Electric-acoustic interaction measurements in cochlear-implant users with ipsilateral residual hearing using electrocochleography

Expanded implantation criteria in cochlear-implant (CI) treatment has increased the relevance of combined electric and acoustic stimulation (EAS; Kiefer et al., 2005; Skarzynski and Lorens, 2010; von Ilberg et al., 2011). Less traumatic surgical techniques (von Ilberg et al., 1999; Lenarz et al., 2009) and softer electrode designs (Hochmair et al., 2015) allow the preservation of substantial residual hearing during and after implantation (Fraysse et al., 2006; von Ilberg et al., 2011; James et al., 2005; Kiefer et al., 2005). While several studies have shown that the combined EAS can improve speech intelligibility, especially in noisy environments (Büchner et al., 2009; Gifford et al., 2013), only a few studies have investigated the underlying interaction mechanisms between electric and acoustic stimulation in humans (Imsiecke et al., 2018; Krüger et al., 2017; Lin et al., 2011). However, this topic is increasingly gaining importance since every day more patients with substantial residual hearing receive a CI. From a clinical perspective, the fundamental understanding of EAS interaction could be important to optimize EAS fitting and consequently improve speech understanding with these devices. EAS fitting includes the setting of the crossover frequency between electric and acoustic stimulation as well as the selection of the gains to be applied to the acoustic component (hearing aid) and the determination of the comfort and threshold levels of electric stimulation (e.g., Incerti et al., 2013).

Lin et al. (2011) investigated electric-on-acoustic and acoustic-on-electric ipsilateral masking using a psychoacoustic experiment. Masking of acoustic responses by electrical stimulation was investigated in a group of six EAS subjects. For five of these subjects implanted with a relatively short electrode array (10 mm), no masking due to electrical stimulation was observed. However, in a single subject implanted with a relatively long electrode array (24 mm), masking of acoustic tones at relatively high acoustic frequencies (500 Hz, 625 Hz, and 750 Hz) was observed when the apical electrodes were simultaneously stimulated. Additionally, masking of electrical responses by acoustic stimulation was investigated for two subjects implanted with a 24-mm and with a 10-mm electrode array. The elevation of electrical detection thresholds caused by simultaneously presented acoustic stimulation was observed for apical and middle electrodes independent of the acoustic masker frequency. Krüger et al. (2017) demonstrated an asymmetry between the effect caused by an electric masker on the detection threshold for acoustic probes and the effect of an acoustic masker on the detection threshold for electric probes in a population of five EAS subjects. Furthermore, the masking was characterized as a function of the relative location between the electric and the acoustic stimulus (electric-acoustic frequency difference; EAFD) in the cochlea. Here, the insertion angles of the CI electrodes were estimated from cone-beam computed-tomography (CBCT) data and were converted into corresponding tonotopic frequencies. The relative difference between these tonotopic frequencies and the frequencies of the acoustic stimuli was defined as the EAFD. For the electric masker, this analysis showed that electric-on-acoustic EAS masking decayed exponentially with increased EAFD and virtually disappeared for EAFDs larger than two octaves. In contrast, acoustic-on-electric masking remained almost constant across the tested EAFD range of six octaves. Imsiecke et al. (2018) extended these masking experiments using non-simultaneous presentation of the electric and acoustic stimuli in a psychoacoustic forward-masking experiment and confirmed the asymmetry observed in the simultaneous presentation experiments. A possible explanation for the asymmetry between masking modalities is the fact that each experiment measured interaction effects toward different directions in the cochlea in relation to the masker stimulus. In the electric-on-acoustic masking experiment, the acoustic probes were presented to more apical cochlear locations than the electric masker stimulus. In contrast, in the acoustic-on-electric masking experiment, the electric probe was presented at cochlear locations more basal than the acoustic masker stimulus. Typically in EAS subjects, fewer functional hair cells are available in locations more basal than the deepest inserted electrode, therefore the mechanisms involved in electric-on-acoustic and acoustic-on-electric masking are probably different. More details about potential explanations for this asymmetry are discussed in Krüger et al. (2017).

Even if EAS masking has been shown before (Imsiecke et al., 2018; Krüger et al., 2017; Lin et al., 2011), its basic mechanisms cannot be determined from psychophysical experiments alone. Electrophysiological experiments may help to determine the fundamental mechanisms of EAS interaction and may be able to explain part of the observed psychoacoustic EAS masking. While EAS masking generally refers to changes in hearing detection-threshold levels for an acoustic stimulus due to the presence of electrical stimulation or vice versa, EAS interaction is not limited to masking only. EAS interaction describes the dependency of electric and acoustic stimulation in a more general context, including inhibition or facilitation for various probe levels.

