Highly precise nanofiber web-based dry electrodes for vital signal monitoring

Abstract

A great variety of dry electrodes have been developed over many decades in order to overcome the drawbacks of conventional Ag/AgCl gel electrodes, but their applications are still restricted due to the low accuracy of the dry electrodes. To improve their accuracy, highly elastic thermoplastic polyurethane (TPU) and poly(styrene-b-butadiene-b-styrene) (SBS) were electrospun into nanofiber webs containing Ag nanoparticles. Then, these webs were silver plated using electroless plating to prepare silver plated nanofiber web (AgNFw)-based dry electrodes that allowed for the detection of high fidelity biopotential signals. The electrode properties of AgNFw dry electrodes were investigated thoroughly using agar phantom systems. Human subject tests were also carried out to determine the real performance of AgNFw dry electrodes on a human body in terms of biopotential recording and electrical impedance tomography (EIT) measurements. The experimental results revealed that AgNFw dry electrodes can exhibit performances comparable to Ag/AgCl gel electrodes, indicating the newly designed AgNFw dry electrodes have superior accuracy to most existing dry electrodes.

---

Introduction

Ubiquitous health care systems, brain computer interface (BCI) and human computer interface (HCI) techniques, bionic engineering, and electrical impedance tomography (EIT) systems require special electrodes that provide both high precision recording of biopotentials and long-term monitoring capability.^1–12 To be specific, ubiquitous health care systems measure electrophysiological signals such as electrocardiography (ECG) and electroencephalography (EEG) to monitor the human body and diagnose various ailments.^8,9,13 In addition, BCI techniques have utilized EEG signals to help people with communication-based disabilities spell words.^3,4 Electromyography (EMG) signals are generally used to diagnose muscle disease and can be employed to handle artificial limbs in bionic engineering as well.⁷ Electrooculography (EOG) is also vital in the diagnosis of ophthalmological problems, while HCIs have been developed to handle a wheelchair and manage a computer mouse using specific algorithms.^10,12,14 EIT imaging techniques have a wide range of clinical applications such as lung imaging for ventilation monitoring, as well as stomach, head and cardiac imaging.⁵ Signals detected by electrodes in an EIT system, however, cannot be strictly considered as biopotential signals because the signals are not generated by the human body. Rather, in EIT, a specific high frequency AC current is artificially injected into body and then a generated potential is detected.

Because of the extremely small signal amplitude of electrophysiological signals, which range from 100 μV (EEG) to 1 mV (ECG), the quality of the measurement of biopotential signals is highly dependent on the electrodes involved.^6,15 Traditionally, the conventional Ag/AgCl gel type electrode have been used to measure such signals to ensure the signal fidelity because the Ag/AgCl gel electrode has been well-characterized, medically verified and studied for many decades.^15,16 Therefore, many studies have been published related to developing new electrodes that use the Ag/AgCl gel electrode as a standard for comparison. However, this kind of Ag/AgCl gel electrode cannot be applied in long-term settings because it can degrade the electrode performance over time as the gel dries; it can also cause skin irritation and be uncomfortable to remove.^{1,2,5–7,16–18}

In order to overcome such drawbacks of Ag/AgCl gel electrodes, many types of dry electrodes have been reported. These electrodes can be roughly classified into two categories: metal-based electrodes and polymer-based electrodes. It is well-documented that solid metal dry electrodes cause discomfort and trigger large motion artifacts as the electrode slips.^8,16–19 Polymer-based electrodes can be classified into two types: composite (foam or rubber-based) electrodes and fabric electrodes. Such electrodes have usually drawn attention from the standpoint of both comfort and usability because they are flexible and can easily adapt to skin topography while still providing good contact with the skin. One obvious downside of composite electrodes is that they have a relatively high electrical resistance because they are normally fabricated by adding some conductive additives into the polymers to endow them with electric conductivity.^{1,2,6,7,18,20} Compared to composite electrodes, fabric electrodes may have better conductivity that is more comparable to metal electrodes because most fabric electrodes are produced using an electroless plating process. However, the performance of fabric electrodes can be degraded due to surface abrasion, poor coating durability, and variations in electric performance with repeated usage.^6,17,21 Apart from fabric and composite electrodes, microscale pin-type dry electrodes were developed to provide high quality signal, but these may have a high risk of inflammation resulting from the skin being punctured by metallic pins. This invasive method is not stable during long term measurements since the skin regenerates and encapsulates the electrode with scar tissue to isolate it from bodily fluids.^2,18,22 Moreover, some special, highly integrated dry electrodes have been reported such as stretchable electrode arrays and epidermal sensors.^23,24 These electrodes can provide conformal contact with the skin and can stick easily. However, they could not measure desirable or specific biopotentials for clinical purposes because they could not be placed in the proper, clinical-standard locations (e.g., the 12-lead ECG) due to the short electrode-to-electrode distance. In summary, the dry electrodes mentioned above have some disadvantages, but it is evident that they have at least addressed the main problems of Ag/AgCl gel electrodes, which cannot be used for long-term biopotential monitoring.

Nevertheless, Ag/AgCl gel electrodes are still universally applied in most clinical and research fields, and billions of gel type ECG clinical electrodes are manufactured every year, whereas dry electrodes are limited to niche or non-medical applications such as fitness monitoring and toys. This is because the real performances of such dry electrodes still could not reach those of Ag/AgCl gel electrodes or they still need more comprehensive and quantitative assessment.^15,16 Today, detecting an electrophysiological signal is not difficult, and a great number of studies regarding dry electrode technology have been published. Recent work has shown that ECG can be obtained through an air gap even up to 40 cm using noncontact dry electrodes, but the quality of the vital signals recorded from these devices is typically quite poor.²⁵ Therefore, one of main issues relating to dry electrodes should be how to obtain more precise measurements and prevent signal distortion by improving the signal fidelity such that it is the same as or better than that of Ag/AgCl gel electrodes.

In our previous study, six kinds of electrodes including two kinds of nanofiber web-based electrodes (PEDOT coated PVDF nanofiber web-based electrodes and silver plated PVDF nanofiber web-based electrodes), three kinds of fabric based electrodes with different plating materials (PET–Cu–Ni fabric electrodes, PET–Cu–Ni–Au fabric electrodes and PET–Cu–Ni–carbon fabric electrodes) and conventional Ag/AgCl gel electrodes were examined using agar phantom systems. We observed that the two nanofiber web-based electrodes were definitely better than any other fabric-based electrodes in terms of contact impedance, step response, and noise. We attributed this to the unique nanofiber structure, which was ultra-fine, thin, and was not as bulky as the fabric electrodes.⁶ Nevertheless, the ECG signal fidelity of nanoweb electrodes did not satisfy our expectations, as it was about 93% of that of a Ag/AgCl gel electrode. However, the potential of nanofiber web-based electrodes was demonstrated.

