ArticlePDF Available

iGLU: An Intelligent Device for Accurate Noninvasive Blood Glucose-Level Monitoring in Smart Healthcare

Authors:

Abstract

In the case of diabetes, fingertip pricking for blood sample is inconvenient for glucose measurement. Invasive approaches like laboratory tests and one touch glucometers enhance the risk of blood related infections. To mitigate this important issue, this article introduces a novel Internet-of-Medical-Things (IoMT) enabled edge-device for precise, noninvasive blood glucose measurement. The device called "Intelligent Glucose Meter" (i.e., iGLU) is based on near-infrared (NIR) spectroscopy and a machine learning model of high accuracy. iGLU has been validated in a hospital and blood glucose values are stored in an IoMT platform for remote monitoring by endocrinologists.
1
iGLU: An Intelligent Device for Accurate
NonInvasive Blood Glucose-Level Monitoring
in Smart Healthcare
Prateek Jain Amit M. Joshi
Malaviya National Institute of Technology Malaviya National Institute of Technology
Saraju P. Mohanty
University of North Texas
Abstract—In case of diabetes, ﬁngertip pricking
for blood sample is inconvenient for glucose mea-
surement. Invasive approaches like laboratory test
and one touch glucometer enhance the risk of blood
related infections. To mitigate this important issue,
in the current paper, we propose a novel Internet-
of-Medical-Things (IoMT) enabled edge-device for
precise, non-invasive blood glucose measurement. The
novel device called “Intelligent Glucose Meter” (i.e.
iGLU) is based on near-infrared (NIR) spectroscopy
and machine learning (ML) model of high accuracy.
iGLU has been validated in a hospital and blood
glucose values are stored IoMT platform for remote
monitoring by endocrinologist.
I. INT ROD UC TI ON
SMart healthcare system comprises of ambient
intelligence, quality of service and also offers
continuous support of the critical diseases moni-
toring [1], [2]. This system is most demandable for
remote monitoring of diabetic patients with low cost
and rapid diagnosis [3]. Traditional blood glucose
measurement is unable to serve everyone’s need
at remote location. Despite having good diagnos-
tic centers for clinical test facility in urban area,
medical services are not approachable to everyone
at remote location [4], [5]. It is necessary to monitor
blood glucose of diabetic patients where diagnosis
facility are not easily available. The instant di-
agnose of blood glucose and frequent monitoring
are the recent challenges in smart healthcare. The
process ﬂow of blood glucose diagnosis in smart
healthcare is depicted in Figure 1.
Non Invasive Measurement
Fasting and Postprandial
Blood Glucose Values (mg/dl)
Internet Connectivity
Internet Connectivity
Connecting to the
available Endocrinologist
Blood Glucose
Prescribed Medicine and
Precautions for Instant
Treatment
Cloud Storage
Nearest Hospital
For Treatment
Endocrinologist
Fig. 1: Blood glucose diagnosis in smart healthcare.
IoMT enabled handheld non-invasive glucose
measurement end-device has strong potential for
rapid monitoring as well as to facilitate the interact
with endocrinologist to the remote located diabetic
patients where diagnosis centers and hospitals are
not easily available. According to this environment,
patients measure their glucose without pricking
blood and directly store to the cloud where nearby
endocrinologist can monitor the glucose data of
each patient. The prescription would also be pro-
vided by endocrinologist to the remote located
patient for further treatment. The ubiquity of di-
abetic patients has become double from 2010 over
the world. The estimated diabetes dissemination
from 2009 is 290 million and is expected to affect
450 million people by 2030. Hence, it is essential
2
to develop the glucose measurement device for
rapid diagnosis of diabetes. People will be more
conscious for their glucose level with frequent mon-
itoring. Invasive method for glucose measurement
is not advisable in case of continuous monitoring.
Therefore, it required to design the non-invasive
device for clinical tests, which is beneﬁcial for
health care. In proposed work, optical detection is
involved. Blood glucose is predicted by machine
learning based computation model.
II. STATE-OF-ART IN BLOOD GLUCOSE-LEVEL
MEASUREMENT
Blood glucose measurement is possible using
invasive, minimally invasive and non-invasive meth-
ods (Figure 2). Frequent pricking, as needed in
invasive methods, for glucose measurement causes
trauma. Therefore, the semi-invasive approach has
the advantage of continuous glucose monitoring
without multiple times pricking. However, non-
invasive methods can completely eliminate prick-
ing which opens door to painless and continuous
glucose monitoring (CGM).
A. Invasive Methods
A low-invasive amperometric glucose monitoring
biosensor has been proposed using ﬁne pointed glu-
cose oxidase immobilized electrode which doesn’t
require more than 1mm in length to be inserted
in skin [6]. A fully implanted ﬁrst-generation pro-
totype sensor has been presented for long-term
monitoring of subcutaneous tissue glucose [7]. This
wearable sensor which is integrated as an implant
is based on a membrane containing immobilized
glucose oxidase and catalase coupled to oxygen
electrodes, and a telemetry system.
B. Minimally Invasive Methods
Implantable biosensors have been deployed for
continuous glucose monitoring [8]. Wearable min-
imally invasive microsystem has been explored for
glucose monitoring [9]. A microsystem has been
presented for glucose monitoring which consists
of microfabricated biosensor ﬂip-chip bonded to a
transponder chip [10]. A method has been discussed
to reduce the frequency of calibration of minimally
invasive Dexcom sensor [11]. An artiﬁcial pancreas
has been represented along with glucose sensor
to control diabetes [12]. But, approaches based
semi-invasive devices have not been tried for real
time application. These wearable microsystems are
neither painless nor cost effective solutions.
C. Non-invasive Methods
To make the painless system, photoacoustic spec-
troscopy has been introduced for non-invasive glu-
cose measurement [13]. However, utilization of
LASER makes the setup costly and bulky. An
enzyme sensor has been explored for glucose mea-
surement in saliva [14]. Glucose detection is possi-
ble using Intensity Modulated Photocurrent Spec-
troscopy (IMPS) spectroscopy that connects the
electrodes to the skin which is affected by sweat
[15]. High precision level is not possible through
these methods as sweat and saliva properties vary
for individuals. The blood glucose measurement has
also been explored using Raman spectroscopy in
laboratory [16]. The experimental setup for Raman
spectroscopy required a large area and will not be
portable. Glucose measurement has also been done
from anterior chamber of the eye which limits it’s
usage of continuous monitoring [17]. Blood glucose
can been estimated using photoplethysmography
(PPG) signal [18], [19].
PPG signal analysis is not based on principle
of glucose molecule detection. Therefore, speciﬁc
wavelengths are not required for glucose estimation.
Hence, iGLU is more precise compared to PPG sig-
nal analysis based system for glucose measurement.
In this way, long NIR wave for optical detection has
been considered for glucose measurement which
is not comparatively precise glucose measurement
system as long wave has shallow penetration [20].
Therefore, small NIR wave is preferred for glucose
detection (Figure 3).
