1iGLU: An Intelligent Device for AccurateNonInvasive Blood Glucose-Level Monitoringin Smart HealthcarePrateek Jain Amit M. JoshiMalaviya National Institute of Technology Malaviya National Institute of TechnologySaraju P. MohantyUniversity of North TexasAbstractIn case of diabetes, fingertip prickingfor blood sample is inconvenient for glucose measurement. Invasive approaches like laboratory testand one touch glucometer enhance the risk of bloodrelated infections. To mitigate this important issue,in the current paper, we propose a novel Internetof-Medical-Things (IoMT) enabled edge-device forprecise, non-invasive blood glucose measurement. Thenovel device called Intelligent Glucose Meter (i.e.iGLU) is based on near-infrared (NIR) spectroscopyand machine learning (ML) model of high accuracy.iGLU has been validated in a hospital and bloodglucose values are stored IoMT platform for remotemonitoring by endocrinologist.I. INTRODUCTIONSMart healthcare system comprises of ambient intelligence, quality of service and also offerscontinuous support of the critical diseases monitoring [1], [2]. This system is most demandable forremote monitoring of diabetic patients with low costand rapid diagnosis [3]. Traditional blood glucosemeasurement is unable to serve everyones needat remote location. Despite having good diagnostic centers for clinical test facility in urban area,medical services are not approachable to everyoneat remote location [4], [5]. It is necessary to monitorblood glucose of diabetic patients where diagnosisfacility are not easily available. The instant diagnose of blood glucose and frequent monitoringare the recent challenges in smart healthcare. Theprocess flow of blood glucose diagnosis in smarthealthcare is depicted in Figure 1.? Non Invasive Measurement?Fasting and PostprandialBlood Glucose Values (mg/dl)Internet ConnectivityInternet Connectivity? Connecting to theavailable Endocrinologist? Blood GlucoseReadings of the Patient? Prescribed Medicine andPrecautions for InstantTreatmentCloud Storage?Nearest HospitalFor Treatment? EndocrinologistFig. 1: Blood glucose diagnosis in smart healthcare.IoMT enabled handheld non-invasive glucosemeasurement end-device has strong potential forrapid monitoring as well as to facilitate the interactwith endocrinologist to the remote located diabeticpatients where diagnosis centers and hospitals arenot easily available. According to this environment,patients measure their glucose without prickingblood and directly store to the cloud where nearbyendocrinologist can monitor the glucose data ofeach patient. The prescription would also be provided by endocrinologist to the remote locatedpatient for further treatment. The ubiquity of diabetic patients has become double from 2010 overthe world. The estimated diabetes disseminationfrom 2009 is 290 million and is expected to affect450 million people by 2030. Hence, it is essential2to develop the glucose measurement device forrapid diagnosis of diabetes. People will be moreconscious for their glucose level with frequent monitoring. Invasive method for glucose measurementis not advisable in case of continuous monitoring.Therefore, it required to design the non-invasivedevice for clinical tests, which is beneficial forhealth care. In proposed work, optical detection isinvolved. Blood glucose is predicted by machinelearning based computation model.II. STATE-OF-ART IN BLOOD GLUCOSE-LEVELMEASUREMENTBlood glucose measurement is possible usinginvasive, minimally invasive and non-invasive methods (Figure 2). Frequent pricking, as needed ininvasive methods, for glucose measurement causestrauma. Therefore, the semi-invasive approach hasthe advantage of continuous glucose monitoringwithout multiple times pricking. However, noninvasive