Hence, in the present study, the nanofiber web-based dry electrodes were completely redesigned for highly precise biopotential recording. In order to achieve this objective, thermoplastic-polyurethane (TPU) and poly(styrene-b-butadiene-b-styrene) (SBS) were selected to replace the previous poly(vinylidene fluoride) (PVDF) materials. TPU and SBS are more hydrophilic than PVDF, which is very important in the silver plating process; TPU and SBS also have better mechanical properties and are much less expensive than PVDF. Another disadvantage of the PVDF nanofiber web is the low dimensional stability because of its low elasticity. This makes it difficult to handle PVDF especially when separating the electrospun nanofiber web from the collecting substrate for further silver-plating processes, which degrades the quality of the electrode. In this study, the electroless silver plating process was also upgraded and optimized step by step from the silver plating solution to the silver plating equipment. Next, silver plated nanofiber webs (AgNFw) were fabricated into electrodes of a certain size and shape, and then their intrinsic electrode properties including contact impedance, step response and noise characteristics were investigated thoroughly using agar phantom systems. Human subject testing was also carried out to determine the realistic performance of AgNFw dry electrodes on the human body in terms of ECG, EOG, EEG, EMG and EIT measurements. In the last stage of this paper, the detecting mechanism and some advantages of AgNFw dry electrodes are fully discussed to explain why only the nanofiber web structure is capable of meeting the requirements.

Experimental section

AgNFw dry electrode

Preparation of nanofiber web containing silver nanoparticles. A 10% (w/v) TPU (Estane R190A, Lubrizol, USA) solution for electrospinning was prepared by dissolving TPU in a binary solvent system of tetrahydrofuran (THF, Aldrich, USA)/N,N-dimethylformamide (DMF, Aldrich, USA) (2/3 v/v). A ternary-solvent system of toluene (Aldrich, USA)/THF/DMF (3/5/2 v/v/v) was used to prepare a 9% (w/v) SBS (Scientific Polymer Products, USA) solution.²⁶ Then, a very small amount of silver nitrate (Dae Jung, Korea), was added to each TPU and SBS solution at 1% based on the weight of the polymer to form silver nanoparticles. These nanoparticles were reduced on the nanofiber web surface using DMF. This is the key point of fabricating AgNFw, because the reduced silver nanoparticles on the nanofiber surface can act as catalysts to promote silver plating on the nanofiber surface. Electrospinning was carried out at an injection rate of 1.5 mL h⁻¹ with a 27 gauge needle and applied voltages of 18 kV for TPU and 24 kV for SBS.

Electroless silver plating process. The fundamental principle of the electroless silver plating technique is the same as the well-known silver mirror reaction, but the concentration of each solution was diluted to slow the reaction rate in order to improve the uniformity and electrical conductivity of the AgNFw. First, three solutions were prepared: a 0.05 M silver nitrate solution, a 0.4 M sodium hydroxide solution (Sam Chun Chemical Co., Korea), and a 0.12 M dextrose solution (Duksan Co., Korea). Then, the silver nitrate solution was mixed with the sodium hydroxide solution at a ratio of 10/5 v/v with stirring. When a brown precipitate formed, the ammonia solution was added dropwise until it dissolved to form a water-soluble diamine silver complex compound. The diamine silver complex solution was poured into a custom-designed silver plating box where Ag nanoparticle-containing nanofiber webs were pre-installed (see ESI Fig. S1†). Then, the dextrose solution was added into the silver plating box until the final ratio of silver nitrate, sodium hydroxide and dextrose solutions reached 10/5/1 v/v/v. After that, diamine silver complex cation ([Ag(NH₃)₂]⁺) was reduced gradually by dextrose to silver, which was deposited preferentially on the nanofiber web surface. The reduction was conducted for 1 h, and during this time the silver plating equipment was rotated at a speed of 10 rpm to obtain even silver plating. It is important to note that the two nanofiber webs should be immersed into isopropanol for 12 h prior to silver plating to further enhance their wettability towards the aqueous electroless silver plating solution. This provides uniform silver plating from the surface to deep inside the nanofiber web. Fig. 1 depicts a schematic of the manufacturing process used to make AgNFws.

Fig. 1 Schematic of the manufacturing process of AgNFws with images corresponding to the steps: (a) TPU or SBS transparent solution with silver nitrate, (b) colored solution due to the silver nanoparticles reduced by DMF, (c) electrospinning to fabricate a silver nanoparticles embedded nanofiber web, (d) nanofiber web dipped in silver plating solution, (e) silver plated nanofiber web. (f) A TPU–AgNFw dry electrode (left) and a Ag/AgCl gel electrode (right).

Fabrication of silver plated nanofiber web dry electrodes. Both the size and shape of the AgNFw dry electrodes were strictly controlled. A diameter of 9 mm was chosen, which is the same as that of the metal part of the Ag/AgCl gel electrode (3M Red Dot, USA). An eyelet structure was adopted for the fabricated AgNFw dry electrodes as shown in Fig. 1(f). A polyurethane sponge was placed between the metal eyelet and nanofiber web to provide the AgNFw dry electrode with appropriate elasticity and softness, which provided tight skin-electrode contact.

Electrode properties on agar phantom system

Three intrinsic electrode properties of each AgNFw dry electrode were assessed using an agar phantom system including contact impedance, step response and noise characteristics. All gel in the Ag/AgCl gel electrodes was removed for quantitative comparison with dry electrodes in agar phantom system experiments.

Contact impedance. A bioelectrical impedance spectroscopy (BIS) system was employed to evaluate electrode contact impedances. Several frequencies of 10, 50, 100, 1k, 5k, 10k 50k, 100k, 250k, and 500k Hz were selected to measure the contact impedances of the electrodes. A three-electrode setup (Fig. 2(a)) was used to measure the impedance, Z₃, which is equivalent to the sum of the agar impedance and electrode contact impedance, at each selected frequency.²⁷ Similarly, a four-electrode setup (Fig. 2(b)) was used to measure the impedance, Z₄, which is equivalent to the agar impedance, at each frequency selected.²⁸ Therefore, the electrode contact impedance (technically the electrode impedance plus the electrode-agar surface impedance) can be readily calculated by subtracting the measured four-electrode impedance from the three-electrode value at each frequency, i.e., Z_M = Z₃ − Z₄. Provided that each agar phantom-electrode contact system is modeled as a parallel RC circuit, the measured electrode contact impedance, Z_M, can be expressed according to the following equation:here, R_M and X_M are the equivalent resistance and capacitive reactance of the electrode, respectively. These were measured using the BIS device, and f is the frequency of the injected current. Thus the actual R and C values for each electrode were obtained by rearranging eqn (1) as follows:

Fig. 2 Experimental setup for (a) three-electrode contact impedance, (b) four-electrode contact impedance.

Since the R and C values vary as a function of frequency, the R and C values were averaged over the selected frequencies using a least-squares algorithm in order to compare each type of electrode more comprehensively. We performed calculations as described in ref. 6.

Step response. Step response characteristics of each electrode were evaluated utilizing the experimental four-electrode setup shown in Fig. 2(b). A current pulse with a 10 mA amplitude, a 100 ms period, and a 50% duty cycle was injected between the outer pair of electrodes. Simultaneously, voltage pulses between the inner pair of electrodes were measured using a data acquisition system, MP36 (Biopac Systems, USA) at a 1 kHz sampling frequency.⁶

Noise characterization. Similarly, the voltage between the two inner pairs of electrodes was measured without an injected current using the setup in Fig. 2(b). Any non-zero voltage measurement was considered to be noise. The sum of the noise power spectral density of each electrode was analyzed and compared.

