Edited by: Rick Friedman, University of California, San Diego, United States
Reviewed by: Oliver Adunka, The Ohio State University, United States; Andrea Canale, University of Turin, Italy
This article was submitted to Neuro-Otology, a section of the journal Frontiers in Neurology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Several studies have demonstrated the advantages of the bilateral vs. unilateral cochlear implantation in listeners with bilateral severe to profound hearing loss. However, it remains unclear to what extent bilaterally implanted listeners have access to binaural cues, e.g., accurate processing of interaural timing differences (ITDs) for low-frequency sounds (<1.5 kHz) and interaural level differences (ILDs) for high frequencies (>3 kHz). We tested 25 adult listeners, bilaterally implanted with MED-EL cochlear implant (CI) devices, with and without fine-structure (FS) temporal processing as encoding strategy in the low-frequency channels. In order to assess whether the ability to process binaural cues was affected by fine-structure processing, we performed psychophysical ILD and ITD sensitivity measurements and free-field sound localization experiments. We compared the results of the bilaterally implanted listeners with different numbers of FS channels. All CI listeners demonstrated good sensitivity to ILDs, but relatively poor to ITD cues. Although there was a large variability in performance, some bilateral CI users showed remarkably good localization skills. The FS coding strategy for bilateral CI hearing did not improve fine-structure ITD processing for spatial hearing on a group level. However, some CI listeners were able to exploit weakly informative temporal cues to improve their low-frequency spatial perception.
The research performed in bilateral cochlear implant (CI) listeners clearly showed that they outperform users with a unilateral CI. Apart from improved speech understanding in noise (
CI technology mainly aimed at improving speech perception. Thus, most available pulse-encoding strategies [e.g., continuous interleaved sampling, or CIS (
Studies have suggested that bilateral CI listeners benefit from the FSP stimulation strategy in a speech discrimination test in noise (
Here, we performed two experiments to test sound localization and ILD/ITD processing in experienced bilateral CI users with and without FSP. In order to quantify the effect of the number of bilateral fine-structure channels, we assessed performance for three subgroups: (i) without FSP; (ii) with two bilateral FSP channels; or (iii) with four bilateral FSP channels. In principle, if FSP improves the ITD perception of bilateral CI users by providing reliable cues (
Twenty-five bilaterally deaf patients implanted with bilateral CIs (BICI) participated in the experiments. Their ages ranged from 22 to 77 years (53 ± 16.3 years). In the Netherlands, bilateral CIs in adults is not reimbursed. Therefore, in order to acquire bilaterally implanted adult subjects, cooperation was sought with the ENT Clinic of St. Elisabeth-Hospital of Ruhr-University in Bochum in our neighboring country Germany, where reimbursement of bilateral cochlear implantation for adults has been standard care for years. All included subjects had been implanted and were recruited at the University Clinic of Bochum (Germany) and traveled to Radboud University in Nijmegen (the Netherlands) to be assessed in the sophisticated spatial hearing labs at the Radboud University in Nijmegen. All research protocols and informed consent forms were approved by the Medical Ethical Committee in Bochum prior to the start of the experiments.
Demographic and general information about subjects.