Prior works related to glucose monitoring have
been discussed which represent wearable and non-
wearable approaches. Raman spectroscopy, photoa-
coustic spectroscopy and invasive approach based
systems are not wearable. Minimally invasive de-
vices which have been discussed, are implantable.
Other approaches based non-invasive device are
wearable. Here, iGLU is non-invasive, optical de-
tection based wearable device for continuous glu-
cose monitoring with IoMT framework.
3
Blood Glucose
Monitoring System
Invasive Minimally Invasive Non Invasive
In-vivo In-vitro
Photoacoustic
Raman
Spectroscopy Polarimetry Photoplethysmography
(PPG) Signal
Long Wave Near-
Infrared (NIR)
Detection
Implanted Micro
System
Glucose Control
System
Electrochemical
Biosensor
Implanted Sensor Amperometric
Near-Infrared (NIR)
Spectroscopy -
iGLU
Bio-impedance
Fig. 2: An overview of various blood glucose-level measurement devices or systems.
LED Detector
Logged voltage values
after absorption and
reflectance of light from
glucose molecule
Specific
Wavelengths
needed for glucose
molecule detection
NIR Spectroscopy
LED Detector
Logged signal for pulse
analysis and features
extraction
Specific
Wavelengths are
not required
PPG Signal Analysis
Fig. 3: PPG signal versus NIRS based glucose
estimation
D. Consumer Electronics for Glucose-Level Moni-
toring
Several devices have been developed for non-
invasive blood glucose measurement. Some prod-
ucts such as glucotrack, glutrac, glucowise, DiaMon
Tech and device from CNOGA medical are not
commercialized. Glutrac is multi-parameter health
test device with smart healthcare. However, they
have limitations in terms of precise measurement.
The cost of product is also high which varies in
the range of 300-400 USD. Therefore, the cost
effective solution for non-invasive blood glucose
measurement is needed.
III. NOVE L DEV IC E IGLU TO ADVANC E TH E
STATE-OF-ART IN WEARABLE FOR
CONTINUOUS BLO OD GL UC OS E MONITORING
Non-invasive measurement reduces the possibil-
ity of blood-related diseases. However, this ap-
proach have some limitations such as large set-
up, measuring object (ratina) and skin properties
(including dielectric constant and sweat level).
Therefore, portable non-invasive precise glucose
measurement device for continuous monitoring is
needed. An initial example is a non-invasive glu-
cose measurement using NIR spectroscopy and
Huber’s regression model [21]. There are several
glucose monitoring systems which neither provide
precise measurement nor cost effective solution.
These systems are not enabled for smart healthcare.
The following questions are resolved in iGLU for
the advancement of smart healthcare: (1) How can
we have a device that automatically performs all
the tasks of blood-glucose monitoring at the user
end without internet connectivity and stores the
data in cloud for future use by the patient and
healthcare providers? (2) Can we have a device that
can perform automatically to avoid hassle and risky
ﬁnger pricking all the time monitoring is needed?
telligent Glucose Meter” (i.e. iGLU) for noninva-
sive, precise, painless, low-cost continuous glucose
monitoring at the user-end and stores the data
on cloud in an IoMT framework. A non-invasive
device has been proposed with precise and low
cost solution. The proposed device is also integrated
with IoMT where the data is accessible to caretaker
for point of care. The device will be portable after
packaging to use everywhere. The device is fast
operated and easy to use for smart healthcare. The
ﬂow of proposed iGLU is represented in Figure 4.
of-art in smart healthcare include the following:
1) A novel accurate non-invasive glucometer
(iGLU) by judiciously using short NIR waves
4
Post-
processing
Computation
Model
Estimated
Blood
Glucose
Value
Cloud
Storage
Analog-to-
Digital
Converter
Infrared
Emitters
(940nm,
1300nm)
Transmitted
Wave
Attenuated
Wave
Vibrations
(Stretching, Wagging,
Bending)
Infrared
Detector
Infrared
Detector
Fig. 4: A conceptual overview of iGLU.
with absorption and reﬂectance of light using
speciﬁc wavelengths (940 and 1300 nm) has
been introduced. The wavelengths are judi-
ciously selected after experimental analysis
which has been done in material research
center MNIT, Jaipur (India).
2) A novel accurate machine learning based
method for glucose sensor calibration has
been presented with calibrated and validated
healthy, prediabetic and diabetic samples.
3) The proposed non-invasive blood glucose
measurement device has been integrated in
IoMT framework for data (blood glucose val-
ues) storage, patient monitoring and treatment
on proper time with cloud access by both the
patient and doctor.
IV. PROPOSED NON-INVASIVE BLOO D
GLU CO SE MEASUREMENT DEVICE (IGLU)
The proposed device based on NIR spectroscopy
with two short wavelengths is designed and im-
plemented using three channels. Each channel is
embedded with emitter and detector of speciﬁc
wavelength for optical detection. The data is col-
lected and serially processed by 16 bit ADC with
sampling rate of 128 samples per second. The
logged data is calibrated and validated thorough
existing regression techniques to analyse the op-
timized model. The ﬂow of data acquisition for
proposed iGLU is presented in Figure 5.
A. The Approach for Glucose Molecule Detection
Glucose molecule vibrates according to its
atomic structure at speciﬁc wavelengths. It is an-
Start
Power Supply ON for Emitters and Detectors of
Connect the Arduino Uno Board to PC through
USB Port
Upload the Program of Data Processing through
ADS 1115 using Arduino 1.8.5 Software
Check
Mode
Placing of objects (Fingers or Ear Lobes) in
Clips for Glucose Measurement
Run the Tera Term Application on PC for Data
Logging
Re-connect the
Arduino Uno Board to
PC through USB Port
Sampling and Averaging of Collected Data from
Three Channels for Calibration and Validation of
Regression Models
Stop
Fig. 5: Process ﬂow data acquisition for iGLU.
alyzed that absorbance and reﬂectance are sharper
and stronger in short wave NIR region [22]. The
absorption peak of glucose spectra at 1314 nm has
been analyzed [23]. The non-invasive blood glucose
measurement using 850, 950 and 1300 nm has
been implemented [15]. The 940 nm wavelength for
detection of glucose molecule has been identiﬁed
[24]. NIR spectra of sucrose, glucose and fructose
are elaborated with CH2, CH and OH stretching at
930, 960 and 984 nm, respectively [25].
B. Proposed Module for Data Acquisition
Proposed iGLU uses NIR spectroscopy to im-
prove the accuracy. A 2-Layer PCB has been devel-
oped to embed infra-red emitters (MTE1300W -for
1300 nm, TSAL6200 -for 940 nm, TCRT1000 -for
940 nm) and detectors (MTPD1364D -for 1300 nm,
3004MID -for 940 nm, TCRT1000 -for 940 nm).
The hardware is designed for data acquisition from
emitters, detectors and ADC with 5V DC supply.
