sensors Article
Microwave-Based Microfluidic Sensor for Non-Destructive and Quantitative Glucose Monitoring in Aqueous Solution Thomas Chretiennot, David Dubuc * and Katia Grenier LAAS-CNRS, Université de Toulouse, CNRS, Toulouse 31031, France;
[email protected] (T.C.);
[email protected] (K.G.) * Correspondence:
[email protected]; Tel.: +33-561-336-292 Academic Editor: Ferran Martín Received: 28 July 2016; Accepted: 11 October 2016; Published: 19 October 2016
Abstract: This paper presents a reliable microwave and microfluidic miniature sensor dedicated to the measurement of glucose concentration in aqueous solution. The device; which is integrated with microtechnologies; is made of a bandstop filter implemented in a thin film microstrip technology combined with a fluidic microchannel. Glucose aqueous solutions have been characterized for concentration ranging from 80 g/L down to 0.3 g/L and are identified with the normalized insertion loss at optimal frequency. The sensitivity of the sensor has consequently been estimated at 7.6 × 10−3 dB/(g/L); together with the experimental uncertainty; the resolution of the sensor comes to 0.4 g/L. These results demonstrate the potentialities of such a sensor for the quantitative analysis of glucose in aqueous solution. Keywords: microwave; microfluidic; sensor; glucose; glycaemia
1. Introduction In its recent edition of the Diabetes Atlas, the International Diabetes Federation estimated that the number of adults living with diabetes has reached to 366 million, representing 8.3% of the global adult population [1]. This number is projected to increase to 552 million people by 2030, or 9.9% of adults, which equates to approximately three more people with diabetes every 10 s. These figures explain why diabetes is predicted to become the seventh leading cause of death in the world by the year 2030 [2]. Nevertheless, a large-scale study [3] has proved that, with a constant blood glucose monitoring, patient can avoid any complication. The precise and regular knowledge of the blood-glucose level is consequently mandatory and, in conjunction with appropriate treatments, levels must be maintained in the range of 0.5 to 2 g/L. Currently, main glucose-monitoring techniques, which are based on electrochemical reaction, have demonstrated high accuracy and strong reliability [4]. However, these solutions lead to the destruction of blood samples. Major efforts are consequently spent on developing techniques to measure the glucose concentration in blood non-invasively. Different electromagnetic liquid sensors have been developed for fluid characterization [5–8], for bio-liquids analysis [9–11], and more specifically for glucose monitoring applications [12–15] and demonstrated that microwaves are appropriate for aqueous solution analysis. These sensors are either based on cumbersome resonant structures, more sensitive to glucose variation or either miniature, for on-chip system integration [14,15] but requiring further resolution (both sensitivity and repetitiveness) improvement. This paper presents the experimental demonstration of a miniature microwave sensor, which may be envisioned for glucose monitoring in aqueous solution without compromise on its sensitivity and resolution, and its application as a potential non-invasive blood glucose analysis method.
Sensors 2016, 16, 1733; doi:10.3390/s16101733
www.mdpi.com/journal/sensors
Sensors 2016, 16, 1733
2 of 7
sensitivity and resolution, and its application as a potential non-invasive blood glucose analysis method. Sensors 2016, 16, 1733
2 of 7
2. Sensor Description 2. Sensor Description The biosensor is made of a quarter-wave length stub implemented in a thin film microstrip technology. The stubisismade connected to a microstrip feedline at implemented one end and to first film electrode of an The biosensor of a quarter-wave length stub inthe a thin microstrip inter-digitated capacitor (IDC) at the other end. The second electrode of the IDC is grounded with a technology. The stub is connected to a microstrip feedline at one end and to the first electrode of an via. Such a structure behaves as a stopband filter and can conveniently be characterized in inter-digitated capacitor (IDC) at the other end. The second electrode of the IDC is grounded with a via. transmission. Such a structure behaves as a stopband filter and can conveniently be characterized in transmission. The implemented in thin film film microstrip microstrip (TFMS) (TFMS) technology The device device is is implemented in aa thin technology constituted constituted of of aa 20 20 µm µm thick SU8 layer. The photoresist SU8 layer, which acts as the TFMS substrate, is patterned to allow thick SU8 layer. The photoresist SU8 layer, which acts as the TFMS substrate, is patterned to allow via via connection of IDC. the IDC. linetop (onoftop the layer) SU8 layer) and ground (on the bottom connection of the BothBoth stripstrip line (on theofSU8 and ground plane plane (on the bottom of the of