J Appl Physiol 107: 253–260, 2009.
First published March 5, 2009; doi:10.1152/japplphysiol.90960.2008.
Application of muscle biopotential measurement for sustained, noninvasive
blood glucose survey
Alexander Vol, Orna Gribova, Sylvia Berman, Yariv Siman-Tov, and Shai Efrati
Research and Development Unit, Assaf Harofeh Medical Center, Zerifin, Affiliated to Sackler Faculty of Medicine,
Tel Aviv University, Tel Aviv, Israel
Submitted 26 July 2008; accepted in final form 5 March 2009
diabetes; biopotential; bioelectricity; muscle metabolism
diabetes mellitus are confined to
repeated daily measurements of blood glucose levels to maintain a proper glycemic control. Current methods for selfmonitoring of blood glucose are invasive, painful, uncomfortable, and still allow only occasional, from time to time,
measurements. Continuous real-time surveillance would provide a helpful tool for improvement of glycemic control, thus
decreasing the incidence of hypo/hyperglycemia (11).
Most of the continuous blood glucose monitoring systems,
available or in development at present, are based on measurements of glucose content in interstitial fluid using an electrochemical enzymatic sensor. Glucose levels in interstitial fluid
are assessed either by a needle sensor inserted subcutaneously
(11, 22), or by implanting the whole device subcutaneously
(12), or by extracting interstitial fluid across the skin using the
applied electrical potential method (iontophoresis) (2, 13).
Currently, no established, continuous, noninvasive method for
blood glucose monitoring is available.
Biopotential is electric potential that is constantly generated
by living tissues, such as nerves and muscles. Currently,
biopotential measurements are amply used in medicine for
ECG, EEG, and EMG estimation. It has become common
knowledge that changes in extracellular electrolyte concentrations alter the biopotential levels. For example, hyper- or
hypokalemia have a unique effect on the ECG and EMG
recordings (15, 20). Since changes in extracellular glucose
levels would inevitably affect extracellular osmolality, extracellular pH, and membrane permeability, it seems reasonable
to expect concomitant changes in electrical spike activity and
in membrane potential levels (16, 17, 18, 29).
In myocardium, these changes in electrical spike activity
have been suggested to affect the ECG recording (24). However, there exist some major tissue-specific limitations for the
measurements of glucose-induced changes in membrane potential levels. Namely, transmembrane glucose flow in skeletal
muscles is increased, compared with smooth muscles, due to
the differences in intracellular glucose content capacity (1, 3,
8). One might propose that, to avoid these problems, changes
in membrane spike activity should be evaluated in peripheral
muscles, with higher ECG recording frequencies, the latter
varying from 0.2 Hz to 1 kHz.
The aim of the present study was, first, to verify the association between blood glucose concentration (BGC) and biopotential of peripheral muscles; and, second, to translate such
association into a simple, noninvasive technique for perpetual
monitoring of BGC.
A MAJORITY OF PATIENTS WITH
MATERIALS AND METHODS
Address for reprint requests and other correspondence: S. Efrati, R&D Dept.,
Assaf Harofeh Medical Center, Zerifin 70300, Israel (e-mail: [email protected]).
Animals. This experimental protocol received approval of the local
Committee for Animal Experimentations. The animals used in this
experiment were maintained in Assaf Harofeh Medical Center Animal
Facilities at specific pathogen-free conditions, according to the National Institutes of Health Guide for the Care and Use of Laboratory
Animals. The study included 58 male Wistar rats. The mean weight of
the rats was 339.6 ⫾ 67.9 g. In 12 rats, diabetes was induced by
intraperitoneal injection of streptozotocin (STZ), 5 mg/kg body wt, in
a single 0.5-ml bolus. Following STZ injection, the diabetic state of
the animals was evaluated by blood glucose measurements using the
ACCU-CHECK device. Glucose levels exceeding 150 mg/dl after
overnight food deprivation were chosen to serve as a cutoff point.
Because, in the present investigation, we aimed to study animals with
different blood sugar levels, all of the animals concomitantly received
the 5 mg/kg STZ injection. Following 7 days, blood sugar levels of all
of the animals exceeded the fasting 150 mg/dl cutoff point. Some of
these rats was used on the same day as a group representing the
http://www. jap.org
8750-7587/09 $8.00 Copyright © 2009 the American Physiological Society
253
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
Vol A, Gribova O, Berman S, Siman-Tov Y, Efrati S. Application of muscle biopotential measurement for sustained, noninvasive
blood glucose survey. J Appl Physiol 107: 253–260, 2009. First
published March 5, 2009; doi:10.1152/japplphysiol.90960.2008.—
Biopotential, the electric potential generated by living tissues, is
affected by changes in extracellular electrolyte and glucose concentrations. We aimed to apply correlation between blood glucose concentrations (BGC) and biopotential of peripheral muscles for noninvasive blood glucose measurement. The study included 58 Wistar rats.
In part of them, diabetes was induced by streptozotocin injection.
Group 1, comprising 19 normal and 5 diabetic rats, received glucosechallenging protocol (intraperitoneal injection of 1 g/ml glucose).
