sensors Article

Intelligent Medical Garments with GrapheneFunctionalized Smart-Cloth ECG Sensors Murat Kaya Yapici 1,2,3, * and Tamador Elboshra Alkhidir 4 1 2 3 4

*

Faculty of Engineering and Natural Sciences, Sabancı University, Istanbul 34956, Turkey Sabancı University Nanotechnology Research and Application Center, Istanbul 34956, Turkey Department of Electrical Engineering, University of Washington, Seattle, WA 98195, USA Department of Electrical and Computer Engineering, Khalifa University, Abu Dhabi 127788, UAE; [email protected] Correspondence: [email protected]; Tel.: +90-216-483-9000

Academic Editors: Edward Sazonov and Subhas Chandra Mukhopadhyay Received: 17 January 2017; Accepted: 22 March 2017; Published: 16 April 2017

Abstract: Biopotential signals are recorded mostly by using sticky, pre-gelled electrodes, which are not ideal for wearable, point-of-care monitoring where the usability of the personalized medical device depends critically on the level of comfort and wearability of the electrodes. We report a fully-wearable medical garment for mobile monitoring of cardiac biopotentials from the wrists or the neck with minimum restriction to regular clothing habits. The wearable prototype is based on elastic bands with graphene functionalized, textile electrodes and battery-powered, low-cost electronics for signal acquisition and wireless transmission. Comparison of the electrocardiogram (ECG) recordings obtained from the wearable prototype against conventional wet electrodes indicate excellent conformity and spectral coherence among the two signals. Keywords: graphene; electrocardiogram; conductive textile; mHealth; ECG electrode; wearable sensor; smart fabric; medical garment; wristband; arrhytmia

1. Introduction Electrocardiography is a non-invasive method to monitor cardiac activity and can reveal vital information on cardiovascular disorders, including heart rhythm abnormalities, collectively known as arrhythmias or dysrhythmias. Some arrhythmias, such as ventricular fibrillation, are emergency conditions which may lead to sudden cardiac arrest and death. On the other hand, others like atrial fibrillation, although still being serious conditions, are more subtle to detect and could develop over time with prolonged consequences, like blood clotting and stroke [1]. Therefore, to properly interpret the heart activity and diagnose possible rhythm disorders in advance, it is useful to have continuous ECG recordings for extended periods of time. For instance, Holter monitors have been used to record ECG for up to 48 h [2]. Typically, the ECG signal is acquired by mounting conductive biopotential electrodes in multiple locations across the body, and for this purpose Ag/AgCl electrodes have been widely adopted. The conventional Ag/AgCl electrodes, also referred to as “wet” electrodes, contain gel to reduce the skin-electrode contact impedance and are sustained by an adhesive layer to improve the contact to skin. However, in some cases there have been reports of skin irritation and allergic reaction due to long-term use of gel-based “wet” electrodes. In addition, the electrode performance degrades with time due to drying of the gel [3]. Therefore, standard Ag/AgCl electrodes are not favorable for long-term monitoring applications. Instead, “dry” electrodes freed from gel have been sought after, which are more suitable for long-term monitoring applications and can provide the desired comfort level to be continuously worn as part of wearable health monitoring devices. Sensors 2017, 17, 875; doi:10.3390/s17040875

www.mdpi.com/journal/sensors

Sensors 2017, 17, 875

2 of 12

So far, different types of dry electrodes have been reported which can be categorized into two types as contact and non-contact electrodes. In contact electrodes, direct physical coupling is established between the skin and the electrode, where the coupling efficiency may be further enhanced by the presence of sweat or moisture [4,5]. On the other hand, non-contact operation relies on capacitive coupling between the electrode and the skin via separation by dielectric material or air [6]. Examples of dry electrodes include microneedle structures based on silicon micromachining [7] and 3D printing technology [8], flexible sensors with thin-film metallic coatings [9], nanomaterial-doped polymer electrodes [10], and capacitive electrodes built on printed circuit boards (PCB) [11]. An alternative technology to develop dry electrodes is based on utilizing conductive textiles. Textiles are one of the most frequently used materials in daily life, and they are ideal for building wearable health monitoring devices due to their inherent flexibility, and soft and comfortable texture. Moreover, electronic components can be readily attached to conductive textiles, which greatly simplifies the development of wearable, intelligent medical garments for long-term monitoring of essential physiological signals. The seamless integration of textile-based biopotential electrodes to low-power, small-form-factor wireless transmission modules and handheld devices would make remote cardiac monitoring possible with minimal disturbance to the daily routine of the patient. To this end, the abundance and variety of clothing accessories such as snap fasteners, elastic belts, and bands further support the integration of electronics on textiles to realize wearable medical garments. For instance, tight vests, belts and shirts worn around the upper chest region were used to sense ECG signals with the help of the textile-based electrodes embedded inside the garment [12–17]. Some ideal properties of conductive textiles for wearable garments include high electrical conductivity, high mechanical durability and stability against repeated use, a high degree of comfort and flexibility, low-cost, and large-area manufacturability. To realize conductive textile electrodes, different fabrication methods have been investigated, including screen printing techniques [18,19], metallic threading [20,21], the addition of conductive polymers to fibers [22], metal coating by physical vapor deposition [23], and electroplating [24]. In an effort to further improve conductive textile technology and its applications, we have recently merged the outstanding material properties of graphene with soft fabrics using a low-cost, scalable technology and demonstrated graphene-clad textile biopotential electrodes with excellent performance [25]. In this paper, we capitalize upon the superior properties of graphene-clad conductive textiles and report the development of a fully-wearable, cloth-based, smart medical garment for ECG monitoring. The prototype is based on elastic bands which house built-in graphene-clad textiles as the ECG sensing electrodes, electronic circuitry for ECG readout and transmission, and lithium-ion batteries for power. We demonstrate successful ECG signal acquisition, processing, and wireless transmission, all achieved on the same wearable textile platform. The unique system-level design of the prototype allows ECG monitoring using minimal number of elegant clothing accessories to be comfortably worn around the wrists or neck, and alleviate the shortcomings of conventional wet-based electrodes in long-term monitoring applications. 2. Materials and Methods 2.1. Preparation and Integration of Graphene-Clad Textiles into Wearable Garments In typical twelve-lead electrocardiography ten conductive electrodes are placed across the chest and limbs; however, for ambulatory monitoring or wearable applications, as few as three electrodes may be used to acquire the ECG signal [26]. In this work, two types of textile-based, wearable ECG sensors were developed as demonstrators of intelligent medical clothing in the form of a wristband and neckband which can be easily carried on the body. In the case of wristbands, ECG measurement is achieved by placing two wristbands, one on each wrist (Figure 1a). Each wristband contains a single sensing electrode, with one of the wristbands containing an additional electrode as a reference node.