the SU8 layer) metallization are realized in gold (0.3 µm thick). The feedline is 54 µm wide to in SU8 layer) metallization are realized in gold (0.3 µm thick). The feedline is 54 µm wide in order order match acharacteristic 50-ohm characteristic impedance. All gaps the10IDC 10 The µm length wide. The matchto a 50-ohm impedance. All gaps in the IDCinare µm are wide. and length width and width of the capacitor are 150 µm and 130 µm respectively. of the capacitor are 150 µm and 130 µm respectively. A microfluidic channel channel made made of of polydimethylsiloxane polydimethylsiloxane (see further details) details) is is placed placed A microfluidic (see [16,17] [16,17] for for further on on the the IDC, IDC, as as it it corresponds corresponds to to the the area area where where the the electric electric field field is is of of highest highest intensity intensity at at the the resonant resonant frequency and it enables the most efficient electric field/fluid coupling [7]. Such microfluidic frequency and it enables the most efficient electric field/fluid coupling [7]. Such microfluidic channel channel is in charge of the fluid conveying over the sensing area and presents the great advantage is in charge of the fluid conveying over the sensing area and presents the great advantage to to be be compatible with microfluidics capabilities for lab-on-a-chip [18]. The cross-section of the compatible with microfluidics capabilities for lab-on-a-chip [18]. The cross-section of the microfluidic microfluidic µmFigure per 50 1µm. Figure 1 shows top device view offabricated the deviceinfabricated the channel is 50channel µm per is5050µm. shows a top view ofa the the clean in room clean room at LAAS-CNRS. Inserts1 in Figurethe 1 provide the detailed architecture of the interdigitated at LAAS-CNRS. Inserts in Figure provide detailed architecture of the interdigitated capacitor capacitor equipped with the microfluidic channel and the IDC dimensions. equipped with the microfluidic channel and the IDC dimensions.
Figure 1. biosensor consisting consisting of of aa quarter quarter wave-length wave-length stub stub implemented implemented in in aa thin thin film Figure 1. Fabricated Fabricated biosensor film microstrip technology. Inlets provide the detailed architecture of the inter-digitated capacitor (IDC) microstrip technology. Inlets provide the detailed architecture of the inter-digitated capacitor with the microfluidic channel and theand IDC dimensions. (equipped IDC) equipped with the microfluidic channel the IDC dimensions.
Figure 2 shows a 3D full wave simulation of the stub at the resonant frequency of 7.5 GHz (see ® . The ®. The been realized with Ansys HFSS color codecode confirms that Figure 33 in inSection Section3). 3).Simulations Simulationshave have been realized with Ansys HFSS color confirms the electric fieldfield of highest intensity is concentrated in theinIDC. that the electric of highest intensity is concentrated the IDC.
Sensors 2016, 16, 1733 Sensors 2016, 16, 1733
3 of 7 3 of 7
Figure Figure 2. 2. Simulated Simulated electric electric field field intensity intensity at at the the resonance. resonance.
Figure Figure 3. 3. Measured Measured S21 S21 spectra spectra in in the the frequency frequency range range [0; [0; 55] 55] GHz GHz for for deionized deionized water water (reference (reference liquid liquid in in this this study) study) and and aa glucose glucosesolution solutionat at80 80g/L. g/L.
3. Experimental 3. Experimental Results Results
microprobes. The The block block Measurements are are made made on on wafer wafer directly directly with withRadioFrequency RadioFrequency (RF) (RF)microprobes. Measurements diagramof ofthe theRF RFmeasurement measurement setup is presented in Figure 4. Measurements are preceded by a diagram setup is presented in Figure 4. Measurements are preceded by a SOLT SOLT calibration process in the [0; 55] GHz frequency band which sets the reference planes at calibration process in the [0; 55] GHz frequency band which sets the reference planes at the tipsthe of tipsmicroprobes. of the microprobes. Fluidare samples are then injected one-by-one thanks pump, to a syringe pump, the Fluid samples then injected one-by-one thanks to a syringe characterized characterized the liquid is stabilized over sensing During zone and evacuated. During the when the liquidwhen is stabilized over the sensing zone andthe evacuated. the measurement, we check measurement, we check with a microscope that any bubble remains in the sensing area. with a microscope that any bubble remains in the sensing area. Temperature is also kept constant at ◦ C thanks to Temperature is aalso kept constant at 20 °C thanks to to a thermally in order to avoid 20 thermally controlled chuck in order avoid anycontrolled shift in thechuck dielectric properties of anyglucose shift in the dielectric propertiestemperature of the glucose solutions due to significant temperature the solutions due to significant variations. variations.