Group 2, 24 normal and 6 diabetic rats, received insulin-challenging
protocol (three 30 IU insulin injections with 15-min intervals). Four
control rats, group 3, were injected with 2-ml saline. BGC were
measured by a standard ACCU-CHEK-Sensor Meter and compared
with those estimated by biopotential sensor, further designated as
GlucoSat, placed around proximal parts of the tails of the anaesthetized animals. GlucoSat results were calculated using the following
biopotential equation: BGC(t) ⫽ k1 ⴱ F1(t) ⫹ k2 ⴱ F2(t) ⴱ k3 ⴱ
F3(t) ⫹ k4, based on an experimental model involving estimation of
pH, muscle metabolism, and tissue conductance, where t is time,
k1– k4 are coefficients, and F1–F4 are functions. Mean biopotential
system measured BGC was 181.7 ⫾ 4.3 mg/dl, not differing statistically from 187.9 ⫾ 4.3 mg/dl estimated by ACCU-CHEK. Pearson’s
correlation coefficient (r2) was 0.961 (P ⬍ 0.00001), indicating
strong, direct correlation between the results. Within the nondiabetic
group, r2 was 0.944 (P ⬍ 0.00001), while, within the diabetic group,
r2 was 0.974 (P ⬍ 0.00001). No significant, adverse skin reactions
were concomitantly observed in any experimental group. Biopotential
measurements may be used for continuous, noninvasive estimation of
changes in BGC. Further studies are needed to evaluate the applicability of this method to humans.
254
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
J Appl Physiol • VOL
attached tissue currents, with amplification of the signal to minimize
the signal-to-noise ratio.
Data analysis. The raw voltage data obtained from three electrodes
(V0, V1, V2; see Fig. 1) were averaged each 10 s within a nonoverlapping window. Within this window, the standard deviations
were calculated (stdV0, stdV1, stdV2). The same raw data were
analyzed by discrete Fourier analysis, wherein the mean spectrum
calculations were based on 15,000 data samples, with the mean values
being 15,000/20,000 ⫽ 0.75 s. These spectra were averaged each 10 s
(avSp).
The areas under the spectra were integrated in different frequency
ranges corresponding to the standard spike durations of each type of
muscle fiber (4, 10, 27): 1) 2–1,200 Hz for the smooth muscle
(Smooth); 2) 2,200 – 6,000 Hz for the skeletal muscle type 1 (Sk1);
and 3) 6,500 –10,000 Hz for the skeletal muscle type 2 (Sk2).
The following equation was used for estimation of BGC:
BGC共t兲 ⫽ k1 ⴱ F1共t兲 ⫹ k2 ⴱ F2共t兲 ⫹ k3 ⴱ F3共t兲 ⫹ k4
where t is time. In this equation, F1, F2, and F3 constituted three
distinct functions based on the above-described model, which can be
calculated from the measured parameters of the biopotential-measuring system.
1) F1 ⫽ (24/0.03) ⴱ 10[⫺1.17 ⫺ (⫺12.5 ⫺ 1,000 ⴱ V2)/59.2]. This function is used for calculation of tissue pH. It is based on the standard
glass pH electrode equation, as well as on Henderson-Hasselbalch
expression (9). Despite the fact that skin is not an exact equivalent of
the glass electrode, the basic electrochemical kinetic laws are widely
applied to the skin in electrophysiological measurements. Therefore,
one may postulate that the measured potential V2 reflects interstitial
fluid acidity. Since our experiments have been performed on young
and otherwise healthy Wistar rats, one might also assume that, under
normal blood glucose conditions, these animals should exhibit normal
physiological parameters.
2) F2 ⫽ (Sk1/Smooth)1/2 ⴱ sin[(V1 ⫺ V0)/V0]. The function is
used for estimation of hemodynamics and of muscular metabolism.
Fig. 1. Biopotential measuring system scheme. Three working electrodes
(electrode-V0, electrode-V1, electrode-V2) are applied on the skin to measure
voltage V0, V1, and V2, respectively. The ground electrode is connected to the
analog ground of the measuring system National Instruments (NI) Data
Acquisition card (DAQ 6016). One of the working electrodes, electrode-V2, is
directly connected to the measuring system NI DAQ 6016, while the two
working electrodes are, on the one side, connected electrochemically to the
AgCl reference electrode embedded in saturated KCl solution and, on the other
side, with each other. These two working electrodes are also interconnected by
a resistor (R) of 9.4 k⍀. The NI DAQ 6016 transfers the recorded data through
USB cable to a personal computer.
107 • JULY 2009 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
starting point of the experiment (moderate blood sugar levels). Within
14 days after STZ injection, blood sugar of the remaining rats
gradually increased up to 200 –250 mg/dl (high blood sugar levels).
Thus we were able to proceed with the experiment step by step, each
time including groups of animals with higher glucose levels. To use
animals of a comparable age in each study group, rats of different
ages, ranging from 4 to 6 mo, were initially included in the study.
Study protocol. All of the rats, diabetic and normoglycemic, were
anesthetized with 1.5–2.5% halothane inhaled via insufflation mask. A
small incision was made on the tail of each anesthetized animal. We
used the animal’s capillary blood, similar to most of the finger-applied
tests performed on diabetic patients, i.e., the first appearing blood drop
was allocated, and the second or third drop was directly applied onto
the ACCU-CHEK glucometer strip. Blood glucose measurements
were performed either when rat blood sugar levels were experimentally challenged by glucose injection, or when rats were similarly
challenged with insulin. Rats receiving a sham (saline) injection
served as unchallenged controls. 1) In the glucose-challenging protocol (19 normal and 5 diabetic rats), there was a single intraperitoneal
injection of 50% glucose solution in a 2-ml bolus (a total of 1 g
glucose/animal). 2) In the insulin-challenging protocol (24 normal rats
and 6 diabetic rats), there were three sequential (every 15 min)
intraperitoneal injections of 30 units of 100 IU/ml actrapid human
biosynthetic insulin. 3) In the control group protocol (4 rats), there
was a single intraperitoneal injection of 2 ml of saline.
