Wearable garments for the measurement of vital signs and biosignals are believed to be a primary part of more personalized, pervasive healthcare to tackle the problems arising with the aging population. During the last decade, several wearable systems including textile electrodes have been reported [1
] and are even being used in commercial products, e.g., Hexoskin T-shirts and Equivital vests. However, there still are issues with the textile–electronic integration, limiting the mass production of such sensorized garments and delaying the market growth forecasted for wearable technology. Manufacturing solutions for interconnecting sensors and wearable instrumentation, such as intarsia knitting with conductive yarn [4
], have been proposed. However, these solutions are inadequate for the mass production of embedded cables in the garments, and there are still other key areas where practical solutions for textile–electronic integration are required.
Interconnections between leads and instrumentation are required to be both electrically and mechanically robust. In addition to such common requirements, when connecting electronics in sensorized garments, the interconnection must stand mechanical pulls resulting from the weight of the device and the natural wearing of the garment itself. Moreover, given the chances for rain or an excess of sweat during wear, another requirement would be moisture resistance or even waterproofing. Interconnectors fulfilling these requirements are often found in the field of electronics; however, unfortunately, they are usually costly, bulky, and not stretchable (see Figure 1
Most importantly, they are completely foreign to the textile manufacturing process, adding more hurdles to the already challenging process of textile–electronic integration [5
Screen printing in textile manufacturing is a widespread and established process worldwide [6
]; conductive pastes are being continuously developed due to the printed electronics phenomena boosted in the last decade [7
]. While most conductive pastes and inks are based on expensive silver or toxic copper [8
], both with no elastic properties, carbon-based materials provide electrical conductivity and elastic possibilities [9
]. Therefore, combining the elastic and conductive properties of carbon nanotubes (CNTs) [13
] with screen printing technology seems feasible and has the required potential for enabling textile–electronic integration.
Wearable systems including electrocardiogram recordings have high potential for long-term electrocardiogram recordings. Such recordings have high potential for the screening, diagnosis, and monitoring of different medical conditions, e.g., atrial fibrillation [15
], autonomic nervous system imbalance, risk assessment of physical workload [17
], and sport applications, e.g., estimating energy expenditure [18
In this paper, we introduce the concept of being “textile-friendly” to refer to those methods and techniques for integrating electronics with textiles that take into the account the intrinsic features of textiles—soft, drapable, etc.—and utilize common methods currently accepted in textile manufacturing or implement techniques especially well-suited for soft, light, and drapable materials. As an example of textile-friendliness, the authors present a solution for interconnecting soft textile conductive elements and hard electronic instrumentation components with a fabric printed with an elastic and conductive CNT-based paste. Electrocardiogram (ECG) recordings obtained with the prototype interconnector were studied with the aim to assess the functional performance of the textile–electronic interconnection.
2. Materials and Methods
The experimental measurements were taken after informed consent was given, according to ethical approval no. 274-11 granted by the Ethical Review Board of Gothenburg, Sweden.
2.1. Polymer-Based Conductive Elastic Paste
The conductive paste was developed at the University of Borås, and it is based on a stretchable polymer compound formulated with multi-walled carbon nanotubes (MWCNTs) (Chengdu organic chemicals, P.R. China) as the conductive component, proper additives to promote homogienity/printability, and polyurethane (PU) as the binder. The paste manufacturing process was optimized regarding material compositions and mixing parameters (speed and duration), targeting conductivity, homogeneity, and printability parameters including viscosity, dispersion quality, and nominal sheet resistivity in a relaxed state and after elongation. The process started with 10 min high-speed mixing using an overhead mixer equipped with a bladed propeller stirrer, IKA RW 20 Digital Dual-Range Mixers, Germany. After mixing and homogenizing using magnetic stirring at room temperature for 13 h, the prepared mix of the PU, CNTs with water, and additives was filtered through an 80 μm mesh, and the printable MWCNT paste was obtained.
The surface impedance of the developed paste was characterized for different values of thickness and elongation, applying a four-probe method using a Keithley 2000 digital multimeter (Keithley Instruments, Ohio, USA).
The connector was prototyped at the University of Borås using screen-printed fabrics and snap buttons as shown in Figure 2
a–c. The fabrics were printed with the conductive pastes in a semiautomatic screen printer from DEK at Acreo, Printed Electronics Laboratory, Norrköping, Sweden. The woven fabric 4191F (F.O.V. Fabrics AB, Sweden) was coated with PU, and it was stretchable in one direction. The conductive pastes used were the developed MWCNTs and the silver-based Dupont Conductor 5000 (DuPont, Delaware, USA).
The conductive paste was printed over the PU-coated side of the 4191F fabric and male snap buttons were pinned through the circular printed pads as shown in Figure 2
a,b. The snap buttons were intended for garment interconnection, while the square pads on the edge of the fabric were intended for connecting the measurement lead points of the wearable measurement instrumentation.