Sensors 2016, 16, 1733 Sensors 2016, 16, 1733
4 of 7 4 of 7
CCD Camera Syringe pump Device Under Test microprobe
microprobe
Microfluidic)channel)
cable
waste cable
Network Analyzer Figure (RF)measurement measurementsetup. setup. Figure4.4.Block Blockdiagram diagramofofthe theRadioFrequency RadioFrequency (RF)
Figure 3 shows the measured magnitude in dB of the S21 parameters versus frequency for Figure 3 shows the measured magnitude in dB of the S21 parameters versus frequency for different channel filling. Such parameter corresponds to the transmission of the stopband filter, different channel filling. Such parameter corresponds to the transmission of the stopband filter, which is expected to be (highly) sensible to the permittivity characteristics of the liquid loaded in the which is expected to be (highly) sensible to the permittivity characteristics of the liquid loaded in microfluidic channel. When the channel is empty (black curve in Figure 3), the biosensor is the microfluidic channel. When the channel is empty (black curve in Figure 3), the biosensor is characterized by a resonant frequency of 16.4 GHz, associated rejection of −17.6 dB and quality factor characterized by a resonant frequency of 16.4 GHz, associated rejection of −17.6 dB and quality factor of 1.2. Full-wave simulations gave a resonant frequency of 16.8 GHz, associated rejection of −17.7 dB of 1.2. Full-wave simulations gave a resonant frequency of 16.8 GHz, associated rejection of −17.7 dB and quality factors of 1.1. Such a good agreement between simulations and measurements confirms and quality factors of 1.1. Such a good agreement between simulations and measurements confirms the good operation of the device. the good operation of the device. A series of six aqueous glucose mixtures has been measured on our biosensor. The glucose A series of six aqueous glucose mixtures has been measured on our biosensor. The glucose concentration varied from 80 g/L (high concentration) to 5 g/L (medium concentration, five time concentration varied from 80 g/L (high concentration) to 5 g/L (medium concentration, five time higher than physiological concentrations). Figure 3 also shows the measurements corresponding to higher than physiological concentrations). Figure 3 also shows the measurements corresponding deionized water (dashed blue) and to the glucose mixture at 80 g/L (red). This figure clearly to deionized water (dashed blue) and to the glucose mixture at 80 g/L (red). This figure clearly demonstrated, based on the glucose dependent permittivity of the injected aqueous solution, a slight demonstrated, based on the glucose dependent permittivity of the injected aqueous solution, a slight but significant contrast (shift in resonant frequency and associated rejection) between pure DI water but significant contrast (shift in resonant frequency and associated rejection) between pure DI water and 80 g/L glucose aqueous solution. and 80 g/L glucose aqueous solution. In order to exacerbate the influence of glucose concentration on the sensor response, each S21 In order to exacerbate the influence of glucose concentration on the sensor response, each S21 signature associated to a glucose mixture is normalized by the S21 parameter of the reference liquid signature associated to a glucose mixture is normalized by the S21 parameter of the reference liquid (deionized water in this study). Practically speaking, that means that a sample of water must be (deionized water in this study). Practically speaking, that means that a sample of water must be measured directly before each sample of glucose mixture. Figure 5 shows the resulting normalized measured directly before each sample of glucose mixture. Figure 5 shows the resulting normalized |S21,normalized| parameters associated to each glucose mixture. This figure clearly facilitates the |S21,normalized| parameters associated to each glucose mixture. This figure clearly facilitates the distinction of each glucose concentration and points out that at 7.5 GHz, the contrasts between the distinction of each glucose concentration and points out that at 7.5 GHz, the contrasts between different glucose concentrations are maximized. The insert in Figure 5 presents the values of the different glucose concentrations are maximized. The insert in Figure 5 presents the values of |S21,normalized| parameter at 7.5 GHz as a function of the glucose concentration. |S21,normalized| parameter at 7.5 GHz as a function of the glucose concentration. This result reveals a linear relationship between the selected microwave readout This result reveals a linear relationship between the selected microwave readout |S21,normalized| |S21,normalized| parameter at 7.5 GHz and the glucose sensor for glucose concentration up to 80 parameter at 7.5 GHz and the glucose sensor for glucose concentration up to 80 g/L. The high linearity g/L. The high linearity of the sensor’s response (the coefficient of correlation is higher than 0.99) of the sensor’s response (the coefficient of correlation is higher than 0.99) validates that microwave validates that microwave sensing of biomolecules in aqueous solution represents a reliable and sensing of biomolecules in aqueous solution represents a reliable and predictable technique. predictable technique.
