lable at ScienceDirect

Current Applied Physics 14 (2014) 563e569

Contents lists avai

Current Applied Physics

journal homepage: www.elsevier .com/locate/cap

In vitro monitoring of goat-blood glycemia with a microwavebiosensor

Seungwan Kim a, Jongchel Kim a, Kyoungchul Kim a, Jung-Ha Lee b, Arsen Babajanyan c,Barry Friedman d, Kiejin Lee a,*

aDepartment of Physics and Basic Science Institute for Cell Damage Control, Sogang University, Seoul 121-742, Republic of KoreabDepartment of Life Science and Basic Science Institute for Cell Damage Control, Sogang University, Seoul 121-742, Republic of KoreacDepartment of Radiophysics, Yerevan State University, Manoogian 1, Yerevan 0025, ArmeniadDepartment of Physics, Sam Houston State University, Huntsville, TX 77341, United States

a r t i c l e i n f o

Article history:Received 13 June 2013Received in revised form7 January 2014Accepted 20 January 2014Available online 31 January 2014

Keywords:In vitroGoat-bloodD-GlucoseMicrowave cavity sensor

* Corresponding author. Tel.: þ82 27058425.E-mail address: klee@sogang.ac.kr (K. Lee).

1567-1739/$ e see front matter � 2014 Elsevier B.V.http://dx.doi.org/10.1016/j.cap.2014.01.011

a b s t r a c t

We used an electromagnetic microwave cavity sensor for real time measurement of the glycemia in goat-blood for three animals. We could determine the concentration of D-glucose in blood in the range of 90e550 mg/dl at the resonance frequency near 4.76 GHz with a bandwidth of 300 MHz. The change inmicrowave reflection coefficient S11 (due to the variation of D-glucose concentration in blood) was about16.33 dB, 23.92 dB, 7.66 dB and resonance frequency shift was about 21.78 MHz, 36.29 MHz, 20.77 MHz,respectively, for the three different samples. The in vitro results show the measured signal-to-noise ratioof about 32 dB, and the minimum detectable signal level of about 0.025 dB/(mg/dl). The results clearlyshow the sensitivity and usefulness of this microwave sensor for these types of biological investigations.This proposed system provides a unique approach for real contactless glucose monitoring and, it mayserve as a bloodless glucometer for the calibration of different glucose levels.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

Noninvasive diagnosis and monitoring of diabetes has attractedtremendous attention in the past two decades because of theemergence of diabetes as a major epidemic, especially whenassociated with the increased overall obesity of the population [1].The noninvasive measurement of blood glucose in particular is ofgreat individual and economic importance because of the largepopulation of diabetics who require regular and accurate infor-mation regarding their blood glucose concentrations. Noninvasivedetermination of glucose will promote more frequent testing,allows tighter control of diabetes, and delays the onset of diabetescomplications and their associated health care costs. Recently newglucose measurement methods have been developed, exploitingthe effect of glucose on erythrocyte scattering, new photoacousticphenomenon, optical coherence tomography, thermo-opticalstudies on human skin, Raman spectroscopy studies, fluorescencemeasurements, and use of photonic crystals [2e13]. In addition tooptical methods, in vivo electrical impedance results have beenreported [14].

All rights reserved.

Some of these methods measure intrinsic properties of glucose;others deal with its effect on tissue or blood properties. An inter-esting approach for the analysis of the glucose-induced dielectricproperty variations is the use of electromagnetic sensors thusavoiding any direct contact with the medium to be investigated[3,15,16]. The electromagnetic coupling technique based on theimpedance measurement stresses the importance of dielectric pa-rameters of blood. The relation between dielectric parameters andthe glucose concentration in blood has been developed.

Microwave instruments have been used to analyze the dielectricproperty variations and have the ability to make nondestructivemeasurement of parameters inside a volume where direct contactwith the sample cannot be achieved [17e19]. This property of mi-crowave sensors provides the prospect for suitable noninvasivemeasurements of biological tissues. An important issue in nonin-vasive glucose measurement is calibration. One calibrationapproach is to study glucose concentration ranges in vitromeasured signals over the in vivo range to be monitored. Analyticalstudies for in vitro measurement sensors include study of deter-mination of factors affecting accuracy, precision and interferencestudies in different blood samples. Analytical calibration of a non-invasive glucose sensor implies determination of glucose concen-tration from a standard curve generated by using a theoreticalmodel. The aim of this study is to outline the design of a microwave

S. Kim et al. / Current Applied Physics 14 (2014) 563e569564

sensor that could be applied for noninvasive detection and moni-toring of glycemia.

