Cochlear Implant Electrode Placement and Music Perception

Key Points

Question

What is the association of electrode placement and image-guided cochlear implant programming with cochlear implant–aided music perception?

Findings

In this cross-sectional study of 50 adults with cochlear implants, electrode placement was associated with music perception outcomes. Participants also experienced improvements in music perception with the image-guided cochlear implant programming strategy.

Meaning

The study results suggest that image-guided cochlear implant programming can help improve music perception with cochlear implants.

Abstract

Importance

Cochlear implants enable improvements in speech perception, but music perception outcomes remain variable. Image-guided cochlear implant programming has emerged as a potential programming strategy for increasing the quality of spectral information delivered through the cochlear implant to improve outcomes.

Objectives

To perform 2 experiments, the first of which modeled the variance in music perception scores as a function of electrode positioning factors, and the second of which evaluated image-guided cochlear implant programming as a strategy to improve music perception with a cochlear implant.

Design, Setting, and Participants

This single-center, prospective study recruited 50 adult patients with at least 6 months of cochlear implant listening experience and normal cochlear anatomy to participate in experiment 1 from 2013 to 2023. Data analysis was conducted from January to February 2024. Thirty-four of the 50 patients from experiment 1 also completed experiment 2.

Interventions

Cochlear implant programming using a computed tomography–guided electrode selection strategy.

Main Outcomes and Measures

University of Washington Clinical Assessment of Music score, including subtests of pitch discrimination thresholds, isochronous familiar melody recognition, and timbre recognition.

Results

Of 50 participants, 20 (40%) were female, and the mean (SD) age was 57.7 (16.4) years. Experiment 1 suggested that better music perception abilities in the 50 participants were associated with patients who were younger and had a postlingual onset of deafness, as well as electrode arrays with a full scala tympani insertion, higher modiolar distance, and shallower insertion depth. Experiment 2 suggested improvements in melody recognition in the 34 participants using the image-guided cochlear implant programming strategy. Patients with apical electrodes that were deactivated were more likely to demonstrate an improvement in their pitch perception thresholds with the image-guided strategy, likely due to the low-frequency stimuli used in the University of Washington Clinical Assessment of Music.

Conclusions and Relevance

This study identified patient and device factors that were associated with music perception outcomes with a cochlear implant. These findings suggest that a personalized, image-guided approach to programming may improve music perception abilities for patients with cochlear implants.

This cross-sectional study examines the association of electrode placement and image-guided cochlear implant programming with music perception in participants with a cochlear implant.

Introduction

Cochlear implants enable improvements in speech understanding and quality of life for individuals with moderate to profound sensorineural hearing loss.^1,2 However, listeners with cochlear impants are often challenged by more complex listening tasks, such as music perception. For example, cochlear implant listeners exhibit poorer pitch discrimination thresholds (cochlear implant, 3 semitones; individual with normal hearing [NH], <1 semitone), isochronous melody recognition (cochlear implant, 25.1%; NH, 87.5%), and timbre recognition (cochlear implant, 45.3%; NH, 94.2%) than their peers with NH, as assessed by the University of Washington Clinical Assessment of Music Perception (UW-CAMP).^3,4 Further, many listeners with cochlear implants report poor musical sound quality with their cochlear implant, which leads to many of them discontinuing interacting with music altogether.^5,6 This is alarming given the known association between music and quality of life outcomes for listeners with cochlear implants.^7,8,9 These deficits in music perception and sound quality are associated, at least partially, with the poor spectral resolution provided by the cochlear implant.^10,11,12,13

Poor spectral resolution is attributable to having a discrete number of intracochlear electrodes. This is problematic because electrical current spreads within the cochlea and limits the number of independent neural populations that can be stimulated.^14,15 Further, while some cochlear implant recipients can use up to 16 of 22 independent channels using an n-of-m strategy for speech,¹⁵ music requires at least 64 channels, many more than are available with current devices.¹⁶ Thus, it is crucial to develop programming strategies that improve music perception by enhancing access to and the quality of spectral information delivered through the cochlear implant.

Individualized electrode selection is a reprogramming strategy for improving spectral resolution that involves identifying electrodes with overlapping stimulation (ie, channel interaction) using perceptual or physiological metrics and deactivating those electrodes in a patient’s cochlear implant map. Several studies have demonstrated benefits in auditory and speech perception using metrics of absolute thresholds for broad vs focused stimulation,^17,18 masked modulation detection thresholds,^19,20,21 or computed tomography (CT)–based anatomical measurements.^{22,23,24,25,26}

Growing evidence has suggested that electrode positioning factors, such as electrode type, scalar location, electrode-to-modiolus distance, and angular insertion depth, are associated with cochlear implant outcomes, at least in the speech domain.^{27,28,29,30,31,32,33} The most significant positional predictors of speech outcomes for precurved arrays are full scala tympani (ST) insertion and lower modiolar distance, whereas a deeper angular insertion depth was found to be the most significant predictor of speech outcomes for straight arrays.³⁴ Therefore, electrode selection methods that incorporate electrode placement information, such as an image-guided cochlear implant programming (IGCIP) strategy,²³ may be advantageous vs psychophysical methods alone.