Previous animal studies investigated EAS interaction in normal-hearing guinea pigs (Stronks et al., 2010, 2012) and normal-hearing cats (Vollmer et al., 2010) using electrophysiological measurements. Vollmer et al. (2010) showed EAS interaction as reduced spike activity at the level of the inferior colliculus (IC) when an electric stimulus was simultaneously presented along with an acoustic stimulus. In a forward-masking experiment, they showed a suppressive effect of electrical maskers on acoustic probes. Moreover, in a simultaneous masking experiment they observed a suppressive effect of acoustic maskers on electric probes. In both experiments, the spike activity caused by the probe was decreased with increased masker levels. Stronks et al. (2010) showed EAS interaction with acoustically evoked compound action potentials (CAP) generated by synchronized responses to the onset of acoustic stimuli with fast rising amplitudes. Stronks et al. (2010) showed that for high frequencies, the acoustically evoked CAP can be suppressed by extracochlear basal electrical stimulation. On the other hand, low-frequency evoked CAPs were less suppressed, probably because apical regions were not stimulated electrically. The lack of low-frequency evoked CAP suppression was supported by Stronks et al. (2011, 2012) who showed that acoustically evoked CAPs were more suppressed if there was a spatial overlap between electric and acoustic stimuli. Nourski et al. (2007) investigated the effect of an acoustic masker on electric probes at the level of the auditory nerve by analyzing electrically evoked CAPs in guinea pigs. They showed that for a fixed masker the amount of masking, as measured with CAPs, decreased with increasing probe levels. Furthermore, masking increased with increasing masker levels. Although animal studies can demonstrate peripheral interaction, it remains a question how electrophysiological interaction relates to psychophysical masking.

Current CIs are provided with a telemetry system, originally designed to record electrically evoked responses that can be adapted to measure intracochlear electrophysiological responses to acoustic stimulation. This method holds the potential to measure EAS interaction electrophysiologically at the level of the hair cells and the auditory nerve in humans, and may give insights about the mechanisms involved in psychophysical EAS masking.

Electrocochleography (ECochG) is a technique to record cochlear potentials evoked by acoustic stimulation, such as tone bursts or clicks. It includes cochlear-microphonic potentials (CM), auditory nerve neurophonic potentials (ANN), action potentials (AP), and the summating potential (SP). However, the focus of the current study was on the CM and the ANN. While the CMs are understood to represent the transducer current generated by the stereocilia of primarily outer hair cells in response to the movement of the basilar membrane, the ANNs are thought to represent evoked potentials of neural fibers caused by inner hair cells (Campbell et al., 2015; Fontenot et al., 2017).

The difference response (DIF) of two ECochG recordings to acoustic stimuli with alternating polarities emphasizes odd harmonics of the stimulus frequency, which results in a spectral peak at the frequency of the acoustic stimulus. This peak is dominated by the CM but can also include the biggest part of the ANN. The summation response (SUM) of two ECochG recordings to acoustic stimuli with alternating polarities emphasizes even harmonics of the stimulus frequency. The rectification and the asymmetric characteristic of the ANN, caused by the shape of the action potentials, generates a spectral peak at the second harmonic of the acoustic stimulation frequency, which can be observed in the SUM response (Fontenot et al., 2017; Forgues et al., 2014). The second harmonic of the ANN is usually smaller than the first harmonic of the ANN or the first harmonic of the CM. The CM and the ANN cannot be isolated completely by the DIF and the SUM, although the DIF is dominated by the CM, and the SUM is dominated by the ANN.

Extracochlear ECochG recordings during surgery (Dalbert et al., 2015b) and post-operative intracochlear ECochG recordings (Dalbert et al., 2015a) have been proposed to monitor cochlear functionality. Abbas et al. (2017), Kim et al. (2018), and Koka et al. (2017) showed a correlation between psychoacoustic and ECochG thresholds measured through intracochlear CI electrodes for EAS users who have residual hearing in the low acoustic frequencies. Koka et al. (2017) found a higher correlation between DIF thresholds and psychoacoustic thresholds than between SUM thresholds and psychoacoustic thresholds. Furthermore, SUM responses were not present in all subjects. These results indicated that, using the telemetry capabilities of the CI, DIF responses were a better candidate to compare electrophysiological EAS interaction with psychoacoustic EAS masking. Abbas et al. (2017) and Kim et al. (2018) showed that changes in hearing thresholds due to the loss of some residual hearing over time were also seen in DIF and SUM responses. Additionally, Abbas et al. (2017) found that for a subset of subjects, the DIF amplitudes grew with increased stimulation level, but little or no SUM component was identified. Koka and Litvak (2017) extended these works and demonstrated the feasibility of intracochlear ECochG to estimate EAS interaction. They recorded ECochG responses to acoustic stimuli in the presence and in the absence of simultaneous electric stimulation and observed EAS interaction as reduced ECochG responses in the presence of electric stimulation in a group of 12 subjects. They measured psychoacoustic EAS interaction in a subset of six subjects but did not observe a correlation with electrophysiological EAS interaction. Still, the use of ECochG to characterize EAS interaction as a supplement to psychoacoustic experiments holds the potential to distinguish between central and peripheral EAS interaction mechanisms.

The current study extended the previous work of Koka and Litvak (2017) by including a CBCT scan analysis to determine the individual electrode positions for each subject, thereby allowing an assessment of the interaction between electric and acoustic stimulation relative to the respective places of cochlear stimulation. Two experiments were designed to investigate electric-acoustic interaction for EAS users with residual acoustic hearing in the low acoustic frequencies. The first experiment was based on a psychoacoustic procedure designed to investigate the potential influence of an electric masker on the detection of acoustic tones. The second experiment used ECochG recordings as an objective measure of EAS interaction.