P1 | 77 | 71 | 73 | Concerto Flex28 | Sonata Standard | Sonnet | Sonnet | FS4HR | FS4HR | 4 | 4 |
P2 | 62 | 58 | 56 | Sonata Standard | Sonata Standard | Opus2 | Opus2 | FSP | FSP | 2 | 2 |
P3 | 43 | 42 | 39 | Sonata Flex28 | Sonata Flex28 | Opus2 | Sonnet | FS4HR | FS4HR | 4 | 4 |
P4 | 29 | 28 | 29 | Sonata Flex28 | Sonata Flex28 | Sonnet | Sonnet | FS4HR | FS4HR | 4 | 4 |
P5 | 71 | 64 | 63 | Sonata Standard | Sonata Standard | Opus2 | Opus2 | FS4HR | FS4HR | 4 | 4 |
P6 | 76 | 69 | 73 | Sonata Flex28 | Sonata Standard | Opus2 | Opus2 | FS4HR | FS4HR | 4 | 4 |
P7 | 66 | 62 | 64 | Sonata Flex28 | Sonata Standard | Opus2 | Opus2 | FSP | FSP | 2 | 2 |
P8 | 22 | 1 | 6 | Pulsar Standard | Combi40 Standard | Sonnet | Sonnet | FSP | CIS | 1 | 0 |
P9 | 22 | 1 | 5 | Combi40+ Standard | Combi40+ Standard | Opus2 | Opus2 | CIS | CIS | 0 | 0 |
P10 | 56 | 51 | 52 | Sonata Flex28 | Sonata Flex28 | Opus2 | Opus2 | FS4HR | FS4HR | 4 | 4 |
P11 | 49 | 44 | 44 | Sonata Standard | Sonata Standard | Sonnet | Sonnet | FSP | FSP | 2 | 2 |
P12 | 41 | 36 | 39 | Concerto Flex28 | Concerto FlexSoft | Sonnet | Opus2 | FS4HR | FS4HR | 4 | 3 |
P13 | 62 | 55 | 54 | Sonata Standard | Sonata Standard | Sonnet | Sonnet | FS4HR | FS4HR | 4 | 4 |
P14 | 46 | 44 | 43 | Concerto Standard | Concerto Medium | Sonnet | Sonnet | CIS | FSP | 0 | 1 |
P15 | 49 | 46 | 46 | Sonata Flex28 | Sonata Standard | Sonnet | Sonnet | FSP | FSP | 2 | 2 |
P16 | 57 | 55 | 56 | Synchrony Flex28 | Synchrony Flex28 | Rondo | Rondo | FSP | FSP | 2 | 2 |
P17 | 55 | 52 | 48 | Sonata Standard | Sonata Flex28 | Opus2 | Opus2 | FSP | FSP | 2 | 1 |
P18 | 50 | 48 | 47 | Sonata Standard | Sonata FlexSoft | Rondo | Rondo | FSP | FSP | 2 | 2 |
P19 | 52 | 47 | 45 | Sonata Standard | Sonata Standard | Opus2 | Opus2 | FS4HR | FS4HR | 4 | 4 |
P20 | 70 | 63 | 65 | Sonata Flex28 | Sonata Standard | Opus2 | Opus2 | FSP | FSP | 2 | 2 |
P21 | 76 | 71 | 70 | Sonata Flex28 | Sonata Flex28 | Opus2 | Opus2 | FS4HR | FS4HR | 4 | 4 |
P22 | 26 | 18 | 24 | Sonata Flex28 | Sonata Standard | Sonnet | Sonnet | FS4HR | FS4HR | 4 | 4 |
P23 | 67 | 58 | 62 | Concerto FlexSoft | Sonata Standard | Opus2 | Sonnet | FSP | FSP | 2 | 2 |
P24 | 50 | 47 | 40 | Sonata Standard | Sonata Flex28 | Opus2 | Rondo | FS4HR | FS4HR | 4 | 4 |
P25 | 50 | 35 | 36 | Combi40+ Standard | Combi40+ Standard | Sonnet | Sonnet | CIS | CIS | 0 | 0 |
The fitting was performed under the currently applied standard procedures for bilateral CI programming, where each device is first fit independently. Later, narrow-band noises were presented in free field for right/left loudness balancing and CI users indicated their percept of the mid-sagittal plane. Note that all CI users were very experienced with the tested coding strategy and no major changes were done on their fitting for this particular experiment.
Eleven NH listeners (ages 24–37 years) were enrolled in the experiments as controls. All had normal thresholds (within 20 dB of audiometric zero), as determined by a standard pure-tone audiogram (ISO 8253-1:2010). Listeners had no visual and motor disorders and were naive about the purpose of the experiments.
