Engineering, 2013, 5, 553-560
http://dx.doi.org/10.4236/eng.2013.510B114 Published Online October 2013 (http://www.scirp.org/journal/eng)
Copyright © 2013 SciRes. ENG
Noninvasive Blood Glucose Monitoring System Based on
Distributed Multi-Sensors Information Fusion of
Multi-Wavelength NIR*
Bo Zeng, Wei Wang#, Na Wang, Funing Li, Fulong Zhai, Lintao Hu
School of Information Science and Engineering, Lanzhou Universit y, Lanzhou, China
Email: zengbo11@gmail.com, #wangw@lzu.edu.cn
Received 2013
ABSTRACT
In this research, a near infrared multi-wavelength noninvasive blood glucose monitoring system with distributed laser
multi-sensors is applied to monitor human blood glucose concentration. In order to improve the monitoring accuracy, a
multi-sensors information fusion model based on Back Propagation Artificial Neural Network is proposed. The Root-
Mean-Square Error of Prediction for noninvasive blood glucose measurement is 0.088mmol/L, and the correlation coef-
ficient is 0.94. The noninvasive blood glucose monitoring system based on distributed multi-sensors information fusion
of multi-wavelength NIR is proved to be of great efficient. And the new proposed idea of measurement based on distri-
buted multi-sensors, shows better prediction accuracy.
Keywords: Noninvasive Glucose Monitoring; NIR Arrays Signals Fusion; BP-Artificial Neural Network
1. Introduction
Diabetes has become a modern disease and more than
150 million people are suffering from it all around the
world [1]. In order to prevent complication, the tight
blood glucose level control is very essential. Currently,
patients are recommended to monitor their blood glucose
level via an invasive finger-tip stick method, this way
will inevitably bring patients pain and infection [2]. In
order to avoid the weakness of invasive method, a num-
ber of noninvasive measurements occur r ed, like middle-
infrared emission spectroscopy [3,4], Near Infrared (NIR)
spectroscopy [5,6]. The main drawback of infrared mea-
surement is its accuracy and stability. In our past work,
an infrared noninvasive blood glucose measurement sys-
tem based on multi-sensors and Mixture of Experts (ME)
[7] was designed. ME algorithm greatly improved the
precision of noninvasive blood glucose measurement. In
the noninvasive blood glucose monitoring system based
on distributed multi-sensors information fusion of mul-
ti-wavelength NIR, this system was designed on the pur-
pose of continuous blood glucose monitoring for patients
at home and hospital [2]. Another advantage of this sys-
tem was the introduction of distributed multi-sensors id ea,
this change has been proved to be better improvement for
blood glucose prediction accuracy. In this distributed
multi-sensors system, a multi-sensors information fusion
model, Back Prop agation Artificial N eural Network ( BP-
ANN), was appli e d.
2. Noninvasive Blood Glucose Monitoring
Model
The experimental data for this research was obtained via
NIR based on laser arrays as shown in Figure 1. The
NIR light source consists of 3*3 laser diodes arrays op-
erating at output powers of 5 mW [8].
2.1. Hardware Design
Wavelengths selection: According to the specialty of
glucose molecular structure and absorption specialty,
second order times frequency absorption exists between
1100 - 1300 nm, and first order times frequency absorp-
tion exists between 1500 - 1800 nm. Other components
in blood, such as hemoglobin and water contain groups
of hydrogen which can trigger NIR absorption as well.
Water absorption peaks are mainly distributed between
1440 - 1460 nm and between 1940 - 1960 nm, those
regions should be avoided in wavelength selection. Be-
tween 1400 - 1800 nm, water absorption only exists at
1787 nm, fat and protein exist no absorption peak in this
region. Therefore, 1400 - 1800 nm region is suitable for
measurement wavelength selection.
Measurement sites selection: In noninvasive blood
*
The work is supported b y National Nature Science Fund of China (No
30970876, 81141076).
*Corresponding a uthor.
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
554
Figure 1. Framework of the multiwavelength arrays monitoring system.
glucose measurement, the selection of an ideal measure-
ment site is essential. Four factors should be considered :
First, for the convenience of measurement, the measure-
ment site should be exposed outside; Second, in order to
decrease the influence from outside factors, individual
difference like gender, age and physical state should be
low; Third, the measurement site should contain rich
blood and the interference from other components is low;
Last, NIR light can transmit measurement site easily. In
the noninvasive blood glucose monitoring system based
on distributed multi-sensors, left ear lobe, right ear lobe
and the right hand part between thumb and index finger
are selected as distributed measurement sites.