Sensors 2017, 17, 875

3 of 12

On the other hand, a neckband, as illustrated in (Figure 1b), is constructed by adding three pieces of conductive textile, of875which two serve as sensing electrodes and one as a reference Sensors out 2017, 17, 3 of 12 electrode. Sensors 2017, 17, 875

3 of 12 Left Arm (LA) electrode

Right Arm (RA) electrode

Reference

Left Arm (LA) electrode

Right Arm (RA) electrode Left electrode

Reference

Reference

Right electrode

(a)

(b)

Left

Right

Figure level 1. System level integration placement of of textile-based electrodeselectrodes to form wearable, Figure 1. System integration andandplacement textile-based to form wearable, electrode electrode Reference intelligent, medical garments for ECG sensing from the wrist or the neck: (a) wristband; (b) neckband. intelligent, medical garments for ECG sensing from(a)the wrist or the neck: (a) wristband; (b) neckband. (b)

Textile-based, dry electrodes were prepared by dipping an ordinary piece of nylon textile into a graphene (GO) suspension, was arranged ahead of time by followingpiece the protocol for Textile-based, dryoxide electrodes were which prepared by dipping an ordinary of nylon textile into a Figure SystemHummers’ level integration andthe placement of textile-based electrodes to conformal form wearable, the1.modified method. Then, dipped textile was thermally treated to allow graphene oxide (GO) suspension, which was arranged ahead time bythen following the protocol for intelligent, medical for ECG sensing the wrist or theof neck: (a)was wristband; (b) neckband. cladding of GO garments flakes on individual fibers of from the textile. The GO-clad textile dipped into the modified Hummers’ method. Then, the dipped textile thermally treatedresults to allow conformal hydrogen iodide (HI) to reduce and convert the graphene oxidewas to graphene, which ultimately graphene-clad textiles with were higher conductivity. Nylon was chosen as the textile material duethen to dryindividual electrodes prepared dipping an GO-clad ordinary piece of nylon textiledipped into a into cladding ofTextile-based, GOin flakes on fibers of the by textile. The textile was its low surface roughness, which in turn promotes higher conductivity upon coating with graphene. graphene oxide (GO) suspension, which was arranged ahead of time by following the protocol forresults hydrogen iodide (HI)description to reduce and convertofthe graphene oxide to graphene, ultimately Detailed of the preparation graphene-clad textiles was reported earlier in which [25]. the modified The Hummers’ method. Then, the dipped textile was thermally treated to allow conformal prepared samples of graphene-clad textile Nylon were cut was into smaller sizesas of the approximately in graphene-clad textiles with higher conductivity. chosen textile material due to cladding of GO on the individual fibers ofelectrodes the textile. GO-clad textile was piece then ofdipped into 6 cm × 3 flakes cm to form ECG electrodes. The wereThe subsequently glued to a larger its low surface cotton roughness, which in turn promotes higher conductivity upon coating graphene. fabric (Figure 2a), which easythe attachment of electrodes various types of clothing by with hydrogen iodide (HI) to reduce andallows convert graphene oxide totographene, which ultimately results sewing with thread. With this approach, electrodes function as disposable parts of the earlier wearable in [25]. Detailed description of the preparation of graphene-clad textiles was reported in graphene-clad textiles with higher conductivity. Nylon was chosen as the textile material due to garment and can be replaced as needed to maintain system performance. To construct the wearable The prepared samples ofwhich graphene-clad textile were cut into smaller sizeswith of approximately its low surface turn promotes higher conductivity upon coating graphene. medicalroughness, garment, elastic bandsin were chosen; as they ensured firm coupling to the skin for high fidelity 6 cm ×Detailed 3 cm todescription form the of ECG electrodes. electrodes were glued to a larger piece ECG acquisition and preparation helped minimize the susceptibility totextiles possiblesubsequently signal distortions in caseinthe the ofThe graphene-clad was reported earlier [25]. patient is in samples physical motion. Figure 2beasy showsattachment a pair ofwere wristbands, with smaller the wristband forof theapproximately left The prepared of graphene-clad textile cut into of cotton fabric (Figure 2a), which allows of electrodes tosizes various types of clothing arm having two electrodes (one of them for the reference node) separated by 1 cm, and the wristband 6 cm × 3 cm to form the ECG electrodes. The electrodes were subsequently glued to a larger piece of by sewing withforthread. With approach, electrodes as disposable the right arm has this one electrode. Similarly, three textilefunction electrodes separated apart by 3 parts cm wereof the wearable cotton fabric (Figure 2a), which allows easy attachment of electrodes to various types of clothing by attached to a neckband, where the middle electrode was used as a reference node (Figure 2c). The garment and can be replaced as needed to maintain system performance. To construct the wearable electrodes andWith elastic this bands were sandwiched between metallicas snap fasteners toparts establish an wearable sewing with thread. approach, electrodes function disposable of the medical garment, elastic bands chosen; as they ensured firm to the skin for high fidelity electrical between were the of the band facing the skin,performance. to the coupling exterior for garment and can link be replaced asinterior needed to maintain system Tointerfacing constructwith the wearable circuitry. Finally, the faces of the snap fasteners on the inner sides of the bands were covered and ECG acquisition and helped susceptibility to possible signal to distortions case the patient medical garment, elastic minimize bands werethe chosen; as they ensured firm coupling the skin forinhigh fidelity insulated with cotton fabric, in order to avoid direct contact of metal pieces to skin (Figure 2d,e). This is in physical motion. Figure 2b shows a pair of wristbands, with the wristband for the left arm having ECG acquisition helped minimize to possible distortions effectivelyand created an electrode contactthe area susceptibility of ~6 cm2 comparable to standardsignal electrode dimensions.in case the two electrodes (one of them for the reference node) separated by 1 cm, and the wristband for the patient is in physical motion. Figure 2b shows a pair of wristbands, with the wristband for the left right arm having two electrodes (one three of themtextile for theelectrodes reference node) separated by 1by cm,3and wristband arm has one electrode. Similarly, separated apart cmthe were attached to for the right arm hasmiddle one electrode. Similarly, three as textile electrodesnode separated apart by 3The cm electrodes were a neckband, where the electrode was used a reference (Figure 2c). attached to a neckband, where the middle electrode was used as a reference node (Figure 2c). The link and elastic bands were sandwiched between metallic snap fasteners to establish an electrical electrodes and elastic bands were sandwiched between metallic snap fasteners to establish an between the interior of the band facing the skin, to the exterior for interfacing with circuitry. Finally, electrical link between the interior of the band facing the skin, to the exterior for interfacing with the faces of the snap fasteners on the inner sides of the bands were covered and insulated with cotton circuitry. Finally, the faces of the snap fasteners on the inner sides of the bands were covered and fabric, insulated in order with to avoid piecescontact to skinof(Figure 2d,e). effectively an cottondirect fabric,contact in orderof tometal avoid direct metal pieces to This skin (Figure 2d,e).created This electrode contactcreated area of cm2 comparable electrodetodimensions. effectively an~6 electrode contact areato of standard ~6 cm2 comparable standard electrode dimensions. Graphene-clad textile

Wristbands

LA electrode

RA electrode

Reference

Cotton fabric

(b)

(a)

Neckband

1 cm

(c)

Left electrode Reference Right electrode Insulation layer

Snap fasteners

Graphene-clad textile

(d)

Wristbands

(e)

LA electrode

electrode Figure 2. Steps showing the integration of graphene-clad textilesRAinto elastic bands: (a) textile pieces glued on a cotton sheet; (b) sewing the textiles to form wristband; and (c) neckband; (d) insulation of snap fasteners (e) prototype of a complete graphene-clad Reference textile integrated wristband. Cotton fabric

(b)

(a) Neckband

1 cm Left electrode Reference Right electrode

(c)

Insulation layer

Snap fasteners (d)

(e)

Figure 2. Steps showing the integration of graphene-clad textiles into elastic bands: (a) textile pieces

Figure 2. Steps showing the integration of graphene-clad textiles into elastic bands: (a) textile pieces glued on a cotton sheet; (b) sewing the textiles to form wristband; and (c) neckband; (d) insulation of glued on a cotton sheet; (b) sewing the textiles to form wristband; and (c) neckband; (d) insulation of snap fasteners (e) prototype of a complete graphene-clad textile integrated wristband. snap fasteners (e) prototype of a complete graphene-clad textile integrated wristband.