Low-frequency sounds are coded both in place and time in the apical region of the cochlea. However, most implants only extract the envelope of the incoming signal for all electrodes (frequency bands), thus eliminating the fine-structure cues. In contrast, FSP developed by MED-EL modifies the timing of pulse stimulation to code temporal information in the low frequencies.
FSP coding is identical to the CIS-based stimulation (
Based on this initial stimulation protocol, the first version of the FS coding strategy was introduced in 2006. The number of apical channels with CSSS varied from one to three according to the fitting variables for a given CI user. Later, FS4 was developed, ensuring CSSS channels up for the first four apical channels (up to 1 kHz). Finally, FS4-P was released, which allows parallel time coding on the four apical CSSS channels. In our study, and as mentioned before, we adopted our own nomenclature to define the FS coding strategy referring to the bilateral number of CSSS or FS channels (see previous section).
Normally, in bilateral CI fittings, the audio processors are not synchronized with each other. However, the apical electrodes on the FSP strategy are locked to the zero crossing of the acoustic input. Since the acoustic input is highly correlated (or “synchronized”) between ears, theoretically, the pulses can be delivered preserving ITDs. However, the sampling rate is the limiting factor for a proper zero-crossing representation. For the CSSS channels, it is typically situated between 3 and 10 kHz, which can lead to temporal accuracies of 0–333 and 0–100 μs, respectively (
Psychometric experiments were used to measure the ILD and ITD sensitivity of the listeners, together with their potential side bias. The stimuli were generated in MATLAB (The MathWorks Inc., Natick, MA, USA), played through an external sound card (MOTU Ultralite, Cambridge, MA). For bilateral CI users, the stimuli were delivered through the CI processor's audio input (only bypassing their microphones), and NH was tested
For ILD sensitivity testing, the level between the two inputs was changed, while the ITD was fixed at 0 s. During the test, the acoustic power of the signal was kept constant, maintaining the same overall loudness for all ILD magnitudes. This task was evaluated with two frequency ranges: low-pass (LP, 0.15–0.8 kHz), generally covering the four apical channels, and high-pass (HP, 1.5–10 kHz), where medial and basal electrodes are stimulated (electrodes 5–12). The applied ILDs for the BICI users ranged from +10 to −10 dB and were divided into 16 equal steps of 1.25 dB. Each ILD was randomly tested 15 times along the experiment, making a total of 240 trials. The NH listeners were tested over a narrower ILD range between ±5 dB (divided into 16 equal steps of 0.625 dB); in addition, the +10- and −10-dB sounds were presented to obtain the two extreme data points. Also, each ILD was evaluated 15 times, making a total of 270 trials for this group.
For ITD testing, the ILD was kept constant at 0 dB, while the onset time between the ears was systematically varied. For this task, LP (0.15–0.8 kHz) noise bursts were presented. ITDs were varied between ±2 ms for BICI users (in 16 equal 0.25-ms steps) and ±0.2 ms for NH listeners (in 16 equal 0.025-ms steps), together with +0.8 and −0.8 ms as the extreme data points. The same amount of trials as for the ILD task was performed.
To describe the psychophysical ILD and ITD data, we performed a sigmoid fit over the binary left/right responses with the following logistic function (
with
Example of an interaural level difference (ILD) psychometric curve, described by Equation (1), with its two parameters indicated. The
To interpret the ILD and ITD sensitivity, we extracted physiological ranges from a 5° horizontal resolution impulse response library (
Sound localization performance was tested for broadband (0.15–10 kHz), high-pass (1.5–10 kHz), and low-pass (0.15–0.8 kHz) noise bursts of 150 ms. As in the psychophysical tasks, the frequency ranges were selected to cover apical (LP; channels 1–4), medial-basal (HP; channels 5–12), and the complete electrode array stimulation (BB; channels 1–12). Sound levels were presented at 50, 60, and 70 dBA and target locations were distributed over the two-dimensional frontal space, between ±75° in azimuth and ±30° in elevation. Stimuli were presented in a dark, anechoic room as described by Van Bentum et al. (
Each sound localization trial started with the presentation of a green fixation LED at straight ahead (0° azimuth and 0° elevation). Using a head-fixed laser pointer, the subjects were instructed to align the laser dot with the fixation LED to ensure the same head orientation at the start of each trial. After the subject pressed a button, the fixation light was turned off within 100–300 ms, followed by the target sound with a 200-ms delay. The subjects were asked to point the laser dot as fast and accurately as possible toward the perceived sound location. The acquisition time of the head movement was 2.5 s.