Circuit design: six channels of laser-driving and photo-
electronic amplification circuit are designed in this
system. Figure 2 shows the laser-driving circuit, LD-
driving and the part of photodiode feedback circuit help
to maintain laser operate with output power of 5 mW and
forward current of 30 mA. By adjusting POT1, forward
current can be adjusted between 0 - 120 mA. Figure 3
shows the photoelectrical signal amplification c ircuit. By
applying an operational amplifier POT1 and some af-
filiated resisters and capacities, the obtained spectral
signal can be amplified by 1000 to 3V, the amplified
signal is transferred to plug seat JP2-1 which connected
to AD converter.
2.2. Software Design
1) Data acquisition and save: Amplified spectral sig-
nal is first conveyed to AD converter, which supports
Labview drive. So the converted data can be displayed
on the screen programmed by Labview language. Figu r e
4 shows the interface of acquired spectral data from six
channels. At each channel, the average value of 10
successive acquired data is calculated as displaying infor-
mation.
2) Distributed multi-sensors information fusion based
on BP-ANN: BP-ANN is one kind of supervised learning
network. There are two phases of positive transmitting
processing and error reverse transmitting processing in
the study processing of BP-ANN.
A three layers BP neural network (Figure 5) is applied
to model fuse multi-sensors information and predict the
blood glucose concentration in experiments. Least Square
function is adopted as error function in the training of BP
network, which writes as:
( )
2
11
2
q
m
rk rk
rk
k
yo
E
= =
=
∑∑
(2)
k
O
is the output of note k; yk is the corresponding de-
sired output; q is the number of output notes, and m is the
number of training samples. For the output weight, the
revise value
jk
w
is:
jkk k
wo
ηδ
∆=
(3)
jk
w is the weight connecting the number k output
note and the number j hidden note:
η
is the study speed,
k
o
is output value at the number k output note;
is
the gradient factor. For the hidden layer, revise value
writes as:
ijj j
wo
δ
∆=
(4)
ij
w
is the weight connecting the number j hidden
note and the number i input note;
is the output at the
number j hi dden note .
Figure 6 shows the training curve in the concentration
prediction of blood glucose, after training of 10 times,
the goal is reached and the performance can reach 0.088
mmol/L.
3) Prediction result display based on Labview: The
acquired data displaying and predicted blood glucose
concentration value interface is shown in Figure 4. On
the interface, 8 measurement methods options are given.
Corresponding blood glucose prediction program will be
launched by clicking any switch, and the predicted value
will be displayed as well.
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
555
Figure 2. Laser-driving circuit.
Figure 3. Photoelectrical signal amplification circuit.
Figure 4. System interface of calculated blood glucose concentration.
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
556
Figure 5. BP framework for multisensor information fusion.
Figure 6. Training curve in human blood glucose concent ration predict i on.
3. Experiments
In both glucose samples experiment and noninvasive
blood glucose measurement experiment, 7 measurement
methods with single, two and three wavelengths were
applied, which were named A, B, C, A + B, A + C, B +
C and A + B + C method respectively.
3.1. Glucose Samples Experiment
Process of glucose experiment: All samples were mea-
sured in a 5 mm quartz cell and the temperature in the
sample cell was controlled at about 37˚C. Two sets of 2 0
samples were prepared with varying concentrations,
ranging from 10 to 200 mg/dL in 10 mg/dL steps. More
than 500 original data sets were obtained, after eliminat-
ing some outlier samples, 200 samples with different
concentrations were left for building calibration model.
In the experiment, the spectrum of empty cell was mea-
sured firstly as a background [9].
Data preparation of calibration model: It is essential
to figure out the significant difference betwe en each
group of spectral data sample. Statistic method of T-test
was used to prove the validity of samples and to check
out the outliers data. In Table 1, a total of 200 glucose
1
x
2
x
3
x
4
x
5
x
6
x
Six spectra input
Hidden layer
Output (y)
Spectra absorption of
wavelength A
Spectra absorption of
wavelength B
Spectra absorption of
wavelength C
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
557
Table 1. Comparison of 20 glucose absorption spectra data.