According to the emitters and detectors, compatible
passive components have been chosen. Architecture
of glucose sensing is shown in Figure 6. Detectors
with daylight blocking ﬁlters are packaged and not
affected by sweat. ADS 1115 with 860 SPS, 16
bit, I2Ccompatible and single ended is controlled
through microcontroller ATmega328P and used to
convert the data (in Volts) from all channel in
decimal form. The noise power and signal-to-noise
5
ratio (SNR) have also been found 0.08 and 25.2 dB,
respectively, which show the minimum noise level.
Fig. 6: Circuit topology of proposed device (iGLU).
C. A Speciﬁc Prototype of the iGLU
Absorption and reﬂectance at 940 nm and ab-
sorption at 1300 nm are implemented for detection
of the glucose molecules. The detector’s voltage
depends on received light intensity. After placing
the ﬁngertip between emitter and detector, the
voltage values are logged. Change in light inten-
sity depends upon glucose molecule concentration.
During experiments, blood glucose is measured
through the invasive device standard diagnostics
(SD) check glucometer for validation of the non-
invasive results. The reading is taken as refer-
enced blood glucose values (mg/dl). During the
process, optical responses through detectors have
been collected from 3 channels simultaneously.
During measurement, the channels data is collected
in the form of voltages from 3 detectors. These
collected voltages correspond to referenced blood
glucose concentration. These voltage values are
converted into decimal form using 4-channel ADS
1115 (Texas Instruments) ADC [26]. Coherent av-
eraging has been done after collection of responses.
Speciﬁcation of a iGLU prototype are presented in
Table I.
The prototype view of proposed iGLU is shown
in Figure 7. The data is collected after ﬁxing three
in such a way that emitters and detectors are placed
beneath the surfaces of pads. Because of this, there
will be enough free spaces between the object
TABLE I: Speciﬁcation of iGLU prototype
Channel 1 Channel 2 Channel 3
Measured (Ideal)
Arduino Supply 4.95V 4.96V 4.95V
(5V) (5V) (5V)
Forward Voltage 0.96V 1.42V 1.40V
(Emitter) (1.1V) (1.5V) (1.5V)
Forward Current 53.4mA 52.8mA 52.9mA
(Emitter) (100mA) (60mA) (60mA)
Reverse Voltage 4.25V 4.16V 4.25V
(Detector) (5V) (5V) (5V)
Output Current 0.45mA 0.5mA 0.52mA
(Detector) (1mA) (1mA) (1mA)
Measurement range 3.2-4.68V 0.8-4.7V 0.5-4.7V
Speciﬁc
Wavelength
1300nm 940nm 940nm
Spectroscopy Absorption Absorption Reﬂectance
and sensors (emitters and detectors). Hence, the
probability of a faulty measurement is minimized.
Fig. 7: Prototype view of proposed device (iGLU).
V. PROPOSED MACHINE-LEARNING (ML)
BASED METHOD F OR I GL U CALIBRATION
Regression models are calibrated to analyze the
optimized computation model for glucose estima-
tion. The detector’s output from three channels
are logged as input vectors for glucose prediction.
The collected data from the samples is required to
convert in the form of estimated glucose values.
It is necessary to develop optimal model for pre-
cise measurement and hence analysis of MAD,
mARD,AvgE and RM SE are performed to
ensure accuracy. The estimated and reference blood
glucose concentration are calculated as BGE st and
BGRef , respectively. A total of 97 samples are
taken for device calibration which include predi-
abetic, diabetic and healthy samples. The baseline
6
characteristics of samples for calibration is repre-
sented in Table II. The proposed process ﬂow of
calibration and validation is shown in Figure 8.
TABLE II: Baseline characteristics of samples
Samples Basic Calibration Validation
Characteristics and Testing
Age (Years) Gender Wise Samples
Male:- 22-77 Male:- 53 Male:- 64
Female:- 17-75 Female:- 44 Female:- 29
Age (Years) Prediabetic
Male:- 22-65 Male:- 18 Male:- 11
Female:- 26-75 Female:- 13 Female:- 10
Age (Years) Diabetic
Male:- 30-68 Male:- 16 Male:- 17
Female:- 30-73 Female:- 14 Female:- 11
Age (Years) Healthy
Male:- 22-65 Male:- 19 Male:- 36
Female:- 17-70 Female:- 17 Female:- 08
Calibration through
Machine Learning
Technique
Error Analysis from
Predicted Glucose
Concentration
Referenced Blood
Concentration (mg/dl)
Prediction from Calibrated
Machine Learning Model for
Validation
Data Set
(Input Voltage values with
Referenced Blood Glucose)
Proposed Machine
Learning based
Regression Model
Input Vectors
(Voltage values from
3 Channels)
MARD (%)
AvgE (%)
RMSE (mg/dl)
Referenced Blood
Concentration (mg/dl)
Fig. 8: The process ﬂow of calibration and valida-
tion of proposed device (iGLU).
Deep Neural Network (DNN) based machine
learning model has been applied for precise blood
glucose prediction (Figure 9) [2]. Proposed DNN
uses sigmoid activation functions and has been
trained through Levenberg-Marquardt backpropaga-
tion algorithm [15]. In proposed model, 10 hidden
neurons and 10 hidden layers are analyzed to esti-
mate the precise blood glucose values. This model
has been used to analyze the non-linear statistical
data which is utilized to calibrate and validate the
model for precise measurement. Here, the voltage
values from three channels are used as inputs of
proposed DNN model. The predicted blood glucose
values are formed through the modeling of three
channels voltage values. Weights of the voltage
values correlate predicted glucose values to the
channels data. The overall accuracy is improved
using 10 hidden layers.
10 Hidden Layers
1
2
3
10
1
2
3
ww
ww
Ʃ
w’
w’
w’
w’
Channel 1
Channel 2
Channel 3
Voltage
Weight of the Voltage Weight of the
Glucose Value
Neural Network for Computation of
Blood Glucose Output Layer
Prediction of Blood Glucose
bias
Input
Output
Sigmoid
Activation
Function
Predicted
Blood
Glucose
(mg/dl)
1
2
3
10
1 10
Fig. 9: The Deep Neural Network (DNN) for pro-
posed work
The Pearson’s correlation coefﬁcient (R) is 0.953.
The error analysis of calibrated machine learning
models is represented in Table III.
TABLE III: Analysis of calibration and validation
of proposed combination and ML model (DNN).
(%) (%) (mg/dl) (mg/dl)
Calibration 6.65 7.30 12.67 21.95
(Validation) 7.32 7.03 09.89 11.56
VI. VAL IDATION OF THE PRO PO SE D IGLU
DEVICE
To validate and test iGLU, 93 healthy, prediabetic
and diabetic samples aged 17-75 are taken follow-
ing medical protocols. A total of 64 males and 29
females are identiﬁed during collection of these 93
samples. All samples are taken in fasting, post-
prandial and random modes. The baseline charac-
teristics and error analysis is represented in Table
II and III, respectively. A 10-fold cross validation
has been performed to validate iGLU.
