Participants

Thirteen (13) adult users of Advanced Bionics CIs were recruited, two of whom were bilaterally implanted and completed the study with both ears for a total of fifteen (15) ears. Demographic information can be found in Table 1, in addition to some experimental details. Inclusion criteria required that it was possible to collect PECAP data that achieved a 10 decibel (dB) or higher signal-to-noise ratio (SNR), as specified for achieving accurate neural excitation profiles with PECAP [26]. Three (3) ears did not achieve this threshold (AB011R, AB018R, and AB020L, indicated by the asterisks * in Table 1) and were excluded from analysis, leaving a total of twelve (12) ears in the primary dataset. Further details for calculating the SNR of the PECAP matrices and other study-specific inclusion criteria for the different parts of the analysis are described below in the “Electrically Evoked Compound Action Potentials” section of the “Materials and Methods” section.

Table 1 Participant demographic information (R right, L left, yrs years, Elecs electrodes, pTP partial tripolar; * indicates that the participant was excluded from analysis due to low PECAP SNR values; Felectrode where pTP stimulation was applied, Belectrode where blurred stimulation was applied, ^electrode that showed high sQP thresholds)Ethical Approval

Ethical approval for the study was obtained from the Massachusetts Eye & Ear (MEE) Internal Review Board (IRB) under Protocol # 2021P001539. All participants attended research sessions at Massachusetts Eye and Ear in Boston and provided written informed consent for their participation in the study.

Focused Thresholds (sQP)

Steered quadrupolar (sQP) stimulation was used to determine focused thresholds along the electrode array for all ears studied. sQP thresholds have been shown to highly correlate with partial tripolar (pTP) thresholds [30], and have previously been hypothesized to reflect some aspects of cochlear neural health [31]. sQP stimulation leverages four adjacent intra-cochlear electrodes where the two middle electrodes serve as active electrodes, the two outer electrodes serve as return electrodes for a fraction of the active current, and the remainder of the current is returned to an extra-cochlear electrode. For this study, unless otherwise specified, 10% of the current was returned to the extra-cochlear electrode, and the other 90% was split evenly between the two flanking intra-cochlear electrodes. The active current is then steered between the two middle electrodes and for electrodes 3–15, is steered entirely towards the more-basal electrode. It is then steered entirely towards the apical electrode to achieve focused stimulation on electrode 2 due to physical limitations. Due to the 4-electrode configuration, focused thresholds could not be obtained at electrodes 1 and 16. The focused threshold for each targeted electrode was set as the lowest current level for which an auditory sensation was perceived.

A rapid threshold measurement procedure based on Békésy tracking was used to determine the sQP threshold for each electrode. It was implemented using custom software written in MATLAB that runs the Bionic Ear Data Collection System (BEDCS) from Advanced Bionics AG (CA, USA) in the background and carries out an adaptive forced-choice procedure in both a forward (apical to basal) and backward (basal to apical) manner. The listener pressed a button to indicate audibility while the frequency of the tone is moving (towards either the apex or base), and the stimulation level of the sQP pulse train either increased (when the button is released) or decreased (when it is pressed). Multiple sweeps in both directions were done to yield repeated observations of threshold for each electrode. Each sQP pulse train was delivered at 997.9 pulses per second (pps) for 200.4 ms (ms). For one participant, AB020L, only 80% of the current was split evenly between the two flanking intra-cochlear electrodes due to compliance limits. The procedure is described in full detail in Bierer et al. [30]. From the focused threshold results, two electrodes between 5 and 12 that contained both the highest and lowest focused thresholds were selected for further current manipulation in the ECAP conditions described below.

