Printed soft skin electrodes for seamless bio-impedance measurements

Bio-impedance measurements are widely used to assess various physiological parameters. Contemporary skin electrodes for bio-impedance measurements are cumbersome and novel electrode designs are needed to allow fast and easy placement, long-term stability and user comfort. This investigation introduces dry, printed, bio-compatible electrode arrays, made of screen-printed carbon and inkjet-printed PEDOT:PSS that measure bio-impedance non-invasively and stably. Two contact impedance measurements yield the lowest normalized values of soft electrodes reported to date. Four contact bio-impedance measurements from the radial, ulnar, common carotid and superficial temporal arteries were performed, demonstrating the ability to capture blood pulsation in different areas with small form factor. Owing to the unique properties of the printed electrodes reported here, we were able to demonstrate for the first time blood pulsation in the face, continuous blood pulsation measurement during simultaneous muscle activation and signal stability over many hours.


Introduction
Non-invasive bio-impedance measurements are used in a wide range of applications, including wound care [1], respiratory activity assessment [2] and monitoring blood pressure [3]. Across these different applications, the need for wireless and wearable systems for long term and stable recordings is becoming widely recognized. Several lingering technological challenges limit the utilization of non-invasive bio-impedance measurements, including hardware miniaturization, data analysis and the large impedance of the skin-electrode interface. In particular, gel electrodes, which are commonly used in skin bio-impedance measurements to reduce the skinelectrode impedance, are notoriously challenging: they are bulky, dry out, tend to be unstable, and may cause skin irritation and discomfort to the user.
In recent years, much attention was directed to the implementation of alternative electrodes. In particular, several approaches based on dry conducting electrodes (e.g. silver, aluminium, copper, carbon) were investigated [4][5][6][7][8]. However, the aforementioned dry electrodes have limited flexibility and do not conform with the skin, resulting in relatively bulky and unstable systems. To improve conformity with the skin and stability, alternative materials and fabrication approaches were explored. For example, graphene-based temporary tattoo electrodes were fabricated using a thin-film process and exhibit high conformity to the skin [9, 10]. Other suggested electrode designs include screen and stencil printing on textiles [11] (used in wet form), electrode arrays printed on composite nanocellulosepolyurethane substrates [1], and inkjet-printed gold electrode arrays [12]. Owing to the limited performances of existing methods, impedance measurements were so far restricted to large and easily accessible regions, such as the forearm, legs and the neck and to short recording sessions [13]. Presently, there are no widely accepted low-impedance soft electrodes compatible with scalable manufacturing for ongoing impedance measurements.
With their excellent manufacturability properties, screen-printed carbon and inkjet-printed poly (3,4- [25,26]. Although PEDOT:PSS films are attracting a lot of attention, their overall performances as skin electrodes are relatively modest, owing to delamination of thin films and instability under high-humidity due to swelling of the hygroscopic, but non-conductive PSS (supplementary figure A1 [27][28][29][30]).
In this paper we explore screen-printed carbon and inkjet-printed (PEDOT:PSS) on polyurethane for the realization of state-of-the-art dry bio-impedance electrodes, specifically impedance plethysmography (IPG). IPG capitalizes on changes in conductivity during pulsation: owing to mechanical changes in blood vessel geometry, blood pulsation is manifested as an impedance change [31]. We demonstrate process optimization, design considerations and bio-impedance measurements from the forearm, neck and face. Both approaches yield soft electrodes that conform with the skin with low contact impedance (CI). We compare the CI results to values measured for standard commercial gel electrodes, and demonstrate that the dry electrodes reach lower specific contact impedance values (normalized by area), compared with gel electrodes and previously reported dry electrodes. Four probe bio-impedance measurements were also performed to characterize the tissue bio-impedance at close proximity to the arteries. We demonstrate long-term (up to 13 h) and stable monitoring of blood pulsation Finally, to demonstrate the robustness of the electrodes, we show how blood pulsation can be detected during muscle action close to the measurement site.

Commercial pre-gelled Ambu
Commercial pre-gelled Ambu electrodes of 40 mm in diameter (Ambu® BlueSensor Q ECG electrodes) were used. These electrodes contain a wet gel area in the silver/silver chloride sensor region, including an adhesive surrounding area to stick them onto the skin, with easy access to a BNC connector for the bioimpedance measurements.