Sensors 2016, 16, 1733 Sensors 2016, 16, 1733 Sensors 2016, 16, 1733
5 of 7 5 of 7 5 of 7
Figure 5. Measured normalized spectra S21 for the highest eight concentrated glucose solutions in Figure Measured normalized spectraS21 S21 the highest eight concentratedglucose glucosesolutions solutionsininthe Figure 5. 5. Measured normalized spectra forfor the highest eight concentrated thethe same frequency range. same frequency range. same frequency range.
In In order to technique tosense sense glucose aqueous solution order demonstratethe theability ability ofofsuch such to glucose ininaqueous solution forfor In order totodemonstrate demonstrate abilityof suchaa technique a technique to sense glucose in aqueous solution blood hyperglycemia applications, further investigations have been carried out considering glucose hyperglycemia applications, further investigations have been carried considering glucose forblood blood hyperglycemia applications, further investigations have been out carried out considering concentration atata aphysiological concentration around g/L. Figure presents the values of concentration physiological concentration around 11 g/L. Figure 6 6presents the values of of glucose concentration at a physiological concentration around 1 g/L. Figure 6 presents the values |S21,normalized| of the glucose concentration for values ranging |S21,normalized|parameter parameterat 7.5GHz GHzas as aaa function function for values ranging |S21,normalized| parameter atat7.5 7.5 GHz as function of ofthe theglucose glucoseconcentration concentration for values ranging from 0.31 g/L to 5 g/L. Various concentrations have been obtained by successive dilution by from 0.31 g/L to 5 g/L. Various concentrations been obtained by successive dilution by a factor from 0.31 g/L to 5 g/L. Various concentrations have been obtained by successive dilution by aa factor factor two a mother solutionand andmeasurements measurements have have been been two of of mother solution and measurements have been repeated repeated10 10times. times. two of aa mother solution repeated 10 times.
Figure 6. Modulus of the normalized S21 parameter at 7.5 GHz as a function of the glucose Figure Modulus normalized S21 parameter at 7.5 as of a the function the glucose concentration. Figure 6.6.Modulus of of the the normalized S21 parameter at 7.5 GHz as a GHz function glucoseofconcentration. concentration.
Once again, the Figure 6 demonstratesa alinear linearresponse responseofofthe thesensor sensorfor forthe the considered considered range of Once again, the Figure 6 demonstrates of Once glucose’s concentrations down to physiological glucose concentration inblood. blood. again, the Figure demonstrates a linear glucose response of the sensorinfor the considered range glucose’s concentrations and6and down to physiological concentration of glucose’s concentrations and down to physiological glucose concentration in blood. Discussionon onthe theChieved Chieved Resolution Resolution 4. 4.Discussion
4. Discussion on the Chieved Resolution From Figure thesensitivity sensitivity ofthe thesensor sensor can can be defined From Figure 6,6,the SSof defined and andevaluated evaluatedasasfollows: follows: From Figure 6, the sensitivity S of the sensor can be defined and evaluated as follows: ∆|S21,normalized|(7.5 GHz)/∆[glucose] = 7.6 × 10−3 dB/(g/L)
(1)
Sensors 2016, 16, 1733
6 of 7
Sensors 2016, 16, 1733
6 of 7
Moreover, repetitive tests enabled the assessment of the experimental uncertainty: δ = 3 × 10−3 dB. Δ|S21,normalized|(7.5 GHz)/Δ[glucose] = 7.6 × 10−3 dB/(g/L) (1) Sensitivity and experimental uncertainty allow the definition of the biosensor resolution (R), that is to Moreover, repetitive tests enabled the assessment of the experimental say the smallest variation of glucose concentration that the sensor is able touncertainty: quantify: δ = 3 × 10−3 dB. Sensitivity and experimental uncertainty allow the definition of the biosensor resolution (R), that 3 R =ofδ/S = 3 ×concentration 10−3 /7.6 ×that 10−the = sensor 0.4 g/L is to say the smallest variation glucose is able to quantify: R = δ/S = 3 × 10−3/7.6 × 10−3 = 0.4 g/L
(2)
(2)
