See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/333647537

Non-invasive Glucose Monitoring System Utilizing Near-Infrared Technology

Chapter · January 2020

DOI: 10.1007/978-981-13-5859-3_71

CITATIONS READS
2 1,896

5 authors, including:

Viet Anh Truong Pham Hien

Ho Chi Minh City University of Technology and Education Ho Chi Minh City International University
64 PUBLICATIONS 836 CITATIONS 55 PUBLICATIONS 256 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Application FACTS devices in power system View project

Full-field and full-range sequential measurement of the slow axis angle and phase retardation of linear birefringent materials View project

All content following this page was uploaded by Pham Hien on 03 December 2019.

The user has requested enhancement of the downloaded file.

 Non-invasive Glucose Monitoring System
Utilizing Near-Infrared Technology

Duc Trinh-Minh Dinh, Viet Anh Truong, An Nhu-Phu Tran,

Hieu Xuan Le, and Hien Thi-Thu Pham

Abstract the result can be more accurate, precise and sensitivity as

Regular monitoring of glucose concentration is essential much as possible.
and urgency, especially diabetics. However, those meth-
ods which involve finger puncturing are invasive, expen-
  
Keywords
sive as well as painful. Also, there are risks of infectious Near-infrared Non-invasive Blood glucose
diseases using these techniques due to the contact of the Diabetes
needle on human skin. This paper proposed a
non-invasive glucose monitoring system utilizing the
near-infrared (NIR) light to measure the glucose concen-
tration in the human blood. The designed system uses a 1 Introduction
980 nm-wavelength NIR LED transmitting through
D-glucose phantom samples, a photo-sensor for analyzing Diabetes is a pathological metabolic condition that occurs
the transmitting optical parameters, a filter & an amplifier when the amount of glucose (sugar) in the blood is too high,
circuit, and a Nano Arduino microcontroller. For calcu- which affects other organs if not diagnosed and left
lating the values of glucose concentration, the R untreated. Nowadays, it has become a significant health
programming with the methodology of the artificial problem globally. Nowadays, the number of people getting
neural network (ANN) was applied. This type of diabetes is rising, and the need for the blood glucose mon-
methodology is considered as one of the most useful itoring is the obvious outcome. Checking the blood pressure
technique in the world of data analysis and because it is regularly not only let people know what state they are in but
adaptive, learns from the provided information and also help them to have suitable prevent diabetes [1–5].
optimizes for better prediction outcomes. The ANN is With the available devices, the users can check their
used to predict the correlation equation between collected blood glucose quickly and regularly. The most popular
voltage and glucose concentration. The obtained glucose device is the invasive type. The user has to make a finger
level is demonstrated directly on the system’s screen or prick by using a test strip, and the device uses their blood to
further sent to the user’s mobile phone. The result calculate the glucose amount. Although giving precise
obtained shows a correlation between the transmittance results but this method has some disadvantages: not suitable
and the concentration of D-glucose solution. The corre- for people who afraid of needles, waste of money and time
lation parameter of the technique is R2 = 0.9957. Despite to buy the test trip, painful and discomfort. Moreover, using
having the acceptable results, there are still some the invasive method could increase the risk of infection.
improvements that could be carried out for more accurate Hence, finding another method is the concern at the moment.
measurement (angle of the LED, a procedure of filtering Indeed, there is a sharp rise of the non-invasive method
and amplifying, usage of an optical instrument…). For for checking the blood glucose level. The majority of these
further investigation and development, it is predicted that methods is the application of the LED and the properties of
light for measurement. These devices can deliver painless
D. T.-M. Dinh  V. A. Truong  A. N.-P. Tran  and comfortable progress for the users. Although this type of
H. X. Le  H. T.-T. Pham (&) method is new, it shows the potential of making a device
Biomedical Engineering Department, International which is more user-friendly, painless and could give reliable
University- Vietnam National University HCMC,
Ho Chi Minh City, 70000, Vietnam results.
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 401

V. Van Toi et al. (eds.), 7th International Conference on the Development of Biomedical Engineering
in Vietnam (BME7), IFMBE Proceedings 69, https://doi.org/10.1007/978-981-13-5859-3_71
402 D. T.-M. Dinh et al.

