Dry electrode geometry optimization for wearable ECG devices

Recently, there has been increasing research in medical technologies particularly for smart, compact, and wearable devices. We have seen such devices finding applications in monitoring athletic performance and ambulatory care.^1–7 The cost of hospital-centric care combined with the demand for long-term recording methods attracts a demand for the development of home care assisting technologies especially in aged-care and neonatal care.^7–9 One of the most important signals we need to monitor frequently is the electrical activity of the heart. As such, electrocardiogram (ECG) plays a very important role in early detection and ongoing treatment of cardiovascular disease (CVD).¹⁰ CVDs, including coronary artery disease and stroke, are one of the major causes of death in the world. In addition to this, at least 20 × 10⁶ people experience a non-fatal heart attack or stroke every year.¹¹ Many of these people then need long and expensive medical care. Regular and repetitive ECG measurements are often needed for patients with the cardiovascular ailments. ECG data recorded over time helps obtain important diagnostic information concerning the activity of the patient's heart.

Conventionally, Ag/AgCl-based electrodes are used to measure ECG signals. The conventional Ag/AgCl electrodes, also referred to as wet electrodes contain a conductive gel to increase signal gain for improved electrical signal acquisition. These electrodes are often uncomfortable, dry out over time, and have been known to cause irritation.^12,13 Therefore, standard Ag/AgCl electrodes are not favorable for long-term ECG signal acquisition. Instead, dry electrodes that are free from the use of conductive gel have been sought after, which are more suitable for stable long-term monitoring and can provide a desired comfort level for the user.^14–18 

Contemporary literature has reported that ECG signal acquisition can be done without the need for electrolytic gels by means of active dry electrodes.^19–25 A non-contact-based sensor was made using printed circuit board technology, which makes use of a metal-based electrode.²⁶ Carbon nano-tubes (CNTs) and its array-based dry ECG electrode were shown to offer better penetration which was shown to remove the epithelial cell layer of the patient and thus reduce measurement noise.^27,28 A flexible polymer-based electrode which was made of a carbon nanotube (CNT) and polydimethylsiloxane (PDMS) composite material was proposed, which was found to function as dry ECG electrodes.²⁹ However CNTs are known to exhibit significant toxicity risks, and hence, their employment in wearable applications is disputed.^30–33 Graphene, which is a two-dimensional (2D) carbon allotrope, has received considerable interest in many scientific disciplines due to its interesting properties including excellent biocompatibility and excellent electrical properties in addition to its exceptional elasticity and stiffness.^34–37 This makes graphene a promising candidate as dry electrodes in wearable ECG sensor applications.^38–41 Graphene has also been directly coated on textiles, making textile-based electrodes that was found to give high quality ECG signals.⁴² In addition to this, there has been significant amount of research done on high-adhesive flexible electrodes particularly for electrophysiological (EP) signal acquisition applications.⁴³ While this offers an alternate opportunity in EP signal sensing, it has its limitations when it comes to practical applications. Further research needs to be done to determine the strength of EP signal acquisition using these electrodes, in addition to insufficient adaptability under different environmental conditions. Textile-based ECG electrodes of various kinds have also been investigated in recent years, due to the ability to achieve unobstructed and continuous ECG monitoring.^{14–16,44–50} These have also been majorly incorporating CNTs, graphene, and adhesive-based electrodes for sensing and likewise succumb to the same issues of toxicity, biocompatibility, and losses of strength in EP signals acquired.

In this work, we aim to develop a compact PoC wearable ECG device for continuous and remote monitoring applications. The performance of such sensors relies on the contact area, conductivity, and stability of the electrical contact. As such, it is important to understand this influence in order to design highly sensitive ECG devices with high signal-to-noise ratios. This can enable simple skin-mountable sensors that are able to pickup even small and intricate features in the signal. To do this, we have investigated various tight space geometry-based dry electrodes and incorporated the most suitable 3-electrode configuration. We have studied different electrode geometries and studied its influence on ECG sensing. The most suitable sensor configuration was integrated with wireless Bluetooth communication, offering remote sensing capabilities, and rendering them portable. This portable ECG device is compact and potentially offer improved comfort for the user without disturbing their day-to-day activities while monitoring their vitals continuously and remotely. The amplitude and time parameters, which are the critical output from the ECG signal acquisition, were compared with an ideal ECG response and were found to be well within the expected range. The as-developed ECG device was further compared with a portable, commercially available Welch Allyn^® ECG device and was found to perform on par with the bulkier commercially available device. To the best of our knowledge, this is the primal work on exploring thin-film-based electrode configuration, structure and mensuration, its effects on ECG sensing, developing an ECG patch based on as-done investigation, and comparing it with standard 12-lead ECG and commercially available portable ECG monitor. These sensors could find potential futuristic applications as standalone wearable ECG sensors or could also be embedded within wearable textiles for PoC and ambulatory care applications.