Biopotential recording on human subjects

Human subject tests were carried out to compare the realistic performance of AgNFw dry electrodes with that of Ag/AgCl gel electrodes on the human body for ECG, EOG, EEG and EMG. All biopotentials were recorded simultaneously using both AgNFw dry electrodes and Ag/AgCl gel electrodes, and the two types of electrodes were aligned as closely as possible to minimize the error caused by position. To achieve this, only a small adhesive part of the Ag/AgCl gel electrode was cut off without gel removal, but a small gap between the gel of the Ag/AgCl gel electrode and the AgNFw dry electrode was still maintained to avoid electrical shortage between the two neighboring electrodes. AgNFw dry electrodes were adhered just utilizing a common adhesive tape. Fig. 3 illustrates the positions of the electrodes used for various biopotential recordings. The physical information of the human subject in this study is as follows; sex: male, race: Korean, height: 175 cm; weight: 70 kg; age: 32.

Fig. 3 Electrode setup for biopotential recording, red dots denote Ag/AgCl gel electrode, blue dots refer to AgNFw dry electrode. (a) ECG measurement, (b) EOG measurement, (c) EEG measurement, (d) EMG measurement.

Electrode setup and ECG recording. Three AgNFw dry electrodes and three Ag/AgCl gel electrodes were placed on the subject's chest in pairs as shown in Fig. 3(a).¹ A commercial, proven wireless ECG data acquisition module, BN-ECG2, (BioNomadix, Biopac Systems, USA) combined with an MP150 (Biopac Systems, USA) was used as the recording system. Factory pre-sets such as bandwidths of 1 Hz to 35 Hz and a fixed gain of 2000 were employed. No filter characteristics other than the factory pre-sets correctly reflected the characteristics of the electrodes. The subject was asked to sit on a chair without moving during the recordings. The correlation between the two signals was computed using a cross-covariance function in Matlab.

Electrode setup and EOG recording. The electrodes were stuck on the subject's face around the eyes in pairs as shown in Fig. 3(b).² MP36 was employed as a recording system, and the pre-set EOG option in the Biopac Student Lab PRO 3.7.3 software was selected, which has a default bandwidth of 0.05 Hz to 35 Hz and a gain of 1000. No other options or filtering processes were conducted. The subject was told to blink his eyes three times and then roll his eyeballs up and down two times immediately during the during the test. The correlation between the two signals was computed as described above.

Electrode setup and EEG recording. The electrodes were adhered to the forehead and the adjacent ear in pairs as shown in Fig. 3(c).² The MP36 was employed, and a pre-set EEG option in the Biopac Student Lab PRO 3.7.3 software was selected, which has a default bandwidth of 0.5 Hz to 35 Hz and a gain of 20 000. The subject was instructed to close his eyes and was not allowed to perform any action (even rolling his eyes) because motion can affect the raw EEG signal. The power spectral density (PSD) and the multiscale entropy (MSE) of the EEG raw signal were also computed in Matlab for further comparison.

Electrode setup and EMG recording. The electrodes were attached on the subject's left forearm in pairs as shown in Fig. 3(d). A wireless EMG data acquisition module, BN-EMG2 (BioNomadix, Biopac Systems, USA) in conjunction with MP150 was used to gather EMG signals. Factory pre-sets including a bandwidth of 10 Hz to 500 Hz, and a fixed gain of 2000 were used. The subject was asked to clench and unclench his left hand repeatedly during the test. The correlations between the two EMG signals as well as the enveloped raw EMG signals were computed and compared.

EIT system and imaging experiments

EIT characteristics of each type of electrode were evaluated using a KHU Mark2 mfEIT system and an EIT algorithm reported in our previous paper.⁵

Agar phantom experiments for static EIT test. Since it is difficult to stick electrodes on the agar phantom using tape or other adhesive materials, an elastic belt design was adopted. The belt was made of an elastic band that was 56 cm long and 3 cm wide, and there were 16 female eyelet connectors in the center of the belt to link electrodes as shown in Fig. 4(a). As seen in Fig. 4(a), the agar phantom used in the static EIT test was composed of three parts with different electrical conductivities. The main white background of the medium electric conductivity contained 0.1 wt% NaCl, the left light blue cylinder of the least conductivity contained 0.02 wt% NaCl, and the right dark blue background having the highest conductivity had 0.5 wt% NaCl. A KHU Mark2 mfEIT system (Fig. 4(b)) was employed to acquire scanned voltage data between neighboring TPU–AgNFw dry electrodes or Ag/AgCl gel electrodes generated by the injected alternating currents in the frequency range between 1 and 500 kHz. Then, time-difference conductivity images were reconstructed based on the truncated singular value decomposition method. Both the system information and the detailed algorithm can be found in our previous paper.⁵

Fig. 4 Electrode setup for EIT test. (a) Agar phantom wrapped with a belt composed of 16 Ag/AgCl gel electrodes or 16 TPU–AgNFw electrodes; the main background of agar (200 mm in diameter) contains 0.1 wt% NaCl, the left light blue cylinder (35 mm in diameter) contains 0.02 wt% NaCl, and the right dark blue one (35 mm in diameter) has 0.5 wt% NaCl. (b) KHU Mark2 mfEIT system. (c) Subject's chest on which 16 TPU–AgNFw electrodes were attached with adhesive tape.

Human experiments for dynamic EIT test. For human lung EIT tests, 16 Ag/AgCl gel electrodes or 16 TPU–AgNFw dry electrodes were equidistantly attached to the subject's thorax in a transverse plane with adhesive tape as shown in Fig. 4(c). Time-difference conductivity images of the chest reconstructed from two types of electrodes illustrated lung ventilation while breathing in and out. These images were displayed and roughly compared.

Results

Electrode properties of AgNFw dry electrodes on agar phantom system

Contact impedance. Contact impedance refers to the impedance of the combination of an electrode with the object it is in contact with.¹ For instance, a skin-electrode contact impedance system consists of electrode impedance, skin impedance and surface impedance caused by the interface. When a biopotential is measured, ionic currents generated within the cells of the body travel down this system and transduce electronic currents; therefore, this system plays an important role in biopotential measurement. Determining and assessing the skin-electrode contact impedance are very crucial in developing and designing an electrode.²⁹ However, as the skin impedance has a high resistance and varies with subject, location and even time, the electrical impedance differences among different types of electrodes may be disguised by the variations in skin impedance when a skin-electrode system is employed to determine the contact impedance of the electrode. Because of this, a stable agar phantom-electrode contact system (where agar phantom is much more electrically homogeneous than human skin) has been employed rather than a real skin-electrode contact system to compare the contact impedance of each electrode used in this study.