Sensors 2017, 17, 875 Sensors 2017, 17, 875

4 of 12 4 of 12

2.2. Performance Evaluation 2.2. Performance To evaluateEvaluation the performance of the wearable elastic bands in biopotential signal acquisition, electrocardiogram performed on bands a voluntary, healthy signal subject using a To evaluate themeasurements performance ofwere the wearable elastic in biopotential acquisition, commercial data acquisition system (PowerLab 26T, AD Instruments, Dunedin, New Zealand). The electrocardiogram measurements were performed on a voluntary, healthy subject using a commercial subjects gave their informed consent26T, for AD inclusion before they participated in the study, and the data acquisition system (PowerLab Instruments, Dunedin, New Zealand). The subjects experimental procedures involving human subjects described in this paper followed the principles gave their informed consent for inclusion before they participated in the study, and the experimental outlined in involving the Declaration ofsubjects Helsinki. Signals from wristfollowed and neck were recorded successfully procedures human described in thisthe paper the principles outlined in the with no prior skin preparation or application of conductive gels. In case of ECG measurement from Declaration of Helsinki. Signals from the wrist and neck were recorded successfully with no prior the wrist, lead-I configuration was followed, where two bands, one on each hand, were fastened by skin preparation or application of conductive gels. In case of ECG measurement from the wrist, lead-I Velcro. Similarly, ECG signal recorded wrapping a single elastic around the configuration was the followed, wherewas twoalso bands, one onby each hand, were fastened by band Velcro. Similarly, neck, where the positions of the sensing electrodes fell roughly below the ear, around the carotid the ECG signal was also recorded by wrapping a single elastic band around the neck, where the arteries. The raw data for ECG signals obtained from andthe neck are shown Figure positions of the sensing electrodes fell roughly below thethe ear,wrist around carotid arteries.inThe raw 3a,b, data respectively. for ECG signals obtained from the wrist and neck are shown in Figure 3a,b, respectively.

Figure Figure 3. 3. Raw Raw electrocardiograms electrocardiograms recorded recorded using using aa commercial commercial data data acquisition acquisition unit. unit. (a) (a) ECG ECG signal signal obtained obtained from from the the wearable wearable wristband; wristband; and and (b) (b) ECG ECG signal signal obtained obtained from fromthe thewearable wearableneckband. neckband.

electrocardiogram is is likely to be distorted with different noise and interference sources A typical electrocardiogram power-line noise noise (50/60 (50/60 Hz) and motion artifacts. Therefore, noise filtering techniques are such as power-line interpretation. For filtering, we have applied needed to obtain a clean ECG signal for proper medical interpretation. the discrete wavelet transform (DWT) on the raw ECG recordings. Discrete wavelet transform (DWT) decomposed into different serves as a beneficial filtering tool for ECG, as it allows the signal to be decomposed frequency bands bands and, and, therefore, therefore, permits permits easy easy identification identification of of noise noise patterns patterns which usually fall within frequency characteristic frequency bands bands [27]. [27]. In Indoing doingso, so,we wehave havedecomposed decomposedthe thesampled sampledsignals signalsusing usinga tree-structure filter bank with 10 levels using Daubechies wavelet families, as they are frequently a tree-structure filter bank with 10 levels using Daubechies wavelet families, as they used in ECG processing [27]. At each level, the incoming signal was passed through a low-pass and byby a factor of of two. Following this,this, the a high-pass filter filter and andoutputs outputsfrom fromeach eachwere weredown-sampled down-sampled a factor two. Following unwanted frequency bands were eliminated from the ECG signal which resulted in denoising and the unwanted frequency bands were eliminated from the ECG signal which resulted in denoising the signal-to-noise signal-to-noise ratio. ratio. improvement of the electrocardiogram was The implementation of discrete wavelet transform to denoise the electrocardiogram ® ® MATLAB (Mathworks, Natick, MA, USA). Figure 4a,b show the filtered ECG signals performed in MATLAB from the thewristband wristband and neckband, respectively. filtering, QRS complex, is a from and neckband, respectively. After After filtering, the QRSthe complex, which is awhich distinctive distinctive feature of an ECG signal, is revealed in both of the signals where the wristband displays feature of an ECG signal, is revealed in both of the signals where the wristband displays a stronger a stronger than the The neckband. Theindifference in ECGisamplitude the larger signal than signal the neckband. difference ECG amplitude attributedistoattributed the larger to biopotential biopotential difference occurring between reciprocal electrodes placed on compared the two wrists compared difference occurring between reciprocal electrodes placed on the two wrists to those placed to those placed around the neck with small separation. around the neck with small separation.

Sensors 2017, 17, 875

5 of 12

Sensors Sensors 2017, 2017, 17, 17, 875 875

55 of of 12 12

Figure 4. 4. Filtered recordings obtained obtained from from (a) (a) wristband; and (b) neckband. Filtered ECG ECG recordings (a) wristband; wristband; and and (b) (b) neckband. neckband. Figure

2.3. System-Level System-Level Design Design 2.3. Upon verifying verifying the the performance performance of of ECG ECG measurement measurement using using commercial commercial data data acquisition acquisition unit, unit, Upon a complete, wearable prototype of a graphene-clad textile embedded band with integrated prototype of a graphene-clad textile embedded band with integrated electronics electronics a complete, wearable prototype was developed. For this this purpose, purpose, low-cost, low-cost, commercially commercially available available discrete was developed. developed. For discrete components components were were used, used, was and the ECG acquisition circuitry was custom-designed. The system-level block diagram of the the the ECG ECG acquisition acquisition circuitry circuitry was was custom-designed. custom-designed. The system-level block diagram of and the wearable prototype is shown in Figure 5, which is comprised of elastic bands embedded with wearable prototype is shown shown in Figure Figure 5, 5, which which is is comprised comprised of of elastic elastic bands bands embedded embedded with wearable graphene-clad textiles, front-end ECG acquisition circuitry, microcontroller, and Bluetooth module. graphene-clad textiles, front-end ECG acquisition acquisition circuitry, circuitry, microcontroller, and Bluetooth module. module. Ideally, ECG acquisition circuitry intended for wearable health monitoring applications should Ideally, ECG ECG acquisition acquisition circuitry circuitry intended intended for for wearable wearable health health monitoring monitoring applications applications should should Ideally, deliver deliver high high performance performance with with aa small small footprint footprint and and with with minimum minimum power power consumption. consumption. Therefore, Therefore, all sub-blocks sub-blocks were designed or selected with the goal goal to to minimize minimize the the total total size sub-blocks were were designed designed or or selected selected with with the size to to fit fit on on elastic elastic bands, bands, all bands, and wristbands were selected as a demonstrator. wristbands were were selected selected as as aa demonstrator. demonstrator. and wristbands To pick-up the activity, aa To pick-up the small small biopotentials occurring during the course ofofnormal normal cardiac activity, To small biopotentials biopotentials occurring occurringduring duringthe thecourse courseof normalcardiac cardiac activity, front-end ECG acquisition circuit, essentially a differential amplifier with high gain, is needed. The ECG acquisition circuit, essentially a differential amplifier withwith highhigh gain,gain, is needed. The afront-end front-end ECG acquisition circuit, essentially a differential amplifier is needed. schematic of the ECG front-end circuit shown in Figure 5 consists of different blocks responsible for schematic of the ECG front-end circuit shown in Figure 5 consists of different blocks responsible for The schematic of the ECG front-end circuit shown in Figure 5 consists of different blocks responsible amplification, analog filtering and suppression of common-mode signals. After amplifying and amplification, analog filtering and suppression for amplification, analog filtering and suppressionofofcommon-mode common-modesignals. signals. After After amplifying amplifying and filtering the ECG signal, an analog to digital (A/D) signal filtering the the ECG ECG signal, signal, an an analog analog to to digital digital (A/D) (A/D) convertor convertor was was used used to to digitize digitize the the acquired acquired signal filtering to to be be sent sent via via aa Bluetooth Bluetooth module module to to aa remote remote PC PC where where offline offline signal signal processing processing and and characterization characterization can be performed. This post-processing approach further eliminates the need for excessive electronic can be performed. This post-processing approach further eliminates the need for excessive electronic components that would be required for comprehensive on-board filtering, which subsequently components that would be required for comprehensive on-board filtering, which subsequently result result components in total reduction of the circuit size and power consumption. reduction of of the the circuit circuit size size and and power power consumption. consumption. in total reduction