The choice of the dry electrode material here is gold (Au), as it gives the added advantage of having an electrode that is chemically inert for all biological processes, highly conductive, and biocompatible. By incorporating gold thin films, we would be able to have light-weight electrodes, with a larger surface area-to-volume ratio, thereby aiding in acquiring ECG signals more effectively. We have used electrodes with thickness of 150 nm. Furthermore, we have explored various dense electrode geometries to accommodate better conformability along the curvilinear nature of the skin, suitable for a compact wearable patch. Such dense geometries also help in minimizing material consumption, while at the same time increasing the points of electrical contacts on the region of interest, which is a major advantage for sensitive signal acquisition such as ECG. Area filling tight-space geometries that include labyrinth and curve designs with varying perimeters benefit in covering a larger areal region on the body, while ensuring minimal consumption of electrode material. For this purpose, we have investigated these fundamental medieval and modern tight-space fill geometries, such as (1) triangular labyrinth, (2) Peano curve, (3) circular labyrinth, (4) hexagonal labyrinth, (5) square labyrinth, and (6) Hilbert curve, as seen in Fig. 1. We have incorporated these variants in our 3-electrode ECG sensor. Standard photolithography processes were employed to fabricate the various ECG electrodes as shown in Fig. 2.

All the electrode geometries have a uniform electrode distance of 3.0 mm between each electrode, with a uniform thickness of 150 nm. The variation in parameters is in terms of the surface area and perimeter. This is of importance in understanding how the electrode's areal coverage over the region of study would impact signal acquisition. Such dense electrode geometries can help accomplish this by reducing material consumed, while having increased points of contact across a tight space on the skin surface. All the electrode design parameters are shown in Table I and Fig. S1. The triangular labyrinth design has the lowest surface area of 100 mm² and a perimeter of 626 mm, offering a resistance of 0.435 Ω. The Hilbert curve design has the highest surface area of 350 mm², a perimeter of 2000 mm, and offers a resistance of 0.164 Ω.

Measurements were taken by placing the electrodes in the posterior of the neck and the anterior of the chest (fourth intercostal space) regions. The reason the fourth intercostal space is chosen as the region of study at the chest region is owing to previously developed well established methods that substantiate the rationale for probing at this plane of region.⁵¹ We have ensured that the placement of the 3-electrode device on the region of measurement follows the tenets of Einthoven's triangle.⁵² Furthermore, we have also studied the same at the posterior of the neck, as this gives us information on the events of heart arrhythmias and palpitations occurring which can be caused by underlying stress, exercises, medication or due to an underlying medical condition. There is increased neural activity at the neck region and electrodes that interface at this region directly have an additional benefit from having large perimeters for a given area- this in turn helps to have better conformability and more areal coverage.^53–57 The ECG signals were acquired from the subject using different positions that include sitting, standing, and lying as shown in Fig. S2 for different stimuli: complete rest, mental stimuli (MS), physical stimuli (PS), and mental + physical stimuli (PS+MS) as shown in Fig. S3. For each specific electrode configuration, a total of 12 measurements (which is a combination of varying body positions and stimuli) were obtained to perform the extensive studies. Among the designs studied, the hexagonal labyrinth, square labyrinth, and Hilbert curve designs were found suitable for ECG signal acquisition at both regions categorically as shown in Table II, with the hexagonal labyrinth being the best candidate due to its relatively better conformability and its ability to establish intimate contact on both the anterior of the chest and posterior of the neck region.

