Continuous
non-invasive
glucose
sensing
has been
been
a topic
topic
of
research
for
over
20
years
(Liu
et
al.,
2005).
Techniques
of
research
for
over
20
years
(Liu
et
al.,
2005).
Techniques
such
as near-infrared
(NIR)
spectroscopy
havebeen
beena stud1. INTRODUCTION
non-invasive
glucose
sensing
has
topic
Diabetes is a metabolic
disorder resulting from the inabil- Continuous
of
research
for
over
20
years
(Liu
et
al.,
2005).
Techniques
of
research
for
over
20
years
(Liu
et
al.,
2005).
Techniques
such
as
near-infrared
(NIR)
spectroscopy
have
been
studDiabetes is
is aa metabolic
metabolic disorder
disorder resulting
resulting from
from the
the inabilinabil- such
as
near-infrared
(NIR)
spectroscopy
have
been
studDiabetes
ied,
but
as
yet,
there
have
been
no
commercialized
prodof
research
for
over
20
years
(Liu
et
al.,
2005).
Techniques
ity
of
the
body
to
regulate
the
level
of
glucose
in
the
blood.
such
as
near-infrared
(NIR)
spectroscopy
have
been
studDiabetes
is
metabolic
disorder
resulting
from
the
inabilsuch
as near-infrared
(NIR)
spectroscopy
have been prodstudbut
as
yet,
have
been
no
commercialized
Diabetes
aa metabolic
disorder
resulting
from
ity of
of the
the is
body
to regulate
regulate
the level
level
of glucose
glucose
in the
the inabilblood. ied,
ied,
but
as
yet,athere
there
have
been
no
commercialized
prodity
body
to
the
of
in
the
blood.
providing
solution
al., 2008).
Products
such
as near-infrared
(NIR)
spectroscopy
have been
studThe
diabetes
is ever
ied,
but
as
have
been
no
commercialized
prodDiabetes
a metabolic
disorder
resulting
from
the
ity
ofprevalence
the is
body
toof
regulate
theworldwide
level
of glucose
glucose
inincreasing,
the inabilblood. ucts
ied,
but
as yet,
yet,aathere
there
have(Poddar
been
noet
commercialized
products
providing
solution
(Poddar
et
al.,
2008).
Products
ity
of
the
body
to
regulate
the
level
of
in
the
blood.
The
prevalence
of
diabetes
worldwide
is
ever
increasing,
ucts
providing
solution
(Poddar
et
al.,
2008).
Products
The
prevalence
of
diabetes
worldwide
is
ever
increasing,
using
NIR
measurements
have
reported
good
correlation
ied,
but
as
yet,
there
have
been
no
commercialized
prodwith
over
366
million
people
expected
to
be
diagnosed
by
ucts
providing
a
solution
(Poddar
et
al.,
2008).
Products
ity
of
the
body
to
regulate
the
level
of
glucose
in
the
blood.
The
prevalence
of
diabetes
worldwide
is
ever
increasing,
ucts
providing
a
solution
(Poddar
et
al.,
2008).
Products
NIR
measurements
have
reported
good
correlation
The
of diabetes
is be
ever
increasing,
with prevalence
over 366
366 million
million
peopleworldwide
expected to
to
be
diagnosed
by using
using
NIR
measurements
have
reported
good
with
over
people
expected
diagnosed
by
with
blood
glucose,
but too
many
of
were
providing
a solution
(Poddar
et the
al., estimations
2008).correlation
Products
2030
(Wild366
et million
al.,
Currently,
individuals
with diausing
NIR
have
reported
good
correlation
The
of 2004).
diabetes
worldwide
is be
ever
increasing,
with
over
people
expected
to
be
diagnosed
by ucts
using
NIR measurements
measurements
have
reported
good
correlation
with
blood
glucose,
but
too
many
of
the
estimations
were
with
over
people
expected
to
diagnosed
by
2030 prevalence
(Wild366
et million
al.,
2004).
Currently,
individuals
with diadiawith
blood
glucose,
but
too
many
of
the
estimations
were
2030
(Wild
et
al.,
2004).
Currently,
individuals
with
not
clinically
acceptable
(Poddar
et
al.,
2008;
Smith,
2017;
using
NIR
measurements
have
reported
good
correlation
betes
need
to
use
invasive
methods,
such
as
finger
pricks,
with
blood
glucose,
but
too
many
of
the
estimations
were
with
over
366
million
people
expected
to
be
diagnosed
by
2030
(Wild
et
al.,
2004).
Currently,
individuals
with
diawith
blood
glucose,
but
too
many
of
the
estimations
were
not
clinically
acceptable
(Poddar
et
al.,
2008;
Smith,
2017;
2030
(Wild
et
al.,
2004).
Currently,
individuals
with
diabetes need
need to
to use
use invasive
invasive methods,
methods, such
such as
as finger
finger pricks,
pricks, Nybacka,
not
clinically
acceptable
(Poddar
et
al.,
2008;
Smith,
2017;
betes
2016).
Different
proteins
and
molecules
within
with
blood
glucose,
but
too
many
of
the
estimations
were
to
monitor
blood
glucose
concentration.
