Monitoring the electrical activity of the human heart is one of the most widely used diagnostic medical procedures. Systems mediating the acquisition of cardiac signals use different types of electrodes, including Ag/AgCl, balloon or clip electrodes, which may be associated with various disadvantages. One of the negative effects may be the risk of an allergic reaction (Kim et al., 2022), which is quite common with Ag/AgCl adhesive electrodes due to the use of adhesive and conductive gel. Another negative phenomenon related mainly to long-term cardiac monitoring is patient discomfort caused, for example, by the adhesive, which can cause discomfort even when removing the electrodes (Sundström et al., 2023). Furthermore, in the case of long-term monitoring, hydrogel electrodes are limited by the tendency of the gel to dry out, or, conversely, during stress measurements, they may begin to peel off due to increased sweat production on the subject being measured Yu et al. (2024). Dry conventional electrodes made of metals offer the great advantage of reusability, but often suffer from increased motion artifacts due to insufficient adhesion to the skin Wan et al. (2024). Textile electrodes have the potential to reduce the negative effects of conventional electrodes mentioned above (Ferri et al., 2022; Ravichandran et al., 2023). In this regard, textile electrodes appear to be an irreplaceable element for advanced epidermal systems Jeon et al. (2025). While the risk of allergy can also occur with textiles (Sanchez Armengol et al., 2022), it tends to be much smaller, nevertheless caution must be exercised when choosing a suitable conductive material. Many types of conductive textiles are now available, including textiles containing metallic filaments (Euler et al., 2022), textiles with printed conductive structures (Tu et al., 2023), or textiles containing conductive polymers (Zhou et al., 2021). The use of textile electrodes has potential not only in electrocardiography (ECG), but also in other medical fields such as electroencephalography (Rahman et al., 2023), electromyography (Ozturk et al., 2021), electrotherapy (Skrzetuska et al., 2021) or thermotherapy (Liu et al., 2024).

The use of textile electrodes not only for ECG purposes is very popular nowadays and quite a lot of studies are dealing with this issue. The topic of the study by Fink et al. (2021) is the design and development of sensor garments for ECG monitoring. The authors fabricated textile electrodes from silver-embroidered fabric and woven fabric containing silver, which they implemented in the chest panel of a T-shirt. Both types of electrodes were tested on a human subject, with visual inspection of the resulting signals (before and after pre-processing). It was found that the embroidered electrodes performed better than the woven electrodes, which, for example, exhibited greater instability in the DC component of the signal (Fink et al., 2021).

A study by Nigusse et al. (2020) discusses the design of a textile electrode for continuous ECG measurements. The authors chose a system of depositing a metal layer (Ag) on cotton and polyester by screen printing, with electrical resistivity of 1.64 Ω 2 for cotton and 1.78 Ω 2 for polyester. According to the authors, the textile electrodes captured clear R-peaks comparable to conventional electrodes in terms of amplitude characteristics, but a higher susceptibility to motion artfacts must be taken into account (Nigusse et al., 2020).

A study by Li et al. (2024) investigates various factors that affect the impedance at the electrode-skin interface on textile electrodes containing Poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), during ECG scanning. In the experiment, 3 sizes (10, 20 and 30 mm circle plan), 3 shapes (triangle, square and circle) and the effect of contact area on impedance were compared and shown in graphs. In terms of size, the larger the electrode is, the lower the impedance it achieves; in the case of shapes, according to the graph, the impedance of the square electrode is higher than that of the other two shapes (the results change subtly with increasing frequency). In the case of partial contact, the impedance is significantly higher than that of the full contact electrode. The authors also compared the signal to noise ratio (SNR) of the different electrodes, and no differences were observed for the sizes and shapes. Significantly worse SNR results were observed in the case of partial electrode contact (Li et al., 2024).