The screen-printed fabrics were cured in a fume hood for 120 min at room temperature for the MWCNT paste and 16 min at 100 °C for the Conductor 5000.
The sister connector using the Dupont 5000 paste was produced to evaluate the influence of the intrinsic electrical conductivity on the functional performance of the connector.
The junction impedance was measured using a Keithley 2000 digital multimeter (Keithley Instruments, Ohio, USA) using the two-probe method over three sets of textile–electronic interconnectors in static conditions and during elongation and folding.
2.3. Sensorized Garment
The connectors were tested with a sensorized T-shirt with sewn textile electrodes in the inner layer of the garment, as shown in Figure 3
c. With the sole purpose of ensuring a high-quality ECG measurement and avoiding any potential source of variability coming from the electrodes or the textile interconnection, commonly and extensively used solutions were selected in the construction of the sensorized garment.
The electrodes were made with conductive Shieldex® P130+B fabric manufactured by STATEX Gmbh (Bremen, Germany) of size 6 × 4 cm [20
]. The position of the electrodes allows for performing a lead I ECG recording, i.e., a lateral bipolar recording
, which is also called “right arm to left arm”. The electrodes were connected through copper cables to the female snap buttons on the sleeve as shown in Figure 3
b for interconnection with the screen-printed connectors. Note that the only elements in contact with the skin are the textile electrodes, see Figure 3
c, which were made with a silverized cotton fabric certificated as biocompatible by the manufacturer; therefore, no additional biocompatibility testing was required.
2.4. Experimental Setup
The evaluation was done by comparing the ECG recordings obtained with the MWCNT-based and the silver-based connectors with ECG recordings obtained with Ag/AgCl repositionable red-dotTM 2670-5 gel electrodes manufactured by 3M of size 4 × 3.2 cm. The recordings were obtained with an ADAS1000 evaluation board (Analog Device, Inc., Massachusetts, United States) configured for simultaneous recordings of two independent continuous measurement channels at sampling frequency of 2 kHz. The data collection was done by a Windows application, and the ECG recordings were saved and imported to MATLAB 2017 (The Mathworks, Inc., Massachusetts, United States) for analysis and comparison.
To allow a full and accurate comparison of the obtained ECG recordings using the MWCNT connector, beyond signal morphology in time and spectral content, the gel electrodes were placed as closely as possible to the textile electrodes to record simultaneous ECG measurements. These measurements were used to perform maximum value on the R-wave in each ECG complex (R-peak) detection, which is critical to any heart rate and heart rate variability application. The experimental recordings were done in two steps:
Simultaneous synchronized ECG recording with gel electrodes and the sensorized garment using the MWCNT connector;
Simultaneous synchronized ECG recording with gel electrodes and the sensorized garment using the Ag connector.
The recordings were performed 10 times for a minimum of 20 s each. The subject remained seated during the experiment with a normal breathing pace at rest.
2.5. ECG Processing
The measurements were preprocessed for powerline interference, baseline wander, and white noise removal. Removed components in each step were analyzed for comparison. The powerline interference was removed by zero-phase forward–backward filtering with an infinite impulse response notch filter ([email protected]
Hz, 2 Hz bandwidth). The baseline wander was estimated by the lowest coarse approximation of the discrete wavelet transform. The symlet wavelet and a decomposition level of 10 were used according to [21
]. The remaining noise including white noise was removed via the wavelet denoising method. With the purpose of comparing the morphological patterns of recordings, the SURE [22
] thresholding rule and hard thresholding method were used to minimize the denoising error in waveform complex formed by the Q, R and S waves on a ECG complex (QRS) complex analysis [23
]. R-peaks were detected on the preprocessed recordings using a Pan-Tompkins QRS detector [24
2.6. Material Characterization
The morphology of the printed conductive paste surfaces was characterized by scanning electron microscopy using a Hitachi s-4800 FE-SEM. The acceleration voltage was 5 kV, and the surfaces were coated with chromium before microscopy.
2.7. Signal Comparison
Powerline interference and the baseline wander level and R-peak amplitude were compared among each setup. The powerline interference level was measured by calculating the Root Mean Square (RMS) value of the removed 50 Hz component. The difference between the maximum and minimum values on baseline in a 10-second window was used to indicate the baseline wander level for a short period.
A visual ECG morphology comparison was implemented on ensemble-averaged ECG complex data, including waverform complex formed all waves present on a ECG complex except U (PQRST) waveforms, representing the heart’s electrical activity [25
]. All PQRST complexes in a 15-second segment were extracted, summed with the alignment of their R-peak, and then averaged [26
]. The averaged signals, with non-repetitive interference removed, can indicate systematic morphological differences.
