Non Invasive Blood Glucose Estimation Based on

Photoplethysmography and Pulse Oximeter Principle

G. S. Cardoso
1Abstract— Diabetes Mellitus is a worldwide epidemic that also uses the principle of pulse oximetry to calculate an
poses a major challenge to health systems across the globe. In this estimation of blood glucose. Philip et al. [7] designed a non
context, the intensified monitoring of glycemic levels in pre- invasive system to measure blood glucose using a German
diagnosed or already diagnosed patients allows the prevention of sensor. The sensor has an infrared and a red LED to emit light.
acute and chronic complications. The proposed work describes the The estimative of blood glucose is done using neural networks,
development and in vitro evaluation of an instrument for the
continuous and non-invasive monitoring of arterial blood glucose
which ones were trained with PPG features, invasive glucose
concentration using the method of spectroscopy and measurements and bioimpedance. Sarkar et al. [8] propose a
photoplethysmography to provide greater convenience to users. non invasive glucometer using a single pair of 940nm
The system developed for the measurement of glycemic wavelength emitter/receptor. Then a regression analysis is
concentrations of this work was designed with two pairs of optical carried out between the PPG signal peak-to-peak voltage and
sensors like pulse oximetry sensors. The selected emitters were the glucose value that produces the PPG. Enric Monte Moreno
805nm and 660nm wavelengths, due to the glucose light absorption et al [9] proposed a system for simultaneous estimation of blood
at these wavelengths. The blood glucose concentration calculous is glucose and blood pressure using PPG and machine learning
done using a polynomial curve. This curve is fitted using collected algorithm. These papers propose multiple techniques for both
data from human tests with the system and commercial
processing and acquisition of PPG signal.
glucometer as reference to blood glucose concentration.

Keywords— Photoplethysmography, non invasive blood

In this Work, we propose a system consisting of two
glucose measurement. wavelengths emitters at 660nm and 805nm and a photodiode
S5972 to condition the PPG signal. The algorithm used to
I. INTRODUCTION estimate the concentration of blood glucose is an adaptation of

A ccording to the World Health Organization (WHO),

around 422 million people have diabetes in the world [1].
In 2012, 1.5 million people died from diabetes [1]. Diabetes
Beer-Lambert law to calculate an R factor. Then the obtained
data are used to fit a polynomial curve, which relates the
concentration of blood glucose and the R factor. Finally, this
does not have cure, but it has treatment. The treatment is based curve is used to estimate the concentration of blood glucose.
on keeping a healthy diet, physical activities and regular
glucose monitoring during the day to keep the glucose level at The paper is structured as follows: In Section II we describe the
normal ranges [2]. The last one is a great inconvenient because Beer-Lambert law, the PPG signal generation, the absorbance
most available home glucometers are invasive. These invasive graphs that supports our emitters wavelengths selection,
glucometers need a small amount of blood by pricking the spectroscopy used to glucose measurement, the architecture of
finger and extracting it. This is necessary for each measurement proposed system, hardware description and digital processing.
during the day. Sometimes this method discourages patients In section III we discuss the results and finally section IV
because of the pricking finger pain [3]. concludes the proposed research.

In order to avoid the pain, non invasive glucose measurement II. METHODOLOGY
methods are developed. There are many possibilities to do non
invasive glucose measurement like a method that use the Beer-Lambert Law
correlation between the exhaled acetone in the air provided by
the breath and the glucose levels [4]. Measurements using tears The Beer-Lambert law relates the intensity of a monochromatic
analysis through the use of contact lens [5]. Also, a widely light (Io) that falls upon a medium. Part of this light is
reviewed method is the glucose measurement using the transmitted through the medium and another part is absorbed.
Photoplethysmography (PPG) signal. This type of signal is The remaining Intensity (I) of light that propagates through the
obtained by illuminating a portion of the skin by a light- medium decays exponentially following the equation:
emitting diode at an specific wavelength and measuring any
changes in the light absorption with a photoreceptor (the most
used is photodiodes). (1)

There are several papers on literature that study use the PPG Where ε(λ) is the extinction or absorption coefficient of the
signal to obtain an estimative of the concentration of blood absorbent substance for a specific wavelength λ, c is the
glucose. Paul et al. [6] worked on a glucometer based on a concentration of the absorbent substance, which is constant for
single pair of 940nm wavelength emitter/receptor. This work the medium, and d is the length of the optical path through the