In the case of textile electrodes, reusability is assumed, which raises the question of hygiene. A study by Doci et al. (2023) investigates the effect of washing and abrasion on textile electrodes. The authors used denim (cotton) as a base, followed by a PEDOT:PSS-based material and a conductive silicone-based material. In the experiment, electrodes with conductive embroidery, with PEDOT:PSS and with conductive silicone were tested. The results show that after 5 washing cycles, the specific resistance of the electrodes increased from 20 to 30 Ω 2 to 25–40 Ω 2 , in the case of abrasion testing, a value of 50 Ω 2 (Doci et al., 2023) was reached.

A study by Alizadeh-Meghrazi et al. (2021) attempts to develop a standardized procedure for measuring impedance at the textile electrode-skin interface (agar model). Silver, carbon fiber, Polydimethylsiloxane (PDMS) based materials were used for the experiment, Ionic Liquids (IL) and PEDOT:PSS. The effect of size, contact pressure, ratio of the mentioned substances and long-term use was tested. The authors observed significant differences in impedance, but according to them, these differences are not of major importance on the quality of the signals obtained. In addition to impedance testing, ECG signals were also analyzed using parameters such as SNR. An interesting result is that the tested electrode, which achieved a very low impedance (comparable to Ag/AgCl), however, achieved a significantly lower SNR (Alizadeh-Meghrazi et al., 2021).

The SNR parameter was also used in a study by Hossain et al. (2022) that tested electrodes in the form of wristbands made by weaving, knitting and crocheting with yarns containing carbon nanotubes (CNT), cotton and thermoplastic polyurethane (TPU) in the first case and CNT, spandex and TPU in the second case. The authors used Ag/AgCl electrodes as reference electrodes for evaluation. The authors mention that the SNR of the signals from the textile electrodes was quite high, but the signals were sufficiently telling. Another parameter is the impedance of the electrodes and testing their resistance. In the case of impedance, the authors found that knitted electrodes had a lower impedance value than crocheted electrodes. The authors also found that the impedance decreases with increasing electrode size (7.1 cm², 5.8 cm², 3.2 cm² used) and when compared to conventional electrodes, the impedance value is higher, but this reportedly does not directly affect their sensing performance. It has also been found that the impedance of textile electrodes decreases with increasing frequency. In terms of durability, the study found that after 10 wash cycles, the amplitude of the R and S waves decreased while the ECG features were still discernible (Hossain et al., 2022).

Studies focusing on the topics related to textile electrodes face several problems, including, for example, conducting the experiment on a limited number of subjects (often only the authors) or the lack of appropriate evaluation parameters (visual inspection, etc.). The goal of this experiment is to perform a comprehensive design, fabrication, and testing of textile electrodes on a larger sample of people using a commercial ECG monitor used in medical offices, followed by a statistical evaluation of the material that worked best. We had two main research questions. How does the type of conductive fabric affect the stability and quality of the measured ECG signal? Furthermore, are textile electrodes capable of preserving the morphology of the measured ECG signal (including testing individual parts of the cardiac cycle, i.e., P wave, QRS complex, and T wave) comparable to the ECG signal measured using conventional adhesive electrodes? This study is based on the hypothesis that the quality of ECG signals acquired using dry textile electrodes is primarily determined by the mechanical adaptability of the material, including properties such as thickness and elasticity.

The main contributions of the work are as follows:

Testing of different textile materials for ECG monitoring.

The rest of the paper is structured as follows: Section Materials and Methods describes the overall methodology of the experiment and section Results contains a description and explanation of the results. Discussion and conclusion are included in last two sections.

This section deals with a detailed description of the measurement system used in this experiment, covering the whole process from data collection to data processing and preparation for analysis. It first discusses the design and calibration of the sensors that were used to capture relevant information from the experimental environment, and then describes the transfer of the measured values to the data acquisition system where they were collected and stored. The section also explains how the data was subsequently processed, including removing noise, correcting for outliers and providing validation, while also focusing on the methods of filtering, analysis and possible transformation of the signals. The measurement system was designed to ensure maximum accuracy and reliability of the results, and this section provides a detailed overview of all the key steps that led to correct and consistent data processing.

