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.
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,
13,
14,
15,
16,
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.
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 interface the circuit with the microcontroller and Bluetooth module. All system parts were neatly 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 wristbands with integrated electronics for ECG monitoring.
The functionality of the prototype system was verified on a voluntary subject by real-time monitoring and transmission of their ECG signals to a personal computer, where the raw ECG 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, the characteristic QRS complex is easily identified in the filtered ECG signal.
To further benchmark the performance of the wearable prototype and compare it with conventional pre-gelled electrodes, ECG signals were acquired using Ag/AgCl (3M
TM Red Dot
TM 2560, 3M, St. Paul, MN, USA) electrodes while keeping the same experimental conditions including the measurement locations (left and right wrist) and the measurement circuitry. Since the miniaturization of the overall system is a key design consideration for wearability, the front-end circuit was developed for single channel data acquisition rather than simultaneous recording from multiple sources. Therefore, to record the ECG biopotentials from the same location and with the same circuit, the electrocardiograms were acquired immediately after one another with a gap of less than two minutes, which also ensured that no significant physiological changes occurred in the healthy test subject.
Figure 8a,b show the raw and filtered ECG signals obtained with conventional electrodes. For the filtering of signals obtained from conventional Ag/AgCl electrodes, the same DWT procedure outlined earlier was followed.
Even though the gap between subsequent ECG recordings is small, the existence of a time-delay prevents direct comparison of the signals in the time-domain. Therefore, to gain an understanding of the correlation between the recorded signals, ECG waveforms recorded from the Ag/AgCl electrodes and the wearable prototype were divided into P-QRS-T intervals (
Figure 9a) and shifted in the time-axis to align the R-peaks (
Figure 9b). Then, the cross-correlation between all 25 possible combinations of P-QRS-T segments were calculated. Comparison of the results of conventional electrodes with the wearable wristband show that the signals conform very well in time domain and display an average cross-correlation of 88% for the entire waveform and a maximum of 97% was achieved between two P-QRS-T segments (
Table 2).
On the other hand, frequency response characteristics of the signals were also compared by plotting their power spectra which were estimated by using the Welch periodogram with a Hamming window available in MATLAB
®. As illustrated in
Figure 10a,b, the power spectrum of the ECG signal obtained from the wearable wristband is in excellent agreement to that of the Ag/AgCl electrodes 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 50 Hz interference.
4. Discussion
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 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 have been primarily attributed to the displacements between the textile electrode and the skin which effectively cause variations in the electrode-skin impedance and epidermal biopotentials [
33]. In the 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 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 electrodes and the skin [
33].
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 the 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 the stability of the electrode-skin interface and motion artifacts may be more pronounced. Hence, to evaluate the response of the graphene textile embedded ECG wristband against motion artifacts, we have performed electrocardiogram recordings when the wrist is twisted in a circular pattern. The red trace in
Figure 11 shows an electrocardiogram recorded for a duration of eight seconds in which during the initial four seconds the wrist is in stationary condition and the typical ECG pattern with the P-QRS-T complex is clearly distinguished. This is followed by a rapid twisting of the wrist for a duration of approximately two seconds which caused artifacts and distortions in the signal. After 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 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 acquired using a data acquisition unit. The blue trace in
Figure 11 shows the electrocardiogram where the distortions that occurred during twisting of the wrist are removed. Effectively, the elastic band-based unique design of our wearable prototype allows for the minimization of possible distortions under the presence of motion and coupled with the wireless data transmission capability further post-processing can be done to achieve high quality ECG acquisition. Real-time removal of distortions due to motion artifacts, baseline wander, power-line or high-frequency noise could also be possible by embedding the related filtering algorithms to the on-board microprocessor in the prototype system.
The graphene textile embedded wristbands provided stable electrocardiogram recordings 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 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 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 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 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 the burden on the input stage of the analog front-end.
5. Conclusions
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 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 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 on wristbands which housed all of the electronic circuitry needed for portable acquisition and wireless transmission of ECG signals to a remote personal computer. The performance of the prototype wearable ECG wristband is in excellent match to that of conventional Ag/AgCl electrodes; with the 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 graphene-clad, textile-based prototype medical garments can be further developed for application to other biopotential monitoring procedures including EEG (electroencephalography) and EMG (electromyography).