Using microwave signal analysis, it is possible to detect theglucose concentrations and levels due to indirect measuring of thedielectric permittivity or direct measuring of the microwavereflection coefficient S11. The estimated parameter is the complexdielectric permittivity 3, which is one of the main material char-acteristics of the medium. The basic principle for measurement ofglucose concentration at microwave frequencies is described in ourrecent works [20,21].

In this paper, we monitor the glycemia in three different goat-bloods by measuring the microwave reflection coefficient S11 us-ing a microwave cavity sensor. Along with the measurements,simulations were done for the three different D-glucose concen-trations in glucose aqueous solution. The change of the D-glucoseconcentration is directly related to the change in the reflectioncoefficient due to an electromagnetic interaction between theresonator and blood sample.

We modeled the microwave cavity sensor for glucose concen-trationmeasurements by using HFSS 5.4 simulation softwarewhichaccurately predicts response of the microwave signal from theresonator with samples.

2. Experimental

2.1. Structure and model of sensor

The fabricated glucose sensor based on the design of a wave-guide cavity resonator is shown in Fig. 1(a) and a detaileddescription is given in Ref. [21]. To increase the sensitivity andselectivity of the sensor we modified the microwave source sizeand position in the resonator according to the computer analysis ofthe Q-factor of the resonant cavity. The outer sizes of the cavitystayed the same; just the size and position of the microwave sourceand the diameter of the tube for the sample was changed. First, weincreased the tube diameter from 6.8 mm to 14.4 mm and thesource diameter decreased from 5.3 mm to 4 mm. Secondly,the source position changed from the front side (same side as thesample entrance) to the back side of the cavity. Ah same time thesource position in the xy plane shifted with respect to the cavitywall with the l/3 and 2l/3 proportions (l/2 and l/2 proportions forthe sensor in Ref. [21]). As a result, the resonator Q-factor increasedfrom 110 to 648 for new the configuration. This growth of Q-factorof about 6 times increased the sensitivity and selectivity of themeasurement. Now the sensor is able to determine smaller changesin the electromagnetic characteristics of the sample. Note to

Fig. 1. (a) The schematic view of experimental system for the microwave cavity sensor withshown.

improve the sensor Q-factor we used computer simulation due tothe complexity of an analytical analysis of the sensor structure. Todetermine the glucose concentration changes, we measured themicrowave reflection coefficient S11 and the resonant frequencyshift of the microwave resonator using a network analyzer (Agilent8753ES). Subsequent changes in electromagnetic coupling betweenthe cavity and the sample cause changes in the magnitude of S11and shift the resonant frequency. This allows characterization of theelectromagnetic properties (dielectric, conducting, volumetric etc.)of the sample. At resonance, the mode we used was TE011, which isthe dominant mode, and the sensitivity of the device is highest forthis operating mode. The resonant frequency and microwavereflection coefficient minima for DI water were f ¼ 4.76 GHz andS11 ¼ �12.17 dB, respectively. Fig. 1(b) shows the sensor model as itappears in the HP HFSS interface. Note that the microwaves used inthe model have a frequency of order 4.76 GHz and a wavelength oforder 6.7 cm; hence a boundary sphere of 13 cm was used in themodel. The parameters of the coax, not identical to the experi-mental values, were selected to give the transmission line animpedance of 50 U. All metallic surfaces in the model were chosento be perfect conductors (Al); it is possible to treat the metallicboundaries more realistically, a wave impedance boundary condi-tion with the relevant conductivity. However, a limited amount ofnumerical “experimentation” indicated for the cavity the perfectconducting boundary condition is adequate.