Seven CI users with ipsilateral residual hearing in the low acoustic frequencies participated in this study. Before starting the testing, post-operative audiograms were measured for each participant, as shown in Fig. 1. The audiograms present the unaided air-conduction pure-tone thresholds measured via headphones (HDA 200, Sennheiser electronic GmbH & Co. KG, Wedemark) and a clinical audiometer (Audio 4000, Homoth Medizinelektronik GmbH & Co. KG, Kaltenkirchen, Germany) at the test appointment. All subjects were implanted with an Advanced Bionics Ultra HiFocus implant combined with a fully inserted SlimJ electrode array. Table I contains a detailed description of each subject's demographic data. The stated CI experience is relative to the activation of the implant (initial fitting of the speech processor), which was done 4–6 weeks after implantation. At the time of the test appointment, there was no indication of residual fluid post-surgery, measured as the difference between air and bone conduction thresholds, and thus losses were sensorineural in nature.

The hardware instrumentation setup for both the psychoacoustic and electrophysiological experiments, adapted from Koka et al. (2017), is shown in Fig. 2. The clinical programming interface (CPI-2, Advanced Bionics, Valencia, CA) in combination with the portable speech processor (PSP, Advanced Bionics, Valencia, CA) was used to connect the subject's implant with a computer to provide electrical stimulation and get access to the telemetry of the implant. The CPI-2 delivered a trigger signal to a digital-to-analog converter (NI DAQ 6216 BNC, National Instruments Corporation, 11 500 Mopac Expwy, Austin, TX), which was connected via USB to the same computer and to an audio amplifier (Sony PHA-2, Sony Corporation, New York, NY) to obtain synchronized acoustic stimulation. An insert earphone (ER2, Etymotic Research, Elk Grove Village, IL) connected to the audio amplifier was used for stimulus presentation. The transmission of the sound from the insert earphone to the basilar membrane introduces a negligible delay with respect to the used stimuli durations. The Bionic Ear Data Collection System (BEDCS, Advanced Bionics, Valencia, CA) was used to generate the electric and acoustic stimuli as well as to collect the telemetry data from the CI. The levels of the acoustic stimuli were calibrated in the subject's ear canal using a microphone amplifier (ER-7C Series B Clinical Probe Microphone System, Etymotic Research, Inc., Elk Grove Village, IL) with a probe tube (ER3–14C, Etymotic Research, Inc, Elk Grove Village, IL) connected to an analog-to-digital converter of the NI DAQ 6216.

Psychophysical and electrophysiological measurements were performed in a sound booth using the same insert earphones for acoustic stimulation. Electrical stimuli were directly presented at single electrodes. The acoustic stimuli were calibrated in the subject's ear canal. Psychophysical and electrophysiological measurements were successively performed during the same session without replacing the probe tube.

The electric stimuli for the psychophysical-masking and the electrophysiological-interaction measures consisted of unmodulated biphasic pulse trains delivered through the most apical intracochlear electrode of the CI. The pulse duration for each pulse phase was set to 25 μs. The pulse trains were delivered at 1019 pulses per second (pps). These electric stimuli were presented at the most comfortable level (MCL) for each subject.

Sinusoidal tones with a sampling rate of 20 kHz were used as acoustic probe stimuli. Depending on the residual hearing for each subject, a subset of audiometric frequencies (125 Hz, 250 Hz, 500 Hz, 750 Hz, 1 kHz, 1.5 kHz, 2 kHz, 3 kHz, 4 kHz) was selected for testing.

In the psychophysical experiment, the acoustic probe level was adapted to determine the detection threshold. The maximum allowed stimulation level was limited to the individual MCL to prevent overstimulation. In the electrophysiological experiment, the acoustic stimuli were presented at MCL to maximize the response amplitude for the ECochG recording. To determine the MCLs, a loudness scaling was performed for each acoustic and electric stimulus by successively increasing the probe level, starting from an inaudible level until MCL was reached. MCL was defined as number six of a ten-scale loudness table ranging from “extremely soft” to “extremely loud.” The mean of two repetitions was used as the MCL for the experiments.

CBCT was used to measure the insertion angle (α) for each electrode for each subject and to determine the EAFD according to Krüger et al. (2017). The insertion angles were determined from post-operative CBCT scans using the posterior margin of the round window, the modiolus, and the target electrode contact (most apical electrode, electrode number 1 of the SlimJ electrode array), as shown in Fig. 3.

The insertion angle was converted into a frequency $fe$ corresponding to the tonotopic location of the cochlea using the Greenwood equation (Greenwood, 1990) and the correction maps for electric stimulation by Stakhovskaya et al. (2007). The difference between $fe$ and an acoustic pure tone stimulus $fa$ was expressed as EAFD in octaves according to Eq. (1),

Positive EAFDs indicated that the corresponding cochlear location of $fa$ was more apical than the corresponding location of $fe$⁠. Negative EAFDs indicated that the corresponding cochlear location of $fa$ was more basal than the corresponding location of $fe$⁠.

