Visual Evoked Potentials Used to Evaluate a Commercially Available Superabsorbent Polymer as a Cheap and Efficient Material for Preparation-Free Electrodes for Recording Electrical Potentials of the Human Visual Cortex

The aim of this study was to investigate the use of inexpensive and easy-to-use hydrogel “marble” electrodes for the recording of electrical potentials of the human visual cortex using visual evoked potentials (VEPs) as example. Top hat-shaped holders for the marble electrodes were developed with an electrode cap to acquire the signals. In 12 healthy volunteers, we compared the VEPs obtained with conventional gold-cup electrodes to those obtained with marble electrodes. Checkerboards of two check sizes—0.8° and 0.25°—were presented. Despite the higher impedance of the marble electrodes, the line noise could be completely removed by averaging 64 single traces, and VEPs could be recorded. Linear mixed-effect models using electrode type, stimulus, and recording duration revealed a statistically significant effect of the electrode type on only VEP N75 peak latency (mean ± SEM: 1.0 ± 1.2 ms) and amplitude (mean ± SEM: 0.8 ± 0.9 µV) The mean amplitudes of the delta, theta, alpha, beta, and gamma frequency bands of marble electrodes were statistically significantly different and, on average, 25% higher than those of gold-cup electrodes. However, the mean amplitudes showed a statistically significant strong correlation (Pearson’s r = 0.8). We therefore demonstrate the potential of the inexpensive and efficient hydrogel electrode to replace conventional gold-cup electrodes for the recording of VEPs and possibly other recordings from the human cortex.


Introduction
Visual evoked potentials (VEPs) are changes of the electrical potential elicited by visual stimuli and recorded using electrodes mounted on the forehead and the scalp above the inion. VEPs are part of electroencephalogram (EEG) and are extracted by stimulus correlation and averaging. VEPs are used to measure the functional integrity of the visual pathways from retina via the optic nerves to the visual cortex [1]. A typical VEP waveform using pattern-reversal stimulation consists of a negative

Electrodes
Water beads are made of an acrylic sodium salt of cross-linked polyacrylic acid (PAA) ([-CH 2 -CH(CO 2 Na)-] n ), a hydrogel, which is able to absorb water up to 500 times its weight. After swelling, the beads consist of up to 99.9% water (Figure 1a), rendering them electrically conductive. The conductance of a marble electrode with a diameter of 1 cm is about 1250 µS (0.8 kΩ, measured with the Diagnosys Espion e 2 , 500 nA at 50 Hz) and therefore similar to that of tap water. Water beads are commonly used for watering plants or for decoration purposes and are available in home improvement stores and garden centers as well as online for about €2 per 1000 pieces. The marble electrodes are always damp but, in contrast to a sponge, they do not loose water when squeezed. The stiffness of a marble electrode is about κ = 0.4 N/mm.
Top hat-shaped holders for the marble electrodes were manufactured from plastic (Perspex) in a workshop of the University Eye Hospital, Tuebingen. The inner diameter of the top hat corresponds to the diameter of the marble electrodes, and its height is about half the diameter of the marble electrode. A conventional gold-cup electrode clipped into the upper end of the hat (opposite to the brim) connects the leads with the marble electrode ( Figure 1b). These holders allow the marble electrodes to be positioned on the scalp using a commercially available electrode cap ( Figure 2). When mounted on the head, the marble electrode is pressed at the same time onto the skin and the gold-cup electrode, ensuring a tight contact.
The holder and the skin electrode can be reused, while the marble electrode can be disposed of after use. The marble electrodes were soaked in pure water for about six hours, until they were swollen to their maximal size of about 1 cm in diameter.   (a) Water beads before and after soaking for several hours in water. When fully swollen, the marble electrodes consist of up to 99.9% water and therefore become electrically conductive. (b) Top hat-shaped holder manufactured in a workshop of the University Eye Hospital, Tuebingen, and a marble electrode. The holders allow the marble electrode to be mounted at the scalp, while the connection between the marble and the amplifier is realized using a conventional gold-cup electrode.
Top hat-shaped holders and gold-cup skin electrodes can be reused; the marble electrode is disposed of after use. (a) (b) Figure 1. (a) Water beads before and after soaking for several hours in water. When fully swollen, the marble electrodes consist of up to 99.9% water and therefore become electrically conductive. (b) Top hat-shaped holder manufactured in a workshop of the University Eye Hospital, Tuebingen, and a marble electrode. The holders allow the marble electrode to be mounted at the scalp, while the connection between the marble and the amplifier is realized using a conventional gold-cup electrode.
Top hat-shaped holders and gold-cup skin electrodes can be reused; the marble electrode is disposed of after use.

