evaluation of cnt/mosfet based active electrode with ecg

Çukurova Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi, 35(3), ss. 699-709, Eylül 2020Çukurova University Journal of the Faculty of Engineering and Architecture, 35(3), pp. 699-709, September 2020

Ç.Ü. Müh. Mim. Fak. Dergisi, 35(3), Eylül 2020 699

Evaluation of CNT/MOSFET Based Active Electrode with ECG,EMG, and EEG Signals

Mustafa İSTANBULLU*1, Mutlu AVCI1

1Çukurova University, Faculty of Engineering, Department of Biomedical Engineering, Adana

Geliş tarihi: 21.09.2020 Kabul tarihi: 23.10.2020

Abstract

In this study, the electrical circuit equivalents of the previously designed novel CNT/MOSFET basedactive electrode and commonly used wet electrode are simulated by applying different biopotentials.Responses of the electrode models are compared within the certain skin surface circumstances. Fouriertransforms and total harmonic distortion calculations of obtained biopotentials are examined to assess theelectrodes. The simulation results demonstrate that the CNT / MOSFET based electrode predominates thewet electrode and can measure biopotentials with high quality.

Keywords: CNT, MOSFET, Biopotential electrodes, Fourier transform, Total harmonic distortion

KNT/MOSFET Tabanlı Aktif Elektrotun EKG, EMG, EEG İşaretleri ileDeğerlendirilmesi

Öz

Bu çalışmada, önceden tasarlanmış olan özgün KNT/MOSFET tabanlı aktif elektrotun ve yaygın olarakkullanılan ıslak elektrotun elektrik devresi eşdeğerlerine, farklı biyopotansiyeller uygulanarakbenzetimleri gerçekleştirilmiştir. Elektrotların giriş sinyallerine verdiği yanıtlar, belirli cilt yüzeyikoşullarında karşılaştırılmıştır. Elektrotları değerlendirmek için elde edilen biyopotansiyellerin Fourierdönüşümleri ve toplam harmonik bozulmaları incelenmiştir. Simülasyon sonuçları, KNT/MOSFETtabanlı elektrotun ıslak elektrottan daha iyi sonuçlar verdiğini ve biyopotansiyelleri yüksek kalitedeölçebileceğini göstermektedir.

Anahtar Kelimeler:KNT, MOSFET, Biyopotansiyel elektrotlar, Fourier dönüşümü, Toplam harmonikbozulma

*Sorumlu yazar (Corresponding author): Mustafa İSTANBULLU, [email protected]

Evaluation of CNT/MOSFET Based Active Electrode with ECG, EMG, and EEG Signals

700 Ç.Ü. Müh. Mim. Fak. Dergisi, 35(3), Eylül 2020

1. INTRODUCTION

Many organs and tissues in the human bodyexhibit their health states or functions withelectrical signals called biopotentials. The signalsobtained from the activities of the heart(electrocardiogram), brain (electroencephalogram),or muscle (electromyogram) are examples ofbiopotentials. These electrophysiological signals orthe other biopotentials measured from the humanbody enable the identification of the diseases.

Biopotentials are small-amplitude signals in themicrovolts or millivolts, and their frequencies arebetween DC to several hundred hertz [1]. Thequality of the measuring electrode is crucial toacquire these small amplitude signals from the skinsurface with high precision [2]. Artefacts andnoises on a biopotential signal generally depend onthe measuring electrodes as well as the skinsurface. The surface of the human skin contains ahigh number of dead cells and hairs, leading toincreased resistivity and decreased contact surfacearea with the measuring electrode [3]. Therefore,the surface of the skin must be prepared beforeperforming the measurement. Abrasion of theoutermost stratum corneum layer decreases thecontact resistivity arising from dead cells. Besides,applying a conductive electrolytic gel on the skinsurface also decreases the resistivity and increasesthe surface area between the electrode and skin.The electrolytic gel between the skin surface andelectrode, diffuses into the skin and enablesforming a highly conductive layer.