Blood glucose was first assessed 20 min before the appropriate
intraperitoneal injection, to serve as a baseline. Subsequently, blood
glucose measurements were performed after each 5-min period within
a 1-h period. Concomitantly, biopotential values were continuously
recorded using a sampling rate of 20 kHz.
Two blood samples, 5 ␮l each, were withdrawn from each animal
at the starting point of the experimental period, and two additional
samples, at the end of the measuring period, to be used for calibration
purposes (calculation of k1, k2, k3, and k4, as detailed in the
following Data analysis section). At the beginning and at the end of
the study protocol, the skin areas directly contacting with the biopotential sensors were examined and photographed from a close distance.
Biopotential measuring system. The biopotential measuring system
consisted of a GlucoSat sensor, data-acquisition card, and personal
computer (scheme 1). The GlucoSat sensor consisted of four passive
electrodes made of two biocompatible materials (silver and platinum):
three of them were the working electrodes, and the last one was the
ground electrode. All four electrodes were placed on the rat tail skin
as follows: two of them were connected to the standard reference
electrode, 2-mm AgCl (EP-2, WPI), embedded in saturated KCl
solution. The platinum electrode was directly connected to the analog
input of the National Instruments (NI) data-acquisition card (DAQPAD-6016). The fourth electrode was used as a ground electrode and
was connected to the analog ground of the NI data-acquisition card.
The reference electrode was connected by wire to the analog input of
the NI data-acquisition card. Through the NI data-acquisition card, the
electrical signals were transferred to the computer by USB connection, with the sampling rate being 20,000 Hz.
In contrast to EEG and EMG, this measurement system demonstrated low input impedance. Thus the effect of static electricity and
piezoelectricity created by epidermal cells (17) was attenuated. Furthermore, this system proved capable of measuring and recording a
frequency range of 0 –10 kHz. Hence, the spectral range of this system
was much wider compared with any other available biopotential
medical system. As shown in scheme 1, the two working electrodes
were connected by a resistor of 9.4 k⍀. The electrodes were in touch
with two distinct surface areas: one being a part of the skin surface,
and the other of the electrolyte surface. The electrolyte surface was
embedded in a saturated KCl solution. The surface area of the skin
electrode was 10,000 times wider than the electrolyte surface area.
This configuration of the measuring system served as a shunt to the
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
255
The function is based on average spike duration of Sk1 and Smooth
(10, 27).
3) F3 ⫽ V2/V0. The function reflects tissue conductance. This
function is based on a biopotential-to-biocurrent ratio, which represents the voltage drop on the measuring resistance (the basic Ohm
law).
All of the functions used in this equation are nondimensional. The
k1– k4 coefficients have been calculated in mg/dl, using standard
linear regression analysis based on the four direct blood glucose
measurements, with two of them obtained at the beginning and the
other two at the end of the study.
Statistical analysis. Matlab version 7.0 software (MathWorks) was
used for data analysis. The data are presented as means ⫾ SD or,
where appropriate, means ⫾ SE. The differences between the results
were evaluated by Kruskal-Wallis test within ANOVA. Differences
yielding P ⬍ 0.05 were considered statistically significant. Correlations between the standard parameters were evaluated by Pearson’s
correlation coefficient (r2), and P value was subsequently calculated.
P ⬍ 0.05 was considered statistically significant.
All of the rats, except one, completed the study protocol.
One rat that belonged to the glucose-challenging protocol died
26 min after the first intraperitoneal glucose injection for an
unknown reason. No significant adverse reactions (edema,
irritation and redness, any changes in skin coloration, humidity, or secretory functions) were noticed on the skin that had
been in contact with the biopotential sensors (Fig. 1).
Scatter plot analysis results demonstrating the measurements
of blood glucose values by the ACCU-CHEK vs. GlucoSat, are
shown in Fig. 2. The plot points are comparable, not only when
applied to a whole population (Fig. 2A), but also within each
groups, i.e., when normal and diabetic rats are analyzed separately (Figs. 2, B and C, respectively). Spearman correlation
coefficient (r2), applied to estimate correlation between the
ACCU-CHEK and the biopotential results of total blood glucose measurements, was 0.961 (P ⬍ 0.00001) (Fig. 2A).
Within the nondiabetic rat group, r2 was 0.944 (P ⬍ 0.00001)
(Fig. 2B), whereas, within the diabetic rat group, r2 was 0.974
(P ⬍ 0.00001) (Fig. 2C). Table 1 demonstrates the examples of
representative correlation coefficients, indicating strong, direct
correlation between the individual measurements. The absolute
mean values of the differences between the ACCU-CHEK
measured glucose and the biopotential estimated glucose were
14.2 ⫾ 30.2 mg/dl in the nondiabetic group and 14.5 ⫾ 38.1
mg/dl in the diabetic rat group.