Screen-printed carbon electrodes
Screen-printed carbon electrodes for electrophysiological and bio-impedance measurements were fabricated, as described previously [16,19]. First, electrodes traces were screen-printed with silver ink (Creative Materials, 125-13 T) on a 50 and 80 µm polyurethane sheet (breathable transparent medical grade polyurethane (PU) film on paper carrier from Del-Star Technologies, Inc.; figure 1(a)). Following silver printing, the films were dried on a heater at 50 • C for 15 min. Subsequently, carbon electrodes (124-50 T and C200, Creative Materials) were printed in alignment with the silver traces. Furthermore, the printed electrodes were dried again on the heating plate at 50 • C for 15 min. After printing, traces were passivated with a double adhesive 80 µm PU film (from Delstar EU94DS) which was cut to leave the carbon electrodes exposed. For impedance measurements, the arrays were bonded to metallic traces on a custom-made printed circuit board (PCB) which was designed to support BNC connections.

Inkjet-printed PEDOT:PSS electrodes
A PixDro LP50 inkjet printer (Süss MicroTec SE) was used to fabricate silver-PEDOT:PSS dry electrodes on PU substrates ( figure 1(b)). The substrates were of the same type as the ones used for the screen-printed samples (80 µm). Before printing, the substrates were heated in a vacuum oven (120 • C for 10 min, 5 mbar, Memmert VO) and subsequently treated in an Ar plasma oven (Pico, Diener) for 30 s to optimize ink wetting. The silver ink (Silverjet, Sigma-Aldrich) was filtered through a 0.45 µm polyvinylidene fluoride (PVDF) filter and deposited with a Sapphire QS-256/10 AAA printhead (drop volume of 10 pl, Fujifilm) set to a resolution of 1000 dpi. The deposited wet silver films were cured on a hotplate at 120 • C for 5 min resulting in a sheet resistance of 5 ± 2 Ω/□ and a layer thickness of 370 ± 130 nm. The latter was extracted from scanning electron microscopy (SEM, Zeiss Auriga System) images that can be found in the SI ( figure B1).
Before the PEDOT:PSS deposition, samples were treated with an additional 60 s of Ar plasma. Next, a commercially available PEDOT:PSS ink (Clevios F HC Solar, Heraeus) was filtered (0.45 µm PVDF), degassed in an ultra sonic bath for 20 min and printed with a Dimatix Materials Cartridge (10 pl, Fujifilm) that was set to a resolution of 1200 dpi. The samples were then placed on a hotplate at 120 • C for 10 min to obtain PEDOT:PSS layers of 100 ± 20 nm thickness (SEM images in figure B1). The sheet resistance of PEDOT:PSS on bare PU and on silver was measured to be around 150 ± 10 Ω/□ and 2.3 ± 0.4 Ω/□, respectively. The utilized PEDOT:PSS ink includes a low percentage of (3-Glycidoxypropyl)methyldiethoxysilane in its formulation. Similar to published work on thermal crosslinking of PEDOT:PSS by addition of (3-Glycidyloxypropyl)trimethoxysilane, we achieved PEDOT:PSS films exhibiting a higher resistance to water and delamination than pristine PEDOT:PSS after the annealing process [32,33]. For details of the crosslinking process, see the corresponding section in the SI. The final passivation steps, with double-sided adhesive and mounting on a PCB, were performed analogously to the screen-printed carbon electrodes.

Data collection
Two and four-terminal measurements were performed with an impedance analyzer (MFIA by Zurich Instruments) on the skin of three healthy volunteers (ages 27-28). IPG measurements were performed using a four-electrode array to exclude the significantly larger CI from the measurement (two inner sensing electrodes and two outer excitation electrodes, figure 1). The array was connected to a custom-made PCB with four BNC connections, which were connected to the impedance analyzer. AC current with a frequency between 10 and 100 kHz was applied between the outer electrodes, and the voltage was measured between the inner electrodes.

Data analysis
All IPG data were analyzed with python. The signals were processed using a second-order Butterworth band-pass filter with low and high frequency cut-offs at 0.5 and 6 Hz. The characteristic points of the pulse shape that were calculated were: the peak, trough and maximum slope (MS). The average unnormalized CI was evaluated over the different layouts (i.e. different shapes, geometries and sizes).