Participants
Twelve healthy volunteers (nine female, three male; age 22-54 years, mean 36.6 years) with best-corrected visual acuity and no history of eye or neurological diseases were recruited from the staff of the Centre for Ophthalmology of the University of Tuebingen. All volunteers gave informed consent. The study followed the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of the Faculty of Medicine, University of Tuebingen.

Visual Stimulation
The VEP recordings were performed monocularly with one eye covered with an eye patch. The checkerboard stimulus was presented using a 21" CRT monitor (Model V999, Elonex, Birmingham, UK) [24] at a distance of 150 cm. Checkerboards of two check sizes-0.84 • and 0.25 • -were presented with a contrast of 80% and two reversals per second, according to the International Society for Clinical Electrophysiology of Vision (ISCEV) guidelines [2]. Each stimulus was presented three times, resulting in a total stimulation time of 2 × 61 s.

Data Acquisition
Electrodes were mounted according to the International 10-20 system [2,25]: active electrode above the inion at Oz, reference electrode at Fz, and ground electrode at Cz.
In the first session, marble electrodes were used as active, reference, and ground electrodes. No skin abrasion was performed. In the subsequent second session, the skin was cleaned using abrasive paste, and gold-cup electrodes were applied using conductive paste.
VEPs were recorded using an Espion e 2 (Diagnosys Ltd., Cambridge, UK) with a sampling frequency of 1000 Hz and digitally band-pass filtered (1.25-100 Hz). No notch filter was used. Post-trigger time was 300 ms. Automated baseline correction was applied by averaging and subtracting a 20 ms pretrigger period. Three averages, consisting of 64 single sweeps, were recorded for each check size [2].
Cursor positions for N75 and P100 [2] were determined automatically as maximum or minimum value, respectively, within the expected time frames using the built-in peak-finding algorithm of the Espion acquisition software and manually adjusted if necessary. Peak times and amplitudes of N75 and P100 were exported for further analysis using a custom-developed software [26,27].
The impedance between the electrodes mounted at Cz (ground electrode) and Oz (active electrode) and at Cz and Fz (reference electrode) was measured before the start and after the end of the recording for either the marble electrodes or the gold-cup electrodes using the Espion acquisition software. Because the top hat-shaped holders contain a gold-cup electrode for connecting the marble electrode with the amplifier, the following components contribute to the impedance measurement for the marble electrodes: gold-cup electrode-marble electrode-skin-marble electrode-gold-cup electrode, while the following components contribute to the impedance measurement for the gold-cup electrodes: gold-cup electrode-skin-gold-cup electrode.

Statistical Analysis
Linear mixed-effects models, fit by restricted maximum likelihood estimates (REML), were used to assess the significance of the electrode type in explaining variations in electrode impedance, mean amplitude of different frequency bands, and VEP N75 and P100 peak times and amplitudes. For all models, the variance inflation factors (VIF) of the predictors were calculated and assured to fall well below the common threshold value, indicating no collinearity between them. Prior to utilizing the results of the models, the normal distribution of the model residuals was confirmed visually, and the homoscedasticity of the variances of the residual was ensured using the Brown-Forsythe test and reported in case of violations.
To increase the statistical power of the analysis despite the small number of subjects, the alpha level was raised to 0.5 for all statistical tests, except otherwise stated.
All statistical analyses were carried out using JMP 14.2.0 (SAS Institute Inc., Cary, NC, USA).