Figure 3 demonstrates representative examples of simultaneous ACCU-CHEK and GlucoSat evaluations of BGC, including the measurements of F1(t), F2(t), and F3(t) in two
diabetic rats during hyperglycemia (Fig. 3A) and during hypoglycemia (Fig. 3B). As earlier delineated, the following equation was used for biopotential estimation of the BGC:
BGC共t兲 ⫽ k1 ⴱ F1共t兲 ⫹ k2 ⴱ F2共t兲 ⫹ k3 ⴱ F3共t兲 ⫹ k4
where F1, F2, and F3 are functions based on the model
described in MATERIALS AND METHODS. The constant factors k1,
k2, k3, and k4 were individually calculated using the calibration function. The light gray line represents ACCU-CHEK
measurements of glucose concentrations. The dark gray line
represents continuous estimation of blood glucose via the body
biopotential (GlucoSat measurements). Calculation of the three
functions within the equation used for BGC estimation yielded
J Appl Physiol • VOL
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
RESULTS
Fig. 2. Scatchard plot for the blood glucose concentrations measured directly
by ACCU-CHEK vs. those estimated by GlucoSat. Data are shown for whole
study population (A), nondiabetic rats (B), and diabetic rats (C).
the following results: k1 value ranged from ⫺193 to ⫹138
normalized arbitrary units; k2 ranged from ⫺2,485 to ⫹2,087
normalized arbitrary units; k3 ranged from ⫺1,160 to
⫹1,872,355.4 normalized arbitrary units; and k4 ranged from
⫺15 to ⫹36 normalized arbitrary units. No significant association was found between any function parameters and animal
weight, age, or severity of diabetes. The mean value of measured biopotential voltage was 2.211 ⫾ 0.031 mV for the V0
and V1 and ⫺10.21 ⫾ 0.079 mV for V2.
107 • JULY 2009 •
www.jap.org
256
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
Table 1. Representative examples of individual rat
correlation coefficients, as obtained by the two different
methods of glucose measurement
Rat No.
Correlation Coefficient
P Value
Challenge Pattern
1
26
27
32
33
35
36
39
52
59
0.774
0.9648
0.9332
0.912
0.93
0.729
0.967
0.966
0.94
0.87
0.0407*
0.0000000004*
0.000003*
0.000087*
0.000011*
0.04*
0.000001*
0.00000001*
0.0000004*
0.00008*
Glucose
Glucose
Glucose
Glucose
Insulin
Glucose
Insulin
Insulin
Insulin
Insulin
*Significant direct correlation (P ⬍ 0.05 in each comparison).
Fig. 3. Representative examples of simultaneous ACCU-CHEK and GlucoSat evaluation of blood glucose concentration in two diabetic rats during
hyperglycemia (A and B) and during hypoglycemia (C and D). a: the light gray line represents ACCU-CHEK measurements of glucose concentrations; the dark
gray line represents continuous estimation of blood glucose via body biopotential. b, c, and d: functions F1, F2, and F3, respectively, according to which GlucoSat
evaluation of blood glucose concentration was calculated.
J Appl Physiol • VOL
107 • JULY 2009 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
Mean BGC values are demonstrated in Fig. 4. Since, for
technical reasons, the numbers of experimental rats significantly differed between the groups, we preferred to exhibit the
mean glucose values ⫾ SE, not SD, thus adjusting the variables
for the amount of the animals per group. When estimated by
ACCU-CHEK, mean blood glucose levels of the whole, undi-
vided experimental rat population were 188.0 ⫾ 4.4 mg/dl,
ranging from 19 to 600 mg/dl. When assessed by GlucoSat,
these total mean blood glucose levels were found to be 181.7 ⫾
4.3 mg/dl, ranging from 10.2 to 614 mg/dl [P ⫽ nonsignificant
(NS) compared with the respective ACCU-CHEK-obtained
values]. When evaluated separately, i.e., within each experimental group designated in MATERIALS AND METHODS, the results
were as follows. Nondiabetic rats challenged with glucose
injection demonstrated mean blood glucose estimated by
ACCU-CHEK of 235.3 ⫾ 6.9 mg/dl, ranging from 19 to 580
mg/dl, while GlucoSat measurements yielded mean glucose
value 234.6 ⫾ 6.7 mg/dl, ranging from 10.2 to 614 mg/dl (P ⫽
NS compared with the respective ACCU-CHEK-obtained values). Challenge of nondiabetic rats with insulin yielded glucose
values of 131.1 ⫾ 9.2 mg/dl when estimated by ACCU-CHEK
and 119.38 ⫾ 2.98 mg/dl when assessed by GlucoSat. Similarly, in unchallenged, control nondiabetic animals, the respective glucose values were 156.2 ⫾ 6.0 and 156.5 ⫾ 5.6 mg/dl
(Fig. 4, P ⫽ NS). For diabetic rats challenged by glucose, the
respective values were 330.2 ⫾ 22.4 and 331.4 ⫾ 22.0 mg/dl
(Fig. 4, P ⫽ NS). Mean blood glucose levels, as estimated by
ACCU-CHEK in insulin-challenged diabetic rats, were
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
Fig. 4. Mean blood glucose concentration values, as evaluated by ACCUCHEK vs. GlucoSat methods.