Results
Carbon and PEDOT:PSS electrodes of different shapes (square and circle) and different interelectrode spacing were printed using screen and inkjet-printing on thin PU films. The PEDOT:PSS films were tailored specifically to achieve stability (see section 2) and adhesion, which is mandatory for stable recordings from the skin. In figure 1(d), we show two such arrays after placement on the forearm. As a reference, we also show an the commercial gel electrodes placed at the same region. Even though the printed electrodes are used in their dry form, their compact and flexible structure result in a conformal attachment to the skin, as evident in the stable twoand four-terminal measurements we describe below.
CI of four different electrode types (dry PEDOT:PSS, dry carbon, carbon with gel, and standard wet electrode (Ambu)) was measured and normalized by the electrode area ( figure 2(a)). As CI depends on skin properties, an accurate comparison of CI of different electrodes mandates measurements at the same location. We used the location of the inner forearm (over the radial and ulnar arteries), as this area is most commonly used for wearable bio-impedance measurements. For completeness, we show CI measurements at low frequencies (10-1000 Hz) in supplementary figure E1.
Unnormalized, the commercial gel electrode has the lowest CI, followed by the gelled carbon. Carbon electrodes appear to have a lower CI compared to the PEDOT:PSS. By normalizing the CI data by area, we note the impact of size on the CI (figures 2(a2), (b2) and (c2)): While the wet electrode had lower CI compared to the dry electrodes before normalization, after normalizing, its specific impedance is in fact higher than the dry electrodes for frequencies above 10 kHz. At 10 kHz, the unnormalized CI of carbon, PEDOT:PSS and the gel electrodes is ∼6.4, ∼13 and 0.51 kΩ respectively (values for carbon and PEDOT:PSS electrodes were averaged over a range of geometries, shapes, and sizes). The areas included in the CI measurements ranged from ∼13 to ∼28 mm 2 . To the best of our knowledge the dry carbon electrodes described here have the lowest CI values reported to date for soft electrodes (with the largest electrode (∼28 mm 2 ), reaching CI of 3.1 kΩ at 10 kHz).
To evaluate the stabilization of the CI, we measured it every 15 min after placement on the skin (figures 2(b) and (c)). We observed a large jump in CI between '0 min' (right when placing the array) and '15 min' (15 min after placing the array). After 15 min, the changes were less significant. We hypothesize that the large jump represents the time it takes the material to properly adhere onto the skin surface (improving charge transfer between the electrode and the skin interface).
We now turn to four-terminal impedance measurements to demonstrate IPG. In IPG, the measured impedance reflects artery volume [31]: As blood pulsates through arteries, the artery volume changes. Since blood is conductive, the artery can be modeled as a cylindrical resistor. When the artery expands (as a result of the pulse), it effectively becomes equivalent to a cylindrical resistor with a larger radius. In terms of impedance, this translates to decrease in the impedance of the artery, and the impedance of the overall measured segment. Figure 3 shows four-terminal IPG measurements with commercial wet electrodes, wet and dry carbon electrodes and dry PEDOT:PSS electrodes. Baseline impedance was in the range of 61 and 140-140 Ω to 170 Ω for the commercial and printed electrodes, respectively, reflecting the difference in their geometries. The change in the IPG signal was 0.03 and 0.13 Ω for the commercial and the printed electrodes, respectively. The smaller dimensions of the printed electrodes contributed to stronger sensitivity. Peaks, troughs and MS points for each pulse were successfully detected and marked automatically. These characteristics are commonly used to evaluate the impedance change, which translates to the volume change of the artery, as well as the pulse transit time which can be used to evaluate blood pressure [34]. Both electrode types measure the same phenomena, albeit the printed electrodes achieve the measurement with dramatically smaller form factor. In particular, the printed electrodes in their dry form offer excellent performances along with long term stability (up to 13 h, supplementary figure D1).
Owing to their soft and dry nature, the electrodes can be conveniently placed at almost any hairfree area of the skin. In particular, the locations of the radial and ulnar arteries (at the forearm), the common carotid artery (at the neck) and the superficial temporal artery (at the face), are of particular interest in many physiological investigations. In figure 4 we show carbon and PEDOT:PSS electrode arrays placed at these locations and their corresponding blood pulsation measurements. In all locations we observe clear pulsation signals. At the forearm, a prominent secondary notch (the dicrotic notch [35], signifying the back reflection of the pressure pulse due to higher vascular resistance) is observed [36]. As expected, at the neck and face this feature appears to be less pronounced and the shape of the pulse is different. It is evident that pulse shape has a strong association with the electrode location [2].
Blood pulsation was also measured during muscle activation to demonstrate electrode stability under dynamic conditions. In figure 5, we show fourterminal measurements of gel (top) and dry carbon (bottom) electrodes during relaxed conditions (left) and fist clenching (right). The electrodes were placed over the same forearm that was clenching the fist, over the ulnar artery. A clear pulse is detectable during fist clenching for both electrodes.
Finally an important parameter that impacts bioimpedance measurements is the electrode design (e.g. shape, dimension and distance) [5,37]. In this investigation, printed electrodes were realized and measured in a range of geometries, shapes and sizes. Although we could identify differences in pulse amplitude and shape the variability between measurements was greater than between sessions proving that physiological variability has greater contribution than variability between electrode arrays. This large variability is a major challenge in assessing the effect of array design on performances and a more methodical investigation is needed, possibly by averaging measurements over many sessions and subjects.