Table II gives the comprehensive list of each electrode and its suitability in different regions of placement across all positions and stimuli. The entire dataset of all the designs studied and their corresponding response, and insets of extrapolated PQRST peaks from the ECG measurements obtained from all the electrode designs across different stimuli for both the anterior of chest and the posterior of the neck regions are shown in Figs. S4–S39. Upon evaluation, the hexagonal labyrinth-based design was found to be the best performing overall for both the chest and neck regions across all the different stimuli. The suitability of the electrodes is evaluated based on the following criteria: (i) how well the electrodes conform to the region of placement, (ii) ability to acquire ECG signals with minimal data loss and minimal noise, (iii) ability to acquire ECG signals on both the anterior of chest and posterior of neck regions, and (iv) ability to acquire ECG signals under different stimuli.

The hexagonal labyrinth design was found to have the best response among the investigated designs, matching the performance of commercially available portable ECG monitor when attached to the chest area across the fourth intercostal space. This was evaluated based on how well the electrodes were able to conform to the skin of the subject and with good quality ECG signal acquisition. Hexagonal labyrinth geometry was found to offer one of the least resistances with a large perimeter while having a relatively smaller surface area as seen from Fig. S1 and Table I, with the ability to have more point of contact with increased areal coverage on the site of inspection. The hexagonal labyrinth-based electrode design was also found to have one of the least resistances of 0.295 Ω among all the electrode designs. Measurements were taken by placing the ECG electrodes on the regions of study-posterior of the neck, and anterior of the chest along the fourth intercostal plane and enclosing it with a highly breathable commercial Allevyn bandage. This particular dressing was used owing to its antimicrobial effects and its ability to offer additional conformability of the ECG sensor along the curvature of the skin. The non-sticky yet adhesive nature of this dressing helps in enhanced locking of the ECG device in place. Figure 3 shows the schematic of how the ECG electrodes were sandwiched between the epidermis of the skin (target region) and the Allevyn^® bandage.

A typical functioning of the heart involving sequences of depolarization and repolarization and tracing the deflections of these waves as shown in Fig. 4. The mechanism and the sequence are as follows: (i) the first step in this mechanism is the initiation of atrial depolarization by the SA node which causes the P wave. (ii) This is followed by the completion of the atrial depolarization, wherein the impulse is delayed at the AV node constituting the P-R interval. (iii) Ventricular depolarization begins at the apex, causing the QRS complex (atrial repolarization occurs). (iv) following this, ventricular depolarization is complete, yielding the S–T interval. (v) then ventricular repolarization begins at the apex, causing the T wave (vi) finally followed by the completion of the ventricular repolarization.⁵⁸ 

From the perspective of a 3-lead wearable ECG sensor which is applicable to our device, we can potentially look at key cardiac parameters that is indicative of cardiac activity and events. These include the following (a) flatline ECG which can be asystole (characterized by absence of pulse and electrical activity) and pulseless electrical activity (PEA), where there is electrical activity of the heart, but the pulse is absent. Asystole is a serious medical emergency and needs to be immediately intervened by cardiopulmonary resuscitation (CPR) (b) absence of P-wave which could indicate sinus dysfunction or the presence of fibrillation or flutter waves, (c) heart rate variability (HRV) from R–R wave data, and other ventricular activity abnormalities, such as (d) prolonged QT wave intervals, (e) indication of ST wave elevation/depression, and (f) T-wave depression.

Initial measurements were made using wired configuration as shown in Fig. S40, wherein thin strands of copper wires that were attached to the electrodes were attached to snap buttons and then connected to snap cables. This was then interfaced with AD8232 controlled by Arduino Uno. Figure 5 shows ECG response of the hexagonal labyrinth design-based sensor placed on the anterior region of the chest along the fourth intercostal space to the right of the sternum in the lying posture under the following stimuli—rest, mental stress, physical stress, and physical + mental stress. Figures 6 and 7 show the same testing conditions under the sitting and standing postures, respectively. The sensor was able to obtain high quality signals in all these different postures under different stimuli.

ECG signal interpretation is a structured assessment of the waves and intervals present/absent in the acquired ECG signal. The ECG signal essentially consists of the temporal parameters along the x-axis and the amplitude parameters along the y-axis. Some of the key temporal parameters are the PR, QT, ST waves, and the P-wave interval.