To test the device stability, an experiments have
been performed from multiple measurements of
same sample by couple of times. For this exper-
imental work, a volunteer has been recruited to
measure blood glucose through iGLU and invasive
method with time intervals of 5 minutes.
A value of 10 mg/dl deviation are considered in
observations during 7 iterations of blood glucose
7
measurement. During analysis, 2-4 mg/dl deviation
has been observed (Figure 10(a)). A different vol-
unteer has also been taken for another experimental
analysis to validate the accuracy of iGLU (Figure
10(b)). Measurement has been done with time inter-
val of 60 minutes using 7 iterations. Variations (low
to high) in reference blood glucose values between
8:00 AM-10:00 AM, 10:00 AM-2:00 PM and 2:00
PM-4:00 PM represent the glucose intakes in the
form of food. During analysis, 5-10 mg/dl deviation
represents the stability of iGLU. It was observed
that the effect of ﬁngers or earlobes changes is
negligible. CEG analysis is used to analyze the
accuracy of predicted glucose values from proposed
device. CEG categorizes the devices in terms of
precise measurement and elaborates the zones by
the difference between referenced and predicted
glucose values [27]. The predicted values are in the
zone A and B; then the device will be desirable.
During analysis, all predicted glucose values found
in zone A and B (Figure 11).
8 : 0 0 A M 8 : 0 5 A M 8 : 1 0 A M 8 : 1 5 A M 8 : 2 0 A M 8 : 2 5 A M 8 : 3 0 A M
120
122
124
126
128
130
132
134
136
138
140
B l o o d G l u c o s e C o n c e n t r a t i o n ( m g / d l )
T i m e
R e f e r e n c e d B l o o d G l u c o s e C o n c e n t r a t i o n ( m g / d l )
P r e d i c t e d B l o o d G l u c o s e C o n c e n t r a t i o n ( m g / d l )
W i t h d i f f e r e n t f in g e r c o m b i n a t i o n
F r o m e a r l o b e
(a) Time interval of 5 minutes
8 : 0 0 A M 9 : 0 0 A M 1 0 : 0 0 A M 2 : 0 0 P M 3 : 0 0 P M 4 : 0 0 P M 5 : 0 0 P M
100
110
120
130
140
150
160
170
180
190
200
210 R e f e r e n c e d B l o o d G l u c o s e C o n c e n t r a t i o n ( m g / d l )
P r e d i c t e d B l o o d G l u c o s e C o n c e n t r a t i o n ( m g / d l )
W it h d i f f e r e n t f i n g e r c o m b i n a t i o n
F r o m e a r l o b e
B l o o d G l u c o s e C o n c e n t r a t i o n ( m g / d l )
T i m e
(b) Time interval of 60 minutes
Fig. 10: Predicted and reference blood glucose con-
centration for validation of iGLU on single sample.
Reference Concentration [mg/dl]
0 100 200 300 400
Predicted Concentration [mg/dl]
0
100
200
300
400
A
D
E C
C E
D
B
B
Clarke's Error Grid Analysis
A (90%)
B (10%)
(a) Validation
Reference Concentration [mg/dl]
0 100 200 300 400 500
Predicted Concentration [mg/dl]
0
100
200
300
400
500
A
D
E C
CE
D
B
B
Clarke's Error Grid Analysis
A (94%)
B (06%)
(b) Testing
Fig. 11: CEG analysis of predicted glucose values
VII. CONCLUSIONS AND FUTURE DIRECTIONS
troscopy NIR technique based non-invasive glucose
monitoring low cost (approximately 20-25 USD)
device iGLU for real-life application. The error
margins for iGLU are improved compared to other
non-invasive approach based systems. After CEG
analysis, 100% samples come in the zone A and B.
During analysis of possible combinations with pro-
posed ML model, iGLU is found more optimized
compared to other measurement device.
In the future research on iGLU, we will involve
more features of IoMT. Glucose-level measurement
from serum is a immediate next goal to further
improve accuracy of iGLU. Integration of stress
measurement along with blood-glucose level is also
in pipeline. A closed feedback from healthcare
providers-end to the end-user side for control of
effects when needed to ensure remote healthcare
when there may be shortage of healthcare providers
can be more effective.
ACK NOWLEDG ME NT
The authors would like to thank Dispensary
MNIT Jaipur (India). We would also thank Dr.
Navneet Agrawal (diabetologist) and his team for
the support at Diabetes, Obesity and Thyroid Centre
Gwalior (India).
REFERENCES
[1] S. Ruan, “Intelligent systems for smart health care: Lever-
aging information for better well-being,” IEEE Consumer
Electronics Magazine, vol. 8, no. 2, pp. 71–71, March
2019.
[2] P. Sundaravadivel, K. Kesavan, L. Kesavan, S. P. Mo-
hanty, and E. Kougianos, “Smart-Log: A Deep-Learning
Based Automated Nutrition Monitoring System in the
IoT,” IEEE Trans. Consum. Electron, vol. 64, no. 3, pp.
390–398, Aug 2018.
[3] S. P. Mohanty and E. Kougianos, “Biosensors: A Tutorial
Review,” IEEE Potentials, vol. 25, no. 2, pp. 35–40,
March 2006.
[4] S. P. Mohanty, U. Choppali, and E. Kougianos, “Every-
thing you wanted to know about smart cities: The Internet
of things is the backbone,” IEEE Consumer Electronics
Magazine, vol. 5, no. 3, pp. 60–70, July 2016.
[5] P. Sundaravadivel, E. Kougianos, S. P. Mohanty, and
M. K. Ganapathiraju, “Everything You Wanted to Know
about Smart Health Care: Evaluating the Different Tech-
nologies and Components of the Internet of Things for
Better Health,” IEEE Consumer Electronics Magazine,
vol. 7, no. 1, pp. 18–28, January 2018.
8
[6] J. Li, P. Koinkar, Y. Fuchiwaki, and M. Yasuzawa, “A ﬁne
pointed glucose oxidase immobilized electrode for low-
invasive amperometric glucose monitoring,” Biosensors
and Bioelectronics, vol. 86, pp. 90–94, 2016.
[7] J. Y. Lucisano, T. L. Routh, J. T. Lin, and D. A. Gough,
“Glucose monitoring in individuals with diabetes using a
long-term implanted sensor/telemetry system and model,”
IEEE Transactions on Biomedical Engineering, vol. 64,
no. 9, pp. 1982–1993, 2017.
[8] A. Sun, A. G. Venkatesh, and D. A. Hall, “A multi-
technique reconﬁgurable electrochemical biosensor: En-
abling personal health monitoring in mobile devices,
IEEE Transactions on Biomedical Circuits and Systems,
vol. 10, no. 5, pp. 945–954, Oct 2016.
[9] G. Wang, M. D. Poscente, S. S. Park, C. N. An-
drews, O. Yadid-Pecht, and M. P. Mintchev, “Wearable
microsystem for minimally invasive, pseudo-continuous
blood glucose monitoring: The e-mosquito,” IEEE Trans-
actions on Biomedical Circuits and Systems, vol. 11,
no. 5, pp. 979–987, Oct 2017.