methods can completely eliminate pricking which opens door to painless and continuousglucose monitoring (CGM).A. Invasive MethodsA low-invasive amperometric glucose monitoringbiosensor has been proposed using fine pointed glucose oxidase immobilized electrode which doesntrequire more than 1mm in length to be insertedin skin [6]. A fully implanted first-generation prototype sensor has been presented for long-termmonitoring of subcutaneous tissue glucose [7]. Thiswearable sensor which is integrated as an implantis based on a membrane containing immobilizedglucose oxidase and catalase coupled to oxygenelectrodes, and a telemetry system.B. Minimally Invasive MethodsImplantable biosensors have been deployed forcontinuous glucose monitoring [8]. Wearable minimally invasive microsystem has been explored forglucose monitoring [9]. A microsystem has beenpresented for glucose monitoring which consistsof microfabricated biosensor flip-chip bonded to atransponder chip [10]. A method has been discussedto reduce the frequency of calibration of minimallyinvasive Dexcom sensor [11]. An artificial pancreashas been represented along with glucose sensorto control diabetes [12]. But, approaches basedsemi-invasive devices have not been tried for realtime application. These wearable microsystems areneither painless nor cost effective solutions.C. Non-invasive MethodsTo make the painless system, photoacoustic spectroscopy has been introduced for non-invasive glucose measurement [13]. However, utilization ofLASER makes the setup costly and bulky. Anenzyme sensor has been explored for glucose measurement in saliva [14]. Glucose detection is possible using Intensity Modulated Photocurrent Spectroscopy (IMPS) spectroscopy that connects theelectrodes to the skin which is affected by sweat[15]. High precision level is not possible throughthese methods as sweat and saliva properties varyfor individuals. The blood glucose measurement hasalso been explored using Raman spectroscopy inlaboratory [16]. The experimental setup for Ramanspectroscopy required a large area and will not beportable. Glucose measurement has also been donefrom anterior chamber of the eye which limits itsusage of continuous monitoring [17]. Blood glucosecan been estimated using photoplethysmography(PPG) signal [18], [19].PPG signal analysis is not based on principleof glucose molecule detection. Therefore, specificwavelengths are not required for glucose estimation.Hence, iGLU is more precise compared to PPG signal analysis based system for glucose measurement.In this way, long NIR wave for optical detection hasbeen considered for glucose measurement whichis not comparatively precise glucose measurementsystem as long wave has shallow penetration [20].Therefore, small NIR wave is preferred for glucosedetection (Figure 3).Prior works related to glucose monitoring havebeen discussed which represent wearable and nonwearable approaches. Raman spectroscopy, photoacoustic spectroscopy and invasive approach basedsystems are not wearable. Minimally invasive devices which have been discussed, are implantable.Other approaches based non-invasive device arewearable. Here, iGLU is non-invasive, optical detection based wearable device for continuous glucose monitoring with IoMT framework.3Blood GlucoseMonitoring SystemInvasive Minimally Invasive Non InvasiveIn-vivo In-vitroRaman PhotoacousticSpectroscopy Polarimetry Photoplethysmography (PPG) Signal Long Wave Near- Infrared (NIR)DetectionImplanted MicroSystemGlucose ControlSystemElectrochemicalAmperometric Implanted Sensor BiosensorNear-Infrared (NIR)Spectroscopy iGLUBio-impedanceFig. 