This section deals with the interpretation of the results of this experiment, which was divided into two parts. In the first part, the textile electrodes were tested on synthetic data, and the results were used to assess whether their functionality is sufficient and whether it is appropriate to proceed with the experiment on real subjects. This process enables the comparison of textile electrode behavior on both real-world and synthetic data. The second part deals with testing electrodes on real data. For this purpose, median segments containing the P wave, QRS complex, and T wave were created in parallel with individual MCPs to reveal whether textile electrodes affect the shape and size of individual segments.

The main objective of this experiment was to comprehensively verify the suitability of five selected conductive textiles for ECG measurement. It also sought to determine whether Silitex, with its properties such as high elasticity, durability, and low skin irritation, would achieve high results in a series of tests that the materials underwent in this experiment and would be a suitable candidate for further testing. In previous section, the results for individual textile electrode positions were interpreted even if these positions were neglected, i.e., only the median values of individual materials were taken. Overall, Silitex performed best all evaluation parameters (R: 0.9919, SNR:17.2518, PRD: 13.7334). These results indicate high similarity between signals measured with the presented textile and conventional electrodes even though the V4 position was included in the evaluation, it achieved significantly lower results than the V5 and V6 positions. Looking at the Figure 7a, it can be seen that outliers occur only in the direction of deterioration of the given evaluation parameter. This phenomenon is caused by the high quality of the results. Thus, if, for example, the median correlation value for all materials is higher than 0.99, it is not possible to find an outlier approaching 1 with this data variability. For even better interpretation of the results, it would be appropriate to use statistical tests. During the experiment, the non-parametric Mann-Whitney test was applied due to the failure to achieve normal distribution of the data by Shapiro–Wilk test. However, the test results did not confirm statistically significant differences between the individual groups at the significance level of α = 0.05 . Additionally, a comparison and definition of this experiment in relation to other studies is provided in Table 8. The studies compared are listed and described in the introductory section.

For the purpose of this experiment, the positions of the standardized leads V4, V5 and V6 were chosen in order to test the electrodes as close as possible to the heart of the subject. By looking at the Tables 4–6, it can be seen that the median values show clear differences between the locations, especially for the V4 location, which differs significantly from the other two locations. Visual inspection of the signals revealed that several subjects showed noticeable differences between the textile and reference signals, such as a decrease in the amplitude of the R-wave, or conversely, an increase in the amplitude of the S-wave in the case of the textile electrodes. This phenomenon may be caused by several reasons. The first factor is the anatomy of the human’s body itself. There is much a larger layer of fat or muscle around V4 position, in comparison to position V6, which may affect the presence or absence of various interferences of the ECG signal. Another important factor is the size of the electrodes themselves. Theoretically, the electrodes sense biological signals as a matrix system. By this we mean that an electrode can be interpreted as a set of quantum “microelectrodes” that sense signals from a quantum of points on the human skin. The resulting signal is therefore the sum of the signals of the individual points of electrodes. This theory raises the possibility of incorrect textile electrode sizing, particularly at V4 position. While the electrode may have been positioned correctly, it might also have captured signals from other locations, masking or drowning the intended V4 lead signal. As already mentioned, a visual inspection of each signal was performed and the mentioned phenomenon of changes in amplitude magnitude occurred mainly in subjects on the borderline of malnutrition, suggesting that the size of the textile electrode may have had a negative effect. In this regard, the next phase of the experiment, building on the results of this pilot study, will focus on evaluating the size of the textile electrodes, their placement on the body, the method of connection to the measurement system, and their robustness against motion and other artifacts. With reference to Figure 3, a significantly greater T-wave amplitude and a marginally increased S-wave amplitude are observed for the textile electrode at position V4 when compared to the signal obtained using the reference electrode. This phenomenon was observed across all tested materials in subject 6 and was also noted in several other subjects whose somatotype can be classified as borderline malnutrition. This fact, together with the previously mentioned factors, may play a key role in the production of this phenomenon. It is also necessary to consider the possibility of measurement error, which may have been caused by a slight deviation from the intended axis of measurement. This phenomenon can theoretically be eliminated by measuring the reference and textile electrodes at the same location, which is, however, impossible when measuring with both electrodes at the same time. It would certainly be worthwhile to test this phenomenon in a further experiments to evaluate the effect of the size and placement of the textile electrodes for ECG sensing.