DISCUSSION
The aim of the present investigation was to test a novel
multiparameter model of a precise, accurate, noninvasive,
nonpainful method for sustained surveillance over blood glucose levels, based on biopotential and bioelectricity measurements. For this purpose, serial biopotential estimations were
performed on the tails of anesthetized rats using GlucoSat, the
novel biopotential and bioelectricity sensor, while concomitantly serial blood samples were procured for standard direct
biochemical measurements of rat blood glucose. Comparison
of the data obtained using the proposed novel GlucoSat sensor
system and those obtained by a standard method demonstrated
no statistical differences between the two series of results, thus
validating the high precision and accuracy of the GlucoSatbased method of BGC. The latter was true for diabetic rats, as
well as for normoglycemic control animals, whether they are
challenged or not with insulin or with glucose injection before
glucose assessments. Furthermore, a strong, direct correlation
was observed between total BGC values obtained using a
standard glucose measuring device and GlucoSat-derived BGC
values. This correlation, persisting when the diabetic and
nondiabetic rat populations were analyzed separately, was also
distinctly manifested when the pairs of the results were individually analyzed.
The two main participants involved in regulation of glucose
metabolism are skeletal muscle cells and adipose tissue. Metabolic activity of skeletal muscles is accompanied by electrical
activity, which can be externally detected and monitored. This
measurable parameter, biopotential, is created by two different
components: one is based on the Nernst equation (concentration gradient), and the second represents a constant involvement of electrically active cells (nerves and muscles) (23).
ECG and EMG can be applied for estimation of muscle
activity, whereas EEG is used for measurement of neuronal
J Appl Physiol • VOL
activity in the brain (14). Any change in extracellular glucose
concentration would affect both Nernst potential and muscle
electrical activity (5, 15–18, 20, 24, 29).
In myocardium, these changes in electrical spike activity
have been suggested to affect the ECG recording (24). To the
best of our knowledge, this is the first study testing the
possibility of a practical application of multiparametric relationship between the body biopotential and the blood glucose
levels. The operative mode of the proposed biopotential-measuring sensor, GlucoSat, is based on simultaneous assessment
of both Nernst potential and muscle electrical activity, in a
wide range of frequencies and a minimal signal-to-noise ratio.
One would suggest at least three main possible reasons for such
a favorable outcome.
First, the proposed electrochemical unit is, as described in
MATERIALS AND METHODS, composed of four separate electrodes;
two of them are embedded in saturated KCl solution. As
detailed in the legend to scheme 1, the working electrode
surface contact area with the skin is 10,000 times higher than
the electrochemical unit contact area. Such a high ratio between the electrode contact surfaces significantly amplifies the
signal.
Second, the measurement noise is also reduced because the
actual parameter measured on the surface body area is the tissue
current and not the skin voltage potential, which can be affected
by both static electricity and piezoelectricity (voltage potential
created by tissue deformation).
Finally, the ground electrode, when placed on the skin,
neutralizes static electricity, thus contributing to the improvement of signal-to-noise ratio.
Changes in blood glucose are accompanied by altered muscle metabolism and glucose uptake (1, 3, 8, 26). However, the
relationship between blood glucose, muscle activity, and body
biopotential would be inevitably influenced by a number of
extraneous, subject-specific factors. For example, insulin sensitivity may differ in obese and nonobese persons, and/or in
individuals suffering from type 1 and type 2 diabetes. The
conductive milieu between the muscles and the biopotential
sensor also differs from one individual to another, depending
on epidermis thickness, skin humidity, composition of fats, etc.
To overcome these interindividual differences, we used a
multiparametric measurement model with calibration function.
Four parameters (k1, k2, k3, and k4) were separately calculated. No association between weight, age, or diabetes stage
and any of the calibration parameters has been evident.
The mean value of measured biopotential voltage was, as
already mentioned, 2.211 ⫾ 0.031 mV for the V0 and V1 and
⫺10.21 ⫾ 0.079 mV for V2. These values corresponded to a
measured system current of 0.01 ␮A for V0 and V1, and ⬍0.01
nA for V2. Such currents of the biopotential system are about
10 times lower than the currents used in standard ECG devices,
which use a current of 0.1– 0.5 ␮A, indicating the safety of the
proposed biopotential system.
The existing clinically available blood glucose monitoring
systems are based on invasive electrochemical sensors (11, 22).
Several other methods for continuous blood glucose monitoring, currently under development, are based on physiologically
questionable electromagnetic, acoustic, or impedance technologies (7, 25, 28). As opposed to any of the above-mentioned
methods, the proposed GlucoSat sensor-based method is completely noninvasive and physiologically noninterfering. Unlike
107 • JULY 2009 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
115.2 ⫾ 7.5 mg/dl, ranging from 42 to 600 mg/dl, while
GlucoSat measurements in these animals yielded mean blood
glucose of 105.4 ⫾ 6.4 mg/dl, ranging from 25.2 to 583.0
mg/dl (P ⫽ NS). In saline-injected diabetic rats, these values
were 353.4 ⫾ 19.5 and 336.4 ⫾ 17.8 mg/dl, respectively (Fig.
4, P ⫽ NS).
257
258
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
J Appl Physiol • VOL
application of the methods for trend analyses, learning algorithms, and self-adjustment for each individual patient.
In this respect, even at such a relatively early stage of
research, the proposed method appears very promising as a
novel noninvasive technique for sustained monitoring of blood
glucose levels. Given the real and pressing clinical need for a
robust, noninvasive, and ideally continuous glucose-sensing
technology, our data suggest that the proposed method of
muscle biopotential measurement for sustained, noninvasive
blood glucose survey calls for further, more extensive, investigations.