Head movements were detected automatically from the calibrated head position signals using a custom-made Matlab script that checked for head velocities exceeding 20°/s. Onset and offset of the head movements were detected by the program and visually checked off-line.
The target–response relationship of the BICI users was modeled with a sigmoid (
α
The limits of the response range (in degrees) are determined by the asymptotes of the fit and are referred to as βleft =
Sound localization fitting example as described by Equation (3). The sigmoid is centered around the target where the curve is equidistant from the two asymptotes (in the example,
To obtain an overall measure for the response accuracy, we also computed the mean absolute error (MAE) across trials, according to:
with α
For sound localization analysis, separate
Means and the 95% confidence intervals of the dependent variables are also reported in the
To determine the sensitivity of ILDs and ITDs, we use ω as a measure of cue sensitivity and θ to quantify the right/left bias (see
Interaural level difference (ILD,
To illustrate the data analysis and responses for individual cases, we show the results from five representative BICI listeners with zero, two, and four FSP channels (
In contrast, the ILD sensitivity for the BICI users and the NH example, quantified by ωILD, were more comparable. While P19 performed more poorly with a sensitivity of ωILD = 31.8 dB for HP and ωILD = 30.4 dB for LP, listener P2 yielded ωILD = 7.9 dB and ωILD = 5.8 dB for HP and LP sounds, respectively. These examples illustrate the large variability across listeners, but indicate also that most of them had a well-defined sensitivity to this cue. The threshold value, θILD, which characterizes the balance between the right and left ears, varied across listeners. Note that the LP value for θILD is correlated to θITD as the ITD was measured at 0 dB ILD for the same frequency range for all CI listeners (
An overview of the sensitivity (ω) and bias (θ) for all sound types and subgroups is provided in
Overall bias (θ) and sensitivity (ω) of interaural level difference (ILD) high-pass
Localization performance in the free anechoic field was tested with BB, HP, and LP sounds presented in pseudorandom order at one of the three presentation levels (50, 60, or 70 dB SPL) in the two-dimensional frontal hemifield. The presentation levels did not affect the response gain, γ (
To illustrate the overall type of responses from the BICI listeners and the sigmoid fit analysis,
Target–response relationship of horizontal sound localization for broadband (BB), high-pass (HP), and low-pass (LP) stimuli. Five examples of BICI listeners
Listener P2 is an extreme example of sound lateralization as the responses were directed to the far left and far right, irrespective of the stimulus type presented (
Listeners P11 and P19 showed similar systematic stimulus–response relations for azimuth than did P1. In both cases, the central range (±40°) showed an almost linear target–response relationship, but saturating at the edges. Interestingly, both listeners also yielded systematic localization responses for the LP stimuli, with small MAEs and near-normal gains, albeit with a reduced response range (ΔLOC = ~90–95°;
To quantify the sound localization performance of all listeners,
BICI sound localization gain (γ) compared to the response compression (
Target (
As an overall measure for the localization performance of the bilateral CI groups, we computed the MAE for the different sounds (
Mean absolute error of sound localization performance for all frequency ranges.