2: An overview of various blood glucose-level measurement devices or systems.LED DetectorLogged voltage valuesafter absorption andreflectance of light fromglucose moleculeSpecificWavelengthsneeded for glucosemolecule detectionNIR SpectroscopyLED DetectorLogged signal for pulseanalysis and featuresextractionSpecificWavelengths arenot requiredPPG Signal AnalysisFig. 3: PPG signal versus NIRS based glucoseestimationD. Consumer Electronics for Glucose-Level MonitoringSeveral devices have been developed for noninvasive blood glucose measurement. Some products such as glucotrack, glutrac, glucowise, DiaMonTech and device from CNOGA medical are notcommercialized. Glutrac is multi-parameter healthtest device with smart healthcare. However, theyhave limitations in terms of precise measurement.The cost of product is also high which varies inthe range of 300-400 USD. Therefore, the costeffective solution for non-invasive blood glucosemeasurement is needed.III. NOVEL DEVICE IGLU TO ADVANCE THESTATE-OF-ART IN WEARABLE FORCONTINUOUS BLOOD GLUCOSE MONITORINGNon-invasive measurement reduces the possibility of blood-related diseases. However, this approach have some limitations such as large setup, measuring object (ratina) and skin properties(including dielectric constant and sweat level).Therefore, portable non-invasive precise glucosemeasurement device for continuous monitoring isneeded. An initial example is a non-invasive glucose measurement using NIR spectroscopy andHubers regression model [21]. There are severalglucose monitoring systems which neither provideprecise measurement nor cost effective solution.These systems are not enabled for smart healthcare.The following questions are resolved in iGLU forthe advancement of smart healthcare: (1) How canwe have a device that automatically performs allthe tasks of blood-glucose monitoring at the userend without internet connectivity and stores thedata in cloud for future use by the patient andhealthcare providers? (2) Can we have a device thatcan perform automatically to avoid hassle and riskyfinger pricking all the time monitoring is needed?This article introduces an edge-device called Intelligent Glucose Meter (i.e. iGLU) for noninvasive, precise, painless, low-cost continuous glucosemonitoring at the user-end and stores the dataon cloud in an IoMT framework. A non-invasivedevice has been proposed with precise and lowcost solution. The proposed device is also integratedwith IoMT where the data is accessible to caretakerfor point of care. The device will be portable afterpackaging to use everywhere. The device is fastoperated and easy to use for smart healthcare. Theflow of proposed iGLU is represented in Figure 4.The contributions this article to advance the stateof-art in smart healthcare include the following:1) A novel accurate non-invasive glucometer(iGLU) by judiciously using short NIR waves4PostprocessingComputationModelEstimatedBloodGlucoseValueCloudStorageAnalog-toDigitalConverter(ADS1115)InfraredEmitters(940nm,1300nm)TransmittedWaveAttenuatedWaveVibrations(Stretching, Wagging,Bending)InfraredDetectorInfraredDetectorFig. 4: A conceptual overview of iGLU.with absorption and reflectance of light usingspecific wavelengths (940 and 1300 nm) hasbeen introduced. The wavelengths are judiciously selected after experimental analysiswhich has been done in material researchcenter MNIT, Jaipur (India).2) A novel accurate machine learning basedmethod for glucose sensor calibration hasbeen presented with calibrated and validatedhealthy, prediabetic and diabetic samples.3) The proposed non-invasive blood glucosemeasurement device has been integrated inIoMT framework for data (blood glucose values) storage, patient monitoring and treatmenton proper time with cloud access by both thepatient and doctor.IV. PROPOSED NON-INVASIVE BLOODGLUCOSE MEASUREMENT DEVICE (IGLU)The proposed device