Fig. 5 displays both the resistive part, R, and capacitive part, C, of the contact impedance of each electrode averaged over various frequencies. In terms of the resistive part of each electrode, the R values of the two AgNFw dry electrodes are even lower than that of the metal part of Ag/AgCl gel electrode (P value < 0.01). There are likely two main contributions associated with this result. One is mostly due to the high electrical conductivity of AgNFw dry electrodes themselves. The other is also attributed to the contact surface area.¹⁸ Contact area is one of main parameters influencing contact impedance, and nanofiber webs normally have a high specific surface area owing to the nanoscale fiber. For the capacitive part, C, a higher value of C results in a lower contact impedance with increasing frequency of the injected AC source. Since the metal part cannot create a capacitance at all, the C value of the Ag/AgCl gel electrode results from the contact surface. The two AgNFw dry electrodes exhibited slightly larger mean C values than the Ag/AgCl gel electrode, but these values were not significantly larger, and they might be considered comparable at levels less than 1 nF.

Fig. 5 Averaged contact resistances and capacitances of each electrode at the ten BIS frequencies ranging from 10 Hz to 500 kHz.

Step response. In our previous study, the fabric based electrodes exhibited distinct signal distortion during a step response test due to their porous textile structure, which impacts the capacitive part of the electrodes and causes signal distortion.⁶ When a current pulse with a 10 mA amplitude, a 100 ms period and 50% duty cycle was injected in this test, three electrodes showed comparable voltage waveforms without signal distortion as shown in Fig. 6. This was attributed to the compact structure of the AgNFw dry electrodes, which led to less signal distortion. In addition, this kind of signal is very similar to the EEG α-wave signal, which has a peak frequency of 10 Hz, although the amplitude is much higher than that of EEG.¹ These data confirm that the AgNFw dry electrodes show performance good enough for recording EEG signals, which are much more difficult to acquire than ECG and EOG signals in subject tests.

Fig. 6 Step response of each electrode to 100 ms period, 50% duty cycle square wave.

Noise characterization. The noise character of the electrodes is one decisive parameter for evaluating an electrode. This is particularly true given the small signal amplitude of biopotential signals (especially the EEG signal, which has the magnitude of μV). There are many factors that affect the noise character of the electrode, and the electrode impedance largely governs the noise level. Fig. 7 provides the sum of the noise power density of the test electrodes. Values in Fig. 7 are normalized to the gain of Ag/AgCl electrodes to correct for the different overall gains of the other two AgNFw electrodes caused by their contact impedances. Nevertheless, the TPU–AgNFw dry electrode exhibits lower noise levels than that of the Ag/AgCl gel electrode, whereas the SBS–AgNFw dry electrode indicates a higher noise level.

Fig. 7 Sum of noise power spectral density of each electrode. All values were normalized to correct for overall electrode impedances in each electrode.

Biopotential recording

Agar phantom-electrode contact systems were used to assess the electrode properties of AgNFw dry electrode via comparisons with Ag/AgCl gel electrodes. However, these results are insufficient to represent the true performance in a real-world environment. Since agar phantom is stable, it cannot reflect the real dynamic and complicated conditions encountered on human skin. Therefore, human subject tests were carried out to compare the realistic performance of AgNFw dry electrodes with that of Ag/AgCl gel electrodes. TPU–AgNFw dry electrodes were selected to perform the subject test because they have better noise characteristics than the SBS–AgNFw dry electrode on an agar phantom system.

ECG measurement. ECG is a vital biopotential in clinic/medical applications since it includes meaningful information related to a person's heart and physical performance. Therefore, it is used to diagnose cardiac diseases and is a standard procedure in current cardiac medicine.^1,13,17 Hence, ECG detected even from the dry electrode should be as accurate as that from the standard and proven Ag/AgCl gel electrode. Fig. 8(a) presents the ECG signals acquired simultaneously by both TPU–AgNFw dry electrodes and Ag/AgCl gel electrodes. The correlation between the two ECG signals is 0.99, which indicates that they are almost the same. In the close-up ECG signal shown in Fig. 8(b), P, Q, R, S, and T peaks are obviously observed for both electrodes, and the ECG signal obtained by TPU–AgNFw electrodes exhibited a little higher intensity than that recorded using the Ag/AgCl gel electrodes for the R peak. This is mainly due to the difference in position between the two types of electrodes.

Fig. 8 (a) ECG recording of Ag/AgCl gel electrode and TPU–AgNFw dry electrode, the correlation of the two ECG signals is 0.99. (b) ECG signal comparison of the P, Q, R, S, and T waves between the Ag/AgCl gel electrode and TPU–AgNFw dry electrode.

EOG measurement. Biopotentials measured in EOG can be applied to ophthalmological diagnosis and HCI techniques. Blinking of the eye is controlled by multiple muscles in the upper eyelid including the orbicularis oculi and levator palpebrae superioris muscle with a minor contribution from Muller's muscle. Eyeball rolling is managed by the cortical and subcortical system in conjunction with cranial nerves and sets of eye muscles that control eye blink.³⁰ In addition, parameters used to describe EOG signals include the blink frequency, peak amplitude and duration.³⁰ Obviously there are distinct differences in EOG signals for eye blinking and rolling in terms of the signal's peak amplitude and the duration as shown in Fig. 9(a). Fig. 9(b) displays EOG signals obtained when the subject was asked to close his eyes and just roll his eyeballs. The correlation values of the two EOG signals recorded from the two types of electrodes were both 0.99. One interesting thing is that the EOG signal also mainly originated from the muscle system, but the shapes are much different from the typical EMG signals.

Fig. 9 (a) EOG recorded from Ag/AgCl gel electrode and TPU–AgNFw dry electrode during three blinks of the eyes and then rolling the eyes up and down two times immediately thereafter, the correlation of the two EOG signals is 0.99. (b) EOG recording of two types of electrodes when the eyes were rolled while closed; the correlation value is 0.99.

EEG measurement. EEG is a powerful non-invasive tool widely used for medical diagnosis, neurobiological research and BCI.³¹ EEG is much more difficult to record than the ECG and EOG tested above. This is because the amplitude of EEG, which is less than a hundred microvolts, is much smaller than those of ECG and EOG, and the frequency is also higher. In addition, the main difficulty is that most of the area required to detect EEG is on the skull, which is covered with hair. This application is therefore very challenging for our TPU–AgNFw dry electrodes. To avoid the effect of hair on the skin, all electrodes were placed on non-hairy skin as shown in Fig. 3(c). Fig. 10(a) displays raw EEG signals obtained from the two types of electrodes and shows a high correlation value of 0.97. EEG power spectral density was also computed and displayed in Fig. 10(b) and (c), indicating that the correlation analyses showed a high similarity between the two electrode signals, which were 0.96 and 0.98, respectively. Typical α-waves could be clearly observed while the subject closed his eyes. Moreover, the MSE calculated from raw EEG data was computed and is illustrated in Fig. 10(d).^32,33 MSE analysis of EEG signals is used for diagnostics to identify changes in dynamics of surface EEG in patients with Alzheimer's disease.³⁴ The result showed a correlation value of 0.99, indicating very good similarity.

Fig. 10 (a) A comparison of EEG raw data recorded using a Ag/AgCl gel electrode and a TPU–AgNFw dry electrode, the correlation value is 0.97. (b) Associated power spectral density in a frequency range of 0.5–35 Hz, the correlation value is 0.96. (c) Associated power spectral density in the alpha wave range from 8 Hz to 13 Hz, the correlation value is 0.98. (d) Related multiscale entropy profile, the correlation value is 0.99.