The presented combination of algorithms and their settings is based on our earlier research, which can be traced back to articles and studies from reputable research teams. Additional information regarding the evaluation and signal processing is in our earlier article focusing on vectorcardiography transformations (Jaros et al., 2019).

Part of these studies (Introduction section) mentioned that textile electrodes were more susceptible to noise and motion artifacts. During the presented experiments there was no significant increase in noise for textile electrode signals compared to Ag/AgCl signals. Figure 2 shows the real signals for position V5. It is apparent that the RAW signals from both types of electrodes are noisy at similar intensities, noting the presence of a more noticeable zero-line drift in the signal from the textile electrode. After preprocessing, the signals are virtually identical (see Figure 3 for preprocessed textile and reference signals at all positions). During data acquisition, it was discovered that the tested electrodes are noticeably more susceptible to the formation of motion artifacts in comparison to Ag/AgCl electrodes. Speech and deep breathing caused artifacts across all electrodes, but with greater intensity in the textile signals. While tightening the chest band effectively reduced these artifacts, it negatively impacted comfort, which is the main advantage of textile electrodes. Future studies should evaluate the electrodes during activities such as walking or weight-bearing exercise, and explore optimal elastic band pressure to balance both artifact reduction and patient comfort.

When testing the impedance of conductive textiles at the electrode-skin interface in a dry environment, it is immediately apparent that Balingen is the only conductive textile to achieve the lowest impedance values, standing out from the other materials in the graph. In Table 1, which interprets the data sheets of the textiles used, Balingen material is defined as a charmeuse fabric that exhibits extreme softness. This softness allows the samples to achieve high mechanical conformity, creating a number of contact points with high pressure on the skin, which increases adaptability to skin irregularities. Looking at the graph showing the relationship between frequency and phase shift, it is also apparent that the same fabric achieves the highest phase shift of all fabrics. In wet testing, the impedance is significantly lower, which may be due to the filling of air cavities between individual fabric fibers with a conductive gel, which also helps the fabric to compensate for skin irregularities. Figure 6c shows a very interesting result: the P180B material achieves significantly lower results than the reference wet electrode. The result can be attributed to the fact that, thanks to its properties, the material achieved a larger effective contact area than the reference. When comparing the figures showing the relationship between frequency and phase shift, the difference between the dry and wet scenarios is again visible, with the maximum phase shift for most electrodes being achieved at 4–5 kHz in wet testing (compared to around 9–10 kHz in dry testing). The results of impedance and phase shifts indicate that in the case of conductive textile materials tested in this experiment, the presence of sweat does not necessarily pose a problem causing errors in conductivity and artifacts. On the contrary, human sweat (saline solution) can help increase the sensory properties of textile electrodes, which can assist in the selection of suitable material, for example, for a sports vital signs monitor, which will certainly be exposed to increased perspiration over a long period of time.