based on NIR spectroscopywith two short wavelengths is designed and implemented using three channels. Each channel isembedded with emitter and detector of specificwavelength for optical detection. The data is collected and serially processed by 16 bit ADC withsampling rate of 128 samples per second. Thelogged data is calibrated and validated thoroughexisting regression techniques to analyse the optimized model. The flow of data acquisition forproposed iGLU is presented in Figure 5.A. The Approach for Glucose Molecule DetectionGlucose molecule vibrates according to itsatomic structure at specific wavelengths. It is anStartPower Supply ON for Emitters and Detectors ofThree Channels and ADS 1115Connect the Arduino Uno Board to PC throughUSB PortUpload the Program of Data Processing throughADS 1115 using Arduino 1.8.5 SoftwareCheckModePlacing of objects (Fingers or Ear Lobes) inClips for Glucose MeasurementRun the Tera Term Application on PC for DataLoggingRe-connect theArduino Uno Board toPC through USB PortSampling and Averaging of Collected Data fromThree Channels for Calibration and Validation ofRegression ModelsStopUploading FailedProgram UploadedFig. 5: Process flow data acquisition for iGLU.alyzed that absorbance and reflectance are sharperand stronger in short wave NIR region [22]. Theabsorption peak of glucose spectra at 1314 nm hasbeen analyzed [23]. The non-invasive blood glucosemeasurement using 850, 950 and 1300 nm hasbeen implemented [15]. The 940 nm wavelength fordetection of glucose molecule has been identified[24]. NIR spectra of sucrose, glucose and fructoseare elaborated with CH2, CH and OH stretching at930, 960 and 984 nm, respectively [25].B. Proposed Module for Data AcquisitionProposed iGLU uses NIR spectroscopy to improve the accuracy. A 2-Layer PCB has been developed to embed infra-red emitters (MTE1300W -for1300 nm, TSAL6200 -for 940 nm, TCRT1000 -for940 nm) and detectors (MTPD1364D -for 1300 nm,3004MID -for 940 nm, TCRT1000 -for 940 nm).The hardware is designed for data acquisition fromemitters, detectors and ADC with 5V DC supply.According to the emitters and detectors, compatiblepassive components have been chosen. Architectureof glucose sensing is shown in Figure 6. Detectorswith daylight blocking filters are packaged and notaffected by sweat. ADS 1115 with 860 SPS, 16bit, I2C compatible and single ended is controlledthrough microcontroller ATmega328P and used toconvert the data (in Volts) from all channel indecimal form. The noise power and signal-to-noise5ratio (SNR) have also been found 0.08 and 25.2 dB,respectively, which show the minimum noise level.Arduino UnoBoardGNDO/PVoltage10k ?GNDPhotoDiodeLEDO/P 2Power SupplyNode1300 nmChannel 1100?10k ?GNDGNDPhotoDiodeLED O/P 3Object940 nm(Reflectance)Channel 3A D S 1 1 1 5O/P 2O/P 1O/P 3Power SupplyDataOutput940 nm(Absorption)Channel 2100 ?GNDFig. 6: Circuit topology of proposed device (iGLU).C. A Specific Prototype of the iGLUAbsorption and reflectance at 940 nm and absorption at 1300 nm are implemented for detectionof the glucose molecules. The detectors voltagedepends on received light intensity. After placingthe fingertip between emitter and detector, thevoltage values are logged. Change in light intensity depends upon glucose molecule concentration.During experiments, blood glucose is measuredthrough the invasive device standard diagnostics(SD) check glucometer for validation of the noninvasive results. The reading is taken as referenced blood glucose values (mg/dl). During theprocess, optical responses through detectors havebeen collected from 3 channels simultaneously.During measurement, the channels data is collectedin the form of voltages from 3 detectors. Thesecollected voltages correspond to referenced bloodglucose concentration. These voltage values areconverted into decimal form using 4-channel ADS1115 (Texas Inst