EMG measurement. EMG is also a common biopotential, and it is used in various kinds of applications including diagnosing neuromuscular disorders, monitoring muscular fatigue, investigating muscle function/timing and as a control input for prosthetic devices.³⁵ Measuring EMG was more challenging than the three biopotential signals examined above because it had the highest frequency, and thus a larger bandwidth of 10 Hz to 500 Hz must be used to detect EMG. In addition, the EMG signal depended heavily on the positions of the electrodes. In order to obtain a high correlation value for the EMG signals from the two electrodes, the electrodes were placed on muscles in the positions shown in Fig. 3(d). Nevertheless, the correlation value of the two electrode signals was only 0.95, which is lower than that of any other biopotential signal measured in the human subject test. Fortunately, the raw EMG signals were not used directly. They were normally processed to a certain form before use. Further, the raw EMG data were rectified and digitally low-pass filtered using a fifth-order Butterworth filter, with a cutoff frequency of 5 Hz as shown in Fig. 11(c).³⁶ After applying this filtering process, the correlation value of the two electrodes increased to 0.99. Moreover, to identify the effect of the electrode positions on the correlation value, two sets of Ag/AgCl gel electrodes were placed on the same positions above, and the signals were recorded simultaneously, showing a correlation value of 0.93 (see ESI Fig. S2†), which was lower than that of the TPU–AgNFw dry electrode.

Fig. 11 (a) A comparison of EMG raw data recorded using a Ag/AgCl gel electrode and a TPU–AgNFw dry electrode, the correlation value is 0.95. (b) Rectified and linear envelope of raw EMG signal collected for the TPU–AgNFw dry electrode. (c) A comparison of the linear envelope of EMG signals, the correlation value is 0.99.

EIT test

The basic principle behind EIT is to non-invasively detect admittivity variations within the object or body by means of repeated recording of surface voltages arising from rotating injection of a known small, high frequency AC source through electrodes adhered on the circumference of the object; these measurements are then reconstructed to form images using a certain algorithm.^5,37

Fig. 12 presents static agar phantom images reconstructed at different frequencies obtained by TPU–AgNFw dry electrodes and Ag/AgCl gel electrodes. Compared to the reference image, which was drawn to the same scale as that of an agar phantom, the pattern of all reconstructed images was identified roughly as a whole. The agar phantom system was composed of three different agar components containing different concentrations of NaCl as illustrated in Fig. 4 to give three different electrical conductivities. Hence, all images exhibited variations in color depending on electrical conductivity. In addition, the images obtained using TPU–AgNFw dry electrodes were a little better than those recorded by Ag/AgCl gel electrodes in terms of the shape and the locations of two dots at each frequency. The image acquired by TPU–AgNFw dry electrodes at 10 kHz was the best of the images. This may be due to the superior electrode properties of TPU–AgNFw dry electrodes including improved contact impedance and noise character.

Fig. 12 Static agar phantom images reconstructed at different frequencies by TPU–AgNFw dry electrodes (left) and Ag/AgCl gel electrodes (right). (a) The reference image, where the blue dot refers to the lowest electrical conductivity (0.02 wt% NaCl), the red dot denotes the highest electrical conductivity (0.5 wt% NaCl), and the yellow dot represents medium conductivity (0.1 wt%): (b) at 1 kHz, (c) at 10 kHz, and (d) at 100 kHz.

On the other hand, the results of the reconstructed agar phantom images, which are more or less distorted and dislocated compared to the reference image, reveal the limitation in our EIT system itself to a certain degree. This kind of small discrepancy cannot be avoided yet, because there are many parameters apart from electrode properties that can affect the results of reconstructed images including the frequency of the injected current, the reconstruction algorithm, the calibration method, and the placements of the electrodes.^5,37,38 In this case, one of the main factors influencing the reconstructed images should be frequency because it has a huge impact on the admittivity of the agar phantom. This is also observed in Fig. 12. The images significantly varied as a function of the frequency. Another factor influencing image reconstruction is related to the settings of the electrodes in the preparation stage. To obtain agar phantom images based on differences in conductivity, the electrodes were first attached to the homogeneous agar phantom system, and the baseline data for the background was measured. After that, the attached electrodes were removed and two holes were punched in the main agar phantom. Two kinds of agar/NaCl solutions were poured into each hole, and these were left at room temperature for a long time (several hours) until the agar solution solidified. Finally, the electrodes were attached again and the EIT data of the different agar phantoms were measured. Two installations of electrodes were needed for this process. Therefore, the gap between the neighboring electrodes not only should be equidistant, but the position of each electrode also has to be the same in order to acquire reproducible EIT images. However, this is actually very difficult in this EIT test process, and some deformation and errors in position are unavoidable. In addition to that, NaCl molecules in the agar with higher NaCl concentrations can diffuse into the agar with lower NaCl concentrations while the poured agar solution solidifies due to differences in the chemical potential. This results in a diffused boundary layer, which also may affect the final EIT image shapes of the two small agar circles.

The main attraction of EIT in clinical applications is to monitor regional lung function of an intensive care patient to prevent ventilator-induced lung injury by guiding positive end-expiratory pressure (PEEP) trials. This method can be used at the bedside as a continuous monitoring technique since it only requires small devices that enable radiation-free recording principles that are relatively inexpensive. The advantages of EIT cannot be achieved by using established equipment such as CT or f-MRI.^37–39

As various biopotential signals have already been successfully acquired using the TPU–AgNFw dry electrode, we also compared these with the Ag/AgCl gel electrode in terms of its performance at high frequency in the case of a lung EIT system for recording the subject's lung function.

Fig. 13 displays a series of time-difference pulmonary images representing air filling starting from end-expiration to end-inspiration recorded using TPU–AgNFw dry electrodes and Ag/AgCl gel electrodes, respectively, at alternating injection currents of 1 mA amplitude at a frequency of 100 kHz. In dynamic lung EIT tests of a human subject, the 2D-image corresponding to the end-expiration was set as a background, which is the first image of each line on the far left of Fig. 13. The last image corresponds to end-inspiration. Each series of time-difference lung images present gradual step by step respiration very clearly. However, there are small differences in the two series of lung images. This is mainly because they are not recorded concurrently. Therefore, each image cannot represent the same time situation exactly. In addition, due to the three dimensional current flow, EIT generally captures a relatively thick, lens-shaped volume of lung, which may depend upon the contact area of the electrodes. The contact area of the gel of the Ag/AgCl electrode was larger than that of the TPU–AgNFw dry electrode; this may inevitably induce a certain difference in the two series of lung images no matter how fairly they are tested. This means that the reconstructed images do not exactly represent the actual cross-sectional lung where the electrodes were placed. Furthermore, the cross-section of the subject's chest is not a perfect circle, so the reconstructed lung images may also be deformed when a circular domain is employed in an image reconstruction algorithm. The signal to noise ratio (SNR) of the two EIT measurements were also calculated.⁵ The SNR value of TPU–AgNFw was 40.35 dB, and the Ag/AgCl gel electrodes exhibited an SNR of 42.35 dB. Thus, both SNR levels were similar.