The PR segment serves as the isoelectric line (baseline) of the ECG curve, and it reflects the slow impulse conduction through the atrioventricular node. Furthermore, the amplitude of any wave is measured by using the PR segment as the baseline. Table III shows the time parameters measured from the ECG signals under the lying down posture. Normally, the PR interval ranges between 0.12 and 0.22 s. We found the PR segment to be around 0.17–0.21 s when obtained from the hexagonal labyrinth electrodes. QT duration is measured from the onset of the QRS complex, and the QT interval increases when heart rate slows and decreases when the heart rate increases. QT duration is then used to find the corrected QT duration (QT_c) which is given by $= Q T duration R R interval$⁠. Interpretation of the ST segment helps in understanding and studying possibilities of myocardial ischemia. ST segment analysis helps in studying the possible presence of heart failure, ischemic ST depressions and supraventricular tachycardias. ST interval normally lies between 0.05 and 0.15 s. ST segment values ranged from 0.05 to 0.14 s when tested using the hexagonal labyrinth electrodes.

The amplitude parameters are tabulated in Table IV for the lying down posture measured using the various electrode designs. The key amplitude parameters are the P, R, Q and T-waves. The R-wave should ideally be around 0.2 mV, and we obtained 0.22–0.23 mV for R-wave duration using the hexagonal labyrinth design. The amplitude response for the P-wave should be within 0.25 mV, and we obtained a P-wave response of 0.1 mV from our hexagonal labyrinth design. R-wave should be within 1.6 mV, and we obtained a value of 0.5 mV for the measured R-wave. Q-wave measured was 0.5 mV is about 40% of the R-wave. T-wave which should be between 0.1 and 0.5 mV was measured to be 0.08 mV. The slight deviations are as would be expected in an ambulatory ECG measurement.

Further measurements were taken by attaching the hexagonal labyrinth design-based sensor on the posterior of the neck region in the lying posture under the following stimuli: rest, mental stress, physical stress, and physical + mental stress as seen in Fig. S31. Figures S32 and S33 show the same under the sitting and standing postures respectively. The sensor was able to obtain good quality signals in all these different postures under different stimuli as seen in Tables III and IV and was able to differentiate between the subject being in complete rest and under duress.

Table V shows the time parameters measured from the ECG signals under the lying down posture. Normally the PR interval ranges between 0.12 and 0.22 s. We found the PR segment to be around 0.17–0.21 s when obtained from the hexagonal labyrinth electrodes. ST interval normally lies between 0.05 and 0.15 s. ST segment values ranged from 0.09 to 0.14 s when tested using the hexagonal labyrinth designs. Normal QT wave interval is between 0.34 and 0.44 s, and we found our device to exhibit QT values between 0.38 and 0.41 s. P-wave interval is ideally 0.11 s, and we found the P-wave to lie between 0.10 and 0.12 s.

The amplitude parameters are tabulated in Table VI for the lying down posture measured using the various electrode designs. The key amplitude parameters are the P, R, Q, and T-waves. The R-wave should ideally be less than 0.2 mV. We obtained 0.22–2.25 mV for the R-wave duration using the hexagonal labyrinth electrode. Typically, the amplitude response for the P-wave should be within 0.25 mV. We measured a P-wave response between 0.06 and 0.24 mV from our hexagonal labyrinth design. R-wave should be within 1.6 mV, and we obtained a valued of 1.25 mV for the R-wave measured. Q-wave is about 25% of the R-wave, and we obtained a value of 0.5 mV for the Q-wave which is 22% of the measured R-wave. T-wave which should be between 0.1 and 0.5 mV was measured to be 0.08 mV. The slight deviations are as would be expected on an ambulatory ECG measurement, and close correlation metrics can hence be established for neck region accordingly.