Electrically Evoked Compound Action PotentialsBaseline Conditions

Electrically evoked compound action potentials (ECAPs) were recorded using the forward-masking artefact cancellation technique at comfortable loudness levels for every combination of masker and probe electrode for all active electrodes in the participants’ “MAPs” (sound processor programs) in standard monopolar mode. Amplitudes were calculated from these ECAP waveforms using custom software that identified and subtracted the first negative peak within 400 µs of the offset of the probe pulse from the following positive peak. These ECAP amplitudes constitute the \(_\) matrix described in Garcia et al. that is required for the Panoramic ECAP method to provide estimates of current spread and neural responsiveness for each electrode in an individual CI [26]. Stimulation employed symmetric, cathodic-leading biphasic current pulses presented at approximately 20 pulses per second (pps) in monopolar mode with phase durations of 43.1 µs, inter-phase gaps of 0 µs, the stimulation ground electrode set to IE2 (the extra-cochlear case electrode), and equal masker and probe electrode current levels. A masker-probe interval of 600 µs was selected in order to avoid temporal overlap between stimulation artefacts and the N1 peak of the ECAP waveform. The gain of the amplifier was fixed at a value of 1000, and the amplifier sampling frequency was fixed at 56 kHz. The recording electrode for the ECAPs was set as a default to 3 electrodes apical to the probe electrode for any given ECAP recording. It was switched to 3 electrodes basal to the probe electrode when the probe was at the apical end of the array, and therefore, no electrodes were available 3 apical to the probe (this occurred when the probe was located on electrodes 1, 2, and 3). If the masker electrode was located where the recording electrode would have been by default, the recording electrode was shifted over one electrode from the default to avoid saturating the amplifier by stimulating and recording on the same electrode, and with the assumption that the effect of shifting the recording electrode by one contact would be sufficiently small to be of little relevance in comparison with the effect of the moving masker.

The loudness scaling procedure was conducted with a 10-point loudness scale using custom software written in MATLAB 2018a that leveraged the Bionic Ear Data Collection System (BEDCS) research software (versions 1.18.321 and 1.18.337) from Advanced Bionics AG (CA, USA). The custom software was controlled through a graphical user interface (GUI) used by a research audiologist wherein the stimulating current level was initialized below perceptual threshold and increased until loudness ratings of 5 (soft but comfortable), 6 (most comfortable level or MCL), 7 (loud but comfortable), and 8 (loud) were achieved. Each time an ECAP was recorded, these data and the perceptual loudness level (if indicated) were automatically saved and the ECAP waveform was plotted in real time so that the experimenter could confirm that stimulation and recording were ongoing throughout the session. This loudness scaling procedure was conducted with 10 repetitions of each ECAP recording for every 3rd electrode active in the participant’s MAP (electrodes 1, 4, 7, 10, 13, and 16). For some participants, electrode 16 was switched off and so the loudness scaling procedure was conducted on electrode 15 instead of 16 (see Table 1). The program then recorded ECAPs with the probe and the masker co-located on the same electrode for every electrode active in the participant’s MAP at the level requested by the experimenter (starting at MCL), with the non-scaled electrodes set to current levels that were interpolated on a dB scale between adjacent electrodes, and with 50 repetitions. These data are referred to as the “diagonal” of the measurement matrix. If no ECAPs were observed in the diagonal at MCL (6), the stimulus level was increased to loudness level “7” or “8.” If the experimenter then determined that the stimulation level was comfortable for the user and achieved observable ECAP waveforms, the program then continued on to record the entire PECAP \(_\) matrix at that loudness level (again with 50 repetitions). The criteria for determining whether an ECAP was present or not was based on a comparison between the calculated ECAP amplitude and the noise floor of the recording calculated from the standard deviation of the noise-only recording frames (D, as described below), as well as visual inspection of the experimenter looking for the presence of negative and positive peaks typical of the morphology of ECAP waveforms. The loudness level used for each participant is indicated in “PECAP Level rating” in Table 1.

The custom research software streamlined the process of recording many ECAPs at once in several ways. The forward-masking artefact reduction technique consists of four recording frames: (A) masker alone, (B) masker + probe, (C) probe alone, and (D) system signature. Recording all combinations of masker and probe with the recording electrode placed relative to the location of the probe would involve repeating the measurements of the (C) and (D) frames for each ECAP. Therefore, these frames were skipped in most cases, and the (C) and (D) frames were only recorded when the probe and the masker were located on the same electrode. The (C) and (D) frames were then re-used for the remainder of that row of the \(_\) matrix. This saved a little less than 50% of the recording time that would otherwise have been necessary. Similar approaches in the Cochlear platform have been shown to incorporate minimal error into ECAP amplitude measures [32]. Secondly, the use of a fast USB-to-serial connection converter enabled faster communication between the testing computer and the research hardware (consisting of the Clarion Programming Interface version II and a Platinum Series Processor (PSP)). With these two factors incorporated, the data collection time for the \(_\) matrix amounted to approximately 28 min per participant.