However, the association of electrode placement with music perception is unknown. Given that music requires more independent channels than speech, further investigation is necessary to determine the association of electrode placement and electrode selection with cochlear implant–aided music perception. The current study used 2 experiments to further explore these issues. Experiment 1 modeled the variance in music perception scores as a function of electrode positioning factors, with the hypothesis that electrode arrays with closer modiolar proximity and a full ST insertion would be associated with music perception outcomes, given that these factors have been associated with greater channel independence.¹⁵ Experiment 2 investigated the association of IGCIP with music perception, with the hypothesis that implementing IGCIP would be associated with meaningful improvements in music perception as assessed by the UW-CAMP subtests of pitch, melody, and timbre.

Methods

Experiment 1

All participants provided written informed consent according to the protocol approved by the Vanderbilt institutional review board. Fifty adult cochlear implant users participated. Inclusion criteria required participants to have at least 6 months of cochlear implant listening experience and normal cochlear anatomy. While musical experience was not explicitly measured, participants had no more than amateur-level musical experience.

Electrode placement information was determined by CT imaging using validated cochlear implant position analysis algorithms.²³ Specifically, we measured 1 continuous modiolar proximity variable (mean distance from intracochlear electrodes to the modiolus), 1 categorical scalar location variable (whether the implant was fully positioned in the ST with no electrodes in scala vestibuli), and 1 continuous insertion depth variable (angular depth of the tip of the array). These positional measurements were based off the association between the postsurgical electrode position and the intracochlear structures, independent of the actual number of electrodes or interelectrode spacing. Therefore, these measurements could be generically applied across array types. Besides positional factors, we also controlled for demographic factors, including age at implant, sex, onset of deafness, length of cochlear implant use, and datalogging.

The outcome measures used were the 3 subtests in the UW-CAMP.³ The pitch discrimination subtest used a 2-alternative, forced choice, 1-up 1-down adaptive testing method.³ The stimuli included complex tones from a synthetic piano with identical spectral envelopes. Three base frequencies were tested, including 262 Hz (C4), 330 Hz (E4), and 392 Hz (G4), and a threshold in semitones was provided for each base frequency. The melody subtest included 12 commonly known melodies created in the octave surrounding and higher than 262 Hz (C4) and presented in an isochronous manner. Each melody was presented in a random order 3 times, and an overall score in percentage correct was provided. The timbre subtest included 8 musical instruments that each played a 5-note sequence encompassing the octave higher than 262 Hz (C4). Each instrument was presented in a random order 3 times, and an overall score in percentage correct was provided. All 3 subtests were presented at a 70-dB sound pressure level using a loudspeaker at 0 degrees azimuth and 1 meter from the participant in a single-walled sound booth. All analyses were completed in the R statistical computing environment, version 4.3.3 (R Foundation).³⁵

Experiment 2

The sample size for experiment 2 was estimated based on the mean degree of IGCIP benefit and standard deviation from our previous work in the speech domain.^23,24 With each participant acting as their own control (baseline vs IGCIP) and the assumption that the response within each group was normally distributed with a standard deviation equal to 20, and if the true difference was 10 percentage points, we estimated that we would need 33 participants to be able to reject the null hypothesis that the population means of the 2 programming strategies were not significantly different with 80% power and P < .05.³⁶ Given this, 34 of the 50 participants from experiment 1 (68%) also participated in experiment 2. All participants provided written informed consent according to the protocol approved by the Vanderbilt institutional review board. Like experiment 1, inclusion criteria required that the participants have a minimum of 6 months of cochlear implant–aided listening experience before participation. A Shapiro-Wilk test was completed and showed that the distribution of scores departed from normality for all UW-CAMP subtests. Therefore, a Wilcoxon matched-pairs signed rank test was used, and the median with IQR was used to summarize the data for each subtest. The 95% CIs of the median for ear Wilcoxon matched-pairs signed rank test were also reported.

