Metamaterial based Printed Circuit Antenna for Blood Glucose Level Sensing Applications

— Due to the urgent need to develop technologies for continuous glucose monitoring in diabetes individuals, a potential research has been applied by invoking the microwave techniques. Therefore, this work presents a novel technique based on a single port microwave circuit, antenna structure, based on Metamaterial (MTM) transmission line defected patch for sensing the blood glucose level in noninvasive process. For that, the proposed antenna is invoked to measure the blood glucose through the field leakages penetrated to the human blood through the skin. The proposed sensor is constructed from a closed loop connected to an interdigital capacitor to magnify the electric field fringing at the patch center. The proposed antenna sensor is found to operate excellently at the first mode, 0.6GHz, with S 11 impedance matching less than -10dB. The proposed sensor performance is tested experimentally with 15 cases, different patients, through measuring the change in the S 11 spectra after direct touching to the sensor a finger print of a patient. The proposed sensor is found to be effectively very efficient detector for blood glucose variation with a low manufacturing cost when printed on an FR-4 substrate. The experimental measurements are analyzed mathematically to obtain the calibration equation of the sensor from the curve fitting. Finally, the theoretical and the experimental results are found to be agreed very well with a percentage of error less than 10%.


INTRODUCTION
Viktor Veselago first presented metamaterials (MTM) in 1967 [1].Such structures may be called artificial materials with nontraditional properties [2].These materials possess negative permittivity (ε) and negative permeability (μ) that in turn support the backward wave propagation of electromagnetic waves [3].In 1999 an interesting subwavelength element realized as split ring resonator (SRR) to achieve negative permeability was proposed by Pendry [4].Basically, SRR can be represented as tank of LC circuit possessing equivalent inductance (L) and the capacitance (C) between two concentric rings resonating at specific frequency [4].The SRR has normally a size much less than, around wavelength (λ/10), the guided wavelength.Therefore, many researchers applied SRR as technique to retrieve the characterizations of materials [4].In such process, the changes in the scattering parameters (S-parameters) with respect to a sample under test (SUT) introduction can be transferred through certain algorithm to retrieve the materials characterizations.
Microwave sensing is a reliable method for liquid characterization that has been employed in the past decade [5].SRRs [6], complementary SRRs (CSRRs) [3], open CSRRs (OCSRRs) [4], closed ring resonator [5] and other miniaturized microwave resonators [6] based on transmission lines attracted a considerable attention of researchers for biomedical applications [6].The electromagnetic properties of these structures depend on their operation frequency and quality factor changes with respect to different liquid introductions [7].Through measuring Sparameters coefficients, complex dielectric parameters of SUT, the liquid characterizations can be addressed [8].This has provided a new sensing platform for the biological, pharmaceutical, and fuel industries [9].Most microwave biological sensors are mounted under ultra-thin cylindrical pipes [10] or slotted cylindrical tubes [11].Based on measuring S-parameters magnitudes at certain frequency, materials under test losses can be extracted for quality detection [12].Such technology, however, the resolution of high-losses was found a significant concern due to the issue of skin depth penetration [13].On the other hand, most suggested methods require micro fluidic channel tubes to contain SUT; that add extra losses and difficulty of penetrations [14].Nevertheless, many sensors detection process is based on certain volume of SUT that limits their use for real-time monitoring applications [15].Moreover, an additional size and fabrication cost could be added to permeate the microwave reach to SUT [16].In such technique, at certain liquid height such process is validated to ensure field retardation and phase change to analyze the dielectric characterization [17] in specific for biological solvent detection.
Based on the current development of microwave sensors technologies, fractal based MTM structures attracted researchers to realize effective and efficient sensors to achieve a high selectivity [18].On top of that, MTM realizes excellent performance in the microwave ranges due to their nontraditional properties [19].Moreover, MTM based fractal geometries use remain very excellent candidates in the biomedical aspects due to their size reduction in comparison to traditional microwave structures.For example, a MTM defected patch based monopole antenna was presented in [20] for pollution detections.Authors in [21] proposed a study of using a traditional microstrip transmission line for liquid properties detection.Then, an extended study based on a fractal MTM structure was presented for blood glucose sensing using a microstrip transmission line loaded with carbon nanotube patch in [22].The use of MTM patch based nanoscale structures was introduced for gas detection in [23] and [24].MTM based fractal structures were applied to realize as MTM defect on the ground plane for cancer cell detection based on their electrical properties changes [25].Now, in this paper, a new approach based on a single port antenna element via MTM transmission line structure is proposed for blood glucose level detection.The MTM is constructed from an interdigital capacitor structure with a closed loop ring.The proposed structure is designed for glucose detection.The paper is organized as follows: The geometrical details are discussed in section II.The analysis process and the parametric study are presented in section III.The experimental measurements are explained in section IV after conducting 15 cases for the proposed study.The paper is concluded in section V.