Inclusion criteria for the study related to a reliability metric that was calculated between repeat measures of the recorded \(_\) for each research participant. Garcia et al. showed that the PECAP algorithm can re-produce underlying neural excitation patterns defined as a combination of current spread and neural responsiveness with 90% accuracy down to a signal-to-noise ratio (SNR) of 10 dB [26]. This was shown using computer simulations where the underlying “ground truth” of the neural excitation patterns was pre-defined. Below this SNR, the accuracy of the algorithm to re-create the ground truth dropped below 90%. Therefore, if the \(_\) of any of the 15 ears included in this study was below 10 dB, then they were excluded from analysis. The SNR was determined by calculating the root mean squared error (RMSE) between repeated measurements of ECAPs for each participant. Repeated measures were available for the ECAPs recorded where the masker and the probe were presented on the same electrode (the “diagonal” of \(_\)) as part of the standard PECAP procedure. The SNR could therefore be calculated from this RMSE using a transfer function defined in Garcia et al. [26] and shown below in Eq. 1.

$$f\left(x\right)= -5.70\times ^^+1.15\times ^^+8.44\times ^^-0.012\;^-0.80x+17.46$$

(1)

As indicated in Table 1, the SNR did not reach the 10 dB threshold for three of the 15 ears, and therefore, data from AB011R, AB018R, and AB020L were excluded from analysis.

The \(_\) s for the remaining twelve ears were then submitted to the Panoramic ECAP algorithm described in Eqs. 1–10 in Garcia et al. to estimate current spread and neural responsiveness for each electrode switched on in the participants’ respective MAPs [26]. In brief, the algorithm constructs neural activation patterns based on a convolution of neural responsiveness and Gaussian current spread centered at each electrode, and reconstructs the expected \(M\) given the estimated neural activation patterns. The current spread and neural responsiveness estimates are then allowed to change within an optimization loop that minimizes the least squared error between the expected \(M\) and the measured \(_\).

Blurring Stimulation Condition

The two electrodes with the highest and lowest focused (sQP) thresholds, respectively (as described above), were selected for each research participant to stimulate with blurred current (see Table 1 for details). There were two blurring factors evaluated: first, one electrode was additionally stimulated on each side of the central stimulating electrode constituting “Blur Factor 3,” and second, two electrodes were additionally stimulated on each side of the central stimulating electrode constituting “Blur Factor 5” (graphically depicted in Fig. 1). Each of the electrodes stimulated simultaneously was presented with the same amount of electrical current. The blur factors of 3 and 5 electrodes were selected based on results from Goehring et al. that showed that speech-in-noise recognition thresholds (SRTs) started to increase compared to a stimulation in standard monopolar mode when all channels were blurred by using 4 electrodes simultaneously [21]. In this condition, a stimulating “channel” will refer not directly to a single electrode but to the central stimulating electrode in a group of 3 or 5 simultaneously stimulated electrodes.

Fig. 1

Graphic representing the stimulation configuration when one electrode is stimulated on its own (baseline condition, blur factor = 1), where three adjacent electrodes are stimulated simultaneously (blur factor = 3), and where five adjacent electrodes are stimulated simultaneously (blur factor = 5)

The selection of the channels for the blurring conditions was necessarily limited by the study design; in order for an electrode to be the central blurred electrode, there had to be at least 2 electrodes available on each side to additionally stimulate for the “Blur Factor 5” condition and an additional electrode on each side to serve as the recording electrode for the ECAP. Therefore, if all electrodes were active in a CI listener’s MAP, the most apical channel that could be blurred was centered on electrode 4, and the most basal channel that could be blurred was centered on electrode 13. As the original study design also included a condition wherein 7 electrodes would be stimulated simultaneously, the software further restricted this to between channels centered on electrodes 5 and 12. However, due to time constraints for the data collection, it was not possible to measure the 7-electrode configuration.