Electrode Impedance
A linear mixed-effects model (Equation (1)) was used to assess the effect of the recording duration on the impedance of the marble electrode (Y), with the position (Fz/Oz) (α) and the time point (before/after) (β) as well as their interaction set as categorical effects and the subject set as random effect (ρ). (1)

Frequency Analysis
The correlation of the mean FFT amplitudes for delta, theta, alpha, beta 1, beta 2, beta 3, and gamma frequency bands were assessed by calculating the bivariate correlation coefficient between the conventional gold-cup electrode and marble electrode for each frequency band [28]. Additionally, a linear mixed-effects model (Equation (2)) was used to assess the effect of the electrode type on the log-transformed mean FFT amplitude value (Y) of the delta, theta, alpha, beta 1, beta 2, beta 3, and gamma frequency bands. The electrode type (α) and the frequency band (β) as well as their interaction were set as categorical effects, and the subject was set as random effect (ρ). (2)

VEP Analysis
Linear mixed-effects models were used to assess the significance of electrode type and recording duration in explaining variations in the amplitudes and peak times of N75 and P100 (Y), respectively,

Statistical Analysis
Linear mixed-effects models, fit by restricted maximum likelihood estimates (REML), were used to assess the significance of the electrode type in explaining variations in electrode impedance, mean amplitude of different frequency bands, and VEP N75 and P100 peak times and amplitudes. For all models, the variance inflation factors (VIF) of the predictors were calculated and assured to fall well below the common threshold value, indicating no collinearity between them. Prior to utilizing the results of the models, the normal distribution of the model residuals was confirmed visually, and the homoscedasticity of the variances of the residual was ensured using the Brown-Forsythe test and reported in case of violations.
To increase the statistical power of the analysis despite the small number of subjects, the alpha level was raised to 0.5 for all statistical tests, except otherwise stated.
All statistical analyses were carried out using JMP 14.2.0 (SAS Institute Inc., Cary, NC, USA).

Electrode Impedance
A linear mixed-effects model (Equation (1)) was used to assess the effect of the recording duration on the impedance of the marble electrode (Y), with the position (Fz/Oz) (α) and the time point (before/after) (β) as well as their interaction set as categorical effects and the subject set as random effect (ρ).

Frequency Analysis
The correlation of the mean FFT amplitudes for delta, theta, alpha, beta 1, beta 2, beta 3, and gamma frequency bands were assessed by calculating the bivariate correlation coefficient between the conventional gold-cup electrode and marble electrode for each frequency band [28]. Additionally, a linear mixed-effects model (Equation (2)) was used to assess the effect of the electrode type on the log-transformed mean FFT amplitude value (Y) of the delta, theta, alpha, beta 1, beta 2, beta 3, and gamma frequency bands. The electrode type (α) and the frequency band (β) as well as their interaction were set as categorical effects, and the subject was set as random effect (ρ).

VEP Analysis
Linear mixed-effects models were used to assess the significance of electrode type and recording duration in explaining variations in the amplitudes and peak times of N75 and P100 (Y), respectively, with electrode type (α) and check size (β) set as categorical effects, recording duration (γ) set as continuous factor nested in check size, and subject (ρ) set as random effect (Equation (3)).

Electrode Impedance
Compared to the conventional gold-cup electrodes, whose impedance was kept well below 5 k according to the ISCEV standard [29], the marble electrodes had far larger impedance, ranging from 20 to 80 kΩ.
The  with electrode type (α) and check size (β) set as categorical effects, recording duration (γ) set as continuous factor nested in check size, and subject (ρ) set as random effect (Equation (3)).