Although the self-gelled skin electrodes presentgood performance for short durationmeasurements, signal deterioration occurs on long-duration records, since the gel dries out over aperiod of time. Besides, the conductive gel isallergenic. Gel-free dry electrodes are a promisingalternative for long-duration recordings [4]. Thereare various dry electrodes, from a pure disk metalto silicon-based microfabricated electrodescontaining on-chip amplifiers [5,8]. However, theabsence of conductive gel in dry electrodes causeshigh electrode-skin impedance, high interfacepotential, and low signal quality problems [9,10].

In the literature, dry electrode designs involvecarbon nanotubes to decrease high contactimpedance as an alternative to metallic needleelectrodes [11,12]. Researches show that thevertically aligned carbon nanotube electrodes havelow electrode-skin impedance and can increase themeasurement quality according to dry and wetelectrodes. Since CNT utilizing electrodes do notcontain electrolytic gel, which negatively affectsthe signal quality over time, it maintains the signalquality even in long-term measurements. AlthoughCNTs increase the quality of the measured signal,they should not be designed long enough to reachnerve cells when attached to the skin [13].

In this work, previously designed novelCNT/MOSFET based active electrode [14] issimulated and evaluated in terms of signal qualityfor ECG, EMG and EEG signals. The inputbiopotential signals are selected randomly from theMassachusetts Institute of Technology Beth IsraelHospital database. The biopotential signals are alsoapplied to the circuit equivalents of commonlyused wet electrodes for comparison. Simulationsare performed for two different skin surface cases.The first case is prepared skin and the second caseis unprepared skin surfaces. The simulations ofpreviously designed electrode are performed underthe worst skin case specifications. The obtainedbiopotential signals and their Fourier transformsare compared. Additionally, total harmonicdistortions of obtained signals are calculated. Thesimulation results indicate the high performanceand feasibility of the CNT/MOSFET based activeelectrode for clinical applications.

2. MATERIAL AND METHOD

2.1. The CNT/MOSFET Based Electrode

The novel CNT/MOSFET based electrodeincreases the electrode-skin contact surface area byutilizing CNTs. CNTs constitute a robust interfacecontact that leads to reduce motion artifacts.Additionally, the electrode does not requireconductive gel and abrasion process since itpenetrates the skin surface. The mentioned multi-walled, metallic CNT arrays are grown on the gate

Mustafa İSTANBULLU, Mutlu AVCI


of depletion type n-channel MOSFET. MOSFETin the electrode converts biopotentials fromvoltage to current form and provides highdielectric insulation between the skin andinstrumentation unit. Besides, the MOSFET deviceamplifies the measured signal with low-noise asreported previously. [14]. Figure 1 shows theCNT/MOSFET based active electrode model.

Figure 1. Schematic of CNT/MOSFET basedelectrode

As shown in the figure vertically aligned brush-like CNT arrays are grown on the gate terminal ofthe n-channel depletion type MOSFET. CNTspenetrate the outermost layer of the skin. Theelectrode provides stable contact and decreasedcontact resistance since the CNTs bypassed thedead skin layer. Figure 2 shows the AC equivalentof the electrode. As shown in the figure, circuitequivalent model of individual CNT in contactwith the gate terminal of the MOSFET is given atthe left hand side. Small signal AC equivalent ofthe n-channel depletion type MOSFET is locatedat the right hand side. When the electrode attachedto the skin surface, CNTs acquire biopotentials andtransfer to the MOSFET. The MOSFET convertsthe sensed biopotential to a current flowingbetween source and drain terminals.

The electrode contains several numbers ofCNT/MOSFET hybridized cells in an arraymanner. All drain and source terminals inindividual measuring cell are electricallyconnected to each other as shown in Figure 3.Thus, biopotentials from different sites of the skinacquired simultaneously.

Figure 2. AC equivalent circuit of CNT/MOSFETbased electrode

Figure 3. The complete circuit diagram of theCNT/MOSFET based electrode

2.2. The Electrode-Skin Interface

The skin-electrode interface is in fact a RCnetwork. The circuit equivalent model of a skin-wet electrode interface is given in Figure 1 (ref), inwhich the model for popular Ag/AgCl electrode isused. In Figure 1, Rd and Cd are the resistance andcapacitance associated with the electrode-electrolyte interface, respectively. Rs is theeffective resistance related with the interfaceeffects of the electrolytic gel between electrodeand the skin. Ese potential arises from the ionicconcentration difference across epidermis.Epidermis layer also behave as an electricalimpedance represented by a parallel RC circuit,Rp║Cp. The subcutaneous tissues can becharacterized by a pure resistance, Rm.