ruments) ADC [26]. Coherent averaging has been done after collection of responses.Specification of a iGLU prototype are presented inTable I.The prototype view of proposed iGLU is shownin Figure 7. The data is collected after fixing threefingers in free space of pads. The pads are designedin such a way that emitters and detectors are placedbeneath the surfaces of pads. Because of this, therewill be enough free spaces between the objectTABLE I: Specification of iGLU prototype Channel 1Channel 2Channel 3Measured (Ideal)Arduino Supply4.95V(5V)4.96V(5V)4.95V(5V) Forward Voltage 0.96V 1.42V 1.40V (Emitter)Forward Current(Emitter)Reverse Voltage(Detector)(1.1V)53.4mA(100mA)4.25V(5V)(1.5V)52.8mA(60mA)4.16V(5V)(1.5V)52.9mA(60mA)4.25V(5V) Output Current 0.45mA 0.5mA 0.52mA (Detector)Measurement rangeSpecificWavelength(1mA)3.2-4.68V1300nm(1mA)0.8-4.7V940nm(1mA)0.5-4.7V940nm Spectroscopy Absorption Absorption Reflectanceand sensors (emitters and detectors). Hence, theprobability of a faulty measurement is minimized.Fig. 7: Prototype view of proposed device (iGLU).V. PROPOSED MACHINE-LEARNING (ML)BASED METHOD FOR IGLU CALIBRATIONRegression models are calibrated to analyze theoptimized computation model for glucose estimation. The detectors output from three channelsare logged as input vectors for glucose prediction.The collected data from the samples is required toconvert in the form of estimated glucose values.It is necessary to develop optimal model for precise measurement and hence analysis of MAD,mARD, AvgE and RMSE are performed toensure accuracy. The estimated and reference bloodglucose concentration are calculated as BGEst andBGRef , respectively. A total of 97 samples aretaken for device calibration which include prediabetic, diabetic and healthy samples. The baseline6characteristics of samples for calibration is represented in Table II. The proposed process flow ofcalibration and validation is shown in Figure 8.TABLE II: Baseline characteristics of samples Samples BasicCharacteristicsCalibrationValidationand Testing Age (Years) Gender Wise SamplesMale:- 22-77 Male:- 53 Male:- 64Female:- 17-75 Female:- 44 Female:- 29Age (Years) PrediabeticMale:- 22-65 Male:- 18 Male:- 11Female:- 26-75 Female:- 13 Female:- 10Age (Years) DiabeticMale:- 30-68 Male:- 16 Male:- 17Female:- 30-73 Female:- 14 Female:- 11Age (Years) HealthyMale:- 22-65 Male:- 19 Male:- 36Female:- 17-70 Female:- 17 Female:- 08Calibration throughMachine LearningTechniqueError Analysis fromPredicted GlucoseConcentrationReferenced BloodConcentration (mg/dl)Prediction from CalibratedMachine Learning Model forValidationData Set(Input Voltage values withReferenced Blood Glucose)Proposed MachineLearning basedRegression ModelInput Vectors(Voltage values from3 Channels)?MARD (%)?AvgE (%)?MAD (mg/dl)?RMSE (mg/dl)Referenced BloodConcentration (mg/dl)Fig. 8: The process flow of calibration and validation of proposed device (iGLU).Deep Neural Network (DNN) based machinelearning model has been applied for precise bloodglucose prediction (Figure 9) [2]. Proposed DNNuses sigmoid activation functions and has beentrained through Levenberg-Marquardt backpropagation algorithm [15]. In proposed model, 10 hiddenneurons and 10 hidden layers are analyzed to estimate the precise blood glucose values. This modelhas been used to analyze the non-linear statisticaldata which is utilized to calibrate and validate themodel for precise measurement. Here, the voltagevalues from three channels are used as inputs ofproposed DNN model. The predicted blood glucosevalues are formed through the modeling of threechannels voltage values. Weights of the voltagevalues correlate