Sample 10 20 30 40 50 60 70 80 90 100
xs±
312.0 ±
0.005
311.8 ±
0.008
311.6 ±
0.003
311.4 ±
0.032
311.3 ±
0.020
311.1 ±
0.023
318.8 ±
0.003
310.6 ±
0.003
310.4 ±
0.003
310.1 ±
0.006
p 0.0001 0.0001 0.0001 0.0007 0.0001 0.0002 0.0001 0.0001 0.00005 0.00005
Sample 110 120 130 140 150 160 170 180 190 200
xs±
309.9 ±
0.019
309.8 ±
0.020
309.5 ±
0.002
309.4 ±
0.005
309.2 ±
0.002
309.0 ±
0.019
308.7 ±
0.007
308.6 ±
0.018
308.3 ±
0.003
308.1 ±
0.019
p 0.0044 0.0001 0.00004 0.00003 0.0011 0.0004 0.0007 0.0001 0.0001 0.0001
absorption spectra were given out, the average and
standard deviation of 10 measurement values for each
glucose sample are calculated as well. Comparing with
the differences between 20 groups of glucose NIR ab-
sorption spectra, their differences achieve significant
level (T-test, p < 0.05).
3.2. Human Noninvasive Measurement
Experiment
Process of Human noninvasive measurement experiment:
In order to get training data covering wide variety scope
of human blood glucose concentration, the blood glucose
tests were made on volunteers at certain time before or
after their meal with the interval time of half an hour.
After measurement of blood glucose, multi-sensors were
attached to measurement sites to get responding spectra
information [10]. A total of 142 samples for each type of
measurements were got in this experiment and the blood
glucose concentration ranges from 3.2 mmol/L to 10.8
mmol/L.
Data preparation for BP-ANN training and testing:
Similarly to the data preparation of calibration model,
experimental data in noninvasive blood glucose measure-
ment were given in Table 2. T-test is applied to test the
spectra difference of different groups of blood glucose
concentration, p of T-test is smaller than 0.05, which
demonstrates that their differences reach significant
levels.
4. Results
4.1. Building Calibration Model
In both the glucose samples experiment and noninvasive
blood glucose measurement experiment, 7 kinds of cali-
bration model were built to realize the function of conti-
nuous monitoring. In A + B + C method of glucose expe-
riment, 20 spectra data samples were used to build the
calibration model. Figure 7 shows the relation between
glucose concentration and glucose absorption spectra,
from the figure we can see this validation owns a fine
prediction ability for unknown received spectra informa-
tion. In Table 3, the performances of different measure-
ment methods were compared, in the A + B + C method,
the RMSEP is 4.412 mg/dL and CC is 0.90, which shows
a great improvement in prediction performance to other
single or two wavelength methods. Compare with statis-
tic values of three single wavelength and combination of
two wavelength methods, single wavelength A shows
best sensitivity
4.2. Results of Human Blood Glucose
Noninvasive Measurement
4.2.1. BP-ANN Measurement Model
In the human noninvasive blood glucose measurement
experiment, a total of 142 samples were obtained, 122
couples of blood glucose absorption spectra data and
corresponding blood glucose reference values were used
for the BP network training. As shown in Figure 6, the
network can reach goal after 10 training, RMSEP of this
model is 0.088 mmol/L and CC is 0.9315 in the method
of A + B + C. Comparing with the other six methods, the
BP-ANN greatly decreased the prediction error, which
can be seen in Table 4.
4.2.2. Testing of Measurement Model
The left 20 samples were used as validation data sets to
test performance of the measurement model. 20 spectra
information data were inputted to the BP-ANN multi-
sensors information fusion model, 20 output values of
predicted blood glucose concentration were received at
the output layer, the result was given in Table 5. The
RMSEP between NIR method and standard method is
0.3143 mg/dL. T-test is applied and p is 0.1728, bigger
than 0.05, which demonstrate that the relation between
those two values is well related.