G. S. Cardoso, Instituto Federal Sul-rio-grandense, Pelotas, Rio Grande do Corresponding author: Gustavo dos Santos Cardoso.
Sul, Brazil, gustavo_16a@hotmail.com
medium. Figure 1 shows the light intensity as a function of the amplitude of AC component has 1-2% of the DC value [10].
medium length. The AC component is related to the substances that composes
the arterial blood, i.e., the blood glucose concentration changes
the absorption of the tissue, which implies PPG signal changes

Figure 1. Beer-Lambert law in homogeneous medium

Figure 3. PPG signal
This form of Beer law does not consider physical process which
includes reflection and scattering, it also considers that the
medium is homogeneous. In live tissues there are many
absorbing substances. Due to this, the system has to have a
method to differentiate the arterial blood absorbance among
other absorbing substances like skin and bone pigmentation and
venous blood [10]. Figure 2 shows the light absorption in live
tissues.

Figure 4. PPG signal components

Emitters Selection Criterion

The 805nm emitter was selected because the influence of

Figure 2 - Light absorption in live tissues
oxyhemoglobins and reduced hemoglobins are lower at this
The absorption described by the Figure 2 is divided into two wavelength. Figure 5 shows the oxyhemoglobins and reduced
parts. One of them, in which there are no changes related to the hemoglobin extinction coefficients.
group of absorbers, is composed by the venous blood, skin
pigmentation and by other substances. The other part is
composed by the arterial blood, which one is responsible for
transporting absorbers substances like glucose. Due to
biological characteristics of the cardiac cycle, the medium
absorption is changed according to both an increase and
decrease of the arterial blood and the length of the optical path
[10]. The PPG signal is the variation of the remaining absorbed
light signal [10].

Photoplethysmography

Figure 3 and 4 shows the AC and DC components of PPG

signal. The amplitude of constant voltage offset from the DC
component can be determined by the nature of the material
which the light passes through. The AC or pulsatile component, Figure 5. oxyhemoglobins and reduced hemoglobin extinction coefficients
which synchronous with the heart rate is depending on the
arterial blood volume pulse. The shape of the AC signal is 660nm was selected because the several numbers of researches
indicative of cardiac performance and vessel compliance. The that uses red lights to obtain the PPG signal [7,9,11].
 (5)
System Architecture, Algorithm and Test description

The architecture of the proposed system is shown in Fig. 6. Due to the necessity to generate a polynomial curve a
calibration test with humans was needed.
Forty human volunteers, both male and female, aged 18 to 60,
weighing 50 to 100kg and with different skins colorations.
The volunteers were oriented to maintain their regular eating
habits, without carbohydrate restrictions for the 72 hours prior
to the exam, to not exercise on the day of the exam, to not eat
during the exam and to fast for 8 to 12 hours before the exam.
The test was performed by monitoring the blood glucose level
(glucose concentration in the blood) for 3 continuous hours with
the proposed device and by taking nine blood glucose level
measurements using a commercial personal blood glucose level
monitor of the “fingertip” type (ACCU-CHECK
ADVANTAGE – Roche Diagnóstica Brasil). For each
measurement taken with the commercial personal blood
glucose level monitor it was necessary to puncture the finger to
collect a small blood sample using a disposable lancet. Samples
were collected after fasting and 15, 30, 45, 60 minutes after the
Figure 6. System architecture start of the test.
Half of these data was used to adjust the polynomial curve and
Total arrangement of the system is divided into different the other half is used to validation.
sections. The digital processing module commands the light
emission by the emitters. Only one emitter can be activated at a Accuracy (Arms), average deviation (B), standard deviation
time to not interfere with the measurement of the other. Then (SDR) and precision (Ps), as suggested by Severinghaus [15],
the light is transmitted through a tissue, the remaining were calculated for several conditions with the use of the
transmitted light is captured by the photodiode. The following mathematical expressions:
photodetector converts the light into electrical signal. Then this
electrical signal is processed by an analogical signal processing
section. The output of this section is a PPG signal of one (5)
wavelength. This signal is converted by the digital processing
module into a digital signal. This process is repeated by the
other emitter of different wavelength. Then both 660nm and
805nm digital PPG signals are digitally filtered with a moving (6)
average filter with cut-off frequency at 10Hz. After that the
peak-to-peak average of the PPG signal of one cycle is related
to his own DC term. This DC term is the average of all intensity
of the light transmitted during the acquiring cycle by the
corresponding emitter. Then an R factor is calculated and (7)
related to the actual blood glucose concentration. The R factor
is calculated as following:

(2)
(8)