Following this, we have compared and correlated the nature of response obtained from both regions to ideal 12-lead ECG response under lying conditions (matching standard posture of ideal 12-lead ECG sensing) across different stimuli. As in the case of a typical ambulatory measurement, the response parameters are usually corrected with a certain factor to be correlated with typical 12-lead ECG measurements. Figure 8 shows the temporal and the amplitude parameters of the hexagonal labyrinth design-based sensor and the ideal 12-lead ECG response. As shown in Fig. 8(a), for the chest region, the temporal parameters (PR, QT, ST waves, and P-wave interval) are closely correlated. Figure 8(c) shows the temporal parameters obtained from the neck region, and they were found to be closely correlated as well, with deviations when stimuli are applied, for both positions. The amplitude parameters for the hexagonal labyrinth-based sensor for the chest and neck regions are shown in Figs. 8(b) and 8(d). It can be seen that the P, Q, and T wave parameters are closely correlated, and the differential R-wave parameter is held as the device specific variation and correlated with standard ECG measurements. We can also obtain a clear demarcation between the state of the subject whether the subject is under rest or under a certain duress. These device-specific responses can be taken into consideration during the development phase of an ECG sensor incorporating the specific electrode configuration. Obtaining more dataset across different categories of subjects (number, age, body composition, health conditions) can further give us a comprehensive understanding and correlation metrics to calibrate and associate with this device specific ECG measurements.

Figure S41 shows the Welch Allyn ECG device, the corresponding lead configuration and the ECG measurements taken using the 12-lead configuration. Tables S5 and S6 give the temporal parameters and the amplitude parameters respectively measured using the Welch Allyn device. As shown in Figs. S42 and S43, the obtained temporal and amplitude parameters from the as-developed device were compared with ECG sensing response acquired from commercially available Hillrom Welch Allyn Wireless 12-electrode configuration ECG device for the lying down position under different stimuli.

Comparisons of the hexagonal labyrinth-based 3-electrode ECG sensor with the Welch Allyn monitor were made and their correlation is shown in Fig. 9. The correlation in the amplitude parameters can be seen in Figs. 9(a)–9(d). Being ambulatory device, the devices each exhibit a characteristic device specific slight variation in measurement. This can be used in calibration of the device to match with expected ideal 12-lead ECG response. When a stimulus is applied as shown in Figs. 9(b)–9(d), as expected, there is a significant change in the amplitude response, indicating the subject is under duress. The temporal response parameters across the devices and their correlation are shown in Figs. 9(e)–9(h). As seen in Fig. 9(e), under the rest condition, the PR, QT and ST waves, and P-wave interval seem to be closely correlated.

Under different applied stimuli, as seen in Figs. 9(f)–9(h), the as-developed hexagonal labyrinth-based ECG sensor was found to be more closely correlated with ideally expected ECG values. The Welch Allyn^® device also exhibited closely correlated values for almost all temporal parameters, except for the P-wave interval. These slight deviations can be associated with the inherent electronic system design and the ambulatory device-dependent acquisition. By extending this study to a larger sample of subjects, we can further study and establish the correlation values and the device calibration metrics comprehensively.

In terms of signal acquisition, the as-fabricated ECG sensor was found to be capable of performing in the capacity of a potential wearable ECG sensor patch. It was also found that the margin of deviation in the temporal and amplitude parameters between the herewith reported hexagonal labyrinth-based ECG sensor and the commercially available portable Welch Allyn ECG device can be associated with the fact that the wired 12-electrode Welch Allyn monitor has complex electronics and amplification circuitry that is able to compensate for signal gain and signal-to-noise ratio (SNR). The reported sensor is proposed to be a simpler 3-electrode type, lightweight ECG patch that is compact and offers portable and wireless ECG sensing capabilities. The Welch Allyn device is comparatively clunky and weighs 200 grams. The entire electrode patch weighs 5 grams and together packed with the Allevyn dressing and battery has a net weight of 12 grams, with a compact footprint of 6.5 cm (length) × 6.5 cm (width) × 1.2 cm (height). Furthermore, the Welch Allyn device makes use of thick paste of silver/silver chloride (Ag/AgCl) material (wet electrode gel), whereas our device is based entirely on dry electrodes incorporating gold (Au) thin films with thickness of 150 nm, which is less than the diameter of human hair strand, and no electrode gel playing a role in ECG signal acquisition.