Electrode 1 Electrode 1 V + V++ + INA116 RG1 INA116 RG1

- V - V-

Electrode 2 Electrode 2

ECG front-end circuit ECG front-end circuit C1 C1

ref. ref.

R1 R1

-

RG2 RG2

V + V++ + INA116 INA116 INA116 INA116 ref. R2 ref. R2 - V - VV+ V+

-

Reference Electrode Reference Electrode

VV-

OPA2336 OPA2336

V+ V+

+ +

R5 R5

Microcontroller Microcontroller

A/D converter A/D converter

Vout Vout

C2 C2

Bluetooth module Bluetooth module

R3 R3 R4 R4

PC / Handheld Device PC / Handheld Device

Figure 5. Block diagram of the graphene-clad textile based wearable ECG monitoring system and Figure 5. 5. Block Block diagram diagram of of the the graphene-clad graphene-clad textile textile based based wearable wearable ECG ECG monitoring monitoring system system and and Figure schematic of the ECG front-end acquisition circuit. schematic of the ECG front-end acquisition circuit. schematic of the ECG front-end acquisition circuit.

Since Since ECG ECG signals signals are are relatively relatively weak weak and and contaminated contaminated by by various various noise noise sources, sources, proper proper signal signal Since ECG signals are relatively weak and contaminated by various noise sources, proper signal amplification with minimum noise is needed for successful acquisition of biopotentials. To this amplification with minimum noise is needed for successful acquisition of biopotentials. To this end, end, amplification with minimumare noise is needed for successful acquisition important of biopotentials.of To thisECG end, instrumentation instrumentation amplifiers amplifiers are widely widely employed employed and and constitute constitute an an important part part of the the ECG instrumentation amplifiers are widely employed and constitute an importantinstrumentation part of the ECG front-end front-end front-end circuit. circuit. For For robust robust and and high high performance performance circuit circuit operation, operation, the the instrumentation amplifier amplifier should be carefully selected based on several specifications. First, the instrumentation amplifier should be carefully selected based on several specifications. First, the instrumentation amplifier

Sensors 2017, 17, 875

6 of 12

circuit. For robust and high performance circuit operation, the instrumentation amplifier should be carefully selected based on several specifications. First, the instrumentation amplifier should have high input impedance to minimize the input current offset and therefore reduce the voltage noise at the output [28,29]. Additionally, a small input current noise is desirable for low-noise amplification, and a high common-mode rejection ratio (CMRR) is helpful to mitigate common mode noises, such as power-line noise and electromagnetic interference [30]. Considering these factors, we chose an instrumentation amplifier (INA116, Texas Instruments, Dallas, TX, USA) with a high √ CMRR of ~84 dB, a low input current noise density of 0.1 fA/ Hz and high input impedance of 1015 /0.2 Ω/pF [29]. Another important consideration in the design of ECG front-end circuit is regarding the filtering of primary noise components. ECG signals are severely affected by the interference at power-line frequencies of 50 or 60 Hz. Therefore, we have designed a cascaded bandpass filter, which allows frequencies in the 0.2 to 40 Hz range relevant to ECG signals to pass, while blocking power-line and high-frequency noise components. To achieve the above passband behavior, the value of R1 and C1 were configured to 100 kΩ and 6.8 µF creating a high-pass filter, whereas R2 and C2 value were set to 39 kΩ and 0.1 µF to implement a low-pass filter. Additionally, the amplification of the ECG signal was divided between two stages to prevent output saturation. Considering the range of suitable input voltages for the A/D converter, the amplifier gain (G = 1 + 50RkΩ ) [29] for each stage was set to G = 51 by selecting the values of RG1 and RG2 as G 1 kΩ; which resulted in a total gain of ~68 dB. Alongside the gain and filtering blocks, another common circuit configuration known as the driven-right leg (DRL) was implemented to achieve single supply, battery-powered circuit operation and to further enhance the common-mode rejection [31]. The DRL circuit was formed by a voltage divider where the resistors R3 and R4 were set equal to provide a virtual ground voltage at V+ /2. This voltage was coupled to a buffer amplifier which was constructed using a standard op-amp chip (OPA2336, Texas Instruments, Dallas, TX, USA), and the output of the buffer was fed to the human body through a limiting resistor (R5 ) and reference electrode to improve the common-mode rejection. The acquired ECG signal was sampled at 500 Hz by the built-in, 10-bit analog-to-digital converter of a commercial microcontroller board (Arduino Pro Mini ATmega328, Arduino, Italy), and transmitted via a Bluetooth module (BlueSMiRF Silver, SparkFun, Boulder, CO, USA) to a personal computer. The data rate for Bluetooth transmission was set to 57.6 kbps to ensure a total sampling rate of 500 Hz. The Bluetooth module was interfaced with the computer through RS-232 communication and then the received ECG signal was further filtered by MATLAB® . A lithium-ion polymer battery with 3.7 volts and 2000 mAh capacity was used to power both the ECG front-end circuitry and the Bluetooth module. The specifications of all system blocks and component values are summarized in Table 1. Table 1. System blocks and specifications. Parts Bluethooth module (BlueSMiRF Silver)

ECG front-end circuit

A/D converter (Arduino Pro Mini)

Specifications Power consumption: 26 µA sleep, 3 mA connected, 30 mA transmit Baud rate: 57,600 Bandwidth: 0.2–44 Hz Gain: 68 dB Power supply: 3.3–5 V RG1 –RG2 , R1 , R2 , R3 –R5 : 1 kΩ, 100 kΩ, 39 kΩ, 100 kΩ C1 , C2 : 6.8 µF, 0.1 µF Resolution: 10 bit Power supply: 3.3–5 V

3. Results Using surface mount electronic components, the entire front-end circuit was built on a standard two-layer PCB with approximate dimensions of 5 cm × 5 cm, and thin wires were used to electrically

Sensors 2017, 17, 875

7 of 12

Sensors 2017, 17, 875with the microcontroller and Bluetooth module. All system parts were neatly 7 of 12 fixed interface the circuit onto one of the wristbands by directly sewing with thread through access holes drilled on the circuit fixed onto one of the wristbands by directly sewing with thread through access holes drilled on the Sensors 2017, 17, 7 of 12 boards. Figure 6 875 shows the prototype of the wearable, graphene-clad textile embedded wristbands circuit boards. Figure 6 shows the prototype of the wearable, graphene-clad textile embedded with wristbands integrated with electronics for ECG monitoring. integrated electronics for ECG monitoring.