The
discomfort
not
clinically
acceptable
(Poddar
et
al.,
2008;
Smith,
2017;
2030
(Wild
et
al.,
2004).
Currently,
individuals
with
diabetes
need
to
use
invasive
methods,
such
as
finger
pricks,
not
clinically
acceptable
(Poddar
et
al.,
2008;
Smith,
2017;
Nybacka,
2016).
Different
proteins
and
molecules
within
betes
need
to
use
invasive
methods,
such
as
finger
pricks,
to
monitor
blood
glucose
concentration.
The
discomfort
Nybacka,
2016).
Different
proteins
and
molecules
within
to
monitor
blood
glucose
concentration.
The
discomfort
blood
absorb
different
wavelengths
of 2008;
light
to differing
not
clinically
acceptable
(Poddar
et
al.,
Smith,
2017;
and
possible
complications
caused
by
these
methods
lead
Nybacka,
2016).
Different
proteins
and
molecules
within
betes
need
to
use
invasive
methods,
such
as
finger
pricks,
to
monitor
blood
glucose
concentration.
The
discomfort
Nybacka,
2016).
Different
proteins
and
molecules
within
absorb
different
wavelengths
of
light
to differing
to
glucose concentration.
The
discomfort
andmonitor
possibleblood
complications
caused by
by these
these
methods
lead blood
blood
absorb
different
wavelengths
of
light
differing
and
possible
complications
caused
methods
lead
extents,
allowing
estimation
of concentration
ofto
com2016).
Different
proteins
and
molecules
within
to
reduced
frequency
of monitoring
patients,
reducblood
absorb
different
wavelengths
of
to
differing
glucose
concentration.
The
discomfort
and
possibleblood
complications
caused by
by by
these
methods
lead Nybacka,
blood
absorb
different
wavelengths
of light
light
toeach
differing
extents,
allowing
estimation
of
concentration
of
each
comand
possible
complications
caused
these
methods
lead
to monitor
reduced
frequency
of monitoring
monitoring
by
patients,
reducextents,
allowing
estimation
of
concentration
of
each
comto
reduced
frequency
of
by
patients,
reducpound
(Delbeck
et
al.,
2019).
blood
absorb
different
wavelengths
of
light
to
differing
ing
the
effectiveness
of
diabetes
prevention
(Chowdhury
extents,
allowing
estimation
of
concentration
of
each
comand
possible
complications
caused
by
these
methods
lead
to
reduced
frequency
of
monitoring
by
patients,
reducextents,
allowing
estimation
of
concentration
of
each
compound
(Delbeck
et
al.,
2019).
to
reduced
frequency
of
monitoring
by
patients,
reducing the
the effectiveness
effectiveness of
of diabetes
diabetes prevention
prevention (Chowdhury
(Chowdhury pound
(Delbeck
et
al.,
2019).
ing
extents,
allowing
estimation
of
concentration
of
each
comet
al.,
2013).
The
majority
of
continuous
glucose
monitors
pound
(Delbeck
et
al.,
2019).
to
reduced
frequency
of
monitoring
by
patients,
reducing
the
effectiveness
of
diabetes
prevention
(Chowdhury
pound
(Delbeck
et
al.,
2019).
ing
the
effectiveness
of
diabetes
prevention
(Chowdhury
The
main
difficulty
in
NIR
detection
is
due
to
the
weak
abet
al.,
2013).
The
majority
of
continuous
glucose
monitors
et
al.,
2013).
The
majority
of
continuous
glucose
monitors
The
main
difficulty
in
NIR
detection
is
due
to
the
weak
abpound
(Delbeck
et
al.,
2019).
(CGMs)
require
subcutaneous
insertion
of
a
filament
for
ing
the
effectiveness
of
diabetes
prevention
(Chowdhury
The
main
difficulty
in
NIR
detection
is
due
to
the
weak
abet
al.,
2013).
The
majority
of
continuous
glucose
monitors
et
al.,
2013).
The
majority
of
continuous
glucose
monitors
sorption
characteristics
of
glucose
in
a
spectral
region
dom(CGMs)
require
subcutaneous
insertion
of
a
filament
for
The
main
difficulty
in
NIR
detection
is
due
to
the
weak
ab(CGMs)
require
subcutaneous
insertion
of
a
filament
for
The
main
difficulty
in
NIR
detection
is
due
to
the
weak
absorption
characteristics
of
glucose
in
a
spectral
region
domelectro-chemical
measurement
of
local
interstitial
glucose
et
al.,
2013).
The
majority
of
continuous
glucose
monitors
sorption
characteristics
of
glucose
in
a
spectral
region
dom(CGMs)
require
subcutaneous
insertion
of
a
filament
for
(CGMs)
require
subcutaneous
insertion
of
a
filament
for
inated
by
absorption
from
water,
haemoglobin
and
lipids
The
main
difficulty
in
NIR
detection
is
due
to
the
weak
abelectro-chemical measurement
measurement of
of local
local interstitial
interstitial glucose
glucose sorption
characteristics
of
glucose
in
a
spectral
region
domelectro-chemical
sorption
characteristics
of
glucose
in
a
spectral
region
dominated by
by absorption
absorption from
from water,
water, haemoglobin
haemoglobin and
and lipids
lipids
concentrations
and
require
multiple
calibrations
each
day.