APPENDIX
The present study has been designed based on a number of
postulates. First, any living organism should be considered an open
thermodynamic system. The very existence of such a system is based
on a combination of stability and variability. Speaking in terms of
thermodynamics, such stationary state, homeostasis, would mean that
the system would constantly be at a minimum of free energy. With
respect to a living organism, homeostasis would mean that, for its
basic metabolism, a given organism would consume a minimal
amount of energy and, consequently, also waste a minimal amount of
energy.
According to the basic thermodynamic laws, any perturbation
caused by external forces (for example, changes in blood glucose
levels) would, by altering the energy balance and decreasing the
degrees of freedom, also decrease the free energy of a living organism.
Second, the blood glucose monitoring must not be based on a single
correlation function. Rather, a combination of vital parameters should
be employed. Moreover, the choice of such a combination has to be
based on taking into account both periodical and nonperiodical
changes occurring in a living organism, such as periodical insulin
“blow-outs” into blood, daily and month cycles of hormonal balance,
emotional and physical loadings, etc.
The concept proposed in the present study is based on a multiparametric model using the following parameters reflecting glucose
metabolism.
1) The first is changes of the tissue fluid acidity. It is well
established that increased tissue fluid acidity increases blood oxygen
consumption from 25 to 70 – 80%. It is also known that hyperglycemia
is always accompanied by increased blood and lymph acidity and
augmented osmolality, which may cause acidosis or hyperosmotic
nonketone coma.
2) Skeletal muscle activity under the rest conditions reflects glycogen formation/release. Muscle and liver glycogen metabolism is an
important part of blood glucose control.
Fig. 5. Scheme of the typical dependence electrolyte pH on ion-selective
electrode.
107 • JULY 2009 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
fasting blood glucose level estimations, fluctuations in daily
blood glucose can be detected only by a permanent glucose
monitoring system, because, even with frequent capillary testing, patients will struggle to detect all of the fluctuations in
daily blood glucose, required for guided therapy. For risk of
long-term complications, although not yet proven, one might
also anticipate that a robust, continuous glucose monitoring
system might correlate better than HbA1c levels with urinary
markers of oxidative stress, markers of endothelial dysfunction,
and symptoms of micro- and macrovascular disease (6, 21).
The present study has several limitations. 1) The experiment
was performed on anesthetized, immobilized rats. Application
of the proposed technology to actively working muscle calls
for further investigations. 2) Proposed sensor system is based
on individual calibration, and currently no data are available to
indicate for how long this calibration remains valid. 3) With
respect to future studies, there might appear some yet unaccounted physiological parameter(s) capable of affecting the
individual calibration. 4) The proposed technology is still to be
improved: at present, the time interval between the two separated calibration points must be considered a twilight zone with
respect to the true blood sugar values, since one cannot be sure
what is going on at this time with the biopotential signal.
However, during the time between the two calibrations, blood
sugar measurements by a conventional glucometers should be
performed in parallel. The issue of inconvenience of two-time
point calibration should be amended in the future, mainly by
algorithm improvement and application of the methods for
trend analyses, learning algorithms, and self-adjustment for
each individual patient.
The current technology is not intended to entirely replace
conventional methods of blood sugar determination. Rather,
sustained, noninvasive blood glucose survey by GlucoSat will
be supported and complemented by conventional methods of
invasive glucose measurements. Accuracy of the tests performed by home glucometers, such as Accu-Check or FreeStyle, is, according to the manufacturers, ⫾10%. A researcher
performing a series of single measurements with home glucometers in short time intervals (e.g., each 5 min) knows that
deviation between the results is always much higher. This is
due to significant physiological oscillations of blood glucose
levels in capillary blood, the amplitude of which might, in a
normal state, exceed 50 mg 䡠dl⫺1 䡠s⫺1. In other words, standard
procedures for measuring blood glucose levels by single invasive tests will never cover for imprecision brought about by
physiological factors, simply because contribution of the latter
to the measuring error will always overrate the accuracy of
internal calibration. More so, it is unrealistic for a diabetic
patient to routinely perform numerous single measurements in
very short time intervals by means of a home glucometer.
Thus, despite any calibration precision of the measuring device, it is not surprising that the effect of timely hypo- or
hyperglycemia alert is hardly ever achieved. For this reason, it
has become a trend among the clinicians to look for a method(s) that could ascertain the incessant online surveillance over
blood glucose rather than rely on numerous discrete blood
sugar tests. The method herein proposed does provide such
surveillance based on physiological parameters. However, the
proposed technology is still to be improved, primarily in terms
of introduction of a single-point calibration method. This will
be achieved in the future by algorithm improvement and
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
3) Smooth muscle activity under the rest conditions reflects blood
flow and blood viscosity. Within the normal-to-low range of blood
glucose levels, blood flow increase is inversely correlated with a
decrease in blood glucose levels (thus providing the energy for the
limb/tail basic metabolism). However, “normal,” “low,” or “high”
blood sugar levels are, by definition, different for different subjects.
When the sugar level reaches a certain high point, which is individual
for a given subject, smooth muscle activity changes its course and
starts to increase in a direct correlation with the increase in blood
glucose levels. This is a different thermodynamic state that provides
energy for much higher metabolism of the muscle and adipose tissues,
the main participants in blood glucose lever regulation.