Figure 4 shows the stimulation configuration used for psychoacoustic EAS masking and electrophysiological EAS interaction measurements. Detection thresholds for acoustic stimuli ( $fstim$⁠) were measured for audiometric frequencies (125 Hz, 250 Hz, 500 Hz, 750 Hz, 1 kHz, 1.5 kHz, 2 kHz, 3 kHz, 4 kHz) in the absence (unmasked condition) and in the presence (masked condition) of a simultaneously presented electric pulse train delivered through the CI's most apical electrode (⁠$elstim$⁠) using the case ground (⁠$casegnd$⁠) as a reference. Psychoacoustic masking was characterized in terms of threshold elevation, defined as the difference between thresholds in the masked and unmasked conditions. A three-interval forced-choice (3-IFC) procedure with a two-down, one-up rule was used to determine the detection threshold, defined as 71% correct on the psychometric function (Levitt, 1971). A graphical user interface displayed on a computer screen showed three buttons with numbering 1, 2, and 3. Synchronized with the acoustic and the electric stimuli, the buttons lit up sequentially for 500 ms, separated by a pause of 500 ms in between to indicate the presentation intervals. Randomly centered in one of these intervals, the probe stimulus was presented for 200 ms, including 5 ms Hanning window ramps for onset and offset. If the masker was present, it was delivered at MCL for 500 ms in all three intervals. Unmasked thresholds were estimated by presenting no electrical stimulation in the three intervals. The subject was instructed to indicate the interval in which the probe stimulus was presented. Initially, the adaptive acoustic probe stimulus was set to the MCL and the step-size was set to 8 dB. After each reversal, the step-size was halved until the minimum step-size of 2 dB was reached. Eight further reversals were performed with the minimum step size. The detection threshold was estimated as the mean of the last four reversals.

Each subject was instructed to indicate the interval where the probe stimulus was presented. For training, at least one run was conducted making sure that the subject could reliably perform the task. Finally, one run each for the unmasked and masked conditions was used to estimate the psychoacoustic threshold elevation for each acoustic frequency. During the training and measurement, it was visually checked that the probe level converged from MCL to threshold level. Measurements with standard deviations across the last four reversals larger than 4 dB were discarded.

Electrophysiological interaction was investigated using intracochlear ECochG responses to acoustic pure tones (A), electric pulse trains (E), and simultaneously presented electric and acoustic stimuli (E+A).

Figure 4 shows the stimulation and recording configuration used for the electrophysiological interaction measurements. As in the psychoacoustic experiment, electrical stimulation was delivered through the most apical electrode (⁠$elstim$⁠) using the case ground (⁠$casegnd$⁠) as reference. The second most apical electrode was selected for recording (⁠$elrec$⁠) and the extracochlear ring electrode (⁠$elref$⁠) acted as a reference. As demonstrated by Koka and Litvak (2017), this procedure enables the recording of EAS interaction even in the presence of electrical stimulus artifacts. The gain of the internal amplifier of the Advanced Bionics HiRes90K Ultra implant was set to 1000, and the sampling rate of the internal analog-to-digital converter (ADC) was configured to 9280 samples. This allowed the internal buffer to record up to 54.4 ms for each measurement. Sinusoidal acoustic tones with durations of 50 ms and with 5 ms Hanning window ramps were used to evoke ECochG responses. One ECochG recording consisted of 120 single trials of two acoustic presentations with alternating polarity (rarefaction and condensation). The trials were repeated with a stimulation rate of 0.57 Hz (1.75 s). The rarefaction and the condensation presentations were separated by 350 ms.

The difference response of two ECochG recordings with alternating polarities emphasizes odd harmonics and is dominated by the CM (Fontenot et al., 2017; Forgues et al., 2014). Therefore, the difference response was considered a representation of the CM response (CM/DIF), although it could include part of the ANN. The summation response to alternating polarities emphasizes even harmonics. Under the assumption that the ANN is mostly responsible for the presence of large even harmonic values, the summation response was considered a representation of the ANN response (ANN/SUM).

Subtracting the response E from the response E+A minimized the electrical artifact caused by the electrical stimulation and resulted in the derived acoustic response A' (Koka and Litvak, 2017). The interaction (i.e., the effect of electric stimulation on the acoustic response) was estimated by subtracting response A from response A'. The recorded responses were transformed into the frequency domain through a discrete Fourier transform (DFT) of 404 samples using zero-padding for adequate amplitude estimation. The amplitude spectrum of the CM/DIF responses presents a peak corresponding to the first harmonic of the acoustic probe stimulus. The amplitude spectrum of the ANN/SUM responses presents a peak corresponding to the second harmonic of the acoustic probe stimulus. The difference in the amplitude peaks between the A responses and the A' responses indicates the amount of interaction. Interaction was detected if the significance level (99% of the confidence interval of the noise floor) was exceeded. Otherwise, no interaction was detected. The CI's noise floor was measured by recording the ECochG without acoustic or electric stimulation. The 99% confidence interval for each frequency bin was calculated applying a bootstrap method with 1000 iterations (Koka et al., 2017).

Figure 5 shows recorded time series and their corresponding spectra as used for the electrophysiological EAS interaction measurements based on CM/DIF responses to an acoustic stimulus of 750 Hz. A similar computational technique was applied to analyze the ANN/SUM EAS interaction. Note that this method cannot differentiate between electric-on-acoustic or acoustic-on-electric interaction, and therefore, interaction can only be detected in general.

Figure 6 presents the results of the psychophysical masking and the electrophysiological interaction experiments for each subject across frequency. Figures 6(A) and 6(B) show the unmasked and masked psychoacoustic thresholds and the resulting threshold elevation. Figures 6(C) to 6(F) show the CM/DIF and ANN/SUM amplitudes recorded without and with electrical stimulation as well as the resulting EAS interaction as represented by amplitude attenuation.