II. Sensor Design and Details
The proposed antenna is mounted on an FR-4 substrate of a dielectric constant   =4.In addition to the proposed closed loop in this design, a copper interdigital capacitor (Cint), see Fig1(c), is conducted to maximize the electrical field intensity [12] at the center of the patch where SUT would be positioned.All geometrical details of Fig1 are listed in Table 1.
Table 1; Geometrical details of the proposed antenna sensor.

III. MTM Patch Theoretical Analysis
As mentioned later, the proposed radiating patch consists of a closed loop coupled with Cint.Therefore, the proposed patch is structured in such way to provide maximum fringing from the Cint edges.The Cint unit cell is constructed as interfaced strip lines with an effective length (Ln) to provide the desirable resonant frequency (fr).The resulted capacitance value of the used Cint can be calculated analytically from the following equation [26]: where C is the capacitance in pF, εre is the effective relative permittivity, n is the capacitor figure number, L is the finger length, K and K' are elliptical integral coefficients and they are given as [27]: where, k is the argument and can be calculated as [26]: where: where, W is the finger width and S is the separation distance between fingers.Now, the proposed sensor is analytically decomposed from the equivalent circuit diagram that is shown in Fig2.From Fig2, the proposed Cint is connected to two strip lines.The strip lines provide L-C branch (Lstrip and Cstrip) connection in parallel with Cint.OCSSR structure is connected to Cint through three shoring posts to be presented by (Lshort) in the equivalent circuit diagram.It is good to mention that the effects of the inherent stray inductors (Ls) and stray capacitors (Cs) are proposed in the equivalent circuit diagram.Nevertheless, the feed line equivalently is considered as an inductor (Lfeed).The equivalent representation for the closed loop (CL) is considered as an inductor (LCL) and can be calculated according to the following equation [23]: Based on the circuit model in Fig2, the authors calculated the relative lumped elements with an initial gauss from equations (1 and 6) for the proposed design at the desired frequency band using ADS software package parametrically.The S-parameters are calculated from the circuit model to be shown in Fig2.It is found that the proposed sensor shows a resonance mode at 0.63GHz according to the lumped elements that are listed in Table 2.
Fig2; Equivalent circuit for the proposed resonator.From the proposed circuit model the equivalent impedance is calculated based on the second branch to be described by L2 and C2 be noted as ZT,2 following: Therefore, the resonant frequency, fr,2 is expressed as In Fig. 2, the equivalent circuit of the proposed patch can be represented as two series of L1Cc circuits and one parallel L2C2 circuit and the total impedance, ZT,1 in addition to the band pass resonant frequency, fr,1 which can be given as following: By employing (2) and (4), relation (3) can be rewritten as following: ZT,1 in relation ( 5) has two resonant frequencies, lower and upper frequencies which are represented by ωr,1 and ωr,2 respectively.When ω = ωr,1 a maximum transmission could occur, where minimum ZT,1 is (ZT,1 ≥ jωL2), while zero transmission occurs at ω = ωr,2 where ZT,1 maximum is achieved.In order to confirm prior discussion, an electromagnetic 3D simulation based on CST software packages in invoked to validate the obtained results from the circuit model.In Fig3, the patch is simulated when mounted in closed to a transmission line to evaluate the S11 spectrum.It is found that there are three resonant frequencies, fr,1, fr,2 and f r,3 , all these frequencies are the same resonant frequencies of the proposed patch.Therefore, these frequencies can be used as indicators for characterizing SUT after applying sensitivity analysis at the resonant frequencies.

IV. Sensitivity Analysis
The proposed sensor is invented based on a quasi-static small antenna [12] in which designed with an interdigital capacitor surrounded with closed loop.Due to such combination a current circulation occurs at three legs of connection between the capacitor and the closed loop.The variation in the capacitance of the structure generally concerns on the variation in the permittivity of SUT.Therefore, the performance variation in the proposed structure with introducing different permittivity values is discussed as following: 1-Resonant frequencies analysis: After introducing different SUT in the CST MWS environment, the frequency shifts are recorded to be tested later experimentally.Therefore, SUT must cover the whole area for the efficient perturbation of the E-field.The resonant frequencies of S11 in Fig. 4   From Fig. 4 (a) it can be noted that the resonant frequency, fr,1is about 170 MHz greater than resonant frequency, fr,2.Moreover, the relative change fr,1 with dielectric constant is about 150 MHz greater than fr,2 as depicted in Fig. 4 (b).In another meaning, for these resonant frequencies, fr,1 shows to be a good preference to obtain high sensitivity for supposed dielectric constant as compared to that of resonant frequency, fr,1.