4) Concomitantly, tissue conductivity increases in a direct correlation with blood glucose levels, due to increasing membrane permeability.
Detailed Description of the Used Functions
pH ⫽ 6.1 ⫹ log 共关HCO3⫺兴/关CO2兴兲
Carbonic acid-bicarbonate is an important extracellular buffering
system in mammals, which is partially responsible for maintaining the
pH of blood at 7.4.
The value (⫺1.17) is the factor for conversion of the results
obtained by our reference electrode used in the experiment, to the
results obtained by a standard hydrogen electrode.
The value (⫺12.5 mV) corresponds to a mean value of normal
blood acidity in rats with physiologically normal blood glucose level.
The value 1,000 ⴱ V2 is a conversion factor from volts to
millivolts.
Interference effects are commonly described by the semi-empirical
Nicolsky-Eisenman equation, the extension of the Nernst equation:
E ⫽ E0 ⫹
冋
RT
ln ai ⫹
z iF
兺共k a
zi/zj
ij j
j
册
兲
where E is the emf (V); E0 is the standard electrode potential (V); z is
the ionic valency including the sign, a is the activity, i is the ion of
interest, j is the interfering ions, kij is the selectivity coefficient, R is
the universal gas constant [R ⫽ 8.314472(15) J 䡠 K⫺1 䡠 mol⫺1], T is the
absolute temperature, and F is the Faraday constant (the number of
coulombs/mole of electrons: F ⫽ 9.64853399(24) ⫻ 104 C/mol).
The smaller the selectivity coefficient, the less is the interference
by j.
To see the interfering effect of Na⫹ to a pH electrode, the following
equation is used:
E ⫽ E0 ⫹
RT
ln 共aH⫹ ⫹ kH⫹,Na⫹aNa⫹兲
F
Scheme of the typical dependence electrolyte pH on ion-selective
electrode is shown in Fig. 5: the voltage produced by the probe
(⫺0.059 volt/pH in basic solutions, ⫹0.059 volt/pH in acid solutions)
into pH units.
F2 ⫽ (Sk1/Smooth)1/2 ⴱ sin[(V1 ⫺ V0)/V0]. Sk1 is the integral of
power spectrum of measurements obtained from electrode V0 in the
frequencies range 2,200 – 6000 Hz.
Smooth is the integral of power spectrum of measurements obtained from electrode V0 in the frequencies range 2–1,200 Hz.
J Appl Physiol • VOL
Both skeletal and smooth muscle functions are by origin square
functions, each representing a square value of amplitude. Therefore, in
the F2 formula the Sk1-to-Smooth ratio is calculated as a square root.
Sin[(V1 ⫺ V0)/V0] reflects a pulse wave activity, since glucose
and oxygen are released into intestinal fluid in a pulse manner
dependent on cardiovascular activity of the organism.
F3 ⫽ V2/V0. This function reflects tissue conductance. F3 is based
on a biopotential-to-biocurrent ratio, which represents the measuring
resistance voltage drop (the basic Ohm law).
Electrode V0 reflects biocurrent, whereas V2 reflects biopotential
(based on its configuration, as can be seen from the schematic diagram
of the measuring system).
ACKNOWLEDGMENTS
We acknowledge Dr. Leonid Levin for invaluable contributions to the
development of the mathematical model and algorithm employed in the present
study.
REFERENCES
1. Allen TJ, Hardin CD. Influence of glycogen storage on vascular smooth
muscle metabolism. Am J Physiol Heart Circ Physiol 278: H1993–H2002,
2000.
2. [Anon]. Accuracy of the GlucoWatch G2 biographer and the continuous
glucose monitoring system during hypoglycemia. Experience of the Diabetes Research in Children Network. Diabetes Care 27: 722–726, 2004.
3. Arkinstall MJ, Bruce CR, Clark SA, Rickards CA, Burke LM,
Hawley JA. Regulation of fuel metabolism by preexercise muscle glycogen content and exercise intensity. J Appl Physiol 97: 2275–2283, 2004.
4. Athenstaedt H, Claussen H, Schaper D. Epidermis of human skin:
pyroelectric and piezoelectric sensor layer. Science 216: 1018 –1020,
1982.
5. Bonen A, McDermott JC, Tan MH. Glycogenesis and glyconeogenesis
in skeletal muscle: effects of pH and hormones. Am J Physiol Endocrinol
Metab 258: E693–E700, 1990.
6. Brownlee M, Hirsch IB. Glycemic variability: a hemoglobin A1c-independent risk factor for diabetic complications. JAMA 295: 1707–1708,
2006.
7. Caduff A, Hirt E, Feldman Y, Ali Z, Heinemann L. First human
experiments with a novel non-invasive, non-optical continuous glucose
monitoring system. Biosens Bioelectron 19: 209 –217, 2003.
8. Casey A, Short AH, Hultman E, Greenhaff PL. Glycogen resynthesis in
human muscle fibre types following exercise-induced glycogen depletion.
J Physiol 483: 265–271, 1995.
9. Emde C. Electrochemical aspects of pH electrodes. Dig Dis 1, Suppl:
18 –22, 1990.
10. Farina D, Arendt-Nielsen L, Graven-Nielsen T. Spike-triggered average
torque and muscle fiber conduction velocity of low-threshold motor units
following submaximal endurance contractions. J Appl Physiol 98: 1495–
1502, 2005.