fixed onto one of the wristbands by directly sewing with thread through access holes drilled on the circuit boards. Figure 6 shows the prototype of the wearable, graphene-clad textile embedded ECG circuit 1 cm wristbands with integrated electronics for ECG monitoring. Battery ECG recording 1 cm Leads

ECG circuit Battery

ECG recording Leads

Bluetooth

Right Arm (RA) Bluetooth electrode

Figure 6. Live

Left Arm (LA) and Reference electrodes Left Arm (LA) and Reference demonstration electrodesof

Right Arm (RA) electrode

ECG measurement using the prototype graphene-clad textile

Figure 6. Live demonstration of ECG measurement using the prototype graphene-clad textile embedded wearable wristband with integrated electronics. embedded wearable wristband with integrated electronics. Figure 6. Live demonstration of ECG measurement using the prototype graphene-clad textile

The functionality of the prototype system was verified on a voluntary subject by real-time embedded wearable wristband with integrated electronics. monitoring and transmission of their ECG signals a personal where the raw The functionality of the prototype system wastoverified on computer, a voluntary subject by ECG real-time recording is shown in Figure 7a. For accurate visualization of ECG patterns, the signal was further monitoring transmission their ECG signals a personal computer, where raw ECG Theand functionality of theofprototype system was to verified on a voluntary subject by the real-time ® . As illustrated in Figure 7b, the denoised by and applying the discrete wavelet monitoring transmission their ECGtransform signals toinaMATLAB personal computer, the raw ECG recording is shown in Figure 7a. of For accurate visualization of ECG patterns,where the signal was further characteristic QRS complex is easily identified in the filtered ECG signal. ® recording is shown in Figure 7a. For accurate visualization of ECG patterns, the signal was further denoised by applying the discrete wavelet transform in MATLAB . As illustrated in Figure 7b, To further benchmark performance oftransform the wearable prototype® and compare it with conventional denoised by applying the the discrete wavelet in MATLAB . Assignal. illustrated in Figure 7b, the the characteristic QRS complex is easily identified in the filtered(3M ECG TM Red DotTM 2560, 3M, St. Paul, pre-gelled electrodes, ECG signals were acquired using Ag/AgCl characteristic QRS complex is easily identified in the filtered ECG signal. To further benchmark the performance of theexperimental wearable prototype and comparethe it with conventional MN, USA) electrodes while the same conditions including To further benchmark thekeeping performance of the wearable prototype and itTM withmeasurement conventional TMcompare pre-gelled electrodes, ECG signalsand were acquired using Ag/AgCl (3M Red Dot 2560, St. Paul, locations (left and right wrist) the measurement circuitry. Since the miniaturization the3M, overall pre-gelled electrodes, ECG signals were acquired using Ag/AgCl (3MTM Red DotTM 2560,of 3M, St. Paul, MN, system USA) electrodes while keeping the same experimental conditions including the measurement is a key design consideration for wearability, the front-end circuit was developed for single MN, USA) electrodes while keeping the same experimental conditions including the measurement channel rather than simultaneous circuitry. recording fromthe multiple sources. Therefore, to locations (leftdata andacquisition right wrist) and the circuitry.Since Since the miniaturization ofoverall the overall locations (left and right wrist) and themeasurement measurement miniaturization of the record ECG biopotentials from the same the location and circuit with the same circuit, the system is a key design consideration for the front-end front-end circuit was developed for single system isthe a key design consideration forwearability, wearability, was developed for single electrocardiograms were acquired immediately after one another with a gapsources. of less than two minutes, channel data acquisition rather thanthan simultaneous recording from multiple Therefore, toto record channel data acquisition rather simultaneous recording from multiple sources. Therefore, which also that no significant physiological changes occurred in the healthy subject. record the ensured ECG from biopotentials the and same and with theelectrocardiograms same test circuit, thewere the ECG biopotentials the samefrom location withlocation the same circuit, the Figure 8a,b show the raw and filtered ECG signals obtained with conventional electrodes. For the electrocardiograms were one acquired immediately afterofone another withminutes, a gap of less thanalso two minutes, acquired immediately after another with a gap less than two which ensured that filtering of ensured signals obtained from conventional Ag/AgCl electrodes, the same DWTtest procedure which also that changes no significant physiological changes the healthy subject. no significant physiological occurred in the healthy testoccurred subject. in Figure 8a,b show the raw and outlined earlier was followed. Figure 8a,b show the raw and filtered ECG signals obtained with conventional electrodes. For the

filtered ECG signals obtained with conventional electrodes. For the filtering of signals obtained from filtering of signals obtained from conventional Ag/AgCl electrodes, the same DWT procedure conventional Ag/AgCl electrodes, the same DWT procedure outlined earlier was followed. outlined earlier was followed.

Figure 7. ECG signal acquired from the wrists using the prototype graphene-clad textile embedded wearable garment: (a) raw ECG recording; and (b) filtered ECG signal. Figure 7. ECG signal acquired from the wrists using the prototype graphene-clad textile embedded

Figure 7. ECG signal acquired from the wrists using the prototype graphene-clad textile embedded wearable garment: (a) raw ECG recording; and (b) filtered ECG signal. wearable garment: (a) raw ECG recording; and (b) filtered ECG signal.

Sensors 2017, 17, 875 SensorsSensors 2017, 17, 87517, 875 2017,

8 of 12 8 8 of 12 of 12

Figure 8. ECG signal acquired from the wrists using conventional Ag/AgCl electrodes with the same:

Figure 8. ECG signal acquired from the wrists using conventional Ag/AgCl electrodes with the same: (a) raw ECG recording; and (b) filtered ECG signal. Figure ECG signal acquired the wrists using conventional Ag/AgCl electrodes with the same: (a) raw8.ECG recording; and (b)from filtered ECG signal. (a) raw ECG recording; and (b) filtered ECG signal. Even though the gap between subsequent ECG recordings is small, the existence of a time-delay prevents directthe comparison of thesubsequent signals in theECG time-domain. Therefore, gainexistence an understanding of Even though gapbetween between recordings small,tothe the ofaatime-delay time-delay Even though the gap subsequent ECG recordings isissmall, existence of the correlation between the recorded signals, ECG waveforms recorded from the Ag/AgCl electrodes prevents direct direct comparison comparison of of the the signals signals in in the the time-domain. time-domain. Therefore, Therefore, to togain gainan anunderstanding understandingof of prevents and the wearable prototype were signals, divided ECG into P-QRS-T intervals (Figure 9a) and shifted in the the correlation between the recorded waveforms recorded from the Ag/AgCl electrodes the correlation between the R-peaks recorded(Figure signals, ECG waveforms recorded from the Ag/AgCl electrodes time-axis to align the 9b). Then, the cross-correlation between all 25 possible and the wearable prototype were divided into P-QRS-T intervals (Figure 9a) and shifted in the time-axis and the wearable of prototype divided P-QRS-T intervalsof(Figure 9a) and shifted in the combinations P-QRS-T were segments were into calculated. Comparison the results of conventional to align the (Figure 9b). (Figure Then, the cross-correlation between all 25 between possible combinations of time-axis to R-peaks align R-peaks cross-correlation all 25 possible electrodes with the the wearable wristband 9b). showThen, that thethe signals conform very well in time domain and P-QRS-T segments were calculated. Comparison of the results of conventional electrodes with the combinations P-QRS-T segments were calculated. Comparison results ofof conventional display anofaverage cross-correlation of 88% for the entire waveform of andthe a maximum 97% was wearable wristband show that the signals conform very well in time domain and display an average achieved between two P-QRS-T segments (Table 2). electrodes with the wearable wristband show that the signals conform very well in time domain and

cross-correlation of cross-correlation 88% for the entireofwaveform andentire a maximum of 97% achieved between two display an average 88% for the waveform andwas a maximum of 97% was P-QRS-T segments (Table 2). achieved between two P-QRS-T segments (Table 2). R-peak 1