The similarities between recordings from the textile electrodes with MWCNT/Ag connectors and the corresponding synchronized gel recordings from the gel electrode were examined in both time and frequency domains. The time series linear correlation was measured using Person’s correlation coefficient defined by the covariance of the two series over the product of their standard deviations. Cross-spectrum coherence was used for frequency domain comparison. The spectrum was acquired using Welch’s overlapped averaged periodogram and the magnitude-squared coherence was calculated.
The agreement of the Time between R peak in consecutive ECG complexes (RR Intervals) series extracted from the recordings obtained with the textile electrodes using MWCNT-based and silver connectors and the corresponding synchronized recordings obtained with the gel electrodes was evaluated by calculating the Lin’s concordance correlation coefficient [27
The prepared MWCNT/polyurethane paste showed a good-quality print on the PU-coated fabric used as a substrate for the connector. The MWCNT material in the interconnector makes the interconnector potentially toxic if it contacts the skin. In such a case, if used as an electrode material, for instance, biocompatibility testing would be been necessary to assess the actual toxicity levels and whether recommendations against their use should be made. Fortunately, in the application presented in this work, it is an interconnector placed outside the sensorized garment, actually connected to the recording device; therefore, between that and the fact that the folding of the PU-coated fabric actually packages the conductive material between two layers of PU, toxic material leaving the interconnector is practically impossible.
4.1. Resulting Electrical Properties of the Connectors
The obtained level of sheet resistivity was quite good when compared with the values previously reported by other authors. Super-aligned CNT films showed a resistance of 1 kΩ/sq before stretch and 1.6 kΩ/sq after 60% stretch [29
]. poly(3,4-ethylenedioxythiophene) (PEDOT): poly(styrenesulfonate) (PSS) sandwiched between poly(3-hexylthiophene) (P3HT): phenyl-C61-butyric acid methyl ester (PCBM) and Polydimethylsiloxane (PDMS) (buckled) presented a resistivity of 750 kΩ/sq, which was stretchable up to 25% [30
]. However, long Ag nanowires on Ecoflex showed resistivity of 9−70 kΩ/sq in prepared stretchable electrodes of a very long Ag nanowire percolation network made by Lee et al. [31
The observed nonlinear increase in the surface resistivity with elongation was reported by Xu and Zhu in [32
] and by Suikkola et al. in [33
]; the ability of the MWCNT paste to increase only 11% in surface resistivity for a stretch of 10% is very attractive, especially for this kind of interconnecting application, since it is not expected that a connector needs to stretch far beyond 10%−20%.
The reported increase in the resistance of the connectors suggests that there is certain influence of bending on the printed tracks. The obtained values fall within the range of values reported recently by Du et al. in [34
] for bended tracks screen printed with nanowire-based screen printing paste and by Maheshwari et al. in [35
] for Ag-based paste. Unfortunately, the sample size does not allow for performing any statistical significance analysis, and in order to produce any accurate further statements on eventual crack formation, SEM images of the connectors after the folding has been performed must be obtained.
4.2. Performance as an Interconnector
Despite the differences in volume resistivity of conductive compounds (20.681 Ω/sq/mil for the MWCNT paste and 15 mΩ/sq/mil for the Dupont Conductor 5000 [36
]) and in the total resistance of the interconnector (10.47 kΩ and 7.47 Ω for the MWCNT paste and the Dupont Conductor, respectively), the results have shown that there is no significant difference in the signal quality or in measurement performance related to the use of the textile–electronic connector in either implementation. The textile measuring system with a sensorized garment and connector was able to acquire ECG recordings with similar quality as those acquired using directly connected Ag/AgCl electrodes. The observed baseline wandering and powerline interference introduced by the system could be eliminated by proper filtering done when acquiring the ECG recordings.
The observed significant differences in volume resistivity are actually negligible when comparing these values with the volume resistivity value of copper. Conventional flex printed circuit connectors are made using polyimide substrates and printed copper traces with volume resistivity values as low as 0.67 mΩ/sq/mil [37
]. The increased resistance observed for the MWCNT-based connector should not affect the measuring setup since it is <105 times smaller than the input impedance of the measurement instrumentation used for the electrical characterization [38
]. Such a connector seems completely suitable for biopotential measurement applications but might not be a proper solution for other kinds of measurement modalities. In some cases, not only is voltage sensed but also electrical stimulation is required, e.g., electrical bioimpedance measurements, where low resistance in series with the measurement load is required to facilitate the current flow through the load to minimize measurement artifacts caused by stray capacitances [39
] or electrode mismatch [40
From the interfacing perspective with instrumentation amplifiers and other sensing electronic instrumentation, the resistance increase observed after folding is relatively too small to have any impact on the overall sensing function. In any case, a more adequate characterization study should be performed. In this regard, the authors believe that a folding test is actually very aggressive considering the intended purpose of the connector. Bending is probably a more realistic and adequate test for a connector meant to be attached to a recording device.