The left/right bias was quantified for both psychophysical experiments (as θ for the ILD/ITD tasks and
Relation between the interaural level difference (ILD) (θILD) and the target sound localization bias (
As shown in the presented study, many bilateral CI users have a remarkably good localization performance, which is mainly attributed to adequate ILD processing. However, we also provided evidence that some listeners may have had access to rudimentary ITD information with but also without the FSP strategy. Furthermore, there was a large variability in performance, which so far remains unexplained. Based on our results, we argue that our data suggest that CI users may learn to successfully integrate even rudimentary binaural information.
The bilateral CI listeners in our study were mostly sensitive to ILDs (
However, several studies have demonstrated a life-long plasticity in the human auditory system, which might help CI listeners with post-lingual deafness. Yet, the localization cues provided by the CIs should be unique and consistent for any source location as inconsistent and ambiguous cues cannot induce successful perceptual learning (
Lack of ITD perception may also be related to the neural health of the cochlea. In other words, potentially, ITD perception may be reached in CI users with substantial neural substrate of the cochlear and spiral ganglion at the lower-frequency region. In addition, ITD perception may also be influenced by auditory central processing capabilities. In the presence of residual hearing at lower frequencies, the residual function of the peripheral neural substrate can be assessed objectively by measuring auditory steady-state response (ASSR) on each side, and central processing skills may be assessed using binaural masking level difference (BMLD) as an objective measure of binaural cue integration. Due to the lack of residual hearing in our subjects, these measurements would not have been possible to perform in this study group. However, since at present candidates often have considerable residual hearing pre-implantation, future prospective studies may include psycho-acoustic and objective measures to assess the auditory pathway prior to bilateral implantation. This would expand insights into the variables that determine spatial hearing.
The overall sound localization performance is in accordance with other results reported for bilateral CI listeners (
The deprived ITD sensitivity and the poor ILD representation might underlie this impoverished sound localization performance. Moreover, the weakness of the low-frequency free-field cues seems to be reflected in spatially compressed localization responses (ΔLOC <180°;
In bilateral CI users, asymmetry is quite common due to a mismatch of the bilateral electrode's position or a difference with regard to the time of implantation between the two sides. While one might argue that these asymmetrical factors might affect the listener's spatial hearing performance, our data did not show any strong supportive correlation for this association. Although the exact position of the electrode array and its insertion angle was not measured, since the same electrode design and length were used on each side and with reported full insertion of all electrodes, a comparable insertion angle between both ears is assumed. However, the variation in response bias for the ILD task and the sound localization task might indicate that bilateral CI fitting was not always well-balanced between ears (
Clearly, one might not expect that CI listeners, equipped with a restricted number of frequency channels and a highly limited dynamic range, can approach similar spatial resolution and localization performance as normal-hearing listeners, who can precisely process the encoded information from over 3,000 channels over a huge dynamic range. Yet, binaural integration in bilateral CI recipients might be further improved in the future with optimized bilateral encoding strategies that allow a better synchronization between the two devices, an optimized spectral overlap, and a reliably balanced loudness perception. Moreover, training for spatial hearing is normally not part of CI standard of care, but should be considered as this rehabilitation approach might improve spatial hearing skills in CI users. This way, CI users may truly exploit spatial auditory cues and might map them into a veridical representation of the acoustic environment.
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
The studies involving human participants were reviewed and approved by ENT clinic of St. Elisabeth-Hospital of the Ruhr-University in Bochum, Germany. The patients/participants provided their written informed consent to participate in this study.
SA, MA, and SB designed and performed the experiments. SA analyzed the data and wrote the paper. AE, CV, SB, SD, and JT supported the data collection and provided critical revision of the paper. AV, AS, MA, and EM supervised the findings and wrote the final manuscript. EM was the initiator of this collaborative study. All authors contributed to the article and approved the submitted version.
SB is an employee of MED-EL. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Günter Windau, Stijn Martens, and Ruurd Lof for their valuable technical assistance. We are also grateful to Martin Wozniak and Florian Krieger for their valuable assistance and help on part of the data collection.