predicted glucose values to thechannels data. The overall accuracy is improvedusing 10 hidden layers.10 Hidden Layers1 2 3101 2 3w w ww?wwwwChannel 1Channel 2Channel 3VoltageWeight of the Voltage Weight of theGlucose ValueNeural Network for Computation ofBlood GlucoseOutput LayerPrediction of Blood GlucosebiasInputOutputSigmoidActivationFunctionPredictedBloodGlucose(mg/dl)1 2 3101 10Fig. 9: The Deep Neural Network (DNN) for proposed workThe Pearsons correlation coefficient (R) is 0.953.The error analysis of calibrated machine learningmodels is represented in Table III.TABLE III: Analysis of calibration and validationof proposed combination and ML model (DNN). mARD(%)6.657.32AvgE(%)7.307.03MAD(mg/dl)12.6709.89RMSE(mg/dl)21.9511.56Calibration(Validation) VI. VALIDATION OF THE PROPOSED IGLUDEVICETo validate and test iGLU, 93 healthy, prediabeticand diabetic samples aged 17-75 are taken following medical protocols. A total of 64 males and 29females are identified during collection of these 93samples. All samples are taken in fasting, postprandial and random modes. The baseline characteristics and error analysis is represented in TableII and III, respectively. A 10-fold cross validationhas been performed to validate iGLU.To test the device stability, an experiments havebeen performed from multiple measurements ofsame sample by couple of times. For this experimental work, a volunteer has been recruited tomeasure blood glucose through iGLU and invasivemethod with time intervals of 5 minutes.A value of 10 mg/dl deviation are considered inobservations during 7 iterations of blood glucose7measurement. During analysis, 2-4 mg/dl deviationhas been observed (Figure 10(a)). A different volunteer has also been taken for another experimentalanalysis to validate the accuracy of iGLU (Figure10(b)). Measurement has been done with time interval of 60 minutes using 7 iterations. Variations (lowto high) in reference blood glucose values between8:00 AM-10:00 AM, 10:00 AM-2:00 PM and 2:00PM-4:00 PM represent the glucose intakes in theform of food. During analysis, 5-10 mg/dl deviationrepresents the stability of iGLU. It was observedthat the effect of fingers or earlobes changes isnegligible. CEG analysis is used to analyze theaccuracy of predicted glucose values from proposeddevice. CEG categorizes the devices in terms ofprecise measurement and elaborates the zones bythe difference between referenced and predictedglucose values [27]. The predicted values are in thezone A and B; then the device will be desirable.During analysis, all predicted glucose values foundin zone A and B (Figure 11).8:00 AM 8:05 AM 8:10 AM 8:15 AM 8:20 AM 8:25 AM 8:30 AM120122124126128130132134136138140B lo o d G lu c o s e C o n c e n tr a tio n (m g /d l)T im eR e fe r e n c e d B lo o d G lu c o s e C o n c e n tr a tio n (m g /d l)P r e d ic te d B lo o d G lu c o s e C o n c e n tr a tio n (m g /d l)W ith d iffe r e n t fin g e r c o m b in a tio nF r o m e a r lo b e(a) Time interval of 5 minutes8:00AM 9:00AM 10:00AM 2:00PM 3:00PM 4:00PM 5:00PM100110120130140150160170180190200210R e fe r e n c e d B lo o d G lu c o s e C o n c e n tr a tio n (m g /d l)P r e d ic te d B lo o d G lu c o s e C o n c e n tr a tio n (m g /d l)W ith d iffe r e n t fin g e r c o m b in a tio nF r o m e a r lo b eB lo o d G lu c o s e C o n c e n tr a tio n (m g /d l)T im e(b) Time interval of 60 minutesFig. 10: Predicted and reference blood glucose concentration for validation of iGLU on single sample.Reference Concentration [mg/dl]0 100 200 300 400Predicted Concentration [mg/dl]0100200300400E D AC CB D EBClarkes Error Grid AnalysisA (90%)B (10%)(a) ValidationReference Concentration [mg/dl]0 100 200 300 400 500Predicted Concentration [mg/dl]0100200300400500E D AC CB D EBClarkes Error Grid AnalysisA (94%)B (06%)(b) TestingFig. 