Discussion
Our investigation reports on dry printed electrode arrays for bio-impedance measurements. We demonstrated low CI values and stable four-terminal impedance measurements. We successfully detected pulsation at rest and during muscle activation and long term stability. The results we presented here show state-of-the-art performances and a path toward wearable bio-impedance devices as an alternative to the cumbersome gel electrode.
To assess the performances of the arrays, we measured pulsation from four different arteries: the radial, the ulnar, the common carotid and the superficial temporal arteries. In addition, to test the versatility and stability of the electrodes, we performed IPG measurements during muscle actions. The actions included: fist clenching when measuring from the radial and ulnar arteries and facial expressions while measuring the superficial temporal artery (data not shown).
In this investigation we studied two alternative printing approaches with similar performances. Inkjet-printing of PEDOT:PSS is widely used and is a convenient approach for rapid prototyping and personalized electrode arrays that can be tailored for each subject, depending on particular physiological features. Film adhesion of printed inkjet PEDOT:PSS is challenging and special care was used in this investigation regarding the stability of the array. Further improvement can be made to the electrical connection of the silver traces to read-out electronics. Given the layer thickness of 100 nm, traces are prone to break when stretched or bent against hard surfaces (such as rigid PCBs), increasing their electrical resistance. Screen-printed carbon is easier to implement, and scale-up is straight forward. However, its suitability for rapid prototyping is more challenging.
The results presented here can benefit a wide range of applications. In particular, the measurement of blood flow and pressure is of special significance [38]. Monitoring changes in blood flow is of critical importance in many fields: from medicine [39,40], to psychology [41][42][43][44] and cosmetic surgery [45].
Invasive and non-invasive monitoring of blood flow is widespread in clinical diagnostics. As abnormal blood flow may be episodic in nature and may not be manifested in clinical settings, wearable solutions for on-going monitoring at home is a major need. Two commonly used techniques for wearable and non-invasive blood flow detection are photoplethysmography (PPG) and IPG. Both are sensitive to blood volume changes. Whilst IPG capitalizes on changes in conductivity during pulsation, PPG monitors the increase in blood in the artery segment during pulsation [46], utilizing changes in light absorption of the tissue. At each pulsation, the increase in blood results in a measurable change in the transmitted or reflected light. Both techniques can be used not just to record the pulsation, but also to extract blood pressure, and thus have attracted huge interest. Although offering similar insight, IPG and PPG differ fundamentally: deep penetration to the tissue is possible with IPG, due to the high-frequency current, but PPG is limited to the skin surface. Deep penetration allows higher accuracy in detection of the blood volume change, as well as providing detailed characteristics of the hemodynamic parameters [9, 10, 35].
The main overreaching gap our technology helps address is the realization of fully independent skin devices, suited for ongoing bio-impedance monitoring under realistic conditions, while still maintaining process compatibility with industrial fabrication methods. A fully wearable bio-impedance system requires additional components such as: low power micro-electronics, improved data analysis approaches [10] and proper coupling between the soft electronics and the wireless data acquisition system. The soft and dry electrodes we presented demonstrate an important step towards this goal.
Wearable devices are becoming an important building block as a digital health solution. They allow on-going monitoring of various parameters, ranging from electro-physiology, heart rate, blood pressure, to name just a few examples. Wearable sensors benefit both patients and the medical system compared with traditional practices. In particular, information can be collected quickly and continuously under realistic conditions and visits to the clinic can be minimized. In the new reality imposed by the Covid-19 crisis, home-based patient monitoring is becoming critically important. The lighter and seamless the technology is, the higher are the chances that it will be widely used and accepted. Simultaneously, it must provide medically validated information that will be trusted by the medical community and by medical practitioners.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.