The non-invasive method works base on the light prop- decrease of scattering light. A constant current circuit is
erties. For each substance, there is always a specific wave- designed for emission of NIR to minimize the fluctuation in
length at which the absorbance of that substance is that the current through the NIR LED. For this circuit, the
strongest. Each substance has its wavelength. Using this BC557 PNP transistor is used along with LM358. LM358
property, the system consists of two main components which op-amp plays the role as a current stabilizer to adjust the
are the LED and the photo-sensor. The LED emit the light suitable current goes into the LED. The output of LM358 is
through the skin area where it is put, and the photo-sensor is given to the base of the transistor for controlling the average
placed on the opposite side. By choosing the appropriate power transmitted by the NIR LED. Figure 2 demonstrates
wavelength which absorbed mostly by blood glucose, the the emitter circuit diagram.
amount of light that reaches the photo-sensor is calculated
and analyzed.
Choosing the appropriate wavelength for the system is the 2.2 Receiving System
most important key. To achieve the good result, the wave-
length must be the one that has the most interaction with The OPT101 is used as the optical sensor. The reason for
blood glucose and also not be affected too much by the other using this sensor is because the sensitivity is high enough
substances. There have been many options for the wave- thus it would not be affected too much by another light
length that is proposed by the previous research since glu- source from the environment or scattering light. LM358 is
cose has light absorption peaks at wavelengths of 940, 1150, also used in the designed circuit. Therefore, the signal col-
1450, 1536 nm [1–5]. The wavelength of 940 nm is chosen lected from the optical sensor is then amplified and filtered.
the most in research since, at this wavelength, the attenuation Figure 3 shows the receiver circuit diagram.
of optical signals by other constituents like platelets, red
blood cells or water is at a minimum.
2.3 Black Box System

2 Methodology A black box system is built to keep the sample and the sensor
from getting noise. There is a slit put in front of the LED, and
The proposed system consists of an adapter and a voltage the size of the slit would fit the emitting angle so the intensity
converter for converting 220–24 and 24–5 V for the power transferred from the LED to sample is the most. The purpose
supply. A NIR light source and an optical sensor are the is to focus the light into a line and lead it directly to the
main parts of the system. When the light reaches the optical sample, and it could reduce the noise since most of the
sensor, the signals go through amplifying and filter circuits,
and an Arduino is used to convert the data from analog to
digital signals. After the processing, the value of blood
glucose level is shown on the computer. Figure 1 validates
the operation of the system.

2.1 Emitting System

The NIR LED used in the system is TSAL 6100 with the
wavelength of 940 nm. This LED has an emitting angle and
the power that suitable for the system, and that leads to the

Fig. 1 The block diagram of the system Fig. 2 Emitter circuit

Non-invasive Glucose Monitoring System … 403

Fig. 3 Receiver circuit

sensitivity contact with the sample. All the light leaving the Since one node in hidden layers may connect with several
sample goes into the optical sensor put behind the sample. others nodes, the value of that node can be calculated with
The box is painted black to absorb the scattering light goes out the formula:
from the sample. There is an area in the box for putting the
circuits. Figure 4 shows the illustration of black box system. Y ¼ f ðw1:X1 þ w2:X2 þ    þ Wn:XnÞ ð1Þ

where Y is the output of that node and X1, X2, …, Xn are

the input nodes; W1, W2, …, Wn is the weights corre-
2.4 Analyzing System sponding to each node respectively.
The final node- output node is utilized to compare the
The method for processing the data getting from the LED is accuracy of the model to the training data. The weights is
by using machine learning. ANN method is chosen for
updated continuously until we obtain the minimum in
processing and analyzes the data since it is more appropriate
“Error estimation” between actual output and model output
with the complex data [6]. Figures 5 and 6 show the com- [6, 7].
ponent of ANN were used in this study.
The network consists of one input layer, one output layer, 1X  ðiÞ ðiÞ
2
J ðwÞ ¼ target  output ð2Þ
and several hidden layers. Each input node maps to all 2 i
hidden nodes, and each hidden node maps to all nodes of the
where J(w) is the error/loss, target(i) is the actual output and
next hidden layers till it reaches the output
output(i) is the network output.
Following, Fig. 7 shows the flowchart of the ANN net-
work for better illustration
There were D-glucose samples prepared with a variant of
concentrations. Every time a sample with known concen-
tration is measured, the output voltage is recorded. After
multiple measurements and records for each concentration, a
set of data with corresponding output voltages is created for
ANN. This set of data is used to train the ANN, and from the
trained data, appropriate function and algorithm is generated
to calculate the connection between D-glucose concentration
and the output voltage. The sample for the training data must
be assured with unification (in the way of making the sam-
ple: the weight of 12 sample for each concentration must be
identical and also the amount of water) to reduce the error
between each measurement and increase the accuracy for the
predicted result.
Fig. 4 Black box system
404 D. T.-M. Dinh et al.