R-peak 1

R-peak 2

R-peak 2

R-peak 3

R-peak 3

R-peak 4

R-peak 4

R-peak 5

R-peak 5

(a)

(a)

(b)

Figure 9. (a) ECG segmentation and alignment of P-QRS-T intervals recorded from conventional Ag/AgCl electrodes and the graphene-clad textile embedded wearable prototype; (b) example of an aligned P-QRS-T interval. (b)

Table 2. Cross-correlation values among all possible P-QRS-Trecorded segment pairs. Figure segmentation and of intervals from Figure 9. 9. (a) (a) ECG ECG segmentation and alignment alignment of P-QRS-T P-QRS-T intervals recorded from conventional conventional Ag/AgCl electrodes and the graphene-clad textile embedded wearable prototype; (b) example of Ag/AgCl electrodes and the graphene-clad textile embedded wearable prototype; (b) example ofan an Wearable Prototype aligned alignedP-QRS-T P-QRS-Tinterval. interval. Ag/AgCl

R-peak 1

R-peak 2

R-peak 3

R-peak 4

R-peak 5

R-peak 1 81.9% 94.4% 95.0% 92.3% 84.5% Table Table2. 2. Cross-correlation Cross-correlation values values among among all all possible possible P-QRS-T P-QRS-T segment segment pairs. pairs. R-peak 2

83.5%

87.4%

R-peak 3

93.3%

89.9% Wearable 82.1% Prototype 91.7%

Ag/AgCl 4R-peak 1 R-peak 92.8%2 R-peak 89.3% 1 R-peak 2 Ag/AgCl R-peak R-peak 1 81.9% 94.4% R-peak 5 89.0% 89.9% R-peak 1 81.9% 94.4% R-peakR-peak 2 Average: 87.4% 2 83.5% 83.5% 88.3% R-peak 3 93.3% 89.9% 3 89.3% 93.3% 89.9% R-peakR-peak 4 92.8% R-peakR-peak 5 89.9% 4 89.0% 89.3% 92.8% Average: R-peak 5 88.3% 89.0%

Average:

88.3%

81.0%

97.0%

Wearable Prototype

89.9%

85.5% 92.9%

R-peak 3 R-peak 4 R-peak 5 89.4% R-peak 3 81.7%R-peak91.8% 4 R-peak 5 95.0% 92.3% 84.5% 84.3% 81.3% 84.4% 95.0% 92.3% 84.5% 81.0% Maximum: 81.0% 97.0%97.0%97.0% 85.5% 82.1% 91.7% 82.1% 91.7%81.7% 92.9% 89.4% 84.3% 89.4% 81.7%81.3% 91.8%

84.3%

85.5% 92.9% 91.8% 84.4%

Maximum:84.4% 97.0% 81.3%

Maximum:

97.0%

Sensors 2017, 17, 875 Sensors 2017, 17, 875

9 of 12 9 of 12

On the other hand, frequency response characteristics of the signals were also compared by plotting their their power power spectra which were estimated by using the Welch periodogram periodogram with a Hamming plotting ® ® window available available in in MATLAB MATLAB . As illustrated in Figure 10a,b, the power spectrum of the ECG signal window wearable wristband wristband is is in in excellent excellent agreement agreement to to that that of ofthe theAg/AgCl Ag/AgCl electrodes obtained from the wearable both before and after filtering, where the critical QRS morphology of the ECG lying in the 0–30 Hz frequency range is accurately captured. As shown in the insets, the filtering effectively removes the interference. 50 Hz interference.

Before filtering

After filtering

Figure 10. Power spectral density of the ECG signals obtained from the graphene-clad textile Figure 10. Power spectral density of the ECG signals obtained from the graphene-clad textile embedded embedded wearable prototype and conventional Ag/AgCl electrodes: (a) before filtering; and (b) after wearable prototype and conventional Ag/AgCl electrodes: (a) before filtering; and (b) after filtering. filtering.

4. Discussion 4. Discussion Another important parameter for wearable ECG monitoring applications is robustness of the Another important parameter for wearable ECG monitoring applications is robustness of the electrode and signal acquisition system against physical motion. Textile-based ECG electrodes are electrode and signal acquisition system against physical motion. Textile-based ECG electrodes are known to be prone to motion artifacts, and for this reason we have previously reported a simple known to be prone to motion artifacts, and for this reason we have previously reported a simple adaptive filtering approach for the removal of motion artifacts [32]. The reason for such artifacts adaptive filtering approach for the removal of motion artifacts [32]. The reason for such artifacts have have been primarily attributed to the displacements between the textile electrode and the skin been primarily attributed to the displacements between the textile electrode and the skin which which effectively cause variations in the electrode-skin impedance and epidermal biopotentials [33]. effectively cause variations in the electrode-skin impedance and epidermal biopotentials [33]. In the In the developed prototype, we have noticed that the susceptibility to motion artifacts is significantly developed prototype, we have noticed that the susceptibility to motion artifacts is significantly compensated with the use of elastic bands. This is mainly due to the tight-fitting pressure applied on compensated with the use of elastic bands. This is mainly due to the tight-fitting pressure applied on the skin by the wristbands, where the applied pressure helps maintain a stable electrode-skin contact the skin by the wristbands, where the applied pressure helps maintain a stable electrode-skin contact and also lowers the electrode-skin contact impedance by reducing the air gap between the textile and also lowers the electrode-skin contact impedance by reducing the air gap between the textile electrodes and the skin [33]. electrodes and the skin [33]. The elastic bands physically limit the displacement of the graphene textile electrodes to a large The elastic bands physically limit the displacement of the graphene textile electrodes to a large extent, when they are subject to pure lateral and vertical motions. Therefore, the susceptibility of extent, when they are subject to pure lateral and vertical motions. Therefore, the susceptibility of the the wearable prototype to motion artifacts along these directions are minimized. However, in more wearable prototype to motion artifacts along these directions are minimized. However, in more complex actions such as twisting of the wrist, the resulting torsional strain would directly affect complex actions such as twisting of the wrist, the resulting torsional strain would directly affect the the stability of the electrode-skin interface and motion artifacts may be more pronounced. Hence, stability of the electrode-skin interface and motion artifacts may be more pronounced. Hence, to to evaluate the response of the graphene textile embedded ECG wristband against motion artifacts, evaluate the response of the graphene textile embedded ECG wristband against motion artifacts, we we have performed electrocardiogram recordings when the wrist is twisted in a circular pattern. have performed electrocardiogram recordings when the wrist is twisted in a circular pattern. The red The red trace in Figure 11 shows an electrocardiogram recorded for a duration of eight seconds in trace in Figure 11 shows an electrocardiogram recorded for a duration of eight seconds in which which during the initial four seconds the wrist is in stationary condition and the typical ECG pattern during the initial four seconds the wrist is in stationary condition and the typical ECG pattern with with the P-QRS-T complex is clearly distinguished. This is followed by a rapid twisting of the wrist the P-QRS-T complex is clearly distinguished. This is followed by a rapid twisting of the wrist for a for a duration of approximately two seconds which caused artifacts and distortions in the signal. duration of approximately two seconds which caused artifacts and distortions in the signal. After After termination of the wrist motion, the electrocardiogram assumes its normal pattern. termination of the wrist motion, the electrocardiogram assumes its normal pattern. To compensate for the motion artifacts, we have applied an adaptive filtering algorithm [32] which To compensate for the motion artifacts, we have applied an adaptive filtering algorithm [32] required acquisition of reference signals that are highly correlated with the motion. For this purpose, which required acquisition of reference signals that are highly correlated with the motion. For this purpose, two additional graphene textile electrodes were placed inside wristbands and signals were