Electrode Impedance
Compared to the conventional gold-cup electrodes, whose impedance was kept well below 5 k according to the ISCEV standard [29], the marble electrodes had far larger impedance, ranging from 20 to 80 kΩ.
The   In the correlation analysis, statistically significant positive relationships (alpha level = 0.05) with large correlation coefficients between the conventional gold-cup and the marble electrode mean FFT amplitudes was observed for theta, alpha, beta 1, beta 2, beta 3, and gamma frequency bands as well as for the mains frequency (50 Hz). Table 1 provides a numeric summary of the Pearson's r, the 95% confidence intervals, and significance values. In the correlation analysis, statistically significant positive relationships (alpha level = 0.05) with large correlation coefficients between the conventional gold-cup and the marble electrode mean FFT amplitudes was observed for theta, alpha, beta 1, beta 2, beta 3, and gamma frequency bands as well as for the mains frequency (50 Hz). Table 1 provides a numeric summary of the Pearson's r, the 95% confidence intervals, and significance values. The linear mixed-effects model (n = 168, R 2 = 0.95) revealed statistically significant effects of the electrode type (F(1, 143) = 52.0405, p < 0.0001), the frequency band (F(6, 143) = 349.3580, p < 0.0001), and the interaction between electrode type and frequency band (F(6, 143) = 2.2029, p = 0.0460) on the log-transformed mean FFT amplitude. Figure 5 depicts the estimated least square means and the standard error of means (whiskers) of the FFT amplitudes of the different frequency bands recorded with conventional gold-cup electrodes (blue) and marble electrodes (red).  The linear mixed-effects model (n = 168, R² = 0.95) revealed statistically significant effects of the electrode type (F(1, 143) = 52.0405, p < 0.0001), the frequency band (F(6, 143) = 349.3580, p < 0.0001), and the interaction between electrode type and frequency band (F(6, 143) = 2.2029, p = 0.0460) on the log-transformed mean FFT amplitude. Figure 5 depicts the estimated least square means and the standard error of means (whiskers) of the FFT amplitudes of the different frequency bands recorded with conventional gold-cup electrodes (blue) and marble electrodes (red). Post hoc comparisons using contrasts revealed statistically significant differences between FFT amplitudes recorded with conventional gold-cup electrodes and marble electrodes for all frequency bands. Table 2 lists the corresponding test statistics. Table 2. Results of the post hoc contrast tests comparing the mean FFT amplitude of the different frequency bands recorded using conventional gold-cup electrodes and marble electrodes. The difference in the log-transformed LS mean amplitudes was converted to amplitude ratio.

VEP Results
VEPs could be recorded in all subjects using conventional gold-cup electrodes and marble electrodes. Figure 6 depicts the grand averages of 3 × 64 single traces for all subjects recorded using Post hoc comparisons using contrasts revealed statistically significant differences between FFT amplitudes recorded with conventional gold-cup electrodes and marble electrodes for all frequency bands. Table 2 lists the corresponding test statistics. Table 2. Results of the post hoc contrast tests comparing the mean FFT amplitude of the different frequency bands recorded using conventional gold-cup electrodes and marble electrodes. The difference in the log-transformed LS mean amplitudes was converted to amplitude ratio.

VEP Results
VEPs could be recorded in all subjects using conventional gold-cup electrodes and marble electrodes. Figure 6 depicts the grand averages of 3 × 64 single traces for all subjects recorded using conventional gold-cup electrodes (blue traces) and marble electrodes (red traces). Both electrode types resulted in comparable recordings, and the line noise was mostly eliminated through averaging. conventional gold-cup electrodes (blue traces) and marble electrodes (red traces). Both electrode types resulted in comparable recordings, and the line noise was mostly eliminated through averaging. Recordings were done using conventional gold-cup electrodes (blue traces) and marble electrodes (red traces). Shaded areas indicate ±1 standard deviation. No cleansing or abrasion was used for the marble electrodes. Both electrode types resulted in comparable recordings, and the line noise was mostly eliminated through averaging. Table 3 presents summary statistics for peak times and amplitudes of the N75 and P100 cursors recorded using gold-cup and marble electrodes. Note: M and SD represent mean and standard deviation, respectively. Recordings were done using conventional gold-cup electrodes (blue traces) and marble electrodes (red traces). Shaded areas indicate ±1 standard deviation. No cleansing or abrasion was used for the marble electrodes. Both electrode types resulted in comparable recordings, and the line noise was mostly eliminated through averaging. Table 3 presents summary statistics for peak times and amplitudes of the N75 and P100 cursors recorded using gold-cup and marble electrodes.