2.2. Preparation and biological analysis of samples

The initial test samples based on goat’s blood have the glycemiaof 140 mg/dl, 92 mg/dl, and 110 mg/dl in the blood for the threedifferent animals. These samples will be defined as sample A,sample B, and sample C, respectively. By adding D-glucose to theinitial samples, 5 samples were prepared with 100 mg/dl glycemiadifference for each case. The goat-blood was mixed with sodiumcitrate (C6H5Na3O7) to avoid fast coagulation and then stored in arefrigerator with storage limitation of about 3 days. All sampleswere analyzed by the veterinary services. Table 1 shows the resultsof biological tests for the three blood samples.

3. Theory

3.1. Microwave reflection coefficient

The microwave reflection principle can be explained by theplane-wave solution model. An interpretation of the reflection

sample. (b) Model as it appears in the HP HFSS interface. The surrounding sphere is not

Table 1The biological tests results for the three blood samples.

Test flag Description Sample A Sample B Sample C

ALB2 Albumin (g/dl) 3 3 3.8ALP2S Alkaline phosphatase (U/l) 409 233 256ALTL Alanine transaminase (U/l) 20 8 14AMYL2 Armylase (U/l) 23 30 38ASTL Aspartate transaminase (U/l) 53 59 85CA Calcium (mg/dl) 8.9 5.9 7.7CHO2I Cholesterol (mg/dl) 73 55 118CREJ2 Creatine (mg/dl) 0.9 1.5 1.1GGTS2 Gamma-Glutamyl

transpeptidase (U/l)43.8 49 110.5

GLU2 Glucose (mg/dl) 140 92 110PHOS2 Phosphorus (mg/dl) 8.7 5.5 9.2TP2M Total protein (g/dl) 5 5.1 5.6UREL Blood urea nitrogen (mg/dl) 10 5 12BILTS Bilirubin (mg/dl) 0.1 0.1 0.1

The data for glucose is bolded

S. Kim et al. / Current Applied Physics 14 (2014) 563e569 565

coefficient S11 magnitude in terms ofmedium parameters is writtenas [22]:

S11 ¼ 20 log����jZinj � Z0jZinj þ Z0

����; (1)

where S11 is the reflection coefficient, Z0 is the characteristicimpedance of the probe tip and Zin is the complex input impedanceof the air/tube/glucose/tube/air/Al system and it can be estimatedas [19]:

Zin ¼ ZtZgtAl þ jZt tanðktttÞZt þ jZgtAl tanðktttÞ

yjZaka

�2�Vg=Sg

�þ tt�� k2att

�Vg=Sg

�23g

1� k2a�Vg=Sg

�3t��Vg=Sg

�þ tt�� k2att

�Vg=Sg

�3g; (2)

where Zt and Za (Za¼ 377U) are the characteristic impedance of thesilicon tube and air, kt (kt ¼ 154 m�1 at 4.76 GHz) and ka(ka ¼ 96 m�1 at 4.76 GHz) are the wave vector in the silicon tubeand in air, respectively. tt, Vg, and Sg are the wall thickness(kt¼ 0.3mm), volume (0.44 cm3) and surface area of base (2 cm2) ofthe cylindrical silicon tube, respectively. 3g is the dielectricpermittivity of the glucose solution, and ZgtAl is the compleximpedance of the glucose/tube/Al system and is given by theexpression:

ZgtAl ¼ ZgZtAl þ jZg tan

�kg�Vg=Sg

��Zg þ jZtAl tan

�kg�Vg=Sg

��y jZaka�tt þ Vg=Sg

�1� k2att

�Vg=Sg

�3g; (3)

where Zg and kg are the complex impedance and wave vector ofglucose solution, respectively, and ZtAl is the complex impedance ofthe silicon tube/Al system (this is similar to a dielectric on perfectmetal system [23]) and it can be expressed as:

ZtAl ¼ jZtktttyjZakatt : (4)

Here 3t and mt are the relative dielectric permittivity and therelative magnetic permeability (mt z 1) of the silicon tube. Notethat the changes of plane waves due to passing the space betweenthe probe and sample (air environment inside cavity sensor) wereneglected.