Analysis of Tables 4–6 reveals that all materials performed very well at the V6 lead position, achieving high similarity to the reference signals. When choosing the right material, the sensing capability is of course the most important parameter that should guide the system designer. There are also other very important aspects on the basis of which such a designer might decide to choose this particular material. How is the material handled? What is the durability? Is it comfortable to wear? Can the material be sanitized or washed in any way? Working with the Balingen material was quite challenging. It is a charmeause fabric (Shieldex, 2024) that has a very soft and flexible texture, which made it difficult to sew into electrodes. After testing the electrodes on 20 subjects, it is also possible to notice some wear in the seams of the electrodes, which is due to the flexibility mentioned above. In the case of this material, the use of a non-conductive fabric as a lining could be considered to increase the overall strength. On the other hand, it should be mentioned that the material is extremely comfortable and the manufacturer also declares washability. Another material is Bonn, which is classified as a non-woven fabric (Shieldex, 2024a). The fabric is not flexible and there were no problems in the manufacture of the electrodes. This material did not experience the problems encountered on Balingen, but degradation of the electrode surface was nevertheless observed. The material also did not produce discomfort for the subjects, but it is certainly worth noting that the manufacturer does not recommend washing it at all. According to the manufacturer, P130 + B is a stretch-tricot material (Shieldex, 2021), the texture of which is somewhat similar to that of Balingen, so that the manufacturing of the electrodes was subtly more complex. Lining could be also used to improve the overall strength of the electrodes. The measured subjects reported that it was not an extremely comfortable material, but at the same time they were not uncomfortable. Interestingly, there were virtually no signs of wear and tear after the experiment and the washability is also a significant advantage. According to the manufacturer, the P180 + B material is classified as a mirror-satin fabric (Shieldex, 2024b). Given the minor manufacturing issues with this fine material, a lining could offer a possible solution. P180 + B is probably the most comfortable material used in this experiment. Some of the subjects on whom measurements were taken even reported that they would not mind wearing a shirt made entirely of this material. A slight disadvantage is the slight wear (puckering) of the material in the seam areas, which is not as noticeable as in the Balingen material. The manufacturer also declares that this material can be washed. The last material tested is Silitex, classified as a stretch tricot knit fabric (Shieldex, 2022). This type of material is characterized by high elasticity, a fine structure that is relatively comfortable on human skin, and good mechanical resistance. What makes Silitex unique is that one side of the knit fabric is coated with a layer of conductive silicone, which significantly enhances the overall strength of the electrode and improves the stability of the molded contact used to connect it to the measuring chain during design and production. In the case of this experiment, it is undoubtedly the fabric that was the best to work with. There were no problems around the seams when sewing the electrodes. There was also virtually no wear and tear on the electrodes after the experiment. The only slight disadvantage may be that, for example, compared to P180 + B material, it is not as comfortable a textile, but on the other hand, none of the subjects complained of discomfort. Another advantage of Silitex is its washability. Wear and tear and the possibility of sanitation or any kind of hygienic operation with these materials were also topics of some of the studies mentioned. An important observation of this study is the fact that no adverse skin reactions were observed with any of the tested textile electrodes, while mild skin irritation was reported in some subjects at the sites where the reference electrodes were attached. In the future, it would also be useful to test these particular materials, particularly in the context of sanitation and their wear and tear or resistance to human sweat.

Although advances in materials engineering have introduced very promising alternatives to conductive textiles, such as carbon-based materials, PEDOT:PSS coatings, or other modern systems Lee et al. (2025), our study focused exclusively on silver-plated textiles. While new materials such as liquid metals offer excellent conformability and lower skin-electrode interface impedance due to their softness Ma et al. (2024), silver-containing textiles currently represent a robust commercially available solution for mass production of smart clothing due to established manufacturing processes, cost-effectiveness, and washability. The main objective of this study was to reveal whether these commercially produced fabrics are suitable candidates for smart clothing systems and whether Silitex, with its properties, will dominate in selected samples. The aim was not to evaluate experimental prototypes. However, future studies should certainly take these materials into account and perform in-depth comparative tests.