Skin-contact impedance is an important parameter when considering biopotential electrodes' design. We conducted skin-contact impedance measurements of all fabricated electrodes and have compared them with commercially available Ag/AgCl electrodes to gather insights into how these parameters contribute toward performance. All measurements were conducted using a Fluke PM6306 1 MHz Programmable RCL component tester. The measurements were taken from 50 Hz to 1 MHz. Across all electrodes, with increasing frequency, the impedance decreased proportionally. It can be understood that by using higher frequency acquisition systems, skin-contact impedance can be greatly reduced, leading to further improvisation on ECG signal quality. Table VII shows the skin-contact impedance measurements for the commercial Ag/AgCl-based electrodes and their comparison with our fabricated electrodes. The Ag/AgCl electrodes when tested offered an impedance of 128.0 kΩ at 50 Hz and at 1 MHz it offered an impedance of 437.0 Ω. Our hexagonal labyrinth electrodes offered an impedance of 250.0 kΩ and 251.0 Ω at 50 Hz and 1 MHz, respectively. The highest impedances were offered by the triangular labyrinth electrodes at 6.0 MΩ and 5.0 kΩ for 50 Hz and 1 MHz, respectively, followed by Hilbert curve at 2.6 MΩ and 1.0 kΩ for 50 Hz and 1 MHz, respectively. The lowest impedance was from Peano curve electrodes at 217.0 kΩ and 214.0 Ω for 50 Hz and 1 MHz, respectively. Furthermore, a table comprising the various electrodes' materials, their skin impedance, SNR, surface area and perimeter, drawing comparisons against commercial Ag/AgCl electrodes is given in Table S7.

We further prototyped a fully functional, compact, wearable ECG sensor incorporating thin films of dry electrodes, that is capable of continuous monitoring of cardiac activity. The AD8232 discussed above is set as a peripheral to the DA14531 BLE module. The ECG signal coming out of AD8232 is collected via PIN 07, which is an ADC pin of the BLE module. The firmware of the BLE module is set to achieve an ECG sampling rate of 50 Hz. Once the raw hexadecimal data from the BLE module are received by the smartphone app, it is converted to ASCII data and displayed on the screen of the smartphone as depicted in Fig. 10. Bluetooth wireless measurements were done using a custom designed Bluetooth module. Surface mounted devices (SMDs) were used to design the BLE module. Dialog Semiconductor GmbH DA14531MOD-00F01002 was the BLE 5.1 SMD component of choice, capable of operating at a frequency of 2.4–2.4835 GHz at an output of 2.2 dBm and sensitivity of –93 dBm. The BLE module was powered by a 3.3 V coin battery, offering ∼12 h of battery life. Further battery power optimization would potentially be explored in future works. The BLE signal communication range was ≃40 m.

The advantage of having the capability to attach the as-developed ECG device on both the anterior of the chest and the posterior of the neck region has huge implications in monitoring athletic performance, in neonatal care, rehabilitation, aged care sectors, patient transport, cardiac monitoring of patients with pacemaker, and patients with dementia, where continuous and uninterrupted acquisition of ECG is paramount while ensuring that the device is not easily prone to user impediment. This as-developed 3-electrode, wireless ECG device evidently has the ability of providing key cardiac information and could potentially be used as a screening device in POC settings.

The as-developed water-proof sensor was worn by the user for a period of 7 days to check for biocompatibility. By the end of the 7 days, no skin irritation, rashes, or allergies were noticed as shown in Fig. S44, which makes this sensor a potential candidate for long-term PoC applications which can remain unaffected when the user showers, sweats and during the day-to-day routine of a user.

We report a compact, waterproof, comfortable 3-electrode-based wearable ECG sensor that is dry (electrodes), for continuous monitoring of cardiac parameters. which aids in observing abnormalities and palpitations, capable of differentiating between user's condition of rest/unrest. The hexagonal labyrinth-based electrode design was found to be the best candidate among the tight-space designs evaluated, which benefits in establishing more contact points on the surface of the skin, providing better signal acquisition while at the same time minimizing material consumption. This developed 3-electrode type compact, wearable ECG device was compared with commercially available Welch Allyn 12-lead ECG monitor and was found to operate on par with it. Its light weight, compactness, wireless measurement capabilities, ability to acquire ECG signals from both the chest and the neck area while being able to differentiate between rest and duress measurements under varying stimuli, and its capability to offer high quality ECG sensing within a 3-electrode configuration system makes it possible to have futuristic potential applications in wearable ECG sensing particularly in ambulatory care and PoC applications wherein measurements can be made using the device standalone or embedded within wearable fabrics.