Fig. 13 Series of time-difference pulmonary images representing air filling during inspiration recorded at a frequency of 100 kHz by (a) TPU–AgNFw dry electrodes showing an SNR value of 40.35 dB and (b) Ag/AgCl gel electrodes showing an SNR value of 42.35 dB.

In summary, the TPU–AgNFw dry electrode was thoroughly investigated in terms of biopotential recording and EIT testing using an agar phantom-electrode system and human subject tests, indicating that the TPU–AgNFw dry electrode had performance comparable to Ag/AgCl gel electrodes. More realistic data is provided in the supplementary biopotential recording video and reconstructed dynamic lung EIT video.

Discussion

Biopotentials are electrical potentials generated due to distributed ions in body, and these potentials can be measured by electrodes when they are transferred to the body surface.¹⁸ In this process, the currents should pass through the skin, which has high impedance, as compared to the conductivity of the human body, especially due to the stratum corneum, which is the outer layer of the skin. The stratum corneum consists of many layers of compacted, flattened, non-nucleated dehydrated cells (named comeocytes), which are filled with cross-linked keratin. The stratum corneum may have a thickness of about 10–40 μm, and typically has around a few hundred kΩ to one MΩ per square centimeter resistivity for a 10 Hz input signal frequency.^29,40,41 Therefore, the stratum corneum significantly inhibits the transfer of electrons from the tissue to the electrode.^15,17,41 Consequently, the skin-electrode contact impedance is mainly dominated by the impedance of the stratum corneum since the impedance of the stratum corneum is normally much higher than that of the electrode. One of most important functions of the gel used in the Ag/AgCl gel electrode is that the gel decreases the skin-electrode contact impedance through moisturizing the stratum corneum layer. The impedance of the stratum corneum depends on wetness, therefore provision of an ionic path way between the metal part of the electrode and the skin makes the transduction of ion currents into electron currents easier.^1,19 This is one of the main reasons why the gel is employed in Ag/AgCl gel electrodes even though it is associated with so many problems. AgNFw dry electrodes have no gel to hydrate the stratum corneum. However, hydration can be achieved by means of perspiration from the human body. Although the ion conductivity of sweat is lower than that of the gel, the impedance of the stratum corneum will become low enough for stable signal recording.^18,19 We also observed in this study that it took several minutes to obtain stable and optimal signals depending on skin conditions. Although reducing the impedance of the stratum corneum is crucial for acquiring high quality biopotential from dry-contact electrodes, it does not depend much on the electrodes themselves, but it mostly depends on ambient conditions and human physical conditions.⁴²

Therefore, one of the main characteristics of AgNFw dry electrodes that influence the accuracy of the recorded biopotentials is the low impedance of AgNFws; this is an important criterion for dry-contact electrodes to reduce noise and to avoid signal distortion during monitoring.^1,17 For textile electrodes including AgNFw dry electrodes, the electrode impedance can be considered as a parallel combination of resistance and capacitance because the pores inside the electrodes function as capacitors. Based on the contact impedance of TPU–AgNFw dry electrodes on agar phantom, the pure resistance shown in Fig. 5 was even lower than that of the metal part of the Ag/AgCl electrode, and the capacitance part, which was averaged over frequency, also almost reached the level of Ag/AgCl electrodes. Consequently, TPU–AgNFw dry electrodes exhibited much lower noise power density than Ag/AgCl electrodes and showed no significant signal distortion in step response test. All of these results for TPU–AgNFw dry electrodes guaranteed highly accurate biopotential recordings on the human subjects tested in this study.

To be more specific, the low impedance of TPU–AgNFw dry electrodes was chiefly attributed to the improved electroless silver plating process. As mentioned above, the electroless silver plating process was upgraded and optimized in order to improve the silver plating quality of the AgNFw. All of the solutions were diluted to reduce the silver reduction rate. We further recognized that the silver particles no longer formed a plating on the nanofiber web surface when the silver particles grew to a certain critical size; rather, they merely agglomerated together. This also led to an unnecessary cost increase. Therefore, reducing the concentration of the electroless plating solution resulted in more silver plating on the Ag-catalyzed nanofiber surface, which prevented agglomeration in the solution.

The hydrophilic properties of the nanofiber webs were improved by lengthening the isopropanol dipping time. Although the surface hydrophilicity of TPU and SBS was better than that of PVDF used in the previous study,^5,6 it was not enough for the silver plating solution to permeate deeply into the nanofiber web because of the extremely small sizes of the pores between nanofibers. In addition, specialized and custom-designed silver plating equipment was employed to improve the uniformity of the AgNFw. This equipment was divided into two parts, a rotating device and a silver plating box (see ESI Fig. S1†). Two clips were installed in the silver plating box to fix the nanofiber web so that silver plating can be conducted more evenly and the silver plating box was made of highly hydrophobic material, such as polypropylene, so as to prevent silver from coating the box surface during the silver plating.

For individual metal plated fiber, there are two important factors which affect the impedance of the fiber including the thickness of the plated layer and the surface area-to-volume ratio of the metal coated fiber.¹⁷ In order to reduce the resistive part of the fiber, the plated layer should be thick. By comparing the two FE-SEM images of Fig. 14(a) and (b), it can be clearly observed that the diameters of the TPU nanofiber became much larger after the silver plating process, and the thickness of the silver layer was a little larger than the fiber diameter as shown in Fig. 14(c). More importantly, the metal coated fiber structure normally had large capacitive properties, which may cause signal distortion.¹⁷ Since nanoscale fiber was used in this study, a high surface area-to-volume ratio was easily obtained to increase the capacitive part of the fiber (detailed calculation is referred to ESI file†). In addition to that, the capacitance created by the gap between fibers and the contact resistance caused by fiber–fiber contacts play an important role in the impedance of TPU–AgNFw dry electrodes. In general, the nanofiber web has high porosity, and air occupies about 80% of the volume of the web. Due to the increased diameter of TPU nanofibers after the silver plating process, the gap between fibers can be diminished to a certain degree to further increase the capacitive part of the TPU–AgNFw dry electrode. Moreover, one interesting result obtained from an improved silver plating process is that most fibers are connected by reduced silver as shown in Fig. 14(b) and (d). Certainly this can also further reduce the fiber–fiber contact resistance, contributing to a decrease in the total impedance of TPU–AgNFw dry electrodes.

Fig. 14 FE-SEM images of (a) TPU nanofiber web and (b) silver plated TPU nanofiber web. Cross-sectional TEM images of (c) silver plated TPU nanofiber and (d) two silver plated TPU nanofibers linked by reduced silver.

The skin-electrode contact impedance is also governed by the contact area and skin properties, thus lower impedances were obtained for larger contact areas.¹⁸ The other benefit of using a Ag/AgCl gel electrode is that the gel quickly occupies the pores and wrinkles in the skin to maximize the effective contact area and minimize the generation of capacitance that is formed in the air between the electrode and the skin.⁴¹ It seems that the TPU–AgNFw dry electrode can have tight and good contact to the skin by conforming to the skin topography because it is flexible, has nanoscale fiber diameters that are much smaller than the size of the pores and wrinkles on the skin surface, and has a high specific area. However, it takes a little more time for the TPU–AgNFw dry electrode to conform to the skin contours and irregularities than the gel.