Figure 4. Circuit model of wet electrode-skininterface

The values of resistances and capacitances shownin Figure 4 depend highly upon skin condition andpreparation. Type of electrode itself andelectrolytic gel, attachment status of the electrodeto the skin surface define the numerical values.Experimentally proven typical values of theseparameters are given in Table 1 [15].

Table 1. Typical values of skin-electrodeinterface parameters

For unpreparedskin

1dR M= W 40dC nF=120uR = W 100sR = W

For wellprepared skin

10dR k= W 10dC nF=120uR = W 100sR = W

Well attachedelectrode

500pR = W

100pC nF= 100sR = W

Poor attachedelectrode

2pR k= W 20pC nF=

100sR = W

Figure 4 shows the electrical equivalent ofelectrode-skin interface in the case of wetelectrode is used for recording biopotentials.However, if the CNT/MOSFET based electrode isused, the electrode-skin interface equivalent isreduced much simpler form since the CNTpenetrated and bypassed the outer layer of the skin.The parameters associated with electrode-electrolyte interface (Rd, Cd, Rs, Ese) are theneliminated. In this case, only Rp║Cp. pair and Rmresistance are remain to exist in the electrode-skininterface. Therefore, a decrease is reasonablypredictable in the contact interface impedance andartifacts due to dead cell containing stratumcorneum layer.

2.3. Fourier Transform and TotalHarmonic Distortion

The Fourier transform is a mathematical tool torepresent a function or a waveform into thefrequency domain. All waveforms are the sum ofsinusoids at different frequencies. The discreteFourier transform (DFT) is obtained bydecomposing a sequence of values into elements ofdifferent frequencies. Thus, DFT is an extremelyuseful operation for signal analysis. The DFT of afinite-length signal of length N is given inEquation 1.

1

0

0,1,..., 1N

knk n N

nF x W for k N

-

=

= = -å (1)

where (2 / )j NNW e p-= and 1j = - .

The Fourier transform of a waveform gives thefundamental frequency components, harmonics aswell as the relative amplitudes. Therefore,analyzing the Fourier spectrum of an electrode'ssignal output provides information about therelated electrode quality.

The total harmonic distortion (THD) can becalculated by the ratio of the sum of the harmoniccomponent powers to the fundamental frequencypower. Considering a biopotential electrode, lowerdistortion in an acquired biopotential lead to amore accurate measurement. The biopotentialobtained as a result of the measurement must befree of noise. Since undesired signals mainlyappear by depending on the measuring electrode,THD calculation of obtained biopotential presentsthe electrode quality. The THD of a signal can bedefined as Equation 2.

2

2

1

N

nn

VTHD

V==å

(2)

where Vn is the RMS voltage of the nth harmonicand n=1 for fundamental frequency. THD valuesare usually expressed in percent. In this work,



THDs of each obtained biopotentials for theelectrodes within all cases are calculated. THDvalues are listed in Table 2.

3. SIMULATION RESULTS

In this work, simulations are performed onsimulation program for integrated circuit emphasis(SPICE). Plotting Fourier transforms and THDcalculations are done in the MATLABenvironment. The obtained biopotentials from thecircuit equivalents of the CNT/MOSFET basedactive electrode and the commonly used wetelectrode circuit model are compared. Two casesof skin condition are taken into consideration forcomparison. These are prepared and unpreparedskin cases. The typical numerical values of skin,gel, and electrode parameters are given in Table 1.In any case, measurements with dry electrodes arenoisier, and they contain more artifacts with

respect to wet and CNT/MOSFET basedelectrodes. Thus, a comparison with dry electrodemeasurement is not performed.