Psychoacoustic thresholds, CM/DIF responses, and ANN/SUM responses without and with electrical stimulation could be measured in all subjects. However, despite the fact that unmasked and masked ANN/SUM responses were above noise floor, because of the overall small ANN/SUM responses, the ANN/SUM interaction recording was less successful than the CM/DIF recording. The amplitude attenuation was significant in 67% (32 out of 48) of the total number of recordings using the CM/DIM response. In contrast, the amplitude attenuation based on the ANN/SUM response, was significant only in 32% of the total number of recordings (12 out of 38). For analysis purposes, the amplitude attenuation was set to 0 dB if the interaction could not be detected (i.e., if the amplitude change for CM/DIF or ANN/SUM recordings was not significant).

Figure 7 shows the mean and median with the standard error for psychoacoustic, CM/DIF, and ANN/SUM EAS interaction magnitude for each frequency. The number of subjects tested (N) for each frequency is given as well as the mean of the tonotopic frequencies, derived from the CBCT scans, corresponding to the most apical electrode. The maximum mean psychoacoustic threshold elevation was observed at 750 Hz, which corresponds with the mean location of the stimulated electrode. One could think that the measurement was dominated by subject ID 2, which showed the largest threshold elevation of 15.67 dB at 750 Hz [Fig. 6(B)]. However, the maximum median threshold elevation was also observed at 750 Hz. In contrast, the CM/DIF interaction showed a more constant effect within the frequency range from 125 Hz to 1500 Hz for both the mean and the median amplitude attenuation. The mean ANN/SUM EAS interaction (amplitude attenuation) was negative, the median was not. The negative mean interaction implies that the presence of electrical stimulation caused a higher ANN/SUM response in comparison to the ANN/SUM response recorded without electrical stimulation. However, no ANN/SUM amplitude attenuation was observed in the median. This contrast implies that the observed negative mean interaction likely reflects one or two outliers [Fig. 6(F)].

Table II presents the frequency-specific means and p-values for the psychophysical and electrophysiological EAS interaction experiment using the Wilcoxon signed rank test across subjects to assess if the interaction differed significantly from zero. The mean psychoacoustic threshold elevation grew with increased frequency, reaching a maximum of 5.43 dB at 750 Hz, which is similar to the mean estimated tonotopic frequency corresponding to electrode location of 728 Hz. However, the psychoacoustic threshold elevation was only significant at 750 Hz. In contrast, the CM/DIF amplitude attenuation showed relatively constant mean amplitude attenuation of 1.14 dB from 125 Hz to 1500 Hz. A maximum amplitude attenuation of 1.54 dB was observed at 3000 Hz. However, only four subjects were tested at 3 kHz, and only subject ID 3 and subject ID 5 presented significant interaction. Therefore, the mean amplitude attenuation at 3 kHz was dominated by subject ID 3 and subject ID 5. CM/DIF amplitude attenuation was significant for 125 Hz and 250 Hz (p < 0.05). CM/DIF amplitude attenuation for 500 Hz and 750 Hz almost reached the significance level (p = 0.06). No significant ANN/SUM amplitude attenuation was observed.

Statistical differences between frequencies 125 Hz, 250 Hz, 500 Hz, and 750 Hz were assessed performing a one-way Friedman's analysis of variance (ANOVA) for psychoacoustic, CM/DIF, and ANN/SUM EAS interaction. In the current analysis, the frequency range was limited to 750 Hz and below to obtain an equal number of subjects at each frequency (N = 7). No significant differences between frequencies could be observed for psychoacoustic interaction (N = 7; Chi-Square = 3.174; df = 3; P = 0.366), CM/DIF interaction (N = 7; Chi-Square = 1.456; df = 3; p = 0.692), or ANN/SUM interaction (N = 7; Chi-Square = 6.769; df = 3; p = 0.08).

Table III shows the means and p-values for the psychoacoustic threshold elevation, CM/DIFF amplitude attenuation, and ANN/SUM amplitude attenuation using the Wilcoxon signed rank test across frequency limited to 750 Hz for equally balanced number of subjects (N = 7 for each condition). Table III shows that the CM/DIF amplitude attenuation of 1.13 dB was significantly above zero. The mean psychoacoustic threshold elevation and the mean ANN/SUM amplitude attenuation across frequency were 2.87 dB and −0.65 dB, respectively. In both cases, the observed EAS interaction was not significant.

Krüger et al. (2017) and Imsiecke et al. (2018) reported that psychoacoustic threshold elevation decreased exponentially with increased EAFD. Motivated by these previous observations, group-average EAS interaction across absolute frequencies (Fig. 7) was also analyzed in terms of EAFD, considering the electrode location for each individual subject. Figures 8(A) to 8(C) present the psychoacoustic threshold elevation, the CM/DIF amplitude attenuation, and the ANN/SUM amplitude attenuation across EAFD. If the electrode's $fe$ was more basal in relation to the acoustic stimulus frequency $fa$⁠, the EAFD was defined as positive. In case of negative EAFDs, the electrode's $fe$ was more apical than the acoustic probe frequency. In this case, the electrode array physically overlapped with the tonotopic location of the acoustic stimulus.

If the change in amplitude for CM/DIF or ANN/SUM responses was not significant, it was considered that no interaction could be detected and the amplitude attenuation was set to 0 dB. By doing this, instead of excluding the non-significant data, an overestimation of the mean amplitude attenuation was prevented.