Sensors 2017, 17, 875

10 of 12

Sensors 2017, 17, 875

10 of 12

two additional textile electrodes placed inside wristbands signals were acquired acquired using graphene a data acquisition unit. The were blue trace in Figure 11 shows theand electrocardiogram where using a data acquisition unit. The blue trace in Figure 11 shows the electrocardiogram the the distortions that occurred during twisting of the wrist are removed. Effectively, where the elastic distortions occurred during twisting of the prototype wrist are removed. Effectively, the elastic band-based band-basedthat unique design of our wearable allows for the minimization of possible unique design of our wearable prototype allows for the minimization of possible distortions under the distortions under the presence of motion and coupled with the wireless data transmission capability presence of motion and coupled with the wireless data transmission capability further post-processing further post-processing can be done to achieve high quality ECG acquisition. Real-time removal of can be donedue to achieve highartifacts, quality ECG acquisition. removal of distortions due could to motion distortions to motion baseline wander, Real-time power-line or high-frequency noise also artifacts, baseline wander, power-line or high-frequency noise could also be possible by embedding be possible by embedding the related filtering algorithms to the on-board microprocessor in the the related system. filtering algorithms to the on-board microprocessor in the prototype system. prototype

Figure 11. 11. Electrocardiogram Electrocardiogram recording recording showing showing motion-related motion-relatedartifacts artifacts as as aa result result of of twisting twisting of of the the Figure wrist and artifact removal by adaptive filtering. wrist and artifact removal by adaptive filtering.

The graphene textile embedded wristbands provided stable electrocardiogram recordings The graphene textile embedded wristbands provided stable electrocardiogram recordings throughout the measurements performed in this study which lasted for a duration of approximately throughout the measurements performed in this study which lasted for a duration of approximately one month. We anticipate that, with time, the graphene textiles embedded inside the wristbands one month. We anticipate that, with time, the graphene textiles embedded inside the wristbands could wear out, in which case either the graphene textiles or the entire wristband (excluding the could wear out, in which case either the graphene textiles or the entire wristband (excluding the electronics) could be readily replaced at a low cost. Owing to the scalable fabrication and integration electronics) could be readily replaced at a low cost. Owing to the scalable fabrication and integration of our graphene textiles into clothing, the electrode size can also be readily increased to adjust the of our graphene textiles into clothing, the electrode size can also be readily increased to adjust the electrode-skin contact impedance. Our measurements show that the average impedance values of the electrode-skin contact impedance. Our measurements show that the average impedance values of the graphene textile embedded wristband ranges from 87.5 kΩ to 55 kΩ in the 10–50 Hz range, which are graphene textile embedded wristband ranges from 87.5 kΩ to 55 kΩ in the 10–50 Hz range, which are in good agreement with our measurements on conventional Ag/AgCl electrodes having impedance in good agreement with our measurements on conventional Ag/AgCl electrodes having impedance values from 50.9 kΩ to 20 kΩ in the same frequency range. With increasing electrode size it could be values from 50.9 kΩ to 20 kΩ in the same frequency range. With increasing electrode size it could be possible to further reduce the contact impedance [34], which would especially be helpful to reduce possible to further reduce the contact impedance [34], which would especially be helpful to reduce the the burden on the input stage of the analog front-end. burden on the input stage of the analog front-end. 5. Conclusions 5. Conclusions In this work, we have developed a complete prototype of a wearable garment with integrated In this work, we have developed a complete prototype of a wearable garment with integrated electronics for electrocardiogram monitoring. As ECG electrodes, we have employed graphene-clad electronics for electrocardiogram monitoring. As ECG electrodes, we have employed graphene-clad conductive textiles which are soft and wearable, washable, weavable, and manufactured using a conductive textiles which are soft and wearable, washable, weavable, and manufactured using a low-cost, scalable technology. The textile electrodes were stitched inside elastic bands for easy low-cost, scalable technology. The textile electrodes were stitched inside elastic bands for easy attachment to body locations, such as the wrist and the neck, and ECG signals were successfully attachment to body locations, such as the wrist and the neck, and ECG signals were successfully acquired with minimal disruption to usual dressing habits. A battery-powered prototype was built acquired with minimal disruption to usual dressing habits. A battery-powered prototype was built on on wristbands which housed all of the electronic circuitry needed for portable acquisition and wristbands which housed all of the electronic circuitry needed for portable acquisition and wireless wireless transmission of ECG signals to a remote personal computer. The performance of the transmission of ECG signals to a remote personal computer. The performance of the prototype prototype wearable ECG wristband is in excellent match to that of conventional Ag/AgCl electrodes; wearable ECG wristband is in excellent match to that of conventional Ag/AgCl electrodes; with the with the added benefit of not requiring prior skin preparation or the application of gels. added benefit of not requiring prior skin preparation or the application of gels. We envision the possible application areas of this prototype to include stationary ECG monitoring for in-bed patients at home or possibly at a hospital, as well as monitoring during physical activity, which we are planning to investigate on a larger pool of subjects in future studies. The

Sensors 2017, 17, 875

11 of 12

We envision the possible application areas of this prototype to include stationary ECG monitoring for in-bed patients at home or possibly at a hospital, as well as monitoring during physical activity, which we are planning to investigate on a larger pool of subjects in future studies. The graphene-clad, textile-based prototype medical garments can be further developed for application to other biopotential monitoring procedures including EEG (electroencephalography) and EMG (electromyography). Acknowledgments: This work was supported in part by the Mubadala-SRC Center of Excellence for Energy Efficient Electronic Systems research contract 2013-HJ2440. Author Contributions: M.K.Y. conceived and designed the experiments; T.E.A. and M.K.Y designed the prototype, performed the experiments and analyzed the data; M.K.Y. wrote the paper. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2.

3.

4. 5. 6.

7. 8. 9.

10. 11.

12.

13. 14. 15. 16.