5. Conclusion
In this noninvasive blood glucose monitoring system,
two new measurement ideas were employed, which are
continuous monitoring and distributed multi-wavelength
measurements. The experimental result has proved that
this system has an improved advantage in blood glucose
prediction, and the multi-wavelength information fusion
model has also contributed greatly to the monitoring
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
558
Table 2. Difference comparison of 28 groups of blood glucose absorption spectra data.
Sample 10.8 10.5 10.1 10.0 9.8 9.6
xs±
567.95 ± 0.03 568.42 ± 0.15 568.93 ± 0.49 569.01 ± 0.33 568.98 ± 0.47 569.91 ± 0. 22
p 0.0125 0.0332 0.0318 0.0412 0.0215 0.0358
Sample 9.4 9.1 8.9 8.6 8.3 8.2
xs±
570.24 ± 0.10 570.67 ± 0.32 571.29 ± 0.41 571.44 ± 0.6071 572.06 ± 0.22 572.87 ± 0.49
p 0.0215 0.3025 0.0217 0.0160 0.0321 0.0245
Sample 7.8 7.6 7.4 6.9 6.7 6.5
xs±
573.52 ± 0.25 573.63 ± 0.21 574.03 ± 0.72 575.18 ± 0.43 575.36 ± 0.37 575.57 ± 0.39
p 0.0418 0.044 0.0360 0.0214 0.0124 0.0173
Sample 6.3 5.9 5.7 5.3 4.9 4.7
xs±
576.745 ± 0.36 577.15 ± 0.63 577.89 ± 0.26 578.13 ± 0.31 579.16 ± 0.11 579.58 ± 0.34
p 0.0368 0.0123 0.0451 0.0358 0.0201 0.0426
sa mpl e 4.4 4.2 3.9 3.7 3.5 3.1
xs±
579.80 ± 0.25 580.71 ± 0.23 581.07 ± 0.32 581.13 ± 0.21 581.21 ± 0.03 572.15 ± 0.16
p 0.0103 0.29 0.0412 0.0328 0.01 0.0257
Table 3. Performance comparison with different measurement methods in calibration model.
Statistics value A B C A + B A + C B + C A + B + C
RMSEP (mg/dL) 4.92 8.36 11.41 6.25 8.02 9.75 4.41
CC 0.89 0. 79 0.67 0.82 0.77 0.74 0.90
Figure 7. Relation between glucose concentration and spectra absorption.
020406080100 120 140160 180200
332
333
334
335
336
337
338
339
glucose concent rat i on
glucose abs orpti on spectra
RMSEP=4.41
CC=0 .90
mg/dL
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
559
Figure 8. Rel ation of blood glucose concentration between measurement method and NIR arrays information fusion method.
Table 4. Performance comparison of different measurement methods in measurement model.
Statistics value A B C A + B
RMSEP(mmol/L) 0.175 0.362 0.423 0.2651
CC 0.8522 0.7325 0.6355 0.79342
Statistics value A + C B + C A + B + C A + B + C based on BP
RMSEP(mmol/L) 0.31288 0.39528 0.115 0.088
CC 0.76392 0.6821 0.8938 0.94713
Table 5. Analysis result of blood glucose concentration between comparison method and NIR method.
Numerical order NIR method (mmol/L) Standard method (mmol/L) difference
1 3.3888 3 0.3888
2 3.9451 3.5 0.4451
3 3.8019 3.8 0.0019
4 4.0913 4.0 0.0913
5 4.7504 4.5 0.2504
6 4.2858 4.8 0.5142
7 4.7266 5.0 0.2734
8 5.3335 5.5 0.1665
9 6.0472 65.8 0.2472
10 6.6496 6.0 0.6496
11 6.9600 6.5 0.4600
12 7.2690 6.8 0.4690
13 7.0227 7.0 0.0227
14 7.5781 7.5 0.0781
15 8.0006 7.8 0.2006
16 8.1452 8.0 0.1452
17 8.1090 8.5 0.3910
18 8.8642 9.0 0.1358
19 9.4608 9.5 0.0392
20 10.1804 10 0.1804
RMSEP 0.3143
p 0.1728 (T-test, p > 0.05)
3 4 5678910
2
3
4
5
6
7
8
9
10
11
A ct ual value(mmol /L)
Calabtat i on val ue(mm ol /L)
RMSEP =0. 088
CC=0.9315
V al ue of c al abrat i on
x=y
B. ZENG ET AL.
Copyright © 2013 SciRes. ENG
560
property. In the later work, we hope that new multi-wa-
velength stra tegy can be used to further improve th e pre-
diction accuracy.