[10] M. M. Ahmadi and G. A. Jullien, “A wireless-implantable
microsystem for continuous blood glucose monitoring,”
IEEE Transactions on Biomedical Circuits and Systems,
vol. 3, no. 3, pp. 169–180, June 2009.
[11] G. Acciaroli, M. Vettoretti, A. Facchinetti, G. Spara-
cino, and C. Cobelli, “Reduction of blood glucose mea-
surements to calibrate subcutaneous glucose sensors: A
bayesian multiday framework,IEEE Transactions on
Biomedical Engineering, vol. 65, no. 3, pp. 587–595,
2018.
[12] I. Pagkalos, P. Herrero, C. Toumazou, and P. Georgiou,
“Bio-inspired glucose control in diabetes based on an
analogue implementation of a β-cell model,” IEEE Trans-
actions on Biomedical Circuits and Systems, vol. 8, no. 2,
pp. 186–195, April 2014.
[13] P. P. Pai, A. De, and S. Banerjee, “Accuracy enhancement
for noninvasive glucose estimation using dual-wavelength
photoacoustic measurements and kernel-based calibra-
tion,” IEEE Transactions on Instrumentation and Mea-
surement, vol. 67, no. 1, pp. 126–136, 2018.
[14] M. Yamaguchi, M. Mitsumori, and Y. Kano, “Nonin-
vasively measuring blood glucose using saliva,” IEEE
Engineering in Medicine and Biology Magazine, vol. 17,
no. 3, pp. 59–63, May 1998.
[15] K. Song, U. Ha, S. Park, J. Bae, and H. J. Yoo,
“An impedance and multi-wavelength near-infrared spec-
troscopy ic for non-invasive blood glucose estimation,”
IEEE Journal of Solid-State Circuits, vol. 50, no. 4, pp.
1025–1037, April 2015.
[16] W.-C. Shih, K. L. Bechtel, and M. V. Rebec, “Nonin-
vasive glucose sensing by transcutaneous raman spec-
troscopy,Journal of biomedical optics, vol. 20, no. 5,
p. 051036, 2015.
[17] C. W. Pirnstill, B. H. Malik, V. C. Gresham, and G. L.
Cot´
e, “In vivo glucose monitoring using dual-wavelength
polarimetry to overcome corneal birefringence in the
presence of motion,” Diabetes technology & therapeutics,
vol. 14, no. 9, pp. 819–827, 2012.
[18] E. Monte-Moreno, “Non-invasive estimate of blood glu-
cose and blood pressure from a photoplethysmograph
by means of machine learning techniques,” Artiﬁcial
intelligence in medicine, vol. 53, no. 2, pp. 127–138,
2011.
[19] S. Habbu, M. Dale, and R. Ghongade, “Estimation of
blood glucose by non-invasive method using photo-
plethysmography,S¯
a, vol. 44, no. 6, p. 135, 2019.
[20] S. Sharma, M. Goodarzi, L. Wynants, H. Ramon, and
W. Saeys, “Efﬁcient use of pure component and interfer-
ent spectra in multivariate calibration,Analytica chimica
acta, vol. 778, pp. 15–23, 2013.
[21] P. Jain, R. Maddila, and A. M. Joshi, “A precise non-
invasive blood glucose measurement system using NIR
spectroscopy and Hubers regression model,Optical and
Quantum Electronics, vol. 51, no. 2, p. 51, 2019.
[22] Y. Uwadaira, A. Ikehata, A. Momose, and M. Miura,
“Identiﬁcation of informative bands in the shortwave-
length nir region for non-invasive blood glucose mea-
surement,” Biomedical Optics Express, vol. 7, no. 7, pp.
2729–2737, 2016.
[23] W. Zhang, R. Liu, W. Zhang, H. Jia, and K. Xu, “Discus-
sion on the validity of nir spectral data in non-invasive
blood glucose sensing,” Biomedical optics express, vol. 4,
no. 6, pp. 789–802, 2013.
[24] S. Haxha and J. Jhoja, “Optical based noninvasive glucose
monitoring sensor prototype,” IEEE Photonics Journal,
vol. 8, no. 6, pp. 1–11, 2016.
[25] M. Golic, K. Walsh, and P. Lawson, “Short-wavelength
near-infrared spectra of sucrose, glucose, and fructose
with respect to sugar concentration and temperature,”
Applied spectroscopy, vol. 57, no. 2, pp. 139–145, 2003.
[26] P. Jain and S. Akashe, “Analyzing the impact of boot-
strapped adc with augmented nmos sleep transistors con-
ﬁguration on performance parameters,” Circuits, Systems,
and Signal Processing, vol. 33, no. 7, pp. 2009–2025,
2014.
[27] W. L. Clarke, “The original Clarke error grid analysis
(EGA),” Diabetes technology & therapeutics, vol. 7,
no. 5, pp. 776–779, 2005.
ABO UT T HE AU TH OR S
Prateek Jain is a Research Scholar in ECE
department of MNIT, Jaipur, India. He can be
contacted at: prtk.ieju@gmail.com.
Amit M. Joshi is an Assistant Professor in
Department of ECE, MNIT, Jaipur, India. He can
be contacted at: amjoshi.ece@mnit.ac.in.
Saraju P. Mohanty is the Editor in Chief
of the IEEE Consumer Electronics Magazine and
Professor in the Department of Computer Sci-
ence and Engineering (CSE), University of North
Texas (UNT), Denton, TX, USA. Contact him at
Saraju.Mohanty@unt.edu.
... The CVD cases observed among overall population is 1 to 2 %, while 10% of the old age people are observed to be affected. Smart Healthcare would provide continuous monitoring with personalised solution for the improvement of the quality of life [6]. This paper presents LVH classification on the basis of Romhilt-Estes' criteria using machine learning. ...
... CorrectP rediction T otalno.of prediction (6) or Accuracy = T P + T N T P + T N + F P + F N ...
Preprint
Full-text available
Left ventricular hypertrophy (LVH) is the heart condition where the walls of the left ventricle would be thicker than the normal condition. That obstructs the electrical activity of the heart and hence significantly decreases the pumping efficiency. LVH develops as the response to pressure overload that might arise from high blood pressure, stenosis etc. This condition of the heart can be reversed if diagnosed and treated in time. Its symptoms like short breathing, fatigue, palpitation of the heart, dizziness, chest pain after exercise are often misinterpreted or masked by the process of homeostasis, and it develops silently over the years. The prediction of presence of LVH using effective machine learning model would be helpful to have early diagnosis of the disease. This paper focuses on the prediction of Left Ventricular Hypertrophy (LVH) using the SVM machine learning model with the help of an electrocardiogram (ECG). The proposed methodology could be integrated with the IoT framework for smart healthcare solutions. Data is taken from the UCI machine learning repository i.e Cleveland, Hungary, Switzerland, and the VA Long Beach database consisting of 920 subjects. Data is split into 8:2 for training and testing purposes. Python 3.0 is used to analyze the data and the classification is verified by the 10-fold cross-validation technique. In this study, normal subjects and subjects with a high probability of LVH are classified successfully with greater ROC-AUC value than 0.5 which shows the unbiased classification. The results show that the proposed model with enhanced accuracy of 81+% and high specificity of 92.36% stands promising for futuristic smart healthcare applications.