2-Quality factor analysis:
For general resonators the quality factor, Q may be presented as [34]: where, ω0 is the angular resonant frequency, W is the electric and magnetic stored energy and PL represents the average power dissipated per cycle.The previous equation also can be rewritten as: where fr is the resonant frequency and ∆f represents the relative 3dB bandwidth of the resonator frequency response.The proposed sensor performance is simulated with different values of loss tangent from 0.01 to 0.15 and corresponding εʹr variation from 76 to 90.The relative quality factor is computed and depicted in Fig. 5. Fig. 5; Quality factor variation of the proposed sensor calculated for fr,1 andfr,2.
From Fig. 5, it is fully interesting to observe that the slope of quality factor is identical to resonant fr,1 in comparison to fr,2 slope with loss tangent change.While the slope of quality factor relative to fr,1 is greater than in fr,2 for the same condition.Hence the resonant fr,1 is utilized for characterizing the SUT.

V. Curve Fitting Analysis and Calibration Equations
Now, the obtained results from the previous section are analyzed to evaluate the calibration sensitivity to be compared to the experimental results.In order to describe the tested samples, a numerical type is desired which generally plots the measured parameters (e.g., the resonant frequency and the quality factor) to the relative permittivity of the SUT.

1-Real permittivity calibration effects
The proposed sensor frequency resonance is found to be changed due to samples loading as can be observed in Fig. 6; in which the inverse square of the resonant frequency (fr) is extracted from the S11 spectra.The results with the corresponding real permittivity (εʹr) of SUT are depicted in Fig. 6.
The obtained results in Fig. 6 confirm the results in equation ( 10) that is achieved from the equivalent circuit model.From equation ( 10), the values of L2 and C2 are supposed to be constant due to the solid values of the overall length of OCSRRs and εr of the substrate.It is interesting to note that the inverse square of the resonant frequency is directly proportional to the real permittivity of the SUT.Thus, in order to combine all the above effects, the dielectric constant of the SUT is mathematically expressed in terms of the resonant frequency (fr) as following: ′ = −3.519( −2 ) 2 + 23.84(  −2 ) − 5.007 The above relation is obtained from employing the curve fitting tools, which supplies a numerical model of the proposed sensor to determine the real permittivity of SUT in terms of the measured resonant frequency (fr,1).It should be noted that all SUT has with a fixed 3mm thickness.

2-Imaginary permittivity calibration effects
After founding the numerical relations to determine the dielectric constant of SUT, an identical analysis is completed to find a numerical relation for computing the loss tangent (tanδ) of SUT.As explained earlier that the resonant fr,1 provides a quality factor greater than those obtained at fr,2.Hence, the resonant fr,1 is employed for calculating tanδ of SUT.Therefore, at first, the dielectric constant in the range of 76 to 90 are possessed and the tanδ values combatable for each dielectric SUT is changed from 0 to 0.15, and the relative simulated results of fr,1change is depicted in Fig. 7.The quality factor (QSUT) for each case is determined from the simulated response of S11spectra, after that the inverse of QSUT values and the corresponding tanδ are depicted in Fig. 7.The relation between the tanδ and the QSUT can be specified as following [12]: where εr' and εr" are the real and imaginary parts of the relative permittivity in equation (15).From Fig. 7, it is noted that the alteration of Q -1 SUT with tanδ is linear compound with a rising values depend on εʹr of SUT.Thus, to deduce the tanδ of SUT, which relies on the loaded quality factor as well as the εʹr of SUT, a curve fitting tool is utilized to conclude the numerical model as presented below: After deciding the εʹr from equation ( 14) and tanδ from ( 16), the imaginary part of the complex permittivity can be determined using (15).

VI. Sensor Performance and Measurements
The proposed sensor is fabricated using printed circuit board technology as shown in Fig. 9.The sensor is fabricated from using chemical wit etching process in the laboratory.The FR4 substrate is considered as the plate form layer for the proposed sensor.Now, the proposed sensor performance is measured in terms of S11 spectrum as seen in Fig. 10 without introducing any SUT.The obtained results from measurements are compared to those obtained from simulation results to show excellent agreement as seen in Fig. 9.The measurement results are conducted to PNA8720 network analyzer after applying a single port calibration process.From the measured S11 spectra, the proposed sensor shows a frequency resonance at 0.63 GHz with |S11|=16dB, and a bandwidth from 0.6GHz to 0.65GHz.This frequency is considered to ensure excellent penetration through the human tissues with minimum skin depth loss [6].