A one-sample Wilcoxon signed rank test was performed to assess whether the psychoacoustic threshold elevation or the electrophysiological amplitude attenuation was significantly different from 0 dB. Table IV shows the probe level's mean change in dB and the significance levels (p-values) for the psychoacoustic threshold elevation, the CM/DIF amplitude attenuation, and the ANN/SUM amplitude attenuation for the three conditions considered: across all EAFDs (all EAFD), for negative EAFDs (EAFD ≤ 0), meaning that the acoustic probe stimulus was presented at the same location or more basally with respect to the electrode location, and for positive EAFDs (EAFD > 0), meaning that the acoustic stimulus was presented more apically with respect to the electrode location. As shown in Table IV, psychoacoustic threshold elevation and the CM/DIF amplitude attenuation were significantly above 0 for all conditions (p < 0.05). A mean psychoacoustic threshold elevation of 2.06 dB was observed across all subjects and EAFDs. The largest threshold elevation of 15.67 dB was observed for subject ID 2 at an EAFD of 0.03 octaves. The lowest threshold elevation of −3.34 dB (corresponding with a decrease in psychoacoustic threshold) was observed for subject ID 4 at an EAFD of 0.44 octaves. The mean amplitude attenuation determined by the CM/DIF response was 1.02 dB (max. 4.57 dB at an EAFD of −2.12 octaves by subject ID 5, min. −0.8 dB at an EAFD of 2.42 octaves by subject ID 6) and −0.57 dB but not significant for amplitude attenuation determined by ANN/SUM responses (max. 4.28 dB at an EAFD of −0.71 octaves by subject ID 3, min. −7.85 dB at an EAFD of 0.267 octaves by subject ID 3).

For positive EAFDs, the mean psychoacoustic threshold elevation was 2.61 dB, and the mean CM/DIF amplitude attenuation was 1.17 dB. For negative EAFDs, the mean psychoacoustic threshold elevation was 1.52 dB, and the CM/DIF amplitude attenuation was 0.88 dB. No significant ANN/SUM amplitude attenuation was observed. However, because the number of measurements for positive and negative EAFDs was different, it was not possible to statistically compare interactions between negative and positive EAFDs for both psychoacoustic and electrophysiological measures.

A goal of the current study was to compare electrophysiological and psychoacoustic interaction. The frequency specific analysis of EAS interaction (Fig. 7 and Table II) indicated that psychoacoustic threshold elevation was largest around frequencies corresponding with cochlear locations close to the electrode positions and that EAS interaction was more prominent toward lower frequencies. In contrast, the CM/DIF amplitude attenuation seemed to be lower but constant across a wider frequency range. Krüger et al. (2017) showed significant psychoacoustic threshold elevations for positive EAFDs up to 1.5 octaves, with the largest threshold elevation of 18 dB at an EAFD of around 0.43 octaves. For this reason, the differences between psychoacoustic threshold elevation and CM/DIF interaction were investigated between 0.0 and 1.5 octaves. Two measurements for each subject and condition were available. All subjects except subject ID 4 showed a larger psychoacoustic threshold elevation than CM/DIF amplitude attenuation. The mean psychoacoustic threshold elevation was 3.52 dB (Wilcoxon signed rank test: p = 0.028; N = 14), and the mean CM/DIF amplitude attenuation was 1.1 dB (Wilcoxon signed rank test: p = 0.005; N = 14), and both were significantly above zero. A related-sample Wilcoxon signed rank test demonstrated that the mean psychoacoustic threshold elevation was significantly larger (difference = 2.42 dB, p = 0.048, N = 14) than the mean CM/DIF amplitude attenuation within this EAFD range.

Table V gives subject specific statistics across all EAFDs rather than restricting analysis between 0.0 and 1.5 octave EAFDs. For each subject a Wilcoxon signed rank test was performed across frequency for psychoacoustic, CM/DIF, and ANN/SUM EAS interaction. Subjects ID 1, ID 3, and ID 5 showed significant psychoacoustic and CM/DIF EAS interaction across all EAFDs. Subject ID 7 showed only significant CM/DIF amplitude attenuation. Subjects ID 2, ID 4, and ID 6 showed no psychoacoustic or CM/DIF interaction. No ANN/SUM EAS interaction was observed.

This study investigated EAS interaction through psychoacoustic masking experiments and electrophysiological intracochlear ECochG for CI users with ipsilateral residual hearing. ECochG responses were successfully recorded, and CM/DIF as well as ANN/SUM responses were derived to estimate the interaction between electric and acoustic stimulation. EAS interaction was observed in the psychoacoustic experiments and from the CM/DIF responses of the electrophysiological experiments (Fig. 6). In contrast to previous studies (Imsiecke et al., 2018; Krüger et al., 2017), psychoacoustic masking and electrophysiological interaction were observed for negative EAFDs of up to −3 octaves (i.e., toward basal regions of the cochlea with respect to the electrode position) because the study participants (particularly ID 1, ID 4, and ID 5) had more residual hearing at relative high acoustic frequencies (Fig. 8). This experiment extends the findings reported in previous studies by Krüger et al. (2017) and Imsiecke et al. (2018), where psychoacoustic electric masking could only be measured toward positive EAFDs (i.e., the acoustic stimulation was delivered more apically than the electrical stimulation). For EAFDs around zero, there is potentially more overlap between the electric and the acoustic stimulation, whereas with increased positive or negative EAFDs, there may be a gap between both stimulation modalities, which can reduce the chance of observing interaction.