Effect of Electrode Type on N75 and P100 Peak Times and Amplitudes
Although the residuals of the linear mixed-effects models of N75 amplitude, N75 peak time, and P100 peak time were heteroscedastic, i.e., the variances were unequal, (F(1, 142) = 7.3829, p = 0.0074; F(1, 142) = 5.2284, p = 0.0237; F(1, 142) = 7.8444, p = 0.0058), the models were used for further analysis. Because the groups were balanced, the variance of the residuals did not depend on the electrode type [30], and the ratio of the maximum to the minimum variance between the groups was less than four for all models [31].
A statistically significant effect on the amplitudes or peak times of N75 and P100 was found for the check size used for stimulation in all models. The electrode type was found to have a statistically significant effect on the N75 amplitude and peak time as well as the interaction of electrode type and check size on the P100 amplitude and peak time. For the P100 amplitude, there was a statistically significant interaction between the electrode and check size. For the P100 peak time, there was a statistically significant interaction between the recording duration and check size (Table 4).   Figure 7 shows the least square means of the peak times and amplitudes of N75 and P100 acquired using gold-cup and marble electrodes using the two different stimulus check sizes-0.8 • and 0.25 • -over the recording duration, along with their standard error of means. Figure 7. LS mean plots depicting the peak times and amplitudes of N75 and P100 acquired using gold-cup and marble electrodes using two different stimulus check sizes-0.8° and 0.25°-over the recording duration. The whiskers indicate the standard error of means. The means showed a statistically significant difference for the check size. The effect of the electrode type was statistically significant only for the N75 amplitude and peak time but not for P100. Neither the recording duration nor the interactions between electrode type and recording duration or check size were statistically significant.

Discussion
Visual evoked potentials were successfully recorded in 12 volunteers for two check sizes using marble electrodes and conventional gold-cup electrodes. The marble electrodes had about 10 times higher impedance compared to the gold-cup electrodes.
To reduce the impedance, we tried using saline solution instead of water for soaking the hydrogel in order to increase its electrical conductivity by adding Na + and Cl − ions. However, in the saline solution, the swelling ability was drastically reduced. This effect was also shown by Horkay et al. [32], who investigated the swelling properties of hydrogels in various physiological salt solutions. Furthermore, using saline solution did not result in decreased impedance (data not shown).
Even though ISCEV recommends keeping the electrode impedance lower than 5 kΩ [2,33], this is nowadays less justified because modern amplifiers have very high input impedance, up to gigaohms. The Diagnosys Espion e² system used in this study has an input impedance of one gigaohm. A high impedance of the electrode may cause only problems when using old amplifiers with an input impedance of less than 100 megaohms. This is in line with the findings of Ferree et al. [6] and Kappenman and Luck [34], who found no significant difference between high impedance recordings and those with an impedance less than 5 kΩ. Furthermore, in the frequency range of interest, from 1 to 100 Hz, bioelectrical-generated noise dominates the recording and lowering the impedance by, e.g., skin abrasion, improves the signal-to-noise ratio only by a few percent [3].