The dependence of dielectric permittivity of solute glucoseconcentration is expressed with the molar increment d and givenby Ref. [20]:

3gðuÞ ¼�

300þcd0

�� j�

3000þcd00

� ¼ �300� j 3

000�þc

�d0 � jd00

�; (5)

where c is the concentration of glucose, d ¼ d0�jd00 is the increase inpermittivity when the glucose concentration is raised by 1 unit:d0 ¼ 0.0577 (mg/dl)�1 and d00 ¼ 0.0015 (mg/dl)�1 and30 ¼ 300 � j 3000 ¼ 74:37� j16:518 is the complex permittivity of de-ionized (DI) water at 4.76 GHz [24].

3.2. Resonant frequency shift

Inserting a goat-blood sample in the resonator can be consid-ered as a perturbation source in dielectric resonator cavity. When acylindrical shape sample is placed in the resonator, the resonantfrequency fr of the resonator is changed. The fractional change inresonant frequency can be expressed, using perturbation theory[19,22] as:

����Dff0���� ¼

Z

VþDV

�½mþ Dm�jH0j2 � ½ 3þ D 3�jE0j2

�dv

Z

V

�mjH0j2 þ 3jE0j2

�dv

; (6)

where 3, m and V are the unperturbed dielectric permittivity, sus-ceptibility and volume, E0 and H0 are the unperturbed electric and

magnetic field, D 3, Dm and DV are changes in the dielectricpermeability, susceptibility and volume, respectively. This relationindicates that the resonant frequency changes Df/f0 depending onwhere the perturbation is located in the resonant cavity. Thus, thefractional change in resonant frequency depends on the interactionbetween the resonator and the sample. Then, the fractional changein resonant frequency can be written as [19,22]:

����Dff0���� ¼ 2cd0Vg

3dVd; (7)

where Vd and 3d is the volume (Vd ¼ 148 cm3) and dielectricpermittivity (air: 3d ¼ 1) of the resonator.

4. Results and discussion

4.1. Measurement of concentration

Fig. 2 shows measured microwave reflection coefficient S11profiles for three goat-blood samples; (a) sample A, (b) sample B,and (c) sample C at the resonant frequency of about 4.76 GHz. Asthe frequency of the microwave source is swept, each trace in Fig. 2exhibited aminimum corresponding to the standingwavemode forthat particular resonatoresample combination. The matchedresonance curve for the DI water has a minimum level of�12.17 dB,which is the reference level for the microwave reflection coefficientS11 in our measurements. As the D-glucose concentration increased,

Fig. 2. Measured microwave reflection coefficient S11 profiles for three goat-blood samples: (a) A, (b) B, and (c) C at the resonant frequency of about 4.76 GHz.

S. Kim et al. / Current Applied Physics 14 (2014) 563e569566

the dielectric permittivity of sample increased and both thereflection coefficient S11 (left axis) and the resonant frequency shiftDf/f0 (right axis) decreased as shown in Fig. 3(a). Note that themicrowave reflection coefficient S11 (and the resonant frequencyshift Df/f0 also) is directly related to the sample electromagneticparameters (dielectric permittivity, electric conductivity, magneticpermeability, etc.). Here, this relation thus the electric and mag-netic field distribution behavior shows the reverse dependence onglucose concentration due to the “two chambers” construction andmode structure of the resonant cavity. It is also notable, that thechange behaviors of S11 and Df/f0 (i.e. measurements sensitivity)were different for different animals even for the similar glycemialevel as follows from Fig. 3(a). We suggest this is due to differentstructure of blood for the three animals. From an analysis of Table 1,we conclude that this effect (i.e. sensitivity of measurements) maybe caused by the concentration of the following three componentsin the blood: Cholesterol, Phosphorus, and urea Nitrogen level. Tosolve the problem of variation for different animals, the S11 and thefrequency shift curves are not enough. The resonant frequency shiftshowed less sensitivity on the animal than S11 as follows fromFig. 3(a). For example, for S11 the slopes of sensitivitywere �0.046 dB/(mg/ml), �0.061 dB/(mg/ml), and �0.025 dB/(mg/ml), while for frequency shift the slopes were 2$10�5 (mg/ml)�1,3$10�5 (mg/ml)�1, and 2$10�5 (mg/ml)�1, i.e. the frequency wasalmost stable. The values of the highest sensitivity (both S11 and Df/f0) correspond to the lowest levels for the each three components asshown in Fig. 3(b). We suppose that the change in Nitrogen con-centration (due to high intrinsic conductance) is the main reasonfor the variation of the measured microwave slopes. The problem isthe structure (i.e. other components) of blood for different animals;