Loudness scaling was performed for each of the two channels selected for blurred stimulation for both of the blur factors. This was done to determine the current level required for the level rating selected for the monopolar \(_\) for that participant for both blur factors (this was not always the default level rating of 6 or MCL; see Table 1). ECAPs were then additionally collected for both these channels and at both blur factor 3 and blur factor 5 for each masker-probe combination condition that involved the central electrode of the channel in question. The blurred stimulation was applied both for the case where the blurring channel was centered on the masker and where it was centered on the probe electrode. This only required additional ECAPs to be recorded in blurring conditions for the row and column in the \(_\) matrix that involved the blurred electrodes. This row and column of ECAP data were then inserted into the \(_\) matrix obtained for the baseline condition, replacing the monopolar ECAP amplitudes for these electrodes. The resultant \(M\) matrices will be referred to as the \(_\) matrices. There were four of these \(_\) matrices in total, consisting of both blur factors (3 vs 5 electrodes) and both sQP threshold levels (high vs low). The four \(_\) matrices were then submitted to the PECAP algorithm to calculate estimates of current spread (σ) and neural responsiveness (η) for each electrode.

Focused Stimulation Condition

The procedure described above was also implemented for partial tripolar (pTP) stimulation. This was conducted for the same two electrodes for each participant as in the blurring stimulation condition, with the exception of AB011L and AB013L for whom the basal electrode for stimulating in pTP mode was shifted 1 electrode more basal compared to with the blurred mode (see Table 1).

Partial tripolar mode consists of sending active current to one central electrode and returning part of the current to the two adjacent electrodes and the remainder of the current to an extra-cochlear electrode. For this stimulation, the portion of the return current evenly split between the two adjacent electrodes is referred to as the α-value. Therefore, α = 1.0 indicates that 50% of the current is returned to the electrode immediately apical to the central active stimulating electrode, and 50% of the current is returned to the electrode immediately basal to it. By contrast, α = 0.5 indicates that 25% of the active current is returned to each of the flanking electrodes and 50% of the current is returned to an extra-cochlear electrode (see Fig. 2). For this portion of the experiment, the α-value was determined by starting with α = 1.0 and increasing the current level until either a loudness level of “8” was achieved or the stimulation compliance limit was reached for that electrode. If the latter was reached first, then the α-value was decreased and the procedure repeated until a loudness level of “8” could be achieved. If this was not possible even at α = 0.5, then the focused stimulation portion of the study was skipped for this participant. This was the case for 6 of the 15 ears (AB002R, AB014L, AB016L, AB017L, AB018R, and AB020L; see Table 1). The α-value was varied in this way to find the highest degree of focusing possible within compliance limits, with the ultimate goal of achieving a current distribution as similar as possible to full tripolar stimulation where no current is returned to extra-cochlear electrodes. The α-value for each remaining participant can be found in Table 1.

Fig. 2

Graphic representing the stimulation configuration when in full tripolar mode where all current is returned to the neighboring electrodes (α = 1.0), a partial tripolar mode when 25% of the current is returned to each of the neighboring electrodes and 50% is returned to the extra-cochlear electrode (α = 0.5), and monopolar mode when all current is returned to the extra-cochlear electrode (α = 0.0). Solid black pulses indicate active current and red, patterned pulses indicate return current. (ex, extra-cochlear electrode)

Fig. 3

(a) \(_\) matrix for AB007R, with each cell of the heat map representing the ECAP amplitude in µV of the masker-probe electrode combination indicated by its location. The red lines indicate which cells of the matrix are manipulated in the blur conditions displayed in (b, c). (b) \(_\) for the high sQP threshold electrode (5) for blur factor = 3. The red lines represent the cells of the matrix that were stimulated with 3 electrodes simultaneously (electrodes 4–6) when either the masker or the probe was centered on electrode 5. (c) \(_\) for the high sQP threshold electrode (5) for blur factor = 5. The red lines represent the cells of the matrix that were stimulated with 5 electrodes simultaneously (electrodes 3–7) when either the masker or the probe was centered on electrode 5. (d) ECAP amplitudes for the row of each \(M\) matrix in (a)–(c) representing the case where the probe is presented on electrode 5, including PECAP’s current spread estimate for electrode 5. (sQP, steered quadrupolar stimulation; σ, PECAP’s current spread estimate; e, electrode)