... without needing blood samples) glucose level monitoring can have a significant impact on society. [19] We appreciate that security and privacy issues may arise due to EasyBand being part of H-CPS, thus novel solutions like lightweight blockchain and physical unclonable function (PUF) can be integrated with its architecture This proposed process flow is important in the development of EasyBand wearable as part of Internet-of-Medical-Things (IoMT) smart wearable device which uses technologies to sense and record the details of another similar device. This IoT innovation will help address problems concerning the current health crisis. ...
Article
Full-text available
The Internet of Things (IoT) is a shift in information technology that connects a variety of things. IoT is otherwise known as the Internet of Everything (IoE), a complex system that allows interconnection and communication among IoT devices. IoT applies various system development methods before building IoT-based network devices, Internet-connected devices, wireless sensors, and other related technologies, and it all requires a process flow. Physical or network connections among IOT-enabled devices requires a standard system development approach to provide quality systems. The basis of the study is the previous journals and literature studies. This literature review paper will help identify various IoT system development methodologies (SDMs) that are commonly known. This paper will present a comprehensive knowledge and conceptualization of different structuring processes utilized to create large-scale and complex Internet of Things (IoT) systems [1]. Reputable journals and studies are identified to provide the validity of the information hereon stated. In this paper, a study has been presented on the IoT System Development Methods (SDMs) and their various types. The first section of this paper includes an overview of IoT concepts about SDMs. Succeeding sections focuses on the various IoT System Development Methods (SDMs), most common IoT Methodologies and its illustrations, and lastly the conclusion.
... Many development studies were performed to optimize this procedure to reduce the pain associated with the measurement, resulting in the use of a blood lancet rather than a traditional needle and syringe [26]. However, this method still causes discomfort and increases the risks of blood-related infections [27]. Therefore, there has been a shift in the research focus towards the development of minimally invasive and non-invasive methods in recent years, although invasive blood glucose monitoring is still the most widely spread commercial approach [28,29]. ...
Article
Full-text available
Molecularly imprinted polymers (MIPs) have gained growing interest among researchers worldwide, due to their key features that make these materials interesting candidates for implementation as receptors into sensor applications. In fact, MIP-based glucose sensors could overcome the stability issues associated with the enzymes present in commercial glucose devices. Various reports describe the successful development of glucose MIPs and their coupling to a wide variety of transducers for creating sensors that are able to detect glucose in various matrices. In this review, we have summarized and critically evaluated the different production methods of glucose MIPs and the different transducer technologies used in MIP-based glucose sensors, and analyzed these from a commercial point of view. In this way, this review sets out to highlight the most promising approaches in MIP-based sensing in terms of both manufacturing methods and readout technologies employed. In doing so, we aim at delineating potential future approaches and identifying potential obstacles that the MIP-sensing field may encounter in an attempt to penetrate the commercial, analytical market.
... The CM and resistive TIA based readout circuit in [27] was operated relatively at higher power dissipation. The readout interface designs for electrochemical signal detection and system for monitoring of glucose level have been reviewed in [28][29][30][31][32]. The various features of instrument development and application for electrochemical sensor signal processing were discussed [33]. ...
Article
Full-text available
The paper presents a novel design of programmable current mode readout amplifier and potentiostat circuits for glucose sensing applications. In the proposed design, the threshold voltage of two MOS transistors is used as the reference supply voltage across glucose sensor reference and working electrodes. Another important feature is enhancement of the gain of PTA by transconductance (gm\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g_\textrm{m}$$\end{document}) boosting technique which is useful to detect sensor currents even if a very small amplitude signal is present. The performance of proposed design is analysed using SCL 0.18 μ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}m CMOS process technology with 1.5 V of power supply. The proposed design have programmable gain ranges from 42.18 to 60.38 dB and it offers a IRN of 1.002 nV/sqrt (Hz)–1.688 μ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}V/sqrt (Hz)@ 1–100 MHz. The % THD ranges from 8.26 to 8.34@ 35 nA to 230 nA of sensor current, high DR of 108 dB and PSRR is 100.4 dB. The linear range is found from 35 to 230 nA for the current of glucose sensor i.e. from 5 to 34 MΩ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Omega$$\end{document} approximately for electrode resistance of glucose sensor which is acceptable for the measurement of glucose level ranges from 40 mgdl\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{\textrm{mg}}{\textrm{dl}}$$\end{document} to 250 mgdl\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{\textrm{mg}}{\textrm{dl}}$$\end{document}. The power dissipation of the proposed architecture ranges from 165.11 to 459.50 μ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}Watt.
... Such applications related to wearable "intelligent" devices are expected to significantly enhance patients' quality of life because they promote a healthier lifestyle and physical exercise in which the birth quality of a given population can be significantly improved [22][23][24]. E.g., pregnancy can be benefited from in-time alarming of unwanted conditions or even from not scheduled appointments with the physician [16]. The use of novel, sophisticated materials, and reducedsize electronics contributed to the evolution of miniaturized body-worn sensors [1,9,10,25,26]. ...
Conference Paper
In this paper, a preliminary implementation of a system monitoring the fetus’s heart rate (FHR) has been designed and implemented as a mobile wearable measuring system with remote sensing specifically developed on Node MCU ESP8266 (ESP). In particular, the proposed system uses sensors for heart rate, humidity, temperature, and a transceiver module. The transceiver module is capable of efficient data transmission to a remote server station using an IEEE 802.11 b/g/n protocol—based on the wireless network. A major benefit is that the patient’s data is monitored at distance using an IoT device. Hence, it complies with the health safety distance measures required due to various situations, including that of the COVID-19 pandemic. The proposed implementation has been proven to be efficient in terms of hardware simplicity and cost-effectiveness and is accompanied by preliminary accurate measurements of the FHR.
Article
Measuring vital body signals is essential to measure basic body functions, prevent misdiagnosis, detect underlying health problems and motivate healthy lifestyle changes. Vital body signals are measured at the fingertips because the skin is thin, and the blood vessels are transparent. Visible light is passed at the fingertips, and the pulses generated are still acceptable on the outer nail. However, the body's vital signal measuring device continuously attached to the fingertip causes discomfort to the user. Therefore, in this study, it is proposed to measure the body's vital signals in other body parts. The wrist was chosen to be attached to the body's vital signal measuring device because the measuring device attached to the wrist allows it to continue to be used. This study aims to measure the body's vital signals, especially heart rate, on the wrist so that the correlation level of the measurement data is known. The main contribution of this study is built an electronic system to measure vital body signals, especially heart rate at the wrist with the help of the MAX30102 sensor that uses visible light with 650 - 670 nm. The MAX30102 sensor, which uses visible light with 650 - 670 nm, was selected for measurement. The ratio of the light reflected through the fingertips compared to the wrist. The result of measuring the heart rate signal on the wrist is in the form of a relatively flat wave so that the data sharpening process is carried out using the detrend method. The results showed that the measurement of heart rate signals at the wrist and fingertips of 15 respondents had accuration 85%. The accuration value shows that the data from the heart rate signal at the wrist is closely correlated with the data from the measurement of the heart rate signal at the fingertips. Therefore, measurements of heart rate signals, usually performed on the fingertips, can also be performed on the wrist. From the test results with a strong accuration, measurements are always taken when the hand can measure the place to measure vital signals, which is usually done at the fingertips.