Fig. 3. (a) Measured microwave reflection coefficient minima S11 (left axis) and relative resgoat-blood samples: A, B, and C. The data was estimated from Fig. 2. (b) Measurement senponents: Cholesterol, Phosphorus, and blood urea Nitrogen.

and to fully define all parameters we need more measured datathan only for S11 and Df/f0: the minimum needed is the behavior for3 additional variables. It is simpler to make a calibration for anyanimal before starting the glycemia measurement because it isnecessary to define the changes in glucose concentration not itsabsolute value (and compare to reference (normal) level of eachanimal individually). Note that another reason for the change in theelectromagnetic characteristics of blood is the variation of the NaClconcentration. According to Ref. [25], there is the correction factor(1.6 meq/l and 4 meq/l before and after 400 mg/dl glucose con-centration, respectively) for decrease in serum sodium concentra-tion for every 100 mg/dl increase in plasma glucose concentration.The solute (glucose, NaCl, etc.) concentration changes the totalcomplex dielectric permittivity: 3, of total solution, thus, the char-acteristic impedance: Zin, and finally, the microwave reflection co-efficient: S11. The maximum change in concentration of sodium is37.4 mg/dl for 400 mg/dl increase in glucose concentration. Ourcalculations show that the effect of glucose was 3 times bigger thanthat for NaCl for all three parameters: 0.23 vs. 0.07 for dielectricpermittivity, 1.51 U vs. 0.47 U for characteristic impedance, and0.0124 dB vs. 0.0038 dB for microwave reflection coefficient. Thus,the main factor that changes the electromagnetic properties ofblood in our measurements is glucose concentration. Beside this,the origin of sodium concentration change is also glucose [25], thus,we can correct the measurements result by means of this effect.However, additional data for NaCl will help to more accuratelydetect the glucose concentration in blood and decisively showswhether NaCl is an important confounding factor. Also, the calcu-lation is really only an estimate since the near-field effect isneglected.

onant frequency shift Df/f0 (right axis) dependence on glucose concentration for threesitivity (both for S11 and Df/f0) dependence on the concentration of three blood com-

Fig. 4. (a) Measured microwave reflection coefficient S11 dependence of the sample volume for three goat-blood samples: (a) A, (b) B, and (c) C at the resonant frequency of about4.76 GHz.

S. Kim et al. / Current Applied Physics 14 (2014) 563e569 567

From the linear relationship as a function of glucose concen-tration, DS11/Dc ¼ 0.061 dB/(mg/dl) or DS11/Dc ¼ 4 � 10�4 in thelinear scale. The root-mean-square statistical noise in reflectioncoefficient S11 was about 10�5 in the linear scale [19]. The measuredsignal-to-noise (SNR) was about SNR ¼ 20 log (4 � 10�4/10�5) z 32 dB. The smallest detectable change in concentrationbased on a criterion of SNR of 32 dB (Q ¼ 648) was about 10 mg/dl.

4.2. Sample volume, temperature, and humidity effect

Themicrowave response (reflection coefficient and the resonantfrequency shift) is very sensitive to the change of the sample vol-ume, temperature and humidity. Fig. 4 showsmeasured microwavereflection coefficient minima S11 plotted as a function of the goat-blood samples ((a) sample A, (b) sample B, and (c) sample C) vol-ume at an operating frequency of about 4.76 GHz. The dynamicrange is large in blood samples with the volume range of 0.5e5 mlwith a minimum at 2 ml. Thus, the measurements were done at2 ml volume, at which the volume dependence is weakest and sothe accuracy of the concentration measurements is the highest.

Fig. 5 shows the measured microwave reflection coefficientminima S11 plotted as a function of the goat-blood temperatureswith biggest difference of the glycemia concentrations for samplesA, B, and C at the resonant frequency near 4.76 GHz. The

Fig. 5. Measured microwave reflection coefficient S11 plotted as a function of thetemperatures for three goat-blood samples: A (540 mg/dl), B (492 mg/dl), and C(510 mg/dl) at the resonant frequency of about 4.76 GHz.

temperature influence of S11 of the microwave cavity sensor wasweaker in goat-blood sample at low temperatures (until 27 �C),while for the high temperatures S11 decreased with linear slope0.083 dB/�C�1 vs. 0.0088 dB/�C�1 for the low temperatures (about 9times). This effect can be caused by the viscosity of the liquid, i.e. forthe higher viscosity material (like a solid) the fluctuations of elec-tromagnetic parameters due to temperature variations is smaller.Note that both temperature and humidity had significant effects onthe reliability of blood glucose monitoring. The effect of tempera-ture was greater than the effect of humidity [26].