(CGMs)
require
subcutaneous
insertion
of
a
filament
for
inated
electro-chemical
measurement
of
local
interstitial
glucose
electro-chemical
measurement
of local
interstitialeach
glucose
(Amerov
et
al., 2004).
Conventional
spectroscopy
has
characteristics
of
glucose
in
a
spectral
region
domconcentrations and
and
require multiple
multiple
calibrations
each
day. sorption
inated
by
absorption
from
water,
haemoglobin
and
lipids
concentrations
require
calibrations
day.
inated
by
absorption
from
water,
haemoglobin
and
lipids
et al.,
2004).
Conventional spectroscopy
has
They
are complex
and
havemultiple
significant
drift
and dynamics
electro-chemical
measurement
of local
interstitial
glucose
(Amerov
al.,
2004).
spectroscopy
has
concentrations
and
require
calibrations
each
day. (Amerov
concentrations
and
require
calibrations
each
day.
been
investigated
the
wavelength
range and
of 500
nm
inated
by et
absorption
fromConventional
water,
haemoglobin
lipids
They are
are complex
complex
and
havemultiple
significant
drift
and dynamics
dynamics
(Amerov
et
al.,
2004).
Conventional
spectroscopy
has
They
and
have
significant
drift
and
(Amerov
et
al., through
2004).
Conventional
spectroscopy
has
been
investigated
through
the
wavelength
range
of
500
nm
(Zhou
et
al.,
2018)
concentrations
and
require
multiple
calibrations
each
day.
been
investigated
through
the
wavelength
range
of
500
nm
They
are
complex
and
have
significant
drift
and
dynamics
They
complex
to
2000
nm
with
three
main
bands
of
interest:
2000
nmet
al.,
2004).
Conventional
spectroscopy
has
(Zhouare
et al.,
al.,
2018)and have significant drift and dynamics (Amerov
been
investigated
through
the
wavelength
range
of
500
nm
(Zhou
et
2018)
been
investigated
through
the
wavelength
range
of
500
nm
to
2000
nm
with
three
main
bands
of
interest:
2000
nmThey
are
complex
and
have
significant
drift
and
dynamics
to
2000
nm
with
three
main
bands
of
interest:
2000
nm(Zhou
et
al.,
2018)
(Zhou
et
al.,
2018)
2500
nm
(combination
overtone),
1400
nm-2000
nm(first
been
investigated
through
the
wavelength
range
of
500
nm
Futures in healthcare rely as much on enabling patient to
2000
nm
with
three
main
bands
of
interest:
2000
nmto
2000
nm
with
three
main
bands
of
interest:
2000
nm2500
nm
(combination
overtone),
1400
nm-2000
nm(first
Futures
in
healthcare
rely
as
much
on
enabling
patient
(Zhou
et
al.,
2018)
2500
nm
(combination
overtone),
1400
nm-2000
nm(first
Futures
in
healthcare
much
on
enabling
patient
overtone)
and
750three
nm-1400
nmbands
(second
overtone)
(Yadav
2000
nm
with
main
of interest:
2000
nmself-management
as onrely
newas
innovative
technologies.
A to
2500
nm
(combination
overtone),
1400
nm-2000
nm(first
Futures
in
healthcare
rely
as
much
on
enabling
patient
2500
nm
(combination
overtone),
1400
nm-2000
nm(first
overtone)
and
750
nm-1400
nm
(second
overtone)
(Yadav
Futures
in
healthcare
rely
as
much
on
enabling
patient
self-management as
as on
on new
new innovative
innovative technologies.
technologies. A
A overtone)
and
750
nm-1400
(second
(Yadav
self-management
et
al.,nm
2015).
andnm
Yamakoshi
(2006)
developed
(combination
overtone),
1400overtone)
nm-2000
nm(first
non-invasive
glucose
method
facilitate
more
overtone)
and
750
nm
(second
overtone)
(Yadav
Futures
in healthcare
much would
on enabling
patient
self-management
as sensing
onrely
newas
innovative
technologies.
A 2500
overtone)
andYamakoshi
750 nm-1400
nm-1400
nm
(second
overtone)
(Yadav
et
al.,
2015).
Yamakoshi
and
Yamakoshi
(2006)
developed
self-management
as
on
new
innovative
technologies.
A
non-invasive
glucose
sensing
method
would
facilitate
more
et
al.,
2015).
Yamakoshi
and
Yamakoshi
(2006)
developed
non-invasive
glucose
sensing
method
would
facilitate
more
aetpulse
glucometer
using
spectrometry
values
from
900 nm
overtone)
and
750
nm-1400
nm
(second
overtone)
(Yadav
patient
monitoring,
reducing
complications
and
improving
et
al.,
2015).
Yamakoshi
and
Yamakoshi
(2006)
developed
self-management
as
on
new
innovative
technologies.
A
non-invasive
glucose
sensing
method
would
facilitate
more
al., 2015).