Such a resolution indicates that this thin film miscrostrip biosensor has reached the same Such as a dielectric resolution resonant-based indicates that this thin film miscrostrip biosensor has reached the same performances devices [12,13,15]. performances as dielectric resonant-based devices [12,13,15]. In order to further evaluate the sensor reliability for glucose monitoring applications, measured In order to further evaluate the sensor reliability for glucose monitoring applications, measured glucose concentrations are plotted using the Clarke error grid. The Clarke error grid has been worked glucose concentrations are plotted using the Clarke error grid. The Clarke error grid has been out in 1987 by the so-called biologist Clarke in order to evaluate the reliability of the commercial worked out in 1987 by the so-called biologist Clarke in order to evaluate the reliability of the glucometers [19].glucometers The Clarke[19]. error grid is given Figure 7. X-axis provides realprovides glucosereal concentrations; commercial The Clarke erroringrid is given in Figure 7. X-axis glucose Y-axisconcentrations; provides the measured glucose divided the griddivided into five Y-axis provides the concentrations. measured glucoseClarke concentrations. Clarke thedifferent grid intoareas named A,different B, C, D and Area AA,isB,the where on area the glucose concentration five areasE.named C,optimal D and E.area Area A is the the error optimal where the error on the does glucose20%, concentration not exceed 20%, which for is considered as acceptable forglycaemia. the screening not exceed which is does considered as acceptable the screening of human Asoffar as glycaemia. As far area B is concerned, error on glucose exceeds 20% is area Bhuman is concerned, error onasglucose concentrations exceeds 20%concentrations but is not detrimental forbut patients. detrimental Areas C,such D and are on thethe hazardous ones: such an error the Areasnot C, D and E are for the patients. hazardous ones: an E error glycaemia measurement mayonendanger glycaemia measurement may endanger the patient's health. the patient's health.
Figure 7. Clarke error grid established with with the lowest solutions measured on the Figure 7. Clarke error grid established lowestfive fiveglucose glucose solutions measured on the presented biosensor. presented biosensor.
Figure 7 shows the lowest five glucose solutions that been have measured been measured onglucose our glucose Figure 7 shows the lowest five glucose solutions that have on our biosensor in the Clarke grid. This figure demonstrates the potentialities of our sensor for the in thebiosensor Clarke error grid. Thiserror figure demonstrates the potentialities of our sensor for the determination of the human glycaemia as allconcentrations the measured concentrations lie down in the area A. of thedetermination human glycaemia as all the measured lie down in the area A. Moreover, since the electromagnetic (EM) field can penetrate different tissue materials at at Moreover, since the electromagnetic (EM) field can penetrate different tissue materials microwave frequencies, especially the different layers of human skin, and combined to the microwave frequencies, especially the different layers of human skin, and combined to the non-destructive ability of the technique, the demonstrated results together with others [20] non-destructive ability of the technique, the demonstrated results together with others [20] contribute contribute to establish the potential of the microwave technique for non-invasive blood glucose to establish the potential of the microwave technique for non-invasive blood glucose monitoring. monitoring.
5. Conclusions 5. Conclusions This This paper hashas experimentally the reliable reliableoperation operation a microwave paper experimentallydemonstrated demonstrated the of of a microwave and and microfluidic based sensor dedicated quantification (down to 0.4 microfluidic based sensor dedicatedfor forglucose glucose concentration concentration quantification (down to 0.4 g/L)g/L) in in aqueous solutions. Combined with its non-destructive the microwave is aqueous solutions. Combined with its non-destructive ability, ability, the microwave sensing issensing consequently consequently identified as an attractive technique for blood parameters monitoring. identified as an attractive technique for blood parameters monitoring. Acknowledgments: This work was supported in part by the CNRS and in part by LAAS-CNRS micro and nano
Acknowledgments: Thismember work was supported in part by the CNRS and in part by LAAS-CNRS micro and nano technolgies platform of the French RENATECH network. technolgies platform member of the French RENATECH network. Author Contributions: The authors contributed equally to this work. Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2016, 16, 1733
7 of 7
References 1. 2. 3.
4. 5.
6. 7.
8.
9. 10. 11.
12. 13. 14.
15.
16.
17.
18. 19. 20.