4.3. Simulated model in HP NFSS

Fig. 6 shows simulated microwave reflection coefficient S11profiles for DI water and D-glucose aqueous solution samples withglucose concentration of 110 mg/dl to 510 mg/dl with interval of100 mg/dl at the resonant frequency of about 4.76 GHz. Thechange of the S11 minimum and the resonant frequency shift ofthe 110 mg/dl to 510 mg/dl samples were �7.66 dB and �22 MHzfor measured data and �6.04 dB and �40 MHz for simulated data,respectively. Both measured and simulated results were obtainedfor the same sample volume (2 ml). The inset of Fig. 6 shows

Fig. 6. Simulated microwave reflection coefficient S11 profiles for D-glucose aqueoussolution at the resonant frequency of about 4.76 GHz. The samples volume were 2 ml.Inset shows microwave reflection coefficient minima S11 (left axis) and relative reso-nant frequency shift Df/f0 (right axis) dependence on glucose concentration for bothmeasured and simulated data.

Fig. 7. The simulated images of the (a) electric and (b) magnetic field distributions near 4.76 GHz for electromagnetic interaction between the resonant cavity sensor and theglucose sample with various concentrations. The solid white circle shows the sample location.

S. Kim et al. / Current Applied Physics 14 (2014) 563e569568

measured (circles) and simulated (rhombus) microwave reflectioncoefficient S11 (left axis) and resonant frequency shift Df/f0 (rightaxis) dependence on D-glucose concentration at the resonantfrequency (4.76 GHz). It was found that both S11 and Df/f0decreased as the glucose concentration increased. The S11 trendin the linear approximation varies with slope of DS11/Dc ¼ �0.0251 dB/(mg/dl) for goat-blood samples, while thesimulated data after linear approximation shows DS11/Dc ¼ �0.0142 dB/(mg/dl). However, in the linear approximationthe frequency shift varies with slopes of (Df/f0)/Dc ¼ 2 � 10�5 (mg/dl)�1 for both measured and simulated data. We found a goodagreement between simulation and experiment with the variationof reflection coefficient S11 and the resonant frequency shift Df/f0on the glucose concentration.

4.4. Electromagnetic field distribution

To visualize the electromagnetic field distribution during mea-surement we simulated the electromagnetic field interaction in thetube filled with the D-glucose aqueous solution by using HP HFSS at4.76 GHz. Fig. 7 shows the (a) electric and the (b) magnetic fielddistributions for DI water and for D-glucose samples with glucoseconcentrations of 92 mg/dl to 492 mg/dl with interval of 100 mg/dlat the resonant frequency of about 4.76 GHz. Note that in D-glucoseaqueous solution the change in glucose concentration brings strongchanges in the electric and magnetic fields strength. For example,when the glucose concentration increased from 92 mg/dl to492 mg/dl, the permittivity of D-glucose aqueous solution alsoincreased while the electric (from 94 V/m to 12 V/m at themaximum, i.e. 7.8 times) and magnetic (from 1.9 A/m to 0.2 A/m atthe maximum, i.e. 9.5 times) fields strength decreased. Clearly, the

sample with high glycemia shows weaker electric and magneticenergy intensities. This reverse dependence on glucose concen-tration (the sample with high glycemia stores more electromag-netic energy due to the bigger complex permittivity) was caused bythe “two chambers” construction and the mode structure of ourresonant cavity and is opposite to the case discussed in Ref. [20].Thus, the further growth of glucose concentrations brings strongerelectromagnetic fields on the inner edges of resonant cavity not inthe sample.