Fig. 4

PECAP estimates of current spread (a) and neural responsiveness (c) for the baseline and blurring conditions for AB007R. The black solid lines represent the baseline condition where no manipulations were made to the spread of electrical current and each electrode was stimulated in standard monopolar mode. The purple lines represent the PECAP results where three electrodes were stimulated simultaneously (blur factor 3), and the orange lines represent the case where five electrodes were stimulated simultaneously (blur factor 5). The vertical black dotted lines represent the electrode that showed the highest focused threshold (sQP), and the vertical black dashed lines represent the electrode that showed the lowest focused threshold; these are the electrodes that were blurred for this participant. The colored dotted lines represent the PECAP results from the scenarios where the electrode on which blurred stimulation was applied showed a high focused threshold, and the colored dashed lines represent where it showed a low focused threshold. Signed differences between the baseline conditions and the manipulated conditions for blurred electrodes (represented by triangles) and for the non-blurred electrodes that were stimulated in standard monopolar mode (represented by circles) are displayed for the current spread estimate (b) and for the neural responsiveness estimate (d). The error bars represent one standard deviation above and below the mean. (sQP, steered quadrupolar stimulation)

Fig. 5

Signed differences for PECAP estimates of current spread (a) and neural responsiveness (b). The triangles represent the blurred electrodes and the circles represent the non-blurred electrodes. The purple symbols represent blur factor = 3 and the orange represent blur factor = 5. Dotted lines (oriented towards the left for each participant) represent electrodes that showed high sQP thresholds and dashed lines (oriented towards the right for each participant) represent electrodes that showed low sQP thresholds. Error bars represent one standard deviation above and below the mean. Across participants, asterisks (*) are corrected for multiple comparisons within the ANOVA and represent 95% significance within a Tukey–Kramer comparison. (sQP, steered quadrupolar stimulation; ANOVA, analysis of variance; *p < 0.05)

Fig. 6

Across-electrode signed differences for PECAP estimates of current spread (a) and neural responsiveness (b). The triangles and the dotted lines represent the blurred electrodes and the circles and the solid lines represent the non-blurred electrodes. The purple symbols (oriented towards the left of each group) represent the blur factor = 3 and the orange (oriented towards the right of each group) represent blur factor = 5. The three-way ANOVA showed significant main effects of blurring (circles vs triangles, p < 0.0001) and of electrode location (more-apical vs more-basal, p = 0.0001), and a significant interaction between the two (p < 0.0001) on the current spread estimate (top), suggesting that the effect of blurring was effective at the more-apical electrodes but not at the basal electrodes. No significant effects were found in the three-way ANOVA on the neural responsiveness estimate. Error bars represent one standard deviation above and below the mean. (**p < 0.001)

Fig. 7

PECAP estimates of current spread (a) and neural responsiveness (b) for each of the 8 participants retained in the focused stimulation portion of the study. Shown are the baseline condition (solid black lines), the condition where the electrode stimulated in pTP mode showed a high sQP threshold (light-blue dotted lines), and the condition where the electrode stimulated in pTP mode showed a low sQP threshold (gray-pink dashed lines). The vertical light-blue dotted lines indicate the electrode with the highest sQP threshold for that participant and the vertical gray-pink dashed lines indicate the electrode with the lowest sQP threshold for that participant. (sQP, steered quadrupolar)

Fig. 8

Signed differences for PECAP’s current spread (σ) estimate (a) and neural responsiveness (ƞ) estimate (b) for each of the 8 participants retained in the focusing portion of the study. The black circles represent the non-manipulated or non-focused electrodes, and the associated error bars represent one standard deviation above and below the mean. The blue triangles represent the focused electrodes that showed high sQP thresholds, and the pink triangles represent the focused electrodes that showed the low sQP thresholds

As in the case for the blurring stimulation conditions, ECAPs were only recorded for each masker-probe combination that involved the central pTP electrodes in question, making up one row and column in the \(_\) matrix. This row and column of ECAP data were then inserted into the \(_\) matrix obtained for the baseline condition, replacing the monopolar ECAP amplitudes for these electrodes. The resultant \(M\) matrices will be referred to as the \(_\) matrices. The two \(_\) matrices were then submitted to the PECAP algorithm to calculate estimates of current spread (σ) and neural responsiveness (η) for each electrode.