4.3. Functionalization of Textiles as Textile–Electronic Interconnections
CNT-based products have been used for functionalizing textiles for a decade [41
], and screen printing processes have been shown to be effective for producing conductive textiles [42
] with various purposes such as transmission lines [43
] or radiofrequency antennas [44
]. Combining CNTs with screen printing has been a successful approach to prototyping electronics on rubber substrates producing elastic diodes [45
In the presented approach, the elasticity was preserved by PU, which is used in a twofold role as a binder and coating agent. First, as an elastic binding agent, polyurethane preserves the ability to stretch to the conductive paste; second, as a coating agent, it treats the fabric surface for printing. Coating the surface of a fabric with PU is one of the most common finishing techniques in textile manufacturing. In this case, the PU coating is critical for obtaining a successful print since the coating not only smooths the otherwise coarse surface of the fabric, which is essential for avoiding open-circuit connections, but also facilitates the bonding with the elastic MWCNT paste during the thermal curing.
The results suggest that the presented solution, combining screen printing, a PU–MWCNT blended conductive paste, and stretchable PU-coated commercially available fabrics, is feasible for prototyping textile–electronic interconnectors. Given the widespread and established use of the material and the processes in textile production, the presented approach has the required potential for enabling true series textile manufacturing.
The intended use of this combination of conductive and stretchable fabrics is to produce an elastic interconnection between a measuring device, typically wearable, with textile conductive pads in sensorized garments. Sensing or actuating electronic devices usually have a relatively small number of inputs/outputs (Bodykom 5, Hexoskin 9, Equivital EQ02 11), and an interconnector printed on fabric would suit these wearable devices nicely at Printed Circuit Board (PCB) -level interconnection. If attempting to interconnect other levels of integrations, like an Integrated Circuit (IC) pinout level, this type of technology will face size limitations that will prohibit its use. If the application requires a large number of interconnections, e.g., 16, given the current resolutions achievable with screen printing, it is reasonable to think that the proposed solution will suit such cases similarly to the ball grid array presented in [47
Other authors have presented alternatives for textile electronic interconnections, but while being successful in decreasing the size of the interconnections, such a feature is achieved at the expense of elasticity [48
The stretchability is provided by the conductive elastomer, PU, used in this formulation. Conductive elastomeric materials in the form of inks and pastes are the most common approaches as indicated in [49
4.4. Textile-Friendly Methods
The method of choice to transfer both conductive and elastic properties into the conformable planar surfaces of textile fabrics is printing technologies [50
], especially when the alternative is micro-fabrication processes.
Not only the production cost should be taken into account when manufacturing sensorized garments—we also need to consider the adoption of the applied manufacturing techniques. Methods like soldering which are industrially mature but foreign to the textile manufacturing world, like in [48
], will be more difficult to adopt in textile production lines than common methods already widely spread within textile manufacturing, like additive methods [50
Much effort has been put into obtaining good textile electrodes with good sensing capabilities but very little effort has been devoted to developing methods and materials to support textile-friendly and easy-to-adopt techniques for textile–electronic integration. The use of well-known textile materials together with broadly accepted methods for textile functionalization reduces the risk when proposing a novel application of them combined, such as in this electronic interconnector fully integrated on a textile substrate.
4.5. Impact in Sensing Performace
The fact of producing a better-expressed QRS is an encouraging result because it suggests that the connector does not negatively impact the recordings; however, the authors believe that the better-expressed QRS is a product of using textile electrodes and not related to the use of the printed connector. The recordings suggest that the use of textile electrodes, with bigger areas than normal Ag/AgCl electrodes, produces a skin–electrode interface with lower resistance, increasing the cut-off frequency of the low-pass filter at the skin–electrode interface. Therefore, higher-frequency components are preserved, which in some cases can be seen as advantage for QRS segmentation but in other scenarios might produce noisier recordings. In such a case, there are plenty of digital noise canceling methods to choose from.
5. Conclusions and Future Work
This study has presented positive results, especially from the functional perspective, since the sensing function was actually allowed. Now that that function has been shown to be possible, the next step is to investigate other material aspects further.
Future work will be focused on investigating the stability under certain degrees of deformation; assessing the need for encapsulation, if any; and evaluating the performance in other kinds of biosignal measurements or other applications such as digital interfacing, i.e., discrete voltage signals. In this study, the experimental protocol was selected to target the influence of the interconnection; therefore, sitting at rest was selected to avoid other sources of errors like motion artifacts. Now that interconnection shows promising results, assessment during exercise, as done in [51
], should be done.