11. Garg S, Zisser H, Schwartz S, Bailey T, Kaplan R, Ellis S, Jovanovic
L. Improvement in glycemic excursions with a transcutaneous, real-time
continuous glucose sensor: a randomized controlled trial. Diabetes Care
29: 44 –50, 2006.
12. Garg S, Schwartz S, Edelman SV. Improved glucose excursions using an
implantable real-time continuous glucose sensor in adults with type 1
diabetes. Diabetes Care 27: 734 –738, 2004.
13. Garg S, Potts RO, Ackerman NR, Fermi SJ, Tamada JA, Chase HP.
Correlation of fingerstick blood glucose measurements with GlucoWatch
biographer glucose results in young subjects with type 1 diabetes. Diabetes Care 22: 1708 –1714, 1999.
14. Grob M, Mouginot D. Heterogeneous chloride homeostasis and GABA
responses in the median preoptic nucleus of the rat. J Physiol 569:
885–901, 2005.
15. Lacourciere Y, Cote C, Lefebvre J, Poirier L, Dumont M. Identifying
which treated hypertensive patients without known coronary artery disease
should be tested for the presence of myocardial ischemia by perfusion
imaging. J Clin Hypertens (Greenwich) 9: 921–928, 2007.
16. Levin BE, Kang L, Sanders NM, Dunn-Meynell A. Neuronal glucosensing: what do we know after 50 years? Diabetes 53: 2521–2528, 2004.
17. Lisin R, Ginzburg BZ, Schlesinger M, Feldman Y. Time domain
dielectric spectroscopy study of human cells. I. Erythrocytes and ghosts.
Biochim Biophys Acta 1280: 34 – 40, 1996.
107 • JULY 2009 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
F1 ⫽ (24/0.03) ⴱ 10[⫺1.17 ⫺ (⫺12.5 ⫺ 1,000 ⴱ V2)/59.2]. This function is
used for calculation of tissue pH. It is based on the standard glass pH
electrode equation, as well as on Henderson-Hasselbalch expression.
The concentration of CO2 in blood is commonly addressed in terms
of its partial pressure, e.g., PCO2 ⫽ 40 Torr, which corresponds to 24
meq/l at 37°C.
Partial pressure is converted to concentration using the conversion
factor 0.03 meq 䡠 l⫺1 䡠 mmHg⫺1 at 37°C, so that the corresponding
Henderson-Hasselbalch expression is now:
259
260
NONINVASIVE BLOOD GLUCOSE ASSESSMENT
18. Lyakhov V, Vol A, Raikhlin-Eisenkraft B. Adaptive activity of neural
structures–a thermodynamic approach. Med Hypotheses 59: 289 –296, 2002.
19. Minehira K, Bettschart V, Vidal H, Vega N, Di Vetta V, Rey V,
Schneiter P, Tappy L. Effect of carbohydrate overfeeding on whole
body and adipose tissue metabolism in humans. Obes Res 11: 1096 –
1103, 2003.
20. Miura J, Uchigata Y, Sato A, Matsunada R, Fujito T, Borgeld H,
Tanaka M, Babazono T, Takahashi C, Iwamoto Y. An IDDM
patient who complained of chest oppression with ischemic changes on
ECG in insulin-induced hypoglycemia. Diabetes Res Clin Pract 39:
31–37, 1998.
21. Monnier L, Mas E, Ginet C, Michel F, Villon L, Cristol JP, Colette C.
Activation of oxidative stress by acute glucose fluctuations compared with
sustained chronic hyperglycemia in patients with type 2 diabetes. JAMA
295: 1681–1687, 2006.
22. Monsod TP, Flanagan DE, Rife F, Saenz R, Caprio S, Sherwin RS,
Tamborlane WV. Do sensor glucose levels accurately predict plasma
glucose concentrations during hypoglycemia and hyperinsulinemia? Diabetes Care 25: 889 – 893, 2002.
23. Navarro X, Krueger TB, Lago N, Micera S, Stieglitz T, Dario P. A
critical review of interfaces with the peripheral nervous system for the
control of neuroprostheses and hybrid bionic systems. J Peripher Nerv
Syst 10: 229 –258, 2005.
24. Ostrander LD Jr, Weinstein BJ. Electrocardiographic changes after
glucose ingestion. Circulation 30: 67–76, 1964.
25. Park JH, Kim CS, Choi BC, Ham KY. The correlation of the complex
dielectric constant and blood glucose at low frequency. Biosens Bioelectron 19: 321–324, 2003.
26. Rose AJ, Richter EA. Skeletal muscle glucose uptake during exercise:
how is it regulated? Physiology (Bethesda) 20: 260 –270, 2005.
27. Steedman WM. Micro-electrode studies on mammalian vascular muscle.
J Physiol 186: 382– 400, 1966.
28. Trajanoski Z, Brunner GA, Schaupp L, Ellmerer M, Wach P, Pieber
TR, Kotanko P, Skrabal F. Open-flow microperfusion of subcutaneous
adipose tissue for on-line continuous ex vivo measurement of glucose
concentration. Diabetes Care 20: 1114 –1121, 1997.
29. Worthley LI, Guerin M, Pain RW. For calculating osmolality, the
simplest formula is the best. Anaesth Intensive Care 15: 199 –202, 1987.
Downloaded from http://jap.physiology.org/ by 10.220.33.1 on June 17, 2017
J Appl Physiol • VOL
107 • JULY 2009 •
www.jap.org