REFERENCES
[1] A. A. Mbaideen and M. Benaissa, “Determination of Glu-
cose Concentration from NIR Spectra Using Independent
Component Regression,” Chemometrics and Intelligent
Laboratory Systems, Vol. 105, No. 1, 2010, pp. 131-135.
http://dx.doi.org/10.1016/j.chemolab.2010.11.008
[2] M. Rohrscheib, R. Robinson and R. P. Eaton, “Non-In-
vasive Glucose Sensors and Improved Informatics—The
Future of Diabetes Management,” Diabetes, Obesity and
Metabolism, Vol. 5, No. 5, 2003, pp. 280-284.
http://dx.doi.org/10.1046/j.1463-1326.2003.00275.x
[3] H. V. L. Toal, M. Weidenmuller, A. Xhelaj and W. Man-
tele, “A Novel Approach to Non-Invasive Glucose Mea-
surement by Mid-Infrared Spectroscopy: The Combina-
tion of Quantum Cascade Lasers (QCL) and Photoacous-
tic Detection,” Vibrational Spectroscopy, Vol. 38, No. 1-
2, 2005, pp. 209-215.
http://dx.doi.org/10.1016/j.vibspec.2005.02.025
[4] M. Mueller, M. Grunze, E. H. Leiter, P. C. Reifsnyder, U.
Klueh and D. Kreutzer, “Non-Invasive Glucose Mea-
surements in Mice Using Mid-Infrared Emission Spec-
troscopy,” Sensors and Actuators B: Chemical, Vol. 142,
No. 2, 2009, pp. 502-508.
http://dx.doi.org/10.1016/j.snb.2009.08.048
[5] K. Maruo, M. Tsurugi, J. Chin, T. Ota, H. Arimoto and Y.
Yamada, “Noninvasive Blood Glucose Assay Using a
Newly Developed Near-Infrared System,” IEEE Journal
of Selected Topics in Quantum Electronics, Vol. 9, No. 2,
2003, pp. 322-399.
http://dx.doi.org/10.1109/JSTQE.2003.811283
[6] K. X. Xu, Q. J. Qiu, J. Y. Jiang and X. Y. Yang, “Non-
Invasive Glucose Sensing with Near-Infrared Spectros-
copy Enhanced by Optical Measurement Conditions Re-
production Technique,” Optics and Lasers in Engineer-
ing, Vol. 43, No. 10, 2004, pp. 1096-1106.
http://dx.doi.org/10.1016/j.optlaseng.2004.06.018
[7] W. Wang, L. F. Yan, B. W. Liu and H. Zhang, “Multi-
sensors Information Fusion With Neural Networks for
Noninvasive Blood Glucose Detection,” Lecture Notes in
Computer Science, Vol. 3498, 2005, pp. 760-766.
http://dx.doi.org/10.1007/11427469_121
[8] Y. L. Yeh, “Real -Time Measurement of Glucose Concen-
tration and Average Refractive Index Using a Laser In-
terferometer,” Optics and Lasers in Engineering, Vol. 46,
No. 9, 2008, pp. 666-670.
http://dx.doi.org/10.1016/j.optlaseng.2008.04.008
[9] Q. B. Li, L. N. Li and G. J. Zhang, “A Nonlinear Model
for Calibration of Blood Glucose Noninvasive Measure-
ment Using Near Infrared Spectroscopy,” Infrared Phys-
ics & Technology, Vol. 53, No. 5, 2010, pp. 410-417.
http://dx.doi.org/10.1016/j.infrared.2010.07.012
[10] A. Caduff, M. S. Talary, M. Mueller, F. Dewarrat, J. Kli-
sic, M. Donath, L. Heinemann and W. A. Stahel, “Non-
Invasive Glucose Monitoring in Patients with Type 1 Di-
abetes: A Multisensor System Combining Sensors for Di-
electric and Optical Characterisation of Skin,” Biosen-
sors and Bioelectronics, Vol. 24, No. 9, 2009, pp. 2778-
2784. http://dx.doi.org/10.1016/j.bios.2009.02.001