Fig. 5 ANN diagram

Fig. 6 The layers of ANN

3 Results and Discussion

The results were taken from twelve samples of five different

concentrations (500, 1000, 1500, 1800 and 2000 ppm). The
mean of each sample were chosen to make the correlation
line between the concentration and the output voltage.
Table 1 and Fig. 8 show the collected data (with standard
deviation) and the correlation.
As data at two concentrations 1500 and 1800 ppm shows
no linearity, the 1500 ppm concentration is removed in order
to increase the accuracy of the system (by comparing the R2
of the model while removing 1500 and 1800 ppm respec-
tively). Figure 9 shows the fit line and trend line of the data
after removing 1500 ppm concentration point.
The results illustrate the linear correlation of the output
voltage and glucose concentration chosen from the data
table. It reveals that the more concentration of the sample,
the more output voltage measured. The trend line has the
R2 = 0.9957 comparing with another reliable source is the
UV-VIS 730 machine with R2 = 0.9999.
The results approve that wavelength and the method of
processing chosen for the research are appropriate and have
the potential to develop further. As the few number of
samples, the use of ANN cannot be optimized and the pro-
duced result is not enough sensitivity. Some improvements
Fig. 7 Flowchart of ANN network
 Non-invasive Glucose Monitoring System … 405

Table 1 The collected data table

Samples (ppm) 1 2 3 4 5 6 7 8 9 10 11 12
500 3.936 3.937 3.945 3.945 3.924 3.924 3.915 3.915 3.922 3.921 3.922 3.93
1000 3.955 3.942 3.945 3.958 3.941 3.942 3.94 3.932 3.933 3.939 3.948 3.945
1500 3.971 3.979 3.969 3.979 3.979 3.97 3.963 3.974 3.98 3.976 3.98 3.977
1800 3.976 3.972 3.979 3.978 3.979 3.974 3.975 3.971 3.98 3.972 3.973 3.972
2000 3.985 3.987 3.985 3.984 3.982 3.987 3.983 3.978 3.981 3.987 3.988 3.983

4 Conclusion

The work in this study provides the way for measuring the
glucose concentration non-invasively. For further develop-
ment, the system is expected to be able to work with more
complicated solutions such as a microsphere, animal blood
as well as other liquid composition in the human body.

Acknowledgements This research is funded by International

University, VNU-HCM under grant number SV2017-BME-07.

Conflict of Interest The authors declare that they have no conflict of

interest.

References
Fig. 8 Correlation between the output voltage and concentration
1. Bobade C.D., Mahadev, S.P.: Non-invasive monitoring of glucose
level in blood using near-infrared spectroscopy
Relation between concentration and voltage 2. Narkhede, P., Dhalwar, S., Karthikeyan, B.: NIR based
3.99 non-invasive blood glucose measurement. Indian J. Sci. Technol.
3.98
R² = 0.9957 9(41) (2016)
3. Yadav, J., et al.: Near-infrared LED based non-invasive blood
3.97
glucose sensor. In: 2014 International Conference on Signal
Voltage(V)

3.96 Processing and Integrated Networks (SPIN). IEEE, New York

3.95 (2014)
4. Lawand, K., Parihar, M., Patil, S.N.: Design and development of
3.94
infrared LED based noninvasive blood glucometer. In: 2015
3.93 Annual IEEE India Conference (INDICON). IEEE, New York
3.92 (2015)
0 500 1000 1500 2000 2500 5. Malin, S.F., et al.: Noninvasive prediction of glucose by
Concentration (ppm) near-infrared 1651–1658
6. Guo, D., et al.: Noninvasive blood glucose measurement based on
Fig. 9 Model after removing 1500 ppm data NIR spectrums and double ANN analysis. J. Biosci. Med. 3(06), 42
(2015)
7. Coster, S., et al.: Monitoring blood glucose control in diabetes
should be carried out for more accurate and reliable results in mellitus: a systematic review. Health Technol. Assess. (Winchester,
future: the emitting angle of the LED, the amplifying and England), 4(12), i–iv, 1–93 (1999)
filtering circuit, the structure of the black box, increasing the
size of the data set for the better fit with ANN in predicting
the fit line, usage of another optical instrument.

View publication stats