The entire fabrication was conducted in an ISO 5 class 1000 cleanroom facility. Heidelberg 150 Advanced Maskless Aligner (MLA) was employed for lithography process, with a defocus of −2 and dosage of 120 mJ/cm². Standard photolithography techniques were employed in the fabrication of thin film ECG sensor electrodes. AZ5214E photoresist was used throughout the lithography process and the parameters employed were as shown in supplementary material Table S7. AZ5214E photoresist was spin-coated onto polyimide films at a speed of 3000 rpm, an acceleration of 1000 rpm/s for a duration of 30s. This was further soft baked at 95 °C for 90 s. Gold (Au) was the electrode material of choice. 30 nm of chromium (Cr) as an adhesion layer and 150 nm of Au was deposited using e-beam deposition (PVD75, Kurt J. Lesker) at the rate of 7 Å/s at an ultrahigh vacuum of <10⁻⁶Torr to achieve a uniform and controlled film thickness. Au was deposited with operating current of 76.5 mA and Cr was deposited operating current of 39.5 mA. Au source material was placed in a graphite crucible liner, and Cr source material was placed in a FABMATE^® crucible liner. Allevyn^® Ag Gentle Border, which is a silicone gel adhesive-based antimicrobial hydro-cellular bandage, was used as the dressing layer.

The human subject that volunteered for the ECG testing was a healthy adult male of 28 years of age with no history of cardiac ailments. All measurements were taken in a stress free environment, with the exception of control stimuli used for measurement purpose. The subject was not caffeinated and had 8 h of sleep prior to testing. All ECG measurements were taking by cleaning the region of interest with ethanol wipes before placing the sensor. A body composition analysis of the adult subject was done prior to ECG testing in order to rule out any abnormalities that might hinder normal ECG analysis. The adult human subject weighed 73 kg with a skeletal muscle mass (SMM) of 31.4 kg, soft lean mass of 50.5 kg, fat-free mass of 53.4, body fat mass of 18.1 kg, body mass index (BMI) of 23.3 kg/m², percent body fat (PBF) of 22.1%.

The commercial 12-electrode configuration-based ECG device that was chosen for comparative study was the Hillrom Welch Allyn^® wireless device.

Arduino 1.0.5 IDE (integrated development environment) was used to interface the AD8232 component with our as developed ECG electrodes, operating at 3.3 V and at 8 MHz clock speed. Measurements were started by initializing a 10 s delay, after which continuous measurements were taken for 10 000 ms, for different stimuli under different positions. All the commercial silver/silver chloride-based electrodes were replaced with our labyrinth and curve design-based electrodes and then interfaced with the AD8232 hardware. The GND, 3.3 V of AD8232 was connected to the GND and 3.3 V, respectively, of the Arduino board. The output of the AD8232 was connected to pin A0 of the Arduino, LO- and LO+ of the AD8232 was connected to PIN 11 and PIN 10 of the Arduino, respectively. The SDN pin was kept unconnected.

See the supplementary material for the supporting information on sensor data set and plots, detailed methods, and fabrication process parameters.

This work was performed in part at the Micro Nano Research Facility at RMIT University in the Victorian Node of the Australian National Fabrication Facility (ANFF). We acknowledge personnel and project funding from the Cooperative Research Centres Projects (CRC-P) and ARC Research Hub for Connected Sensors for Health. We thank Ms. Gayatri Chaturvedi for their contribution on the illustrations used in this work.

The authors have no conflicts to disclose.

Peter Francis Mathew Elango: Conceptualization (equal); Formal analysis (lead); Investigation (lead); Methodology (equal); Validation (lead); Visualization (lead); Writing – original draft (lead). Shanmuga Sundar Dhanabalan: Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Md Rokunuzzaman Robel: Formal analysis (supporting); Investigation (supporting); Validation (supporting). Sherly Pushpam Elango: Investigation (supporting); Methodology (supporting). Sumeet Walia: Supervision (supporting); Writing – review & editing (supporting). Sharath Sriram: Formal analysis (supporting); Resources (equal); Supervision (supporting); Visualization (supporting); Writing – review & editing (supporting). Madhu Bhaskaran: Conceptualization (lead); Formal analysis (equal); Investigation (equal); Methodology (equal); Resources (lead); Supervision (lead); Visualization (equal); Writing – review & editing (lead).

The data that support the findings of this study are available within the article and its supplementary material.