Yamakoshi
and
Yamakoshi
(2006)
developed
a
pulse
glucometer
using
spectrometry
values
from
900
nm
non-invasive
glucose
sensing
method
would
facilitate
more
patient
monitoring,
reducing
complications
and
improving
a
pulse
glucometer
using
spectrometry
values
from
900
nm
patient
monitoring,
reducing
complications
and
improving
to
17002015).
nm. Through
PLS
methods,
they
gained
accurate
et
al.,
Yamakoshi
and
Yamakoshi
(2006)
developed
quality
of
life.
a
pulse
glucometer
using
spectrometry
values
from
900
non-invasive
glucose
sensing
method
would
facilitate
more
patient
monitoring,
reducing
complications
and
improving
a
pulse
glucometer
using
spectrometry
values
from
900 nm
nm
to
1700
nm.
Through
PLS
methods,
they
gained
accurate
patient
monitoring,
reducing
complications
and
improving
quality
of
life.
to
1700
nm.
Through
PLS
methods,
they
gained
accurate
quality
of
results
from
glucoseusing
measurent,
but required
further
work
pulse
glucometer
spectrometry
values
fromaccurate
900
nm
1700
nm.
Through
PLS
methods,
they
gained
patient
reducing complications and improving ato
quality monitoring,
of life.
life.
to
1700
nm.
Through
PLS
methods,
they
gained
accurate
results
from
glucose
measurent,
but
required
further
work
quality
of
life.
results
from
glucose
measurent,
but
required
further
work
on
their
instrumentation
to
take
the
sensor
out
of
the
to
1700
nm.
Through
PLS
methods,
they
gained
accurate
results
from
glucose
measurent,
but
required
further
work
The
project
has
received
funding
from
EU
H2020
R&I
programme
quality
of
life.
results
from
glucose measurent,
butthe
required
further
work
on
their
instrumentation
to
take
sensor
out
of
the
The
funding
from
H2020
programme
on
their
instrumentation
to
take
the
sensor
out
of
the
The project
project has
has received
received
funding
from EU
EU
H2020 R&I
R&I
programme
laboratory
(Yamakoshi
et al.,
2009).
results
from
glucose
measurent,
but
required
further
work
(MSCA-RISE-2019
call) under
agreement
#872488
–DCPM
on
their
instrumentation
to
take
the
sensor
out
of
the
The
project
has
received
funding
from
EU
H2020
R&I
programme
on
their
instrumentation
to
take
the
sensor
out
of
the
laboratory
(Yamakoshi
et
al.,
2009).
The
project
has
received
funding
from
EU
H2020
R&I
programme
(MSCA-RISE-2019
call)
under
agreement
#872488
–DCPM
laboratory
(Yamakoshi
et
al.,
2009).
(MSCA-RISE-2019 call) under agreement #872488 –DCPM
on
their
instrumentation
to
take
the
sensor
out
of
the
laboratory
(MSCA-RISE-2019
call)
agreement
#872488
–DCPM
The project has received
funding
from EU
H2020 R&I
programme
laboratory (Yamakoshi
(Yamakoshi et
et al.,
al., 2009).
2009).
(MSCA-RISE-2019
call) under
under
agreement
#872488
–DCPM
laboratory
(Yamakoshi
et
al.,
2009).
(MSCA-RISE-2019
call)
under
agreement
#872488
–DCPM
2405-8963 Copyright © 2020 The Authors. This is an open access article under the CC BY-NC-ND license.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2020.12.388
Jake D. Campbell et al. / IFAC PapersOnLine 53-2 (2020) 15970–15975
15971
Where:
ID :
IE :
:
c:
d:
Detected light intensity
Emitted light intensity
Absorption coefficients
Concentration of molecules
Optical path length
Blood consists of many constituents at varying concentrations and absorption characteristics, meaning absorption
in Beer Lambert’s law is the sum of all components:
A(λ) =
Fig. 1. Equivalent circuit diagram for the sensing diode
within LED-LED detection.
1.1 LED-LED Light Detection
i (λ)ci
(2)
A(λ) represents the total absorbance through blood at a
particular wavelength (λ). By using different absorption
characteristics at varying wavelengths, the concentrations
of each constituent can be calculated (Zhang et al., 2016).
1.3 Detection Wavelength
Conventional methods of light detection use LEDs to emit
light and photodiodes to detect returning light. The current is then amplified and converted to a voltage, and
sampled by a microprocessor through an analog to digital
converter (ADC) (Kennedy, 2015). The sensing methodology used here has a forward-biased LED to emit, and
a reverse-biased LED detect light. Rather than taking
instantaneous measurements with an ADC, the reversebiased LED uses its inherant capacitance to integrate
the incoming light intensity over the sensing period (Stojanovic and Karadaglic, 2013; Lau et al., 2006). Thus,
the output from the sensor represents the inverse of the
detected light intensity.
Lau et al. (2006) demonstrates that the lower photocurrent
producing efficiency of LEDs is an advantage as LED
detection provides approximately an order of magnitude
greater response to light intensity changes compared to
photodiodes. The spectral detection region of the LEDs
is also of slightly lower wavelength of the emission spectrum, creating a narrower detection range(O’Toole and
Diamond, 2008).