11: CEG analysis of predicted glucose valuesVII. CONCLUSIONS AND FUTURE DIRECTIONSThis article introduced a dual short-wave spectroscopy NIR technique based non-invasive glucosemonitoring low cost (approximately 20-25 USD)device iGLU for real-life application. The errormargins for iGLU are improved compared to othernon-invasive approach based systems. After CEGanalysis, 100% samples come in the zone A and B.During analysis of possible combinations with proposed ML model, iGLU is found more optimizedcompared to other measurement device.In the future research on iGLU, we will involvemore features of IoMT. Glucose-level measurementfrom serum is a immediate next goal to furtherimprove accuracy of iGLU. Integration of stressmeasurement along with blood-glucose level is alsoin pipeline. A closed feedback from healthcareproviders-end to the end-user side for control ofeffects when needed to ensure remote healthcarewhen there may be shortage of healthcare providerscan be more effective.ACKNOWLEDGMENTThe authors would like to thank DispensaryMNIT Jaipur (India). We would also thank Dr.Navneet Agrawal (diabetologist) and his team forthe support at Diabetes, Obesity and Thyroid CentreGwalior (India).REFERENCES[1] S. Ruan, Intelligent systems for smart health care: Leveraging information for better well-being, IEEE ConsumerElectronics Magazine, vol. 8, no. 2, pp. 7171, March2019.[2] P. Sundaravadivel, K. Kesavan, L. Kesavan, S. P. Mohanty, and E. Kougianos, Smart-Log: A Deep-LearningBased Automated Nutrition Monitoring System in theIoT, IEEE Trans. Consum. Electron, vol. 64, no. 3, pp.390398, Aug 2018.[3] S. P. Mohanty and E. Kougianos, Biosensors: A TutorialReview, IEEE Potentials, vol. 25, no. 2, pp. 3540,March 2006.[4] S. P. Mohanty, U. Choppali, and E. Kougianos, Everything you wanted to know about smart cities: The Internetof things is the backbone, IEEE Consumer ElectronicsMagazine, vol. 5, no. 3, pp. 6070, July 2016.[5] P. Sundaravadivel, E. Kougianos, S. P. Mohanty, andM. K. Ganapathiraju, Everything You Wanted to Knowabout Smart Health Care: Evaluating the Different Technologies and Components of the Internet of Things forBetter Health, IEEE Consumer Electronics Magazine,vol. 7, no. 1, pp. 1828, January 2018.8[6] J. Li, P. Koinkar, Y. Fuchiwaki, and M. Yasuzawa, A finepointed glucose oxidase immobilized electrode for lowinvasive amperometric glucose monitoring, Biosensorsand Bioelectronics, vol. 86, pp. 9094, 2016.[7] J. Y. Lucisano, T. L. Routh, J. T. Lin, and D. A. Gough,Glucose monitoring in individuals with diabetes using along-term implanted sensor/telemetry system and model,IEEE Transactions on Biomedical Engineering, vol. 64,no. 9, pp. 19821993, 2017.[8] A. Sun, A. G. Venkatesh, and D. A. Hall, A multitechnique reconfigurable electrochemical biosensor: Enabling personal health monitoring in mobile devices,IEEE Transactions on Biomedical Circuits and Systems,vol. 10, no. 5, pp. 945954, Oct 2016.[9] G. Wang, M. D. Poscente, S. S. Park, C. N. Andrews, O. Yadid-Pecht, and M. P. Mintchev, Wearablemicrosystem for minimally invasive, pseudo-continuousblood glucose monitoring: The e-mosquito, IEEE Transactions on Biomedical Circuits and Systems, vol. 11,no. 5, pp. 979987, Oct 2017.[10] M. M. Ahmadi and G. A. Jullien, A wireless-implantablemicrosystem for continuous blood glucose monitoring,IEEE Transactions on Biomedical Circuits and Systems,vol. 3, no. 3, pp. 169180, June 2009.[11] G. Acciaroli, M. Vettoretti, A. Facchinetti, G. Sparacino, and C. Cobelli, Reduction of blood glucose measurements to calibrate subcutaneous glucose sensors: Abayesian multiday framework, IEEE Transactions onBiomedical Engineering, vol. 65, no. 3, pp. 587595,2018.