Before evaluating the electrodes with consideringbiopotentials, circuit equivalent model responsesof the CNT/MOSFET based electrode onunprepared skin, wet electrode on prepared skinwith gel and wet electrode on unprepared skin withgel are simulated. Figure 5 shows the simulationresults of the three electrode models to the 2 Hz,100 mV sinusoidal wave. Bypassing the deadstratum corneum layer with CNT/MOSFET basedelectrode has great effect on signal amplitudes asobvious in the figure. Signal amplitude obtainedfrom the CNT/MOSFET based electrode modelmaintains the 93.27% of the input signal, whereasthe signal amplitudes taken from wet electrodes onprepared and unprepared skin surfaces retain theirinput amplitudes 48.26% and 0.99%, respectively.

Figure 5. Responses of electrode circuit equivalents to the sinusoidal signal

ECG, EMG and EEG signals given in Figure 6 arealso applied to electrodes in order to observe theireffects on the signals. The EMG signal is takenfrom a healthy person with a 4000 Hz samplingfrequency, whereas the EEG signal is obtained bythe 10-20 electrode system with 500 Hz samplingfrequency. The EEG signal, given in the Figure 6(c), is the pre-frontal (Fp1) recording. Fouriertransforms of the input signals are also given inFigure 6.

Signal responses of the CNT/MOSFET basedelectrode, as well as wet electrodes in two cases, to

the ECG, EMG and EEG inputs, are shown inFigure 7, Figure 8 and Figure 9, respectively.Figure 7 shows the captured signal waveforms ofthe CNT/MOSFET based electrode and the wetelectrode models for ECG signal in certain cases.The simulation results, shown in Figure 7, are theCNT/MOSFET based electrode on unpreparedskin (a), wet electrode on prepared skin (b), andwet electrode on unprepared skin (c), respectively.Figure 7 (b) is the first case where the skin is wellprepared, and the electrolytic gel is applied. Theresults show that the amplitude of the input signalis reduced to approximately half of the baseline



shown in Figure 6 (a). In the second case, the skinis not cleaned from hairs and stratum corneumlayer, the signal quality is quite distorted, and the

amplitude of the signal is reduced one-tenth of thebaseline.

Input Signals Fourier Transforms

Figure 6. Input biopotential signals (a) ECG, (b) EMG, (c) EEG and their Fourier transforms (d), (e), (f),respectively

Biopotentials obtained from the wet electrodemodel for the second case show approximately~480 mV half-cell potential, whereas half-cellpotential of the first case is ~290 mV. Theamplitude of the signal obtained by theCNT/MOSFET based electrode is attenuatedslightly and maintains the baseline amplitude atapproximately 93%. Additionally, theCNT/MOSFET based electrode model presents thesmallest half-cell potential (~30 mV).

In Fıgure 8 (a), the EMG signal obtained from theCNT/MOSFET based electrode remains theamplitude of the input signal, whereas the

amplitude of the signal, taken from the wetelectrode on the prepared skin surface with thepresence of conductive gel, is decreasedapproximately 60% as given in Figure 8 (b).Moreover, the signal acquired from the wetelectrode on the unprepared skin is reducedconsiderably, as shown in Figure 8 (c).

Similarly, when the EEG signal is applied to thethree electrode models, comparable results areobtained with that of ECG and EMG inputs. Thesignal output from the CNT/MOSFET basedelectrode almost does not show voltage drop incontrast with the wet electrodes, as given in



Figure 9 (a). The simulation results show that theCNT/MOSFET based electrode has much lowercontact impedance and obtain biopotentials withhigh-quality. Although skin preparation and/or gelapplication are performed in the both wet electrodeacquisitions, the electrophysiological signalqualities are much less than that of theCNT/MOSFET based electrode, as shown inFigure 9 (b) and (c).

The Fourier transforms of obtained biopotentialsignals prove the abovementioned considerations.Figures 7 (d), 7 (e) 7 (f), 8 (d), 8 (e) 8 (f) and 9 (d),9 (e) and 9 (f) show the Fourier transforms of thesimulation outputs for ECG, EMG and EEG signalinputs. The figures annotated with (d) show theFourier transforms for the CNT/MOSFET based

electrode model, whereas (e) and (f) give theFourier transforms for the wet electrode onprepared skin and the wet electrode on theunprepared skin, respectively. Figures 7 (d), 8 (d)and 9 (d) exhibit almost the same Fouriertransform with respect to that of original signalsgiven in Figure 6 (d), (e) and (f). Fouriertransforms for wet electrode at the case-2 showconsiderable decrement on the amplitude ofprincipal frequency components as given in Figure7 (e), 8 (e) and 9 (e). It is obvious from the Fouriertransforms for the wet electrode at case-1, thesignal amplitude is extremely reduced.Additionally, amplitudes of noise components areincreased considerably, as shown in Figure 7 (f), 8(f) and 9 (f).