Recently, the LED-LED sensing method has been employed with success in absorbance measurements for determining the concentration of solutions due to their spectrally sensitive absorption spectrum and their ease of implementation (Shin et al., 2016). Figure 1 shows the basic
circuit diagram used in this study. Each wavelength to be
detected had a pair of LEDs, one emitting and the other
detecting the reflected light.
1.2 Beer-Lambert’s Law
Beer Lambert’s law describes the exponential attenuation
of light travelling through a substance. The attenuation of
light across a path length (d) is from the concentration
of the solute (c) and how much light it absorbs ()
(Mendelson, 1992), yielding:
ID = IE ecd
(1)
To take advantage of the light absorption bands of glucose
at 750 nm-1400 nm, LEDs of wavelength 660 nm, 850 nm
and 940 nm were used. Haxha et al used a 940 nm LED to
monitor changes in glucose level (Haxha and Jhoja, 2016).
In this region, haemoglobin is the dominant absorber, so
any changes in blood oxygen level will skew the results
(Bashkatov et al., 2005). Thus, an 850 nm LED was chosen
because the isosbestic point of oxyhaemoglobin and deoxyhaemoglobin is at 800 nm, within the lower detection
spectrum (Chan et al., 2013). At the isosbestic point,
the absorption by haemoglobin is consistent, regardless
of the oxygen saturation. This wavelength was used to
account for the change in absorption by haemoglobin when
assessing the change in glucose at the same wavelength.
In the second absorption band, glucose has a dominant
absorption peak at 1550 nm as well as at 1650 nm. Figure
2 shows the approximate absorption by glucose, water,
lipids and albumin from 500 nm to 2000 nm. At 1450 nm
absorption is dominated by water, meaning variations in
concentrations of glucose and other blood constituents
cause a minimal change in absorption. At 1550 nm and
1650 nm, glucose is the dominant absorber, so a rise
in glucose level is expected to increase absorption at
these wavelengths. LEDs at 1550 nm and 1650 nm were
chosen as the absorption of lipids vary greatly, potentially
allowing extraction of the lipid concentration, which is also
clinically useful.
When calculating the change in absorption, the displacement of water by glucose must be included (Amerov et al.,
2004). The relatively strong absorption properties of water
in the NIR spectrum causes significant changes in total
absorbance when glucose is added. Upon dissolution of glucose, water molecules are displaced from the optical path
length. At wavelengths where water dominates absorption
(1450 nm), the total absorbance is expected to drop with
increasing glucose. At 1550 nm and 1650 nm, glucose is the
dominant absorber of light, so even with displacement of
water, the absorbance is expected to increase with glucose
concentration.
Jake D. Campbell et al. / IFAC PapersOnLine 53-2 (2020) 15970–15975
15972
glucose concentration. 1650 nm is also expected to increase
with glucose, but not as much as 1550 nm.
2.3 Water Testing
Fig. 2. Absorption spectrum of the major light absorbers
in the NIR range, not to scale. Derived from (Yadav
et al., 2015; Amerov et al., 2004)
The first test conducted was with glucose added to deionised water. To minimize the effect of outside variables
during testing, 40 litres of water was kept at body temperature inside an insulated container. Glucose was added
in increments of 3.60 g to increase the concentration by
0.5 mmol l−1 at each step. The concentration was increased
from 0 mmol l−1 up to 20 mmol l−1 . The target physiological range of blood glucose in a person is between
4 mmol l−1 and 8 mmol l−1 . While blood glucose levels may
climb above 20 mmol l−1 in hyperglycemic patients. Thus,
measuring a range from 0-20 mmol l−1 provides a physiologically relevant range to show a response to glucose.
A sensor was placed in a resealable bag 200 mm below
the surface of the water, minimising the effect of light
reflecting off the surface of the water. Upon each addition
of glucose, the water was stirred for 20 seconds and allowed
to come to a rest. Data were recorded over 30 seconds and
the average over this time was used for calculations. The
test was conducted in a dark room with no light source
other than the sensor.
Table 1. Experimental procedure
Test
Water
Blood 1
Blood 2
Fig. 3. Diagram of the developed glucose sensor.
Volume
l
40.20
1.10
0.94
Gstart
mmol l−1
0
4.9
4.3
Gstep
mmol l−1
0.50
0.50
0.50
Gend
mmol l−1
20.0
16.8
20.6
2.4 Pig Blood Testing
2. METHODS
2.1 Sensor Design
To measure absorption of light at the required wavelengths, a sensor with 12 LEDs was built. Each of the
six wavelengths have an emitter and detector LED. The
sensor was designed in a rosette shape to ensure the
path of each wavelength travelled through the same space.
Figure 3 shows the sensor. The 660 nm and 940 nm
LEDs are contained in the same package to minimise
the footprint. An ATMEGA32U2 running at 16 MHz
was used to power the emitting LED and to time the
discharge of the reverse-biased LED. The LEDs used
were XZM2MRTNI55W-8 (660,940 nm), 15412085A3060
(850 nm), MTSM5014-843-IR (1450 nm), MTSM5015-843IR (1550 nm) and MTSM5016-843-IR (1650 nm).