[12] I. Pagkalos, P. Herrero, C. Toumazou, and P. Georgiou,Bio-inspired glucose control in diabetes based on ananalogue implementation of a ?-cell model, IEEE Transactions on Biomedical Circuits and Systems, vol. 8, no. 2,pp. 186195, April 2014.[13] P. P. Pai, A. De, and S. Banerjee, Accuracy enhancementfor noninvasive glucose estimation using dual-wavelengthphotoacoustic measurements and kernel-based calibration, IEEE Transactions on Instrumentation and Measurement, vol. 67, no. 1, pp. 126136, 2018.[14] M. Yamaguchi, M. Mitsumori, and Y. Kano, Noninvasively measuring blood glucose using saliva, IEEEEngineering in Medicine and Biology Magazine, vol. 17,no. 3, pp. 5963, May 1998.[15] K. Song, U. Ha, S. Park, J. Bae, and H. J. Yoo,An impedance and multi-wavelength near-infrared spectroscopy ic for non-invasive blood glucose estimation,IEEE Journal of Solid-State Circuits, vol. 50, no. 4, pp.10251037, April 2015.[16] W.-C. Shih, K. L. Bechtel, and M. V. Rebec, Noninvasive glucose sensing by transcutaneous raman spectroscopy, Journal of biomedical optics, vol. 20, no. 5,p. 051036, 2015.[17] C. W. Pirnstill, B. H. Malik, V. C. Gresham, and G. L.Cote, In vivo glucose monitoring using dual-wavelength ´polarimetry to overcome corneal birefringence in thepresence of motion, Diabetes technology & therapeutics,vol. 14, no. 9, pp. 819827, 2012.[18] E. Monte-Moreno, Non-invasive estimate of blood glucose and blood pressure from a photoplethysmographby means of machine learning techniques, Artificialintelligence in medicine, vol. 53, no. 2, pp. 127138,2011.[19] S. Habbu, M. Dale, and R. Ghongade, Estimation ofblood glucose by non-invasive method using photoplethysmography, Sadhan ¯ a¯, vol. 44, no. 6, p. 135, 2019.[20] S. Sharma, M. Goodarzi, L. Wynants, H. Ramon, andW. Saeys, Efficient use of pure component and interferent spectra in multivariate calibration, Analytica chimicaacta, vol. 778, pp. 1523, 2013.[21] P. Jain, R. Maddila, and A. M. Joshi, A precise noninvasive blood glucose measurement system using NIRspectroscopy and Hubers regression model, Optical andQuantum Electronics, vol. 51, no. 2, p. 51, 2019.[22] Y. Uwadaira, A. Ikehata, A. Momose, and M. Miura,Identification of informative bands in the shortwavelength nir region for non-invasive blood glucose measurement, Biomedical Optics Express, vol. 7, no. 7, pp.27292737, 2016.[23] W. Zhang, R. Liu, W. Zhang, H. Jia, and K. Xu, Discussion on the validity of nir spectral data in non-invasiveblood glucose sensing, Biomedical optics express, vol. 4,no. 6, pp. 789802, 2013.[24] S. Haxha and J. Jhoja, Optical based noninvasive glucosemonitoring sensor prototype, IEEE Photonics Journal,vol. 8, no. 6, pp. 111, 2016.[25] M. Golic, K. Walsh, and P. Lawson, Short-wavelengthnear-infrared spectra of sucrose, glucose, and fructosewith respect to sugar concentration and temperature,Applied spectroscopy, vol. 57, no. 2, pp. 139145, 2003.[26] P. Jain and S. Akashe, Analyzing the impact of bootstrapped adc with augmented nmos sleep transistors configuration on performance parameters, Circuits, Systems,and Signal Processing, vol. 33, no. 7, pp. 20092025,2014.[27] W. L. Clarke, The original Clarke error grid analysis(EGA), Diabetes technology & therapeutics, vol. 7,no. 5, pp. 776779, 2005.ABOUT THE AUTHORSPrateek Jain is a Research Scholar in ECEdepartment of MNIT, Jaipur, India. He can becontacted at: [email protected] M. Joshi is an Assistant Professor inDepartment of ECE, MNIT, Jaipur, India. He canbe contacted at: [email protected] P. Mohanty is the Editor in Chiefof the IEEE Consumer Electronics Magazine andProfessor in the Department of Computer Science and Engineering (CSE), University of NorthTexas (UNT), Denton, TX, USA. Contact him [email protected].