Figure 7. Simulation results to the ECG signal. (a) CNT/MOSFET based electrode, (b) Wet electrode onprepared skin (c) Wet electrode on unprepared skin, and their Fourier transforms (d), (e), (f),respectively




Figure 8. Simulation results to the EMG signal. (a) CNT/MOSFET based electrode, (b) Wet electrode onprepared skin (c) Wet electrode on unprepared skin, and their Fourier transforms (d), (e), (f),respectively

THD calculations of the obtained biopotentialsfrom each electrode model are given in Table 2. Asshown in the table, CNT/MOSFET based electrodeexhibits THD around 6.7% for ECG, EMG, andEEG signals, whereas THD for wet electrode onprepared skin is around 51.5%. The impedancevalue of dead cells containing the stratum corneumlayer of the skin depends on signal frequency, asstated before. Since the stratum corneum layerremains to exist in unprepared skin, it affects theTHD calculations for different biopotentials withdifferent frequencies. The value of THD for thewet electrode on unprepared skin is 67.09% for

EMG signal, whereas 98.66% and 97.29% forECG and EEG signals, respectively. THD valuesof the wet electrode on unprepared skin is morethan that of CNT/MOSFET based electrode andthe wet electrode on prepared skin, as it isexpected. However, THD values ofCNT/MOSFET based electrode are small andstable for all biopotentials, since the stratumcorneum layer is bypassed employing CNTs. THDcalculations prove the feasibility of theCNT/MOSFET based electrode on biopotentialmeasurements.




Figure 9. Simulation results to the EEG signal. (a) CNT/MOSFET based electrode, (b) Wet electrode onprepared skin (c) Wet electrode on unprepared skin, and their Fourier transforms (d), (e), (f),respectively

Table 2. THD values of the biopotentialsobtained from the electrodes

THD ValuesElectrode Type ECG EMG EEGCNT/MOSFETBased Electrode 6.72% 6.69% 6.72%

Wet electrodeon prepared skin 51.73% 51.27% 51.68%

Wet electrodeon unprepared skin 98.66% 67.09% 97.29%

4. CONCLUSIONS

The stratum corneum layer of the skin consistingof dead cells has negative effects on the

biopotential signal acquisition such as noise andincrease in the electrode-skin contact impedance.To improve the obtained signal quality, thenegative effects of this layer must be reduced witheither the electrode itself or applying theconductive gel. The conductive gel applied to theskin surface dries up in long-term measurementsand causes noise on the obtained biopotentials.Besides, it may cause allergic reactions.

In this work, previously designed novelCNT/MOSFET based active electrode is simulatedand evaluated in terms of signal quality for ECG,EMG and EEG signals.



The effects of stratum corneum layer and theconductive electrolytic gel on the obtainedbiopotentials are shown in the simulation results.SC layer on the human skin exhibitsresistive/capacitive medium properties that lead toa voltage drop at the output of the electrode.Comparison of simulations for prepared andunprepared skin conditions prove thisphenomenon. Although conductive gel reduces thesignificant decrease in signal amplitude, it cannotcompletely prevent the distortion and noise on thesignal, since it cannot completely eliminate theeffect of the SC layer. In contrast with the wetelectrode cases, the CNT/MOSFET basedelectrode simulated for unprepared skin withoutconductive gel. Although it is simulated at theworst case specifications, the CNT/MOSFETbased model obtained the highest quality signals.As shown in the Fourier transforms of thebiopotential signal outputs, the signal to noise ratioof the CNT/MOSFET based electrode is higherthan that of the wet electrodes. THD calculationsof the obtained biopotentials also indicate the greateffect of measuring electrode on the signal quality.The simulation results demonstrate and prove theefficiency of the CNT/MOSFET based electrode.