International Diabetes Federation. Available online: http://www.idf.org/diabetesatlas (accessed on 12 October 2016). World Health Organization. Diabetes Fact Sheet N312. Available online: http://www.who.int/mediacentre/ factsheets/fs312/en (accessed on 12 October 2016). The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N. Engl. J. Med. 1993, 329, 977–986. Hönes, J.; Müller, P.; Surridge, N. The technology behind glucose meters: Test strips. Diabetes Technol. Ther. 2008, 10, S-10. [CrossRef] National Physical Laboratory. A Guide to the Characterisation of Dielectric Materials at RF and Microwave Frequencies; The Institute of Measurement and Control & National Physical Laboratory: Teddington, UK, 2003. Marsland, T.P.; Evans, S. Dielectric measurements with an open-ended coaxial probe. IEEE Proc. 1987, 34, 341–349. [CrossRef] Chrétiennot, T.; Dubuc, D.; Grenier, K. Optimized electromagnetic interaction microwave resonator/microfluidic channel for enhanced liquid biosensor. In Proceedings of the 2013 European Microwave Conference, Nuremberg, Germany, 6–10 October 2013; pp. 464–467. Kapilevich, B.; Litvak, B. Microwave sensor for accurate measurements of water solution concentrations. In Proceedings of the 2007 Asia-Pacific Microwave Conference, Bangkok, Thailand, 11–14 December 2007; pp. 1–4. Li, L.; Uttamchandani, D. A microwave dielectric biosensor based on suspended distributed MEMS transmission lines. IEEE Sens. J. 2009, 9, 1825–1830. [CrossRef] Gennarelli, G.; Romeo, S.; Rosaria Scarfi, M.; Soldovieri, F. A microwave resonant sensor for concentration measurements of liquid solutions. IEEE Sens. J. 2013, 13, 1857–1864. [CrossRef] Shaforost, E.N.; Klein, N.; Vitusevich, S.; Barannik, A.A.; Cherpak, N.T. High sensitivity microwave characterization of organic molecule solutions of nanoliter volume. Appl. Phys. Lett. 2009, 94, 112901. [CrossRef] Kim, J.; Babajanyan, A.; Hovsepyan, A.; Lee, K.; Friedman, B. Microwave dielectric resonator biosensor for aqueuous glucose solution. Rev. Sci. Instrum. 2008, 79, 086107–086110. [CrossRef] [PubMed] Babajanyan, A.; Melikan, H.; Kim, S.; Kim, J.; Lee, K.; Friedman, B. Real-time non-invasive measurement of glucose concentration using a microwave biosensor. J. Sens. 2010, 2010, 452163. Guariti, G.; Hofmann, M.; Weigel, R.; Fischer, G.; Kissinger, D. Determination of sugar concentration in aqueous solutions using ultra-wideband microwave impedance spectroscopy. In Proceedings of the 2013 IEEE MTT-S International, Seattle, WA, USA, 2–7 June 2013; pp. 1–4. Schwerthoeffer, U.; Weigel, R.; Kissinger, D. A highly sensitive glucose biosensor based on a microstrip ring resonator. In Proceedings of the 2013 IEEE IMWS-BIO Workshop, Seattle, WA, USA, 9–11 December 2013; pp. 1–3. Grenier, K.; Dubuc, D.; Poleni, P.E.; Kumemura, M.; Toshiyoshi, H.; Fujii, T.; Fujita, H. Integrated broadband microwave and microfluidic sensor dedicated to bioengineering. IEEE Trans. Microw. Theory Tech. 2009, 57, 3246–3253. [CrossRef] Chrétiennot, T.; Dubuc, D.; Grenier, K. A microwave and microfluidic planar resonator for efficient and accurate complex permittivity characterization of aqueous solution. IEEE Trans. Microw. Theory Tech. 2013, 61, 972–978. [CrossRef] Whitesides, G. The origins and the future of microfluidics. Nature 2006, 442, 368–373. [CrossRef] [PubMed] Clarke, W.L.; Cox, D.; Gonder-Frederick, L.A.; Carter, W.; Pohl, S.L. Evaluating clinical accuracy of systems for self-monitoring of blood glucose. Diabetes Care 1987, 10, 622–628. [CrossRef] [PubMed] Hofmann, M.; Fischer, G.; Weigel, R.; Kissinger, D. Microwave-based noninvasive concentration measurements for biomedical applications. IEEE Trans. Microw. Theory Tech. 2013, 61, 2195–2204. [CrossRef] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