The current study showed that the threshold elevation for acoustic stimuli caused by simultaneously presented electrical stimulation was most pronounced for acoustic stimuli close to the electrode position (Fig. 7). This result is in agreement with previous findings from Krüger et al. (2017) where an exponential decay in threshold elevation across EAFD was shown. In comparison to our previous study (Krüger et al., 2017), the current results on psychoacoustic threshold elevation present higher variability, particularly within the first octave of the positive EAFD range. This variability could be attributed to the poorer residual hearing in the low acoustic frequencies of the study participants. However, due to the relatively small number of subjects, this hypothesis cannot be confirmed. In Krüger et al. (2017), we measured a wider range of positive EAFDs (five octaves in comparison to three octaves), which permitted a better demonstration of the decay of psychoacoustic threshold elevation across positive EAFDs. However, the maximum observed threshold elevation of 15.67 dB at an EAFD of 0.03 octaves [Fig. 8(A)] is in agreement with the findings from Krüger et al. (2017) and Lin et al. (2011), who reported a maximum threshold elevation of 18 dB and approximately 15 dB, respectively. The results are also in agreement with the findings from Stronks et al. (2011, 2012), who showed that acoustically evoked CAPs were more suppressed when there was a larger spatial overlap between electric and acoustic stimuli.

A goal of the current study was to assess how much of the observed psychoacoustic interaction between electric and acoustic stimulation can be explained electrophysiologically at the level of the hair cells or the auditory nerve. In general, across subjects and EAFDs, the psychoacoustic threshold elevation and CM/DIF amplitude attenuation showed similar behavior (Fig. 8). Interaction, as measured through psychoacoustic threshold elevation and CM/DIF amplitude attenuation was significant for positive EAFDs, negative EAFDs, and across all EAFDs (Table IV). The average interaction obtained through CM/DIF and psychoacoustic measures across all EAFDs was 1.02 dB (p < 0.001) and 2.06 dB (p < 0.001), respectively (Table IV). However, the difference between the mean CM/DIF and the mean psychoacoustic interaction through the whole range of tested EAFDs was not significant. Despite this similar behavior, no significant correlation was observed between threshold elevation and CM/DIF amplitude attenuation at an individual or group level under the same combination of stimulated electrode and acoustic frequency, probably because of the small number of subjects (N = 7) and the intra and inter subject variability. These results are in agreement with the findings from Koka and Litvak (2017), where also no correlation between psychoacoustic and CM/DIF or ANN/SUM EAS interaction was found.

Although there was little statistical evidence that the CM/DIF interaction was smaller than the psychoacoustic interaction, there was a general trend in this direction. There was also a difference in the pattern of results, whereby psychoacoustic interaction was frequency dependent and largest for frequencies corresponding to the electrode location, whereas the CM/DIF interaction was more consistent across a broad frequency range.

Krüger et al. (2017) showed that psychoacoustic threshold elevation decreased with increased EAFD and that the threshold elevation was significant for EAFDs between 0.0 to 1.5 octaves. For this reason, in the current study we compared CM/DIF interaction and psychoacoustic threshold elevation within the same range of EAFDs. The results showed a mean psychoacoustic threshold elevation of 3.52 dB (p = 0.028) and a mean CM/DIF amplitude attenuation of 1.1 dB (p = 0.005). Moreover, the difference between psychoacoustic threshold elevation and CM/DIF amplitude attenuation was significant (2.42 dB, p = 0.048). These results may indicate that peripheral interaction as measured through CM/DIF responses cannot explain psychophysical interaction and in particular, it could be that interaction is stronger for neural than for hair cell elements even though it could not be supported by ANN/SUM recordings. It is also possible that at least part of the interaction occurs at a more central location along the auditory pathway.

The observed electrophysiological interactions are in agreement with the findings from Koka and Litvak (2017) who observed CM/DIFF amplitude attenuation from ECochG recordings ranging from 0 to 4 dB and a psychoacoustic threshold elevation ranging from 0 to 20 dB. Imaging data were not available in Koka and Litvak (2017), and therefore, the relation between interaction and electrode location could not be investigated in their study. However, for subjects with MidScala electrode arrays (Advanced Bionics, Valencia, CA), they observed that psychoacoustic threshold elevation was largest between 500 Hz and 750 Hz, a frequency range that most likely corresponds with the location of the most apical electrode. Koka and Litvak (2017) also suggested a decrease in psychoacoustic EAS interaction toward lower frequencies, while CM/DIF interaction seemed to remain constant for the low frequencies.

A reason for the difference between psychoacoustic and CM/DIF EAS interactions could be that electrophysiological measures were obtained from acoustic stimuli presented at MCL to maximize the amplitudes of the CM/DIF and the ANN/SUM responses, whereas psychoacoustic interactions were measured at threshold levels of the probe stimulus. Nourski et al. (2007) and Stronks et al. (2010) showed that the relative level of acoustic and electric stimuli can affect the amount of masking, for both electric-on-acoustic or acoustic-on-electric masking. Nourski et al. (2007) recorded electrically evoked CAP in guinea pigs to assess acoustic masking and showed that the amount of masking decreased with increasing electric probe levels. Stronks et al. (2010) demonstrated the most pronounced suppression (up to 20 dB) of acoustically evoked CAP responses in the presence of electrical stimulation for low acoustic levels measured in guinea pigs. For high probe levels, the CAP suppression through electrical stimulation decreased to approximately 2 dB. These results are in agreement with the ranges of psychoacoustic and CM/DIF EAS interaction observed in the current study. Stronks et al. (2011) also demonstrated that CAP suppression was most pronounced at low acoustic levels and for electric and acoustic stimulation that physically overlapped. Based on these previous works, it is possible that the CM/DIF interaction observed in this study would have been larger if measured at softer levels.