Article
Full-text available
Artificial intelligence (AI) algorithms in combination with continuous monitoring technologies have the potential to revolutionize chronic disease management. The recent innovations in both continuous glucose monitoring (CGM) and the closed‐loop highlight the far‐reaching potential of AI biosensors for individual healthcare. This review summarizes some of the most advanced progress made in CGM biosensing. We will focus on three main applications of AI algorithms in diabetes management: closed‐loop control algorithms, glucose predictions, and calibrations. The challenges and opportunities of AI technologies for CGM in individualized and proactive medicine will also be discussed. The combination of artificial intelligence (AI) and biosensors enables better acquisition of patient data and improved design of wearable sensors for continuous glucose monitoring (CGM). The basic architecture of AI biosensors for CGM is composed of three main elements: wearable CGM sensor, AI‐data processing, and decision support system. Wearable CGM sensors can be used to capture a holistic model of glycemic control to obtain data collection. AI‐data processing can be grouped into interface, data classification, and data analysis. The decision support system and the closed‐loop highlight the far‐reaching potential of AI biosensors for individual healthcare.
Chapter
Full-text available
Artificial Intelligence (AI) is gradually emerging and becoming stronger in various fields such as technology, mathematics, engineering, and physics. AI is considered as a functional future in the scientific and technical world where people communicate, learn, and share ideas and opinions with the help of soft and hard technologies. AI will decrease the learning gaps and will also create the learning interest among students, and increase their learning ability and pace of learning. India is the fourth-largest generator of AI-related scholarly articles and is among the top ten AI patent-producing countries. AI has an important role in India's economy, including agriculture, industry, and various other sectors, which include finance, transportation, public administration, and defence. The National Education Policy (NPE) 2020 recognizes the importance of AI and emphasizes the importance of training everyone for an AI-driven economy. The aim and objective of the current study is to discuss various AI initiatives and programmes in India. The study also highlights the position of India in AI race. Keeping this in mind, the researchers describe in detail various AI initiatives, documents and papers like AI for All Initiatives, AI Gamechangers Program, AI Academic Institutions and Centres in India, State of Artificial Intelligence In India 2021, 75 @ 75 India's AI Journey, Responsible AI #AIFORALL and National Strategy on AI 2018. The study also briefly describes the various AI startups in India
Article
Full-text available
Background The Internet of Medical Things (IoMT) is now being connected to medical equipment to make patients more comfortable, offer better and more affordable health care options, and make it easier for people to get good care in the comfort of their own homes. Objective The primary purpose of this study is to highlight the architecture and use of IoMT (Internet of Medical Things) technology in the healthcare system. Method Several sources were used to acquire the material, including review articles published in various journals that had keywords such as, Internet of Medical Things, Wireless Fidelity, Remote Healthcare Monitoring (RHM), Point-of-care testing (POCT), and Sensors. Result IoMT has succeeded in lowering both the cost of digital healthcare systems and the amount of energy they use. Sensors are used to measure a wide range of things, from physiological to emotional responses. They can be used to predict illness before it happens. Conclusion The term “Internet of Medical Things” refers to the broad adoption of healthcare solutions that may be provided in the home. Making such systems intelligent and efficient for timely prediction of important illnesses has the potential to save millions of lives while decreasing the burden on conventional healthcare institutions, such as hospitals. patients and physicians may now access real-time data due to advancements in IoM.
Article
Full-text available
Diabetes is one of the prominent diseases around the world. Presently, invasive techniques need a finger prick blood sample . A repetitively painful procedure that produces the chance of infection. To resolve this issue, non-invasive measurement approach is proposed. In this paper, an efficient NIR wave based optical detection system is proposed with optimized post-processing regression model. After real-time data analysis, it has been found that the coefficient of determination ($$R^{2}$$) is improved with the value of 0.9084 using proposed regression model. Mean absolute derivative is also increased with 3.87 mg/dl corresponding to predicted blood glucose concentration. Mean absolute relative difference has exceeded to 3.25%, and average error is improved with 3.77% using proposed regression model. Average accuaracy has been analyzed 94–95% for predicted blood glucose concentration.
Article
Full-text available
correct balance of nutrient intake is very important, particularly in infants. When the body is deprived of essential nutrients, it can lead to serious disease and organ deterioration which can cause serious health issues in adulthood. Automated monitoring of the nutritional content of food provided to infants, not only at home but also in daycare facilities, is essential for their healthy development. To address this challenge, this paper presents a new Internet of Things (IoT) based fully-automated nutrition monitoring system, called Smart-Log, to advance the state-of-art in smart healthcare. For the realization of Smart-Log, a novel 5-layer perceptron neural network and a Bayesian Network based accurate meal prediction algorithm are presented in this paper. Smart-Log is prototyped as a consumer electronics product which consists of WiFi enabled sensors for food nutrition quantification, and a smart phone application that collects nutritional facts of the food ingredients. The Smart-Log prototype uses an open IoT platform for data analytics and storage. Experimental results consisting of 8172 food items for 1000 meals show that the prediction accuracy of Smart-Log is 98.6%.
Article
Full-text available
The Internet-of-Things (IoT) has taken over the business spectrum, and its applications vary widely from agriculture and health care to transportation. A hospital environment can be very stressful, especially for senior citizens and children. With the ever-increasing world population, the conventional patient-doctor appointment has lost its effectiveness. Hence, smart health care becomes very important. Smart health care can be implemented at all levels, starting from temperature monitoring for babies to tracking vital signs in the elderly.
Article
This paper presents a system which estimates blood glucose level (BGL) by non-invasive method using Photoplethysmography (PPG). Previous studies have shown better estimation of blood glucose level using an optical sensor. An optical sensor based data acquisition system is built and the PPG signal of the subjects is recorded. The main contribution of this paper is exploring various features of a PPG signal using Single Pulse Analysis technique for effective estimation of BGL values. A PPG data of 611 individuals is recorded over duration of 3 minutes each. BGL value estimation is performed using two types of feature sets, (i) Time and frequency domain features and (ii) Single Pulse Analysis (SPA). Neural network is trained using above mentioned proposed feature sets and BGL value estimation is performed. First we validate our methodology using the same features used by Monte Moreno in his earlier work. The experimentation is performed on our own dataset. We obtained comparable results of BGL value estimation as compared with Monte Moreno, with maximum R² = 0.81. Further, BGL estimation using (i) Time and frequency domain features and (ii) Single Pulse Analysis (SPA) is performed and the resulting coefficient of determination (i.e., R²) obtained for reference vs. prediction are 0.84 and 0.91, respectively. Clarke Error Grid analysis for BGL estimation is clinically accepted, so we performed similar analysis. Using Time and frequency domain feature set, the distributions of data samples is obtained as 80.6% in class A and 17.4% in class B. 1% samples in zone C and Zone D. For Single Pulse Analysis technique (SPA) the distribution of data samples are 83% in class A and 17% in class B. The proposed features in SPA have shown significant improvement in R² and Clarke Error grid analysis. SPA technique with the proposed feature set is a good choice for the implementation of system for measurement of non-invasive glucometer.