2.2 Beer-Lambert’s Law
While the intention of the sensor is to detect glucose in
pulsatile human blood, testing was first done in vitro
to verify a response to changing glucose. The 1450 nm
and 1550 nm wavelengths would be the focus as glucose
has no absorption at 1450 nm and dominant absorption
at 1550 nm. Compared to 1450 nm, the reflected light
intensity at 1550 nm should increase with the addition of
glucose. Thus, the ratio of 1450 nm and 1550 nm should
represent a Beer-Lambert response curve with increasing
The next step in testing the sensor’s response to glucose
was to add glucose to blood under the same conditions as
the test in water. Two litres of pigs blood were used for
testing. To reduce the effects of clotting, 1.8 g of EDTA for
1 l of pig blood was added as an anticoagulant. The fresh
pig blood was kept in an insulated container between 04 ◦C during transport and was raised to room temperature
for testing. An ACCU-CHEK Performa Glucometer was
used to measure the blood glucose content throughout
the test for sensor calibration. Table 1 gives the test
conditions. The sensor was mounted underneath the flask,
in contact with the glass.
The first test was done within 3 hours of obtaining the
fresh blood. Figure 4 shows the setup for both tests. The
first test had 1.1 l of blood and an initial blood glucose level
of 4.9 mmol l−1 . For each measurement, 0.09 g of glucose
powder was stirred into the blood for 30 seconds and
allowed to settle. One minute of data were recorded and
the average reading taken. The second test was conducted
two days after the first to further reduce the blood glucose
content. The blood stayed between 2-8 ◦C during waiting
period and was brought to room temperature for testing.
3. RESULTS
As the sensor integrates reflected light intensity during a
measurement, an increase in absorption is directly proportional to an increase in discharge time of the receiving
Jake D. Campbell et al. / IFAC PapersOnLine 53-2 (2020) 15970–15975
15973
Fig. 4. Experimental setup for pig blood testing.
LED. This inverse relation is in contrast to conventional
photodiode measurement, where an increase in absorption
reduces the detected voltage. Figures 5 and 6 show the
time taken for each LED to discharge. A higher discharge
time results from higher absorption by light travelling
through the solution, and therefore higher concentration.
The value α is used to represent the proportionality between discharge time and light intensity.
4. DISCUSSION
4.1 Water Testing
Testing showed the predicted relationship between glucose
levels and light absorption. An increase in absorption
at 940 nm, 1550 nm and 1650 nm was recorded, while at
660 nm and 850 nm, there was a decrease in absorption.
Figure 5 plots the discharge time of each wavelength at
increasing glucose levels. At 1450 nm and 850 nm, the data
were very noisy with slight upwards and downwards trends
respectively over the range of glucose. At 660 nm there was
a slight decrease and at 940 nm, an increase. This increase
is due to the slight absorption peak at 940 nm.
4.2 Pig Blood Testing
Testing in blood gave different results to just water. Figure
6 shows that at 1450 nm and 1650 nm there was decreased
absorption with increasing glucose. This difference is represented in the change in absorption at 850 nm, as this
wavelength represents the total haemoglobin count. Absorption at 1450 nm, 660 nm, 940 nm and 1650 nm were
consistent between tests.
Table 2. Relative Reflection Intensity (α/ID )
readings for an empty flask
λ(nm)
α/ID
660
0.72e5
860
0.40e5
940
0.21e5
1450
2.63e5
1550
2.17e5
1650
0.54e5
Fig. 5. Change in reflected light intensity in water with an
increase in glucose concentration.
4.3 Wavelength Ratios
The three main wavelength relationships of interest were
between 1450 nm and 1550 nm, 1550 nm and 1560 nm and
940 nm and 850 nm. At 1450 nm and 1550 nm is the greatest difference between glucose absorbance coefficients and
at 940 nm and 850 nm is the isobestic point of haemoglobin
and a slight absorbance by glucose. At 1650 nm and
1550 nm, there is a difference in absorption by albumin and
lipids as well as glucose. In Figure 6, the noise variation in
readings are present in all wavelengths. Taking the ratio of
each wavelength pair removes the constant noise component during each measurement such as ambient lighting.
Figure 7 shows the relationship to glucose. For better
comparison, each response is normalised and centred at
zero. The absorbance at 1650 nm has the largest decrease,
so the ratio of 1550 nm to 1650 nm was also calculated
(R1550/1650 ). This ratio gives the greatest response to the
normalised data, with R1550/1450 second greatest. R940/850
gives the lowest response to increasing glucose.
4.4 Beer Lambert’s Law
As the sensor was placed in a resealable plastic bag, there
was a slight difference in absorption compared to glass.
Testing was done with an empty flask and the baseline
recording is shown in Table 2. All wavelengths except
for 1450 nm and 660 nm had lower responses without
blood. As the testing was done under ambient lighting,
the increase in 660 nm and 1450 nm is due to room lighting
and sunlight.
The response of each ratio in Figure 7 follows the inverted
Beer Lambert’s law. At low glucose concentrations of
glucose, reflected light intensity exponentially decreases.