Hypotheses and Statistical AnalysisPlanned Hypotheses

First, if the PECAP algorithm is able to detect increases in current spread correctly, then the estimated current spread (σ) for the channels on which electrical stimulation was blurred should be higher than in the baseline condition. Likewise, if PECAP is able to attribute the blurred current to the correct channel, then there should be no impact on the current spread estimate for non-blurred channels, and if it is also able to separate its estimates of the two factors characterizing the electrode-neuron interface, then there should not be an effect on the neural responsiveness estimate.

Second, it was hypothesized that simultaneously stimulating five adjacent electrodes (blur factor 5) would produce a greater increase in current spread than simultaneously stimulating three adjacent electrodes (blur factor 3).

Third, we hypothesized that if partial tripolar mode focuses electrical current, then PECAP would predict a decrease in the current spread in the focused condition for the electrodes stimulated in partial tripolar mode compared to the baseline condition. Likewise, if PECAP is able to attribute the focused current to the correct electrode, then there should not be an impact on the current spread estimate for non-focused electrodes, and if PECAP is able to separate its estimates of the two factors characterizing the electrode-neuron interface, then there should not be an effect on the neural responsiveness estimate.

We also investigated whether the effectiveness of both the blurring and the pTP stimulations on PECAP’s current spread estimate differed between electrodes with high vs low focused thresholds. It has been argued that low focused thresholds reflect regions of good neural health [31] and there are ways in which this might moderate the effectiveness of the blurring or pTP manipulations. For example, increasing current spread (blurring) beyond a highly localized region of high neural responsiveness might have a minimal effect on the neural spread of excitation, because the current would spread to a less-responsive region. Alternatively, a broad region of high neural health might mean that MCL can be reached with a narrow current spread, in which case blurring might have a large effect on neural excitation.

Statistical Analysis

Signed differences were calculated for both PECAP’s current spread (σ) and neural responsiveness (η) estimates between the baseline conditions and each of the four blurring conditions as well as the two focusing conditions for each research participant. This was done in order to evaluate and quantify the size of the effect for electrical current manipulations on PECAP estimates of the electrode-neuron interface. Instead of directly assessing the current spread (σ) and neural responsiveness (η) estimates, signed differences in these estimates were chosen as the dependent variables in order to remove between-subject variation representative of the heterogeneity known to be characteristic of most behavioral and electrophysiological assessments of cochlear implant perception. The signed differences were separated into “manipulated” electrodes — representing the central electrodes that were presented with either blurred or partial tripolar stimulation and listed in Table 1 — and “non-manipulated” electrodes — representing the remainder of the electrode array stimulated in standard monopolar mode. The manipulated group contained 1 data point per condition per participant by design, whereas the non-manipulated group contained either 15 or 14 values, depending on how many electrodes were switched on in the participant’s MAP and therefore included in the PECAP measurements. The individual values within the non-manipulated group are each used in the analysis instead of calculating their means in order to retain information about their variance. A positive signed difference would indicate that the effect of current manipulation increased the spread of electrical current (or neural responsiveness) for the manipulated condition compared to the baseline condition. All analyses were done using MATLAB version 2020b (Mathworks, Natick, MA, USA).

For the blurring conditions, two 3-way analyses of variance (ANOVAs) were conducted, one for the current spread and another for the neural responsiveness estimates. Three factors each with two levels were submitted to both ANOVAs: Factor 1: blurring (blurred vs non-blurred electrodes), Factor 2: blur factor (Blur Factor 3 vs Blur Factor 5), and Factor 3: sQP threshold (high vs low). This was done to determine if there was an effect of blurring overall, if there was a difference between the two blurring factors, and whether there was an effect of sQP threshold on the blurring effect, respectively. Residuals were checked for normality prior to reporting results.

For the focusing conditions, two 2-way ANOVAs were conducted, one for the current spread and another for the neural responsiveness estimates. Two factors each with two levels were submitted to both ANOVAs: Factor 1: focusing (focused vs non-focused electrodes), and Factor 2: sQP threshold (high vs low). Residuals were checked for normality prior to reporting results.

A few post hoc statistical tests were also performed and are described in the “Results” section. Note that the statistical tests performed evaluated the presence of effects, but could not confirm their absence.