Article
Recently, intelligent systems have increased opportunities to leverage information for proactive and preventive health care. They also offer a way to incorporate patients into their health care as important players in disease management, thus reducing costs and improving diagnostics and treatment outcomes. By making measurements and analysis automatic, intelligent systems trigger new smart health care, leading to revolutionary human medical care together with progress in computer science, data science, and telecommunications. In this special section, interesting research articles on electroencephalogram (EEG), posture, heart rate, and biomechanics are presented to show how health care can be advanced with smart and intelligent technologies.
Article
Frequent monitoring of blood glucose levels is an essential part of diabetes care, but the invasiveness of current devices deters regular measurement. Noninvasive measurement techniques are painless to implement and rely on changes in sample properties to estimate glucose concentration. However, such methods are affected by the presence of different biomolecules, resulting in an increased estimation error and necessitating calibration to obtain accurate glucose concentration estimates. The use of photoacoustic spectroscopy for continuous noninvasive glucose monitoring is studied through measurements on different sample media. In vitro photoacoustic measurements taken from aqueous glucose solutions, solutions of glucose and hemoglobin, and whole blood samples at multiple excitation wavelengths show amplitude and area-based signal features to rise with the increase in sample glucose concentration. The calibration of photoacoustic measurements from glucose solutions using Gaussian kernel-based regression results in a root mean square error (RMSE), mean absolute difference (MAD), and mean absolute relative difference (MARD) of 7.64 mg/dl, 5.23 mg/dl, and 2.07%, respectively. Kernel-based calibration also performs well on solutions of glucose and hemoglobin, and whole blood samples, resulting in lower estimation errors than that of previous efforts and with glucose estimates being in the acceptable zones of a Clarke error grid (CEG). It allows for individual calibration of photoacoustic measurements in vivo, resulting in an RMSE, MAD, and MARD of 19.46 mg/dl, 10.79 mg/dl, and 7.01%, respectively, with 89.80% of the estimates being within Zone A of the CEG. The improvement in estimation accuracy with dual-wavelength photoacoustic measurements and kernel-based calibration would enable continuous noninvasive glucose monitoring, facilitating improved diabetic care.
Article
A multi-modal spectroscopy IC combining impedance spectroscopy (IMPS) and multi-wavelength near-infrared spectroscopy (mNIRS) is proposed for high precision non-invasive glucose level estimation. A combination of IMPS and mNIRS can compensate for the glucose estimation error to improve its accuracy. The IMPS circuit measures dielectric characteristics of the tissue using the RLC resonant frequency and the resonant impedance to estimate the glucose level. To accurately find resonant frequency, a 2-step frequency sweep sinusoidal oscillator (FSSO) is proposed: 1) 8-level coarse frequency switching (fSTEP = 9.4 kHz) in 10-76 kHz, and 2) fine analog frequency sweep in the range of 18.9 kHz. During the frequency sweep, the adaptive gain control loop stabilizes the output voltage swing (400 mV p-p). To improve accuracy of mNIRS, three wavelengths, 850 nm, 950 nm, and 1,300 nm, are used. For highly accurate glucose estimation, the measurement data of the IMPS and mNIRS are combined by an artificial neural network (ANN) in external DSP. The proposed ANN method reduces the mean absolute relative difference to 8.3% from 15% of IMPS, and 15-20% of mNIRS in 80-180 mg/dL blood glucose level. The proposed multi-modal spectroscopy IC occupies 12.5 mm² in a 0.18 μm 1P6M CMOS technology and dissipates a peak power of 38 mW with the maximum radiant emitting power of 12.1 mW.
Article
Objective: In most continuous glucose monitoring (CGM) devices used for diabetes management, the electrical signal measured by the sensor is transformed to glucose concentration by a calibration function whose parameters are estimated using self-monitoring of blood glucose (SMBG) samples. The calibration function is usually a linear model approximating the nonlinear relationship between electrical signal and glucose concentration in certain time intervals. Thus, CGM devices require frequent calibrations, usually twice a day. The aim here is to develop a new method able to reduce the frequency of calibrations. Methods: The algorithm is based on a multiple-day model of sensor time-variability with second-order statistical priors on its unknown parameters. In an online setting, these parameters are numerically determined by the Bayesian estimation exploiting SMBG sparsely collected by the patient. The method is assessed retrospectively on 108 CGM signals acquired for 7 days by the Dexcom G4 Platinum sensor, testing progressively less-calibration scenarios. Results: Despite the reduction of calibration frequency (on average from 2/day to 0.25/day), the method shows a statistically significant accuracy improvement compared to manufacturer calibration, e.g., mean absolute relative difference when compared to a laboratory reference decreases from 12.83% to 11.62% (p-value of 0.006). Conclusion: The methodology maintains (sometimes improves) CGM sensor accuracy compared to that of the original manufacturer, while reducing the frequency of calibrations. Significance: Reducing the need of calibrations facilitates the adoption of CGM technology both in terms of ease of use and cost, an obvious prerequisite for its use as replacement of traditional SMBG devices.
Article
Objective: The use of a fully implanted, first-generation prototype sensor/telemetry system is described for long-term monitoring of subcutaneous tissue glucose in a small cohort of people with diabetes. Methods: Sensors are based on a membrane containing immobilized glucose oxidase and catalase coupled to electrochemical oxygen detection and telemetry systems, integrated as an implant. The devices remained implanted for up to 180 days, with signals transmitted every 2 minutes to external receivers. Results: The data include signal recordings from blood glucose clamps and spontaneous glucose excursions, matched to reference blood glucose values. The sensor signals indicate dynamic tissue glucose, for which there is no independent standard, and a model describing the relationship between blood glucose and the signal is therefore included. The values of all model parameters have been estimated, including the permeability of adjacent tissues to glucose, and equated to conventional mass transfer parameters. As a group, the sensor calibration varied randomly at an average rate of -2.6%/week. Statistical correlation indicated strong association between sensor signals and reference glucose values. Conclusions: Continuous, long-term glucose monitoring in individuals with diabetes is feasible with this system. Significance: All therapies for diabetes are based on glucose control and require glucose monitoring. This fully implanted, long-term sensor system may facilitate new approaches for improved management of the disease.