VII. Sensing Measurements and Validations
In this section, the proposed sensor measurement operation is based on placing a finger on the interdigital capacitor part to monitor the variation in the S11 magnitude and frequency resonance shift.The field penetration through the finger skin is affected by the blood glucose variation [12].Such variation is attributed to the blood glucose change that could be reflected on the effective permittivity of the blood as discussed [23].Therefore, the effects of touching the proposed sensor by 15 patients at three different times to realize 45 recorders are listed in Table 3 to analyze the sensor performance.The recorded data in Table 3 are collected based on S11 spectra change in terms of S11 magnitude, frequency resonance, phase change, bandwidth, and quality factor.

1-Sensing Process
The S11 spectra of the proposed sensor are obtained according to the samples listed in Table 3.The sensor is designed to ensure that the first resonant position is located around 0.63 GHz.Therefore, the fabricated sensor S11 spectra changes are evaluated after introducing the finger touch.Thus, the prepared design is experimentally tested using the PNA8720 network analyzer.The obtained changes in the S11 spectra are monitored in terms of |S11|, frequency shift, phase change, quality factor, and bandwidth.

2-Sensing Validation
The variation in the S11 spectra of the proposed sensor is measured after placing a finger on it as a non-invasive technique.Therefore, the glucose level is monitored through a normal device glucose meter, PRODIGY Autocode, and the results are recorded in Table 3.Then, the patient finger is placed on the proposed sensor and the frequency resonance shift and S11 magnitude change are listed in Table 3.
Next, the measured glucose level is compared with respect to the relative values of εr' and εr" that are listed in Table 3.Therefore, the measured fr,1 and QSUT values are applied in equation ( 14) to (15) to calculate the relative values of εr' and εr" from the measured data.The calculated values of εr' and εr" are compared to their relatives from measurements in Table 3.
Thus, in Table 4, the relative errors between the measured and calculated εr' and εr" values are calculated.It is found a good agreement between the measured and calculated values.Therefore, from this comparison between the relative values of εr' and εr", the glucose level can be detected according to Table 4.
It is found the maximum error from the total values is less than 10%.

VIII. Conclusion
The proposed sensor is presented to characterize blood glucose level through measuring the relative values of εr' and εr" for different blood samples.The proposed sensor is constructed as a single-port network; therefore, it is designed based on an interdigital capacitor patch to sense the blood glucose level non-invasively.The reason of that, the field fringing from the proposed sensor is found to be magnified and easy to penetrate through the human skin to the blood vassals.It is found that the proposed sensor shows different frequency resonances within the band of interest.However, it is decided to consider only the first frequency resonance (fr,1) for sensing where the maximum sensitivity can be achieved.In this case, fr,1 and QSUT measurement are gathered from different patients.Therefore, from the measured values, an analytical model is synthesized based on curve fitting analysis.In such process, fifteen patients are submitted to the proposed sensor for estimating the level of glucose in the blood, ending with results very similar to the results measured by traditional commercial methods.The measured values of εr' and εr" are found to be agree very well with those obtained from the calculated results based on curve fitting analysis with less than 15% errors.It is found that the proposed sensor is a suitable choice for biomedical applications including blood glucose measurements.Finally, the proposed measurements point out the total error is about 10%.

Declarations
We confirm here that the following sections are

Fig 3 ;
Fig 3; S11 spectra comparison between ADS and CST software packages.
are considered as the reference of unloaded filter (fr,1 and fr,2).The proposed sensor is then loaded with the SUT, where the dielectric constant of the sample is changed randomly in a broad range from 76 to 90.The resonant frequencies (fr,1 andfr,2) corresponding to each sample are extracted, which are also plotted with variation of the dielectric constant as shown in Fig.4(a).For preferable conception the change in the resonant frequencies (∆fr= unloaded (fr,1 and fr,2)corresponding loaded (fr,1and fr,2)) is plotted in Fig. 4 (b).

Fig. 4
Fig. 4 Performance variation in terms of frequency resonance of the proposed sensor with changing εr: (a) fr change and (b) ∆fr -1 .

Fig. 7 .
Fig. 7. Inverse of Q-factor in terms of tanδ for various values of εʹr (Linear relevance between Q - 1 SUT and tanδ for all values of εʹr)

Table 2 ;
Calculated lumped elements of the equivalent circuit.

Table 3 ;
Clarification of the results of measuring different blood samples

Table 4 ;
Comparison relative errors between the measured and calculated εr' and εr" values