In contrast to psychoacoustic and CM/DIF EAS interaction, ANN/SUM EAS interaction was not statistically significant, even though both unmasked and masked ANN/SUM responses were significantly above the noise floor (Table IV). Similarly, Koka and Litvak (2017) did not observe ANN/SUM interaction. In a real environment, recordings are always contaminated by noise, which is considered additive to the signal in a linear scale. In a log scale, however, the determination and comparison of signal amplitudes become less accurate with decreasing signal amplitudes. ANN/SUM interaction was derived from the second harmonic of the summation potential, and these amplitudes were usually smaller in comparison to the CM/DIF or the ANN/SUM amplitudes at the first harmonic. The small amplitudes and the small amplitude changes could be the reason for the non-significant ANN/SUM interaction. However, due to the limitations in separating CM and ANN responses using the difference and summation techniques, it is possible that the observed CM/DIF interaction was caused by the ANN at the level of the auditory nerve, since the CM/DIF contains the largest part of the ANN (Fontenot et al., 2017).

The EAFD was useful to demonstrate differences between psychophysical and electrophysiological EAS interaction and to investigate the potential mechanisms involved. While the psychoacoustic threshold elevation includes possible central and peripheral mechanisms, CM/DIF amplitude attenuation is limited to peripheral mechanisms. Moreover, it is not known to what extent peripheral interaction affects psychophysical EAS interaction. It is known that electrical stimulation of cochlear regions with residual hearing can elicit hair cell depolarization through electrophony. Electrophonic stimulation has been demonstrated in animal studies (McAnally et al., 1993, 1997; Sato et al., 2016; Stronks et al., 2013). Some of these studies (McAnally et al., 1997; Stronks et al., 2013) showed that acoustic evoked CAP responses can be suppressed by electrical stimulation independent of the electrical stimulus location along the cochlea and that the amount of suppression corresponded to the energy distribution given by the pulse duration of the electrical stimulus. In the present study, we observed CM/DIF amplitude attenuation toward the whole range of positive EAFDs corresponding with the apical part of the cochlea, while psychoacoustic threshold elevation was more prominent close to the location of electrical stimulation (Fig. 8). The presence of CM/DIF EAS interaction toward low frequencies [Fig. 8(B)], despite the large EAFD, could indicate electrophonic interaction in accordance with the findings from McAnally et al. (1993, 1997) and Stronks et al. (2013). The subjects' extensive high-frequency hearing loss and the absence of significant CM/DIFF interaction for frequencies above 750 Hz could exclude the presence of electrophonic interaction at these frequencies. The spectrum of pulsatile electrical stimulation, as used in the present study, consisted of a series of harmonics corresponding to the stimulation rate, whose energy distribution is in accordance with the inverse of the pulse duration. The short pulse duration of 25 μs used in the current study led to a relatively flat spectral profile, which could explain the observed constant amplitude attenuation across positive EAFDs. In contrast, psychoacoustic threshold elevation was only observed at the location where electrical stimulation was delivered, which may reflect interaction at the level of the auditory nerve or toward a more central location of the auditory pathway that cannot be measured through CM/DIF or ANN/SUM responses.

At the individual subject level, psychoacoustic EAS interaction could not be entirely explained by the CM/DIF responses derived from ECochG recordings. However, the maximum psychoacoustic masking of 15.67 dB observed at an EAFD close to zero shows that these interaction effects can be perceived. These effects will become more relevant if the development of EAS systems and surgery techniques allow even deeper electrode insertions in subjects with increased residual hearing. If masking between electric and acoustic stimulation has an impact on speech intelligibility (Imsiecke et al., 2019), its estimation through ECochG could be applicable to develop new fitting rules, automated fitting procedures, and new sound coding strategies that take into account electrophysiological EAS interaction.

This work shows that CM/DIF responses derived from intracochlear ECochG recordings can be used to measure EAS interaction. Furthermore, this study compared electrophysiological and psychophysical EAS interactions. By means of clinical imaging, the relative location between electric and acoustic stimulations, termed EAFD, was used to characterize the interaction between electric and acoustic stimulations. Whereas psychophysical interaction was only observed for frequencies close to the electrode position, electrophysiological interaction remained roughly constant across low frequencies. Considering the EAFD, the psychophysical interaction was significantly larger than the electrophysiological interaction in the range between 0 and 1.5 octaves. From the results presented in this work, it can be concluded that, even though peripheral EAS interaction was observed across the whole range of positive EAFDs, the perceived psychoacoustic EAS interaction was dominated by the location at which electrical stimulation was delivered. Previous findings from psychoacoustic and electrophysiological EAS interaction (Imsiecke et al., 2018; Koka and Litvak, 2017; Krüger et al., 2017; Lin et al., 2011) were extended toward negative EAFDs because of the extended high-frequency residual hearing of the study participants in combination with their relatively deep inserted electrodes, meaning more physical overlap between electric and acoustic stimulation.

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy – EXC 2177/1 - Project ID 390895286 and by the DFG - Project number: 396932747 (PI: W.N.). Without the subjects who participated in these experiments, this work would have not been possible; therefore, the authors would like to thank the subjects for their invaluable contribution. The authors would like to thank Advanced Bionics for providing software and hardware equipment.