Where the subscripts numbers only denote that is for different

emitters. Then this factor R can be related to blood glucose where n is the number of measurement pairs used, Gi is the i th
concentration through the following equation [12]: blood glucose level value estimated by the pulse glucometer
hereby proposed, Gri is the ith reference value presented by the
commercial personal blood glucose level monitor of the
(3) “fingertip” type and Gfiti is the ith blood glucose level value,
Where Cg denote the blood glucose concentration. Alternatives adjusted with the polynomial curve.
polynomial curves can be used. The following equations
represent some of these curves [13,14]: Hardware Description

The Hardware used in the analogical processing section is

(4) described as in Figure 8.
 Figure 7. Electrical circuits in analogical processing section

[9] Monte-Moreno, E. Non-invasive estimate of blood glucose and blood

pressure from a photoplethysmograph by means of machine learning
techniques. Artificial intelligence in medicine, v. 53, n. 2, p. 127-138, 2011.
[10] Webster, J. G. Design of pulse oximeters. CRC Press, 1997.
Figure 8. Analogical section structure [11] Pande, M. C. Joshi, A. K. Non-invasive optical blood glucose
measurement. International Journal of Engineering Research and
Applications (IJERA), v. 3, n. 4, p. 129-31, 2013.
Figure 7 shows the electrical circuits used in the proposed [12] Mendelson, Y. Kent, J. C. Variations in optical absorption spectra of adult
system. The function of analogical processing section is to and fetal haemoglobins and its effect on pulse oximetry. IEEE Transactions
condition the PPG signal to acquire it using the digital on Biomedical Engineering, IEEE, v. 36, n. 8, p. 844–848, 1989.
[13] Fine, I.; Weinreb, A. Multiple scattering effect in transmission pulse
processing module. This filter stage has a passband about oximetry. Medical and Biological Engineering and Computing, Springer,
0.05Hz to 5Hz with gain around 155. This passband is selected v. 33, n. 5, p. 709–712, 1995.
due to researches about the PPG physiological composition as [14] NGHIA, L. N. T. Design of A SPO2 pulse oximeter phototype. PhD thesis
explained in Hayes & Smith and Porto [16,17]. — International University HCMC, Vietnam, 2012.
[15] L. Yun-Han, H. Fu-Rong, L. Shi-Ping, and C. Zhe, “Detection Limit of
Glucose Concentration with Near-Infrared Absorption and Scattering
Spectroscopy,” Chin. Phys. Lett., Vol. 25, No. 3, pp. 1117-1119, 2008.
[16] Porto, C. C. Semiologia médica.: Guanabara Koogan, 2009.
[17] Hayes, M. J. Smith, P. R. A new method for pulse oximetry possessing
inherent insensitivity to artifact. IEEE Transactions on Biomedical
Engineering, IEEE, v. 48, n. 4, p. 452–461, 2001.
III. RESULTS AND DISCUSSION Gustavo dos Santos Cardoso was born in Pelotas, Brazil, He
is currently finishing his bachelor’s degree in Electrical
Engineering at the Instituto Federal Sul-rio-grandense
IV. CONCLUSION (IFSUL).

REFERENCES

[1] World Health Organization [Internet].

[2] BRASIL. Ministério da Saúde.Caderno de Atenção Básica n 16-Diabetes
Mellitus. Brasília, 2006.
[3] Grossi, S. A. A., Cianciarullo, T. I., & Della Manna, T. Análise dos ajustes
insulinoterápicos realizados a partir de dois esquemas de monitorização
domiciliar em pacientes com diabetes mellitus do tipo 1. Acta Paulista de
Enfermagem, v. 19, n. 4, p. 420-426. 2006
[4] Todd, C., Salvetti, P., Naylor, K., & Albatat, M. Towards non-invasive
extraction and determination of blood glucose levels. Bioengineering, v.4,
n.4, p.82. 2017
[5] SMITH, John L. The pursuit of noninvasive glucose: Hunting the deceitful
turkey. Revised and Expanded, copyright, p. 70, 2015.
[6] Paul, B., Manuel, M. P., & Alex, Z. C. Design and development of non
invasive glucose measurement system. In: 2012 1st International
Symposium on Physics and Technology of Sensors (ISPTS-1). IEEE,
2012. p. 43-46.
[7] Philip, L. A., Rajasekaran, K., & Jothi, E. S. J. Continous monitoring of
blood glucose using photophlythesmograph signal. In: 2017 International
Conference on Innovations in Electrical, Electronics, Instrumentation and
Media Technology (ICEEIMT). IEEE, 2017. p. 187-191.
[8] Sarkar, K., Ahmad, D., Singha, S. K., & Ahmad, M. Design and
Implementation of a Noninvasive Blood Glucose Monitoring Device. In:
2018 21st International Conference of Computer and Information
Technology (ICCIT). IEEE, 2018. p. 1-5.