A very important part of this experiment is the actual creation of representative cardiac periods, which requires accurate definition of the R-peak positions. As mentioned in section Materials and Methods, a custom detector was used to detect R-peaks in the signals. The proposed detector is based on a DNN processing 2D CWT scalograms, which shows significantly higher success rates compared to, for example, the Pan-Tompkins algorithm. The main drawback of the Pan-Tompkins algorithm was its susceptibility to errors in cases where the amplitude of the S wave exceeded the amplitude of the R wave (low R wave and deep S wave), or where T or P waves were misclassified as R peaks due to amplitude thresholding. In contrast, the proposed detector radically reduced this negative phenomenon by using signal segmentation and subsequent logical evaluation and validation. Thanks to the strict requirement that the R-peak be located between the predicted Q and S waves, the system uses the morphological context of the entire QRS complex. While Pan-Tompkins detectors rely on fixed mathematical heuristics that can fail with highly variable signals, neural networks offer a solution capable of adapting to complex anomalies, provided they are trained on a sufficiently diverse and comprehensive dataset. It is also important to note that although the detector constructed for use in this experiment is significantly more robust than the conventional algorithm mentioned above, it is far from flawless. The trained model is still in the early stages of development, which will continue for a long time. The model was trained on approximately 100 h of diverse ECG signals, including pathological phenomena. In the future, we plan to significantly expand the training dataset, primarily with signals contaminated by various types of noise or other artifacts. This procedure will make it possible to further improve the overall detection of R-peaks and prevent false detections in problematic signals. Another future goal is to adapt the model to abdominal ECG signals, which contain both the mother’s ECG and the fetal signal. This topic involves very complex processing of abdominal signals from which fetal signals are extracted. With a suitably trained model, it is possible to skip the complex extraction of fetal signals and detect both maternal and fetal R-peaks at the same time. Figure 4 shows the training process of the DNN model. The curves show that the validation accuracy (val. ACC) is higher than the training accuracy (train ACC). At first glance, this phenomenon may look like an error, but it is a logical phenomenon. Dropout layers were applied during the model design, which essentially randomly turn off some neurons, increasing the model’s generalization ability by preventing unwanted co-adaptation of neurons. During the validation process, these dropout layers are turned off, often resulting in a better ACC result than during training. In the case of the loss function, the opposite effect can be seen: the validation loss function (val. loss) is higher than the training loss function. This is caused by the strong weighting of minority classes (Q-wave, R-peak, S-wave), which penalizes the loss function even with very low uncertainty in the model prediction.

The potential of textile electrodes lies in several fields of electrocardiography (and not only ECG). This experiment focuses on testing electrodes contained in an elastic chest belt, and although only positions from 3 standardized leads were tested here, the idea of the experiment is a system containing textile electrodes (ideally a chest belt) that can be used for long-term clinical ECG imaging. Such a system should provide several advantages, including ease of installation on the patient, comfort of wear and repeatability of use. A similar system is addressed, for example, in the study by Teferra et al. (2022). Another attractive application is the implementation of textile electrodes in clothing, which is addressed, for example, in the study by Fink et al. (2021). Conversely, these systems may have benefits in non-ambulatory applications, which may include commercial tools for biosignal measurement. The concept of an autonomous sports T-shirt capable of capturing ECG, respiratory activity, and other biosignals during various physical activities and transmitting them wirelessly to a smartphone is an attractive approach to this problem.

This study focused on the sensory evaluation of five materials from the manufacturer Shieldex, which were used to make textile electrodes for ECG measurements. To obtain the dataset, measurements were performed on twenty subjects of variable gender, age and somatotype. The electrode placement was fixed to the positions of the standardized ECG leads (V4, V5, V6), and the textile electrodes were attached using an elastic band, while reference electrodes (Ag/AgCl) were glued directly under the textile electrodes. Evaluation of signal quality was performed based on correlation coefficient, signal-to-noise ratio and percentage root-mean-square difference. Based on the obtained results, it was found that all materials showed very high similarity to the reference data. It was also found that the quality of the signals increases significantly when approaching the V6 position. When the positions of electrodes were neglected, it was found that the P130 + B material achieved the worst results while the Silitex material achieved the best results. Based on these preliminary findings, it can be concluded that materials tested in this study can be used for the fabrication of textile electrodes, but other aspects affecting data acquisition, such as electrode size and placement, must also be considered an explored.