As the only factor in Equation 1 that changes is the
concentration, the observed changes in light absorption
can be attributed solely to the addition of glucose. At
wavelengths with no glucose absorption peaks (660 nm,
15974
Jake D. Campbell et al. / IFAC PapersOnLine 53-2 (2020) 15970–15975
of water molecules. At 1650 nm and 1550 nm, the signal
gave the greatest rise in absorption. The response at
1450 nm was expected to reduce due to the displacement
of water. However, it remained noisy, slightly increasing
with glucose level. This high variation is likely due to
the minimal effect of water displacement due to glucose.
The high absorption of light by water at 1450 nm means
minimal light entering the solution returned to the sensor,
so the majority of the light returned will be reflected off
the surface of the water.
The intensity of reflected light was much higher for 660 nm
and 940 nm as water and glucose absorb minimally at
these wavelengths. Absorption of 660 nm and 940 nm light
in the pig blood were much higher due to the presence
of haemoglobin. It is also expected the glass of the flask
reduced the amount of reflected light reaching the sensor
as the LED discharge times for an empty flask were very
similar to those with blood present. The exact absorption
by the glass is not further explored as the purpose of this
paper is to measure a change in absorption with increasing
glucose concentrations, so a constant effect from glass does
not affect the conclusions.
Fig. 6. Change in reflected light intensity in pigs blood
with an increase in glucose concentration.
Fig. 7. Normalised and centred absorbance ratios of
R1550/1650 , R1550/1450 and R940/850 in pigs blood over
an increasing glucose level.
850 nm, 1450 nm), the reflected light intensity has minimal
change, providing further validation.
The development of a non-invasive blood glucose monitor
has been the focus of intense interest for several decades.
Among possible approaches, NIR spectroscopy has shown
much promise. However, definitive results have remained
elusive. This paper investigated the LEDs as light detectors at selected spectral wavelengths, creating a discrete
spectrometer. The designed glucose sensor was tested in
water and pigs blood.
In water, the change in absorbance was solely due to the
increase in glucose concentration and the displacement
Testing in pig blood yielded different absorption responses
to testing in water. The presence of haemoglobin reduced
the intensity of reflected light in the first absorption band
and the presence of other lipids and albumin in the blood
gave an opposite response than that of water at 1650 nm.
At 1650 nm, glucose, water, albumin and lipids have similar absorption coefficients. As glucose displaces the other
molecules from the optical path, the corresponding increase in absorption by glucose does not account for the
reduced absorption by displacement. Thus, the absorption
in blood at 1650 nm decreases with increasing glucose
levels. In contrast, the increased absorption from glucose at
1550 nm is more than the reduction in absorption from displacement, so the overall absorption at 1550 nm increases
with increasing glucose concentration.
Ratios of absorbance at various wavelengths were calculated to test the response to glucose. Figure 7 shows that
the three ratios investigated all follow Beer Lambert’s law
for light absorption. The wavelength 1450 nm was chosen
for testing as water is the primary absorber. By comparing
the increasing absorbance at 1550 nm to the decreasing absorption at 1450 nm, the effect of glucose concentration can
be seen. Water has minimal absorption in the first absorption band, so the ratio between 940 nm and 850 nm shows
the relationship between glucose and haemoglobin. At the
isobestic point of haemoglobin (850 nm) the decrease in
absorption is due to the displacement of haemoglobin and
at 940 nm, the increase in absorption is from the increased
absorption of glucose. This behaviour can also be seen at
660 nm as absorbance decreases with increasing glucose.
The ratio of 1650 nm to 1550 nm gave the largest change
with increasing glucose in blood. Other than 1650 nm, all
other wavelength absorptions were as expected. Moving
forward, testing will be done in vivo to determine the
absorption response. The sensor will be designed for use
as a pulse glucometer, using the same principles as pulse
oximetry. In vivo testing will isolate the response in
the blood from the pulsatile signal, similar to work by
Yamakoshi et al. (2009).
Jake D. Campbell et al. / IFAC PapersOnLine 53-2 (2020) 15970–15975
5. CONCLUSION
A sensor was developed to measure the light absorption
by glucose in the NIR spectrum. Using discrete spectroscopy at 660 nm, 850 nm, 940 nm, 1450 nm, 1550 nm and
1650 nm, varied responses were found through testing in
both water and pig blood. The results obtained can be explained by Beer Lambert’s law as the only factor affecting
absorbance is the change in glucose concentration. The
glucose concentrations tested were within the expected
physiological range of blood glucose concentrations.
Using LED-LED detection methods, the sensor integrates
the reflected light intensity of the six wavelengths. The ratio of R940/850 increases with glucose due to the absorption
peak at 940 nm and minimal change at the isobestic point
of haemoglobin (850 nm). In the first overtone, R1550/1450
gives an increase in absorption because glucose is the dominant absorber at 1550 nm, while water is the only absorber
at 1450 nm. The final ratio calculated was R1550/1650 which
gave the greatest rise due to the displacement of other
absorbing molecules (lipids, albumin).
Future work on the sensor will involve an increase in output light intensity to increase the amount of light reaching
the sensor. Non-invasive testing on pulsatile human blood
will be completed and compared to the absorption in water
and pig blood.