5. Conclusion

A new type microwave cavity sensor has been developed fordetermination of glycemia in goat-blood samples. The microwavecavity sensor is a novel non-invasive glucometer with a minimum10 mg/dl detectable change in blood. The results show the sensi-tivity and the usefulness of the cavity sensor for sensing andmonitoring of D-glucose concentration. It can be useful for the real-time measurements of glucose concentration, and potentially it isan interesting approach for in vivo measurement of humanglycemia.

Acknowledgments

This work was supported by a Sogang University SpecialResearch Grant 2013 (201314002), Priority Research Centers Pro-gram through the National Research Foundation of Korea (2012-0006690), Parts and Materials International Collaborative R&DProgram (MKE, Korea; N0000481).

S. Kim et al. / Current Applied Physics 14 (2014) 563e569 569

References

[1] IDF, Diabetes Atlas, fourth ed., 2010.[2] D.E. Goldstein, R.R. Little, Endocr. Metab. Clin. North Am. 26 (1997) 475e486.[3] O.S. Khalil, Clin. Chem. 45 (1999) 165e177.[4] O.S. Khalil, Diabetes Technol. Ther. 6 (2004) 660e697.[5] G.S. Wilson, R. Gifford, Biosens. Bioelectron. 20 (2005) 2388e2403.[6] C.D. Geddes, J.R. Lakowicz (Eds.), Topics in Fluorescence Spectroscopy: Glucose

Sensing, Springer Science and Business Media Inc., NY, USA, 2006.[7] F. Musil, A. Smahelova, M. Kusalova, T. Halajcuk, A. Voseckova, V. Pavlik,

V. Velika, Nutrition 24 (2008) 614.[8] J.D. Newman, A.P.F. Turner, Biosens. Bioelectron. 20 (2005) 2435e2453.[9] A. Tura, A. Maran, G. Pacini, Diabetes Res. Clin. Pract. 77 (2007) 16e40.

[10] M.S. Ibbini, M.A. Masadeh, J. Med. Eng. Technol. 29 (2005) 64e69.[11] A. Tura, S. Sbrignadello, D. Cianciavicchia, G. Pacini, P. Ravazzani, Sensors 10

(2010) 5346e5358.[12] A. Rouane, Sens. Trans. J. 93 (2008) 69e81.[13] Y. Hayashi, I. Oshige, Y. Katsumoto, S. Omori, B. Friedman, K. Lee, Med. Eng.

Phys. 34 (2012) 299e304.

[14] A. Caduff, M.S. Talary, M. Mueller, F. Dewarrat, J. Klisic, M. Donath,L. Heinemann, W.A. Staheld, Biosens. Bioelectron. 24 (2009) 2778e2784.

[15] R.J. McNichols, G.L. Cote, J. Biomed. Opt. 5 (2000) 5e16.[16] T. Koschinsky, L. Heinemann, Diabetes Metab. Res. Rev. 17 (2001) 113e123.[17] H.P. Schwan, IEEE Trans. Microwave Theory Tech. 19 (1968) 146e152.[18] S.C. Hagness, A. Taflove, J.E. Bridges, Electron. Lett. 33 (1997) 1594e1595.[19] J. Kim, A. Babajanyan, A. Hovsepyan, K. Lee, B. Friedman, Rev. Sci. Instrum. 79

(2008) 086107.[20] H. Melikyan, E. Danielyan, S. Kim, J. Kim, A. Babajanyan, J. Lee, B. Friedman,

K. Lee, Med. Eng. Phys. 34 (2012) 299e304.[21] S. Kim, H. Melikyan, J. Kim, A. Babajanyan, J. Lee, L. Enkhtur, B. Friedman,

K. Lee, Diabetes Res. Clin. Pract. 96 (2012) 379e384.[22] D.M. Pozar, Microwave Engineering, third ed., J. Wiley & Sons, NY, USA, 2005.[23] E. Silva, M. Lanucara, R. Marcon, Supercond. Sci. Technol. 9 (1996) 934e941.[24] D. Lide, Handbook of Chemistry and Physics, 85th ed., CRC Press, Florida, USA,

2004.[25] T.A. Hillier, R.D. Abbott, E.J. Barrett, Am. J. Med. 106 (1999) 399e403.[26] M.J. Haller, J.J. Shuster, D. Schatz, R.J. Melker, Diabetes Technol. Ther. 9 (2007)

1e9.