Discussion
Visual evoked potentials were successfully recorded in 12 volunteers for two check sizes using marble electrodes and conventional gold-cup electrodes. The marble electrodes had about 10 times higher impedance compared to the gold-cup electrodes.
To reduce the impedance, we tried using saline solution instead of water for soaking the hydrogel in order to increase its electrical conductivity by adding Na + and Cl − ions. However, in the saline solution, the swelling ability was drastically reduced. This effect was also shown by Horkay et al. [32], who investigated the swelling properties of hydrogels in various physiological salt solutions. Furthermore, using saline solution did not result in decreased impedance (data not shown).
Even though ISCEV recommends keeping the electrode impedance lower than 5 kΩ [2,33], this is nowadays less justified because modern amplifiers have very high input impedance, up to gigaohms. The Diagnosys Espion e 2 system used in this study has an input impedance of one gigaohm. A high impedance of the electrode may cause only problems when using old amplifiers with an input impedance of less than 100 megaohms. This is in line with the findings of Ferree et al. [6] and Kappenman and Luck [34], who found no significant difference between high impedance recordings and those with an impedance less than 5 kΩ. Furthermore, in the frequency range of interest, from 1 to 100 Hz, bioelectrical-generated noise dominates the recording and lowering the impedance by, e.g., skin abrasion, improves the signal-to-noise ratio only by a few percent [3].
The statistically significant lower impedance of the marble electrodes mounted on Fz compared to Oz is probably the result of a worse contact at Oz, caused by dense hair [21]. Combing the hair before mounting the electrodes may reduce this problem [35]. Furthermore, it is possible that the contact pressure at Oz may have been lower than at Fz, resulting in a smaller contact area of the marble electrode, which is related to an increase in noise [4].
The difference in the impedance between Fz and Oz is likely the reason for picking up of line noise, which caused a contamination of the single sweeps with a 50 Hz signal. Line noise usually results from electrodes with high impedance, as is the case with marble electrodes, in combination with a differential amplifier [4].
The increased noise caused by the higher impedance of the marble electrodes may have also led to the statistically significant higher mean amplitudes in the Fourier analysis of the different frequency bands of the recordings compared to those of conventional gold-cup electrodes. On average, the mean amplitudes recorded using gold-cup electrodes were between 67% and 82% lower than for marble electrodes. However, the mean amplitudes of the different frequency amplitudes obtained with marble electrodes showed a statistically significant high correlation to those recorded with conventional gold-cup electrodes.
Even though the single traces were strongly contaminated with line noise, these artifacts could be removed almost completely using averaging and therefore only had a small effect on the measured amplitudes and peak times of the VEP. Linear mixed-effect models revealed a good agreement between the cursors obtained from recordings of the marble electrodes and the conventional gold-cup electrodes. A statistically significant difference was found for the N75 amplitude and the peak time between recordings using marble electrodes and conventional gold-cup electrodes. For the P100 peak time and amplitude, the interaction between check size and electrode type showed a statistically significant effect. A statistically significant effect of the recording duration was only found as an interaction with the check size and with the electrode type for the P100 peak time. All effects, except check size, reached statistical significance only because the alpha level was raised to 0.5. Furthermore, the mean differences in amplitudes and peak times between the conventional gold-cup electrode and the marble electrode were in the range of the intrasubject variability published by several groups [36][37][38][39] and those reported by Tello et al. for the repeatability of transient visual evoked potentials [40] and therefore of no clinical relevance.

Conclusions
This study demonstrated the potential of marble electrodes to replace conventional gold-cup electrodes for the recording of visual evoked potentials. Using modern differential amplifiers and averaging, high-quality VEP recordings were obtained without scalp abrasion. The differences in the amplitudes and the peak times between conventional gold-cup and marble electrodes were within the range of intrasubject variability and therefore of no clinical relevance. The ease of use and their low cost may render marble electrodes useful for other application domains, such as simultaneous recordings of the electroencephalogram during functional magnetic resonance imaging, brain-computer interfaces, or transcorneal electrical stimulation. However, further studies are needed to evaluate such applications, especially with regard to the differences in the amplitudes of the Fourier spectrum in the different frequency bands.
In a previous study, we demonstrated the application of marble electrodes for the recording of electroretinograms in small animals (Strasser et al., IOVS2012, Vol. 53, 2462. Additional uses for the application of marble electrodes may be brain-computer interfaces. This is because, in contrast to currently used electrodes, hydrogel-based electrodes, such as marble electrodes, provide a higher wearing comfort and are better tolerated, as Pinegger et al. investigated [5].
As marble electrodes consist of up to 99% water, they do not dry out, rendering them useful for long-time recordings, e.g., for brain-computer interfaces or during functional magnetic resonance imaging.
Avoiding the need for cleansing and abrasion of the skin increases patient comfort and significantly reduces the time for preparation. Additionally, it eliminates the risk of infection during abrasion of the skin [6]. Because the marble electrode is disposed of after the recording, time-consuming disinfection or sterilization of the electrodes can be omitted as well.
Marble electrodes may also be used for transcorneal electrical stimulation [41]. Several companies provide commercial devices (e.g., the OkuStim system, Okuvision GmbH, Reutlingen, Germany), which usually use modified versions of Dawson-Trick-Litzkow (DTL) electrodes [42]. These could be replaced by sterile marble electrodes and therefore ease the application for patients.