Further, it is possible that ionic currents in the human body may be directly detected on appendages to the skin such as hair follicles, sweat ducts, and sebaceous glands without passing through the stratum corneum.⁴³ In addition, every square centimeter of skin area comprises 40–70 hair follicles and 200–250 sweat ducts,⁴¹ and the sizes of the hair follicles is 60–160 μm, and the diameter of the sweat ducts ranges from 5 to 20 μm.⁴⁴ Taking advantage of this follicle transdermal transport mechanism, a new kind of biopotential fiber sensor was developed and an ECG signal was obtained using a biopotential fiber sensor that was comparable to the signal from an Ag/AgCl electrode (and was even better in terms of noise performance).⁴² The biopotential fiber sensor consists of a fiber bundle of about 100 individual fiber strands coated with an Ag/AgCl ink where the diameter of each fiber was about 20 μm. Compared to such a fiber sensor, the fiber diameter of a TPU–AgNFw dry electrode was much smaller, even smaller than the size of sweat ducts, and has a 0.6 cm² contact area. This means that there is a high possibility that some silver plated TPU nanofibers could contact the pores on the skin to record biopotential rather than through the stratum corneum. As long as the TPU nanofibers touch the pores, the skin-electrode contact impedance will be reduced significantly and will likely be able to reach the level of the Ag/AgCl electrode. This would offset the shortage that the impedance reduction in the stratum corneum caused by sweat in the TPU–AgNFw dry electrode is lower than that reduced by the conductive gel in the Ag/AgCl gel electrode.

In contrast to biopotential measurements, the electrodes serve two functions in EIT tests including current injection and receiving the resulting voltages. Since high frequency alternating currents (≥1 kHz) are commonly injected in the EIT test, different aspects of electrodes should be considered. In this case, the capacitive part of the skin-electrode contact impedance can be an important factor, because both injected and detected currents may also pass through capacitive parts as well as the resistive parts of the electrodes due to the reduction in the capacitive reactance at high frequency.

Although positive results were obtained in human subject lung EIT testing, it was difficult to obtain any reconstructed lung image using TPU–AgNFw dry electrodes at frequencies of 10 kHz and 50 kHz. In contrast, Ag/AgCl gel electrodes obtained relatively high quality reconstructed lung images at these frequencies (see ESI Fig. S4 and Fig. S5†). It is evident that this is associated with the different electrode properties of TPU–AgNFw dry electrodes than the Ag/AgCl gel electrode because the EIT tests were conducted for the same subject with the same EIT system under the same conditions.

As stated above, TPU–AgNFw dry electrodes have inherently large capacitive properties resulting from the textile structures such as the coated fibers and the gaps between fibers. Even though the effect of these factors was reduced as much as possible via a silver plating process, they cannot be eliminated completely. This can have a negative influence on the TPU–AgNFw dry electrode in the current injection process since this can waste a part of the injected AC energy. Moreover, the main reason for using a high frequency AC source is to “penetrate” the high skin-electrode contact impedance, especially the stratum corneum. If the skin-electrode contact system is considered as a parallel RC circuit separated by a stratum corneum layer (considered as another parallel RC circuit) as illustrated in Fig. 15, the capacitive reactance decreases with increasing frequency. Thus, much of the current must pass through the capacitive part of the stratum corneum rather than the resistive part at high frequency because of the extremely high resistance of the stratum corneum. Perspiration will wet the stratum corneum resulting in a decrease in the resistance and an increase in the dielectric permittivity, which increases the capacitance of the skin-electrode contact system consisting of a TPU–AgNFw dry electrode. However, sweat is not as effective as that of the gel of Ag/AgCl dry electrode due to the lower conductivity of the sweat.^18,19 This is also one disadvantage of TPU–AgNFw dry electrodes compared to Ag/AgCl gel electrodes. Considering the contact area difference between the two types of electrodes, the larger contact area of the Ag/AgCl gel electrode obviously leads to a higher capacitance and a lower resistance. Therefore, all of these adverse factors of TPU–AgNFw dry electrodes result in a problem in that the injected alternating current must have a much higher frequency to have enough energy to penetrate the skin-electrode system. Fortunately, although such factors of TPU–AgNFw dry electrodes can be eliminated at high frequency (100 kHz), they should be considered carefully in developing and designing a new type dry EIT electrodes because the frequency cannot be increased blindly as it will eventually become harmful to the human body.

Fig. 15 Schematic of skin-electrode contact system with (a) TPU–AgNFw dry electrode, (b) Ag/AgCl gel electrode. (c) Equivalent electric parallel RC circuit.

On the contrary, it is interesting that a satisfactory result was obtained using TPU–AgNFw dry electrodes even at 1 kHz in an agar phantom EIT static test. This is mainly because of the high electric conductivity of the agar phantom system due to the absence of a highly insulated stratum corneum layer. Therefore, the alternating current is easily injected and received even at low frequency.

Conclusions

The performance of elastic polymer-based AgNFw dry electrodes and their performance in a wide range of biopotential recordings and EIT measurements ranging from low frequency to high frequency have been thoroughly examined and compared with that of Ag/AgCl gel electrodes. Moreover, the detection mechanism and some advantages of AgNFw dry electrodes were discussed based on their structure. The results indicated that the performance of the TPU–AgNFw dry electrode compared favorably with Ag/AgCl gel electrodes under static test conditions. All of these positive results may be attributed to the unique nanofiber web structure of the TPU–AgNFw dry electrode.

To quantitatively evaluate the AgNFw dry electrodes, a simple eyelet shape electrode type was selected, but other applications are also available since AgNFws are thin, flexible, washable, and can even be directly integrated into electric circuits. Although motion artifacts with regard to AgNFw dry electrodes were not included in this study, motion artifacts may be more affected by some other factors, especially electrode design, rather than the material itself. Several works aiming at developing prototypes with anti-artifact design by extending some other applications are presently in progress for future publication.

Acknowledgements

This work was supported by the Ministry of Health and Welfare (Grant No. HI14C0743) in South Korea.