The CNT/MOSFET based electrode hasadvantages such as recording biopotentials withhigh signal to noise ratio, eliminating the negativeeffects of the stratum corneum skin layer,providing robust, stable skin contact, no need toboth conductive electrolytic gel usage and skinabrasion process. The simulation results prove theefficiency and importance of the previouslydesigned electrode. This electrode can be used forclinical measurements even for long-termmeasurements without biocompatibility problems.

5. REFERENCES

1. Grimnes, S., Martinsen, O.G., 2015..Bioimpedance And Bioelectricity Basics,Academic Press, 3rd Edt.

2. Plonsey, R., Barr, R.C., 2007. Bioelectricity: AQuantitative Approach, Springer US, 3rd Edt.

3. Webster, J.G., Ed., 2006. Encyclopedia ofMedical Devices and Instrumentation,

Hoboken, NJ., USA: John Wiley & Sons, Inc.4. Neuman M., 2010. Biopotential Electrodes in

Medical Instrumentation Application andDesign, Fourth Edi., Webster J., Ed. Wiley,189–240.

5. Valchinov, E.S., Pallikarakis, N.E., 2004. AnActive Electrode for Biopotential Recordingfrom Small Localized Bio-sources, Biomed.Eng. Online, 3(1), 25.

6. Yoo, J., Yan, L., Lee, S., Kim, H., Kim, B.,Yoo, H.J., 2009. An Attachable ECG SensorBandage with Planar-fashionable CircuitBoard, Proc.-Int. Symp. Wearable Comput.ISWC, 145–146.

7. Wang, L.F., Liu, J.Q., Yang, B., Yang, C.S.,2012. PDMS-Based Low Cost Flexible DryElectrode for Long-Term EEG Measurement,IEEE Sens. J., 12(9), 2898–2904.

8. Griss, P., Tolvanen, H.K., Meriläinen, P.,Stemme G., 2002. Characterization ofMicromachined Spiked BiopotentialElectrodes, IEEE Trans. Biomed. Eng., 49(6),597–604.

9. Mestrovic, M.A., Helmer, R.J.N., Kyratzis, L.,Kumar, D., 2007. Preliminary Study of DryKnitted Fabric Electrodes for PhysiologicalMonitoring, Proc. 2007 Int. Conf. Intell.Sensors, Sens. Networks Inf. Process. ISSNIP,601–606.

10. Gargiulo, G., Bifulco, P., Calvo, R.A.,Cesarelli, M., Jin, C., Van Schaik, A., 2008.Mobile Biomedical Sensing with DryElectrodes, ISSNIP 2008-Proc. 2008 Int. Conf.Intell. Sensors, Sens. Networks Inf. Process.,261–266.

11. Giulio, R., Stephen, D., Esteve, F., Josep, M.P.,Chris, R., Ernest, M., Ravi, S., Carles, G.,2006. A Dry Electrophysiology ElectrodeUsing CNT Arrays, Sensors Actuators A Phys.,132(1), 34–41.

12. Abu-Saude, M.J., Morshed, B.I., 2015.Patterned Vertical Carbon Nanotube DryElectrodes for Impedimetric Sensing andStimulation, IEEE Sens. J., 15(10), 5851-5858.

13. Abu-Saude, M., Consul-Pacareu, S., Morshed,B.I., 2015. Feasibility of Patterned VerticalCNT for Dry Electrode Sensing ofPhysiological Parameters, 2015 IEEE TopicalConference on Biomedical Wireless



Technologies, Networks, and Sensing Systems(BioWireleSS), San Diego, CA, 1-4.

14. Istanbullu, M., Avci, M., 2019. NovelMultiwalled Carbon Nanotube (MWCNT)Modified Metal Oxide Semiconductor FieldEffect Transistor (MOSFET) Based Electrodefor Electrophysiological Measurements onHuman Skin, Instrum. Sci. Technol., 47(5),545–563.

15. Liu, Y.L., Lin, D.L., Lin, Y.D., 2004. SpiceModel for Computer-Aided Design ofBiopotential Amplifier, Biomed. Eng. Appl.Basis Commun., 16(3), 151–156.

evaluation of cnt/mosfet based active electrode with ecg

Documents