Jain, P.K.; Tiwari, A.K. Heart monitoring systems—A review. Comput. Biol. Med. 2014, 54, 1–13. [CrossRef] [PubMed] Fung, E.; Järvelin, M.-R.; Doshi, R.N.; Shinbane, J.S.; Carlson, S.K.; Grazette, L.P.; Chang, P.M.; Sangha, R.S.; Huikuri, H.V.; Peters, N.S. Electrocardiographic patch devices and contemporary wireless cardiac monitoring. Front. Physiol. 2015, 6, 1–10. [CrossRef] [PubMed] Marozas, V.; Petrenas, A.; Daukantas, S.; Lukosevicius, A. A comparison of conductive textile-based and silver/silver chloride gel electrodes in exercise electrocardiogram recordings. J. Electrocardiol. 2011, 44, 189–194. [CrossRef] [PubMed] McAdams, E. Biomedical electrodes for biopotential monitoring and electrostimulation. In Bio-Medical CMOS ICs, 1st ed.; Yoo, H.-J., van Hoof, C., Eds.; Springer: New York, NY, USA, 2011; pp. 31–124. Xu, P.J.; Zhang, H.; Tao, X.M. Textile-structured electrodes for electrocardiogram. Text. Prog. 2008, 40, 183–213. [CrossRef] Ueno, A.; Akabane, Y.; Kato, T. Capacitive Sensing of Electrocardiographic Potential Through Cloth From the Dorsal Surface of the Body in a Supine Position: A Preliminary Study. IEEE Trans. Biomed. Eng. 2007, 54, 759–766. [CrossRef] [PubMed] Yu, L.M.; Tay, F.E.H.; Guo, D.G.; Xu, L.; Yap, K.L. A microfabricated electrode with hollow microneedles for ECG measurement. Sens. Actuators A Phys. 2009, 151, 17–22. [CrossRef] Yang, C.-S.; Yang, B.; Zhu, H.-Y.; Liu, J.-Q.; Peng, H.-L.; Wang, L.-F. MEMS-based flexible capacitive electrode for ECG measurement. Electron. Lett. 2013, 49, 739–740. Kim, H.-L.; Kim, Y.-H.; Kim, Y.-J. Miniature electrocardiography sensor using a flexible printed circuit and MEMS technology. In Proceedings of the IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, Seoul, South Korea, 20–22 August 2008; pp. 545–550. Chen, C.-Y.; Chang, C.-L.; Chien, T.-F.; Luo, C.-H. Flexible PDMS electrode for one-point wearable wireless bio-potential acquisition. Sens. Actuators A Phys. 2013, 203, 20–28. [CrossRef] Chi, Y.M.; Cauwenberghs, G. Micropower non-contact EEG electrode with active common-mode noise suppression and input capacitance cancellation. In Proceedings of the 31st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Minneapolis, MN, USA, 3–6 Septmber 2009; pp. 4218–4221. Fuhrhop, S.; Lamparth, S.; Heuer, S. A textile integrated long-term ECG monitor with capacitively coupled electrodes. In Proceedings of the IEEE Biomedical Circuits and Systems Conference (BioCAS), Beijing, China, 26–28 November 2009; pp. 21–24. Sardini, E.; Serpelloni, M. Instrumented wearable belt for wireless health monitoring. Procedia Eng. 2010, 5, 580–583. [CrossRef] Tseng, K.C.; Lin, B.; Liao, L.; Wang, Y.; Wang, Y. Development of a Wearable Mobile Electrocardiogram Monitoring System by Using Novel Dry Foam Electrodes. IEEE Syst. J. 2014, 8, 900–906. [CrossRef] Lee, Y.D.; Chung, W.Y. Wireless sensor network based wearable smart shirt for ubiquitous health and activity monitoring. Sens. Actuators B Chem. 2009, 140, 390–395. [CrossRef] Lee, S.M.; Sim, K.S.; Kim, K.K.; Lim, Y.G.; Park, K.S. Thin and flexible active electrodes with shield for capacitive electrocardiogram measurement. Med. Biol. Eng. Comput. 2010, 48, 447–457. [CrossRef] [PubMed]

Sensors 2017, 17, 875

17.

18. 19. 20. 21. 22. 23. 24. 25. 26.

27. 28. 29. 30. 31. 32.

33. 34.

12 of 12

Weder, M.; Hegemann, D.; Amberg, M.; Hess, M.; Boesel, L.; Abächerli, R.; Meyer, V.; Rossi, R. Embroidered Electrode with Silver/Titanium Coating for Long-Term ECG Monitoring. Sensors 2015, 15, 1750–1759. [CrossRef] [PubMed] Paul, G.; Torah, R.; Beeby, S.; Tudor, J. The development of screen printed conductive networks on textiles for biopotential monitoring applications. Sens. Actuators A Phys. 2014, 206, 35–41. [CrossRef] Merritt, C.R.; Nagle, H.T.; Grant, E. Fabric-based active electrode design and fabrication for health monitoring clothing. IEEE Trans. Inf. Technol. Biomed. 2009, 13, 274–280. [CrossRef] [PubMed] Rattfält, L.; Lindén, M.; Hult, P.; Berglin, L.; Ask, P. Electrical characteristics of conductive yarns and textile electrodes for medical applications. Med. Biol. Eng. Comput. 2007, 45, 1251–1257. [CrossRef] [PubMed] Stoppa, M.; Chiolerio, A. Wearable electronics and smart textiles: A critical review. Sensors 2014, 14, 11957–11992. [CrossRef] [PubMed] Tsukada, S.; Nakashima, H.; Torimitsu, K. Conductive polymer combined silk fiber bundle for bioelectrical signal recording. PLoS ONE 2012, 7, e33689. [CrossRef] [PubMed] Silva, N.L.; Gonçalves, L.M.; Carvalho, H. Deposition of conductive materials on textile and polymeric flexible substrates. J. Mater. Sci. Mater. Electron. 2012, 24, 635–643. [CrossRef] Oh, T.I.; Yoon, S.; Kim, T.E.; Wi, H.; Kim, K.J.; Woo, E.J.; Sadleir, R.J. Nanofiber web textile dry electrodes for long-term biopotential recording. IEEE Trans. Biomed. Circuits Syst. 2013, 7, 204–211. [PubMed] Yapici, M.K.; Alkhidir, T.; Samad, Y.A.; Liao, K. Graphene-clad textile electrodes for electrocardiogram monitoring. Sens. Actuators B Chem. 2015, 221, 1469–1474. [CrossRef] Sörnmo, L.; Laguna, P. The electrocardiogram—A brief background. In Bioelectrical Signal Processing in Cardiac and Neurological Applications, 1st ed.; Sörnmo, L., Laguna, P., Eds.; Elsevier Academic Press: Burlington, MA, USA, 2005; pp. 411–452. Banerjee, S.; Gupta, R.; Mitra, M. Delineation of ECG characteristic features using multiresolution wavelet analysis method. Measurement 2012, 45, 474–487. [CrossRef] Gargiulo, G.; Bifulco, P.; Cesarelli, M.; Romano, M.; Calvo, R.A.; Jin, C. An ultra-high input impedance ECG amplifier for long-term monitoring of athletes. Med. Dev. Evid. Res. 2010, 3, 1–9. [CrossRef] Hann, M.W. Ultra Low Power, 18 bit Precision ECG Data Acquisition System; Texas Instruments Incorporated: Dallas, TX, USA, 2013. Nemati, E.; Deen, M.J.; Mondal, T. A wireless wearable ECG sensor for long-term applications. IEEE Commun. Mag. 2012, 50, 36–43. [CrossRef] Spinelli, E.M.; Martinez, N.H.; Mayosky, M.A. A single supply biopotential amplifier. Med. Eng. Phys. 2001, 23, 235–238. [CrossRef] Alkhidir, T.; Sluzek, A.; Yapici, M.K. Simple Method for Adaptive Filtering of Motion Artifacts in E-Textile Wearable ECG Sensors. In Proceedings of the 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Milan, Italy, 25–29 August 2015; pp. 3807–3810. Cömert, A.; Honkala, M.; Hyttinen, J. Effect of pressure and padding on motion artifact of textile electrodes. Biomed. Eng. Online 2013, 12, 1–18. [CrossRef] [PubMed] Celik, N.; Manivannan, N.; Strudwick, A.; Balachandran, W. Graphene-enabled electrodes for electrocardiogram monitoring. Nanomaterials 2016, 6, 156. [CrossRef] [PubMed] © 2017 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/).