Notes and references

1. Y.-H. Chen, M. de Beeck, L. Vanderheyden, E. Carrette, V. Mihajlović, K. Vanstreels, B. Grundlehner, S. Gadeyne, P. Boon and C. Van Hoof, Sensors, 2014, 14, 23758–23780.
2. C. T. Lin, L. De Liao, Y. H. Liu, I. J. Wang, B. S. Lin and J. Y. Chang, IEEE Trans. Biomed. Eng., 2011, 58, 1200–1207.
3. L. F. Nicolas-Alonso and J. Gomez-Gil, Sensors, 2012, 12, 1211–1279.
4. R. Chaudhari, R. Deore and B. Gawali, Int. J. Adv. Res. Comput. Eng. Technol., 2012, 1, 90–92.
5. T. I. Oh, T. E. Kim, S. Yoon, K. J. Kim, E. J. Woo and R. J. Sadleir, Physiol. Meas., 2012, 33, 1603–1616.
6. T. I. Oh, S. Yoon, T. E. Kim, H. Wi, K. J. Kim, E. J. Woo and R. J. Sadleir, IEEE Transactions on Biomedical Circuits and Systems, 2013, 7, 204–211.
7. A. C. Myers, H. Huang and Y. Zhu, RSC Adv., 2015, 5, 11627–11632.
8. J. Yoo, L. Yan, S. Lee, H. Kim and H. J. Yoo, IEEE Trans. Inform. Tech. Biomed., 2009, 13, 897–902.
9. C. T. Lin, K. C. Chang, C. L. Lin, C. C. Chiang, S. W. Lu, S. S. Chang, B. S. Lin, H. Y. Liang, R. J. Chen, Y. T. Lee and L. W. Ko, IEEE Trans. Inform. Tech. Biomed., 2010, 14, 726–733.
10. M. Merino, O. Rivera, I. Gómez, A. Molina and E. Dorronzoro, in Proceedings – 1st International Conference on Sensor Device Technologies and Applications, SENSORDEVICES 2010, 2010, pp. 100–105.
11. A. B. Usakli, S. Gurkan, F. Aloise, G. Vecchiato and F. Babiloni, Comput. Intell. Neurosci., 2010, 2010, 1–5.
12. S. Yathunanthan, L. U. R. Chandrasena, A. Umakanthan, V. Vasuki and S. R. Munasinghe, in Proceedings of the 2008 4th International Conference on Information and Automation for Sustainability, ICIAFS 2008, 2008, pp. 283–288.
13. I. B. Bahareh Taji, S. Shirmohammadi and V. Groza, in Proceedings of the 29th Annual International Conference of the IEEE EMBS, Cite Internationale, Lyon, France, August 23–26, 2007, pp. 5739–5742.
14. H. Singh and J. Singh, International Journal of Advanced Engineering Technology, 2012, 3, 115–122.
15. Y. M. Chi, T. P. Jung and G. Cauwenberghs, IEEE Reviews in Biomedical Engineering, 2010, 3, 106–119.
16. N. Meziane, J. G. Webster, M. Attari and a. J. Nimunkar, Physiol. Meas., 2013, 34, R47–R69.
17. P. J. Xu, H. Zhang and X. M. Tao, Text. Prog., 2008, 40, 183–213.
18. A. Gruetzmann, S. Hansen and J. Müller, Physiol. Meas., 2007, 28, 1375–1390.
19. A. Serle and L. Kirkup, Physiol. Meas., 2000, 21, 271–283.
20. K. P. Hoffmann and R. Ruff, Annu. Int. Conf. IEEE Eng. Med. Biol. – Proc., 2007, vol. 4, pp. 5739–5742.
21. S. S. Lobodzinski and M. M. Laks, Cardiology Journal, 2008, 15, 477–480.
22. E. Forvi, M. Bedoni, R. Carabalona, M. Soncini, P. Mazzoleni, F. Rizzo, C. O'Mahony, C. Morasso, D. G. Cassarà and F. Gramatica, Sens. Actuators, A, 2012, 180, 177–186.
23. S. M. Eeg, R. Ma, D. Kim, M. Mccormick, T. Coleman and J. Rogers, 32nd Annu. Int. Conf. IEEE EMBS, Buenos Aires, Argentina, August 31–Sept. 4, 2010, vol. 2, pp. 6405–6408.
24. D.-H. Kim, N. Lu, R. Ma, Y.-S. Kim, R.-H. Kim, S. Wang, J. Wu, S. M. Won, H. Tao, A. Islam, K. J. Yu, T. Kim, R. Chowdhury, M. Ying, L. Xu, M. Li, H.-J. Chung, H. Keum, M. McCormick, P. Liu, Y.-W. Zhang, F. G. Omenetto, Y. Huang, T. Coleman and J. a. Rogers, Science, 2011, 333, 838–843.
25. R. J. Prance, S. T. Beardsmore-Rust, P. Watson, C. J. Harland and H. Prance, Appl. Phys. Lett., 2008, 93, 1–4.
26. M. Park, J. Im, M. Shin, Y. Min, J. J. Park, H. Cho, S. Park, M.-B. Shim, S. Jeon, D.-Y. Chung, J. Bae, U. Jeong and K. Kim, Nat. Nanotechnol., 2012, 7, 803–809.
27. S. Grimne, Med. Biol. Eng. Comput., 1983, 21, 750–755.
28. N. M. Oldham, Am. J. Clin. Nutr., 1996, 64, 405S–412S.
29. J. Rosell, J. Colominas, P. Riu, R. Pallas-Areny and J. G. Webster, IEEE Trans. Biomed. Eng., 1988, 35, 649–651.
30. M. Iwasaki, C. Kellinghaus, A. V Alexopoulos, R. C. Burgess, A. N. Kumar, Y. H. Han, H. O. Lüders and R. J. Leigh, Clin. Neurophysiol., 2005, 116, 878–885.
31. C.-T. Lin, L.-W. Ko, J.-C. Chiou, J.-R. Duann, R.-S. Huang, S.-F. Liang, T.-W. Chiu and T.-P. Jung, Proc. IEEE, 2008, 96, 1167–1183.
32. M. Costa, A. L. Goldberger and C.-K. Peng, Phys. Rev. Lett., 2002, 89, 068102.
33. M. Costa, A. L. Goldberger and C. K. Peng, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 2005, 71, 1–18.
34. A. C. Yang, S. J. Wang, K. L. Lai, C. F. Tsai, C. H. Yang, J. P. Hwang, M. T. Lo, N. E. Huang, C. K. Peng and J. L. Fuh, Prog. Neuro-Psychopharmacol. Biol. Psychiatry, 2013, 47, 52–61.
35. P. Laferriere, E. D. Lemaire and A. D. C. Chan, IEEE Trans. Instrum. Meas., 2011, 60, 3259–3268.
36. A. Barrett, a. E. Donnelly and G. ÓLaighin, in Annual International Conference of the IEEE Engineering in Medicine and Biology – Proceedings, 2007, pp. 4883–4886.
37. I. Frerichs, Physiol. Meas., 2000, 21, R1–R21.
38. B. Grychtol, G. Elke, P. Meybohm, N. Weiler, I. Frerichs and A. Adler, PLoS One, 2014, 9, e103045.
39. J.-M. Constantin, S. Perbet, J. Delmas and E. Futier, Crit. Care, 2014, 18, 164.
40. A. S. Berson and H. V Pipberger, Am. Heart J., 1968, 76, 514–525.
41. E. T. McAdams, J. Jossinet, a. Lackermeier and F. Risacher, Med. Biol. Eng. Comput., 1996, 34, 397–408.
42. S. M. Lobodzinski and M. M. Laks, J. Electrocardiol., 2006, 39, S41–S46.
43. C. Cullander, Solid State Ionics, 1992, 53–56, 197–206.
44. N. Otberg, H. Richter, H. Schaefer, U. Blume-Peytavi, W. Sterry and J. Lademann, J. Invest. Dermatol., 2004, 122, 14–19.

---

Footnote

† Electronic supplementary information (ESI) available: Biopotential recording video, reconstructed dynamic lung EIT video, silver plating equipment and calculation of the surface-to-volume value of the metal coated fiber. See DOI: 10.1039/c6ra00079g

---