REFERENCES
Amerov, A.K., Chen, J., and Arnold, M.A. (2004). Molar
Absorptivities of Glucose and other Biological Molecules
in Aqueous Solutions over the First Overtone and Combination Regions of the Near-Infrared Spectrum. Applied Spectroscopy, 58(10), 1195–1204.
Bashkatov, A.N., Genina, E.A., Kochubey, V.I., and
Tuchin, V.V. (2005). Optical properties of human skin,
subcutaneous and mucous tissues in the wavelength
range from 400 to 2000 nm. Journal of Physics D:
Applied Physics, 38(15), 2543–2555.
Chan, E.D., Chan, M.M., and Chan, M.M. (2013). Pulse
oximetry: Understanding its basic principles facilitates
appreciation of its limitations. Respiratory Medicine,
107(6), 789–799.
Chowdhury, K., Srivastava, A., Sharma, D.N., and
Sharma, D.S. (2013). Challenges & Countermeasures
in Optical Noninvasive Blood Glucose Detection. International Journal of Innovative Research in Science,
Engineering and Technology, 2(1), 6.
Delbeck, S., Vahlsing, T., Leonhardt, S., Steiner, G.,
and Heise, H.M. (2019). Non-invasive monitoring of
blood glucose using optical methods for skin spectroscopy—opportunities and recent advances. Analytical
and Bioanalytical Chemistry, 411(1), 63–77.
Haxha, S. and Jhoja, J. (2016). Optical Based Noninvasive
Glucose Monitoring Sensor Prototype. IEEE Photonics
Journal, 8(6), 1–11.
Kennedy, S.M. (2015). An introduction to pulse oximeters: Equations and theory. University of WisconsinMadison, 20.
Lau, K.T., Baldwin, S., O’Toole, M., Shepherd, R., Yerazunis, W.J., Izuo, S., Ueyama, S., and Diamond, D.
(2006). A low-cost optical sensing device based on
15975
paired emitter–detector light emitting diodes. Analytica
Chimica Acta, 557(1-2), 111–116.
Liu, R., Deng, B., Chen, W., and Xu, K. (2005). Next Step
of Non-invasive Glucose Monitor by NIR Technique from
the Well Controlled Measuring Condition and Results.
Optical and Quantum Electronics, 37(13-15), 1305–1317.
Mendelson, Y. (1992). Pulse oximetry: theory and applications for noninvasive monitoring. Clinical Chemistry,
38(9), 1601–1607.
Nybacka, L. (2016). FTIR spectroscopy of glucose.
O’Toole, M. and Diamond, D. (2008). Absorbance Based
Light Emitting Diode Optical Sensors and Sensing Devices. Sensors, 8(4), 2453–2479.
Poddar, R., Andrews, J.T., Shukla, P., and Sen, P. (2008).
Non Invasive Glucose Monitoring Techniques: A review
and current trends. 48.
Shin, D.Y., Kim, J.Y., and Eom, I.Y. (2016). Spectral
Responses of Light-Emitting Diodes as a Photodiode
and Their Applications in Optical Measurements: LEDs
as a Photodiode in Optical Measurements. Bulletin of
the Korean Chemical Society, 37(12), 2041–2046.
Smith, J.L. (2017). The Pursuit of Noninvasive Glucose.
The Pursuit of Noninvasive Glucose: Hunting the Deceitful Turkey. 5th edition.
Stojanovic, R. and Karadaglic, D. (2013). Design of an
Oximeter Based on LED-LED Configuration and FPGA
Technology. Sensors, 13(1), 574–586.
Wild, S., Roglic, G., Green, A., Sicree, R., and King, H.
(2004). Global Prevalence of Diabetes: Estimates for
the year 2000 and projections for 2030. Diabetes Care,
27(5), 1047–1053.
Yadav, J., Rani, A., Singh, V., and Murari, B.M. (2015).
Prospects and limitations of non-invasive blood glucose
monitoring using near-infrared spectroscopy. Biomedical
Signal Processing and Control, 18, 214–227.
Yamakoshi, K. and Yamakoshi, Y. (2006). Pulse glucometry: a new approach for noninvasive blood glucose measurement using instantaneous differential near-infrared
spectrophotometry. Journal of Biomedical Optics, 11(5),
054028.
Yamakoshi, Y., Ogawa, M., and Tamura, T. (2009).
Multivariate Regression and Classification Models for
Estimation of Blood Glucose Levels Using a New
Non-invasive Optical Measurement Technique Named
“Pulse-Glucometry. The Open Optics Journal, 3(1), 63–
69.
Zhang, Z., Cai, Z., Han, G., Sheng, W., and Liu, J.
(2016). Estimation of glucose absorption spectrum at its
optimum pathlength for every wavelength over a wide
range. Spectroscopy Letters, 49(9), 588–595.
Zhou, T., Dickson, J.L., Shaw, G.M., and Chase, J.G.
(2018). Continuous Glucose Monitoring Measures Can
Be Used for Glycemic Control in the ICU: An In-Silico
Study. Journal of Diabetes Science and Technology,
12(1), 7–19.