Programming the IGCIP condition was completed using our previously published method of automated CT image analysis and processing.²² Electrode deactivation plans were determined by estimating a geometric-based spread of excitation for each electrode by determining the spatial relationship between each electrode and the modiolus in association with the adjacent electrodes. The electrode deactivation plan aimed to retain the largest number of independent active electrodes (≥8) to reduce channel interaction.²² This individualized IGCIP electrode deactivation plan was created using the US Food and Drug Administration–approved software for each cochlear implant manufacturer; thus, it did not require any experimental signal processing changes. IGCIP deactivation plans across the participants included electrodes across the array, with apical electrodes deactivated in 26 patients (76%), middle electrodes deactivated in 29 patients (84%), and basal electrodes in 24 patients (71%). This distribution of deactivations aimed to reduce channel interactions across the entire frequency range, potentially improving spectral information for music perception. See eTable in Supplement 1 for each experiment 2 participant’s IGCIP deactivation plan. Following electrode deactivation, the frequency allocation table was automatically reallocated to the remaining active electrodes. All other parameters remained unchanged from the baseline program. All participants completed the UW-CAMP in the baseline and IGCIP conditions using the same set up described previously in experiment 1. Participants were tested using IGCIP after 2 to 4 weeks of listening experience.

Results

Experiment 1

Table 1 provides participant demographic information and a summary of factors included in the general linear model. We conducted Pearson correlations with 95% CIs to assess the independent associations between the UW-CAMP subtests and the demographic and electrode placement factors. Table 2 shows the Pearson correlation coefficients for factors. For the pitch subtest, better pitch perception (ie, lower thresholds) was associated with full ST placement (r = 0.42; 95% CI, 0.16-0.62) and more hours of use per day (r = −0.35, 95% CI, −0.61 to −0.01). For the melody subtest, better familiar melody recognition (ie, higher percentage correct) was associated with having a straight electrode array (r = 0.34; 95% CI, 0.07-0.57) and a larger mean modiolar distance (r = 0.30; 95% CI, 0.02-0.53). For the timbre subtest, better instrument identification (ie, higher percentage correct) was associated with being female (r = −0.34; 95% CI, −0.57 to −0.07) and having a straight array (r = 0.32; 95% CI, 0.04-0.55) and a larger mean modiolar distance (r = 0.39; 95% CI, 0.12-0.60).

Table 1. Participant Demographic Information.

UW-CAMP scores	Mean (SD) [95% CI]
Pitch	3.6 (2.4) [0.50-9.1]
Melody	22.1 (19.4) [2.78-88.9]
Timbre	40.7 (17.4) [12.5-79.2]
Sex, No. (%)
Female	20 (40)
Male	30 (60)
Age at test, y	57.7 (16.4) [21-85]
Onset of deafness, No. (%)
Prelingual	8 (16)
Postlingual	42 (84)
Manufacturer, No. (%)
Advanced Bionics	11 (22)
Cochlear	27 (54)
MED-EL	12 (24)
Array type, No. (%)
Precurved	25 (50)
Straight	25 (5)
Scalar location, No. (%)
ST	24 (48)
ST-SV or SV	26 (52)
Modiolar distance, mm	0.79 (0.36) [0.14-1.44]
Angular insertion depth, °	436.8 (117.4) [251.9-716.8]
Length of cochlear implant use, mo	45.0 (39.5) [6-175]
Datalogging	12.6 (3.9) [0-18]

Abbreviations: ST, scala tympani; SV, translocated insertion; UW-CAMP, University of Washington Clinical Assessment of Music.

Table 2. Pearson Correlation Coefficients Between Factors and University of Washington Clinical Assessment of Music Subtest Scores.

Characteristica	R (95% CI)
	Pitch	Melody	Timbre
Sex	−0.05 (−0.33 to 0.23)	−0.19 (−0.44 to 0.10)	−0.34 (−0.57 to −0.07)
Age	−0.10 (−0.37 to 0.18))	−0.02 (−0.29 to 0.26)	−0.19 (−0.44 to 0.10)
Onset of deafness	−0.19 (−0.44 to 0.10	0.16 (−0.13 to 0.42)	−0.17 (−0.31 to 0.24)
Manufacturer	0.13 (−0.16 to 0.39)	−0.11 (−0.38 to 0.17)	0.08 (−0.20 to 0.35)
Array type	−0.20 (−0.45 to 0.83)	0.34 (0.07 to 0.57)	0.32 (0.04 to 0.55)
Scalar location	0.42 (0.16 to 0.62)	−0.26 (−0.50 to 0.02)	−0.22 (−0.47 to 0.06)
Mean modiolar distance	−0.24 (−0.49 to 0.04)	0.30 (0.02 to 0.53)	0.39 (0.12 to 0.60)
Angular insertion depth	−0.14 (−0.40 to 0.15))	0.06 (−0.22 to 0.34)	0.05 (−0.24 to 0.32)
Length of cochlear implant use	0.17 (−0.12 to 0.43	−0.15 (−0.42 to 0.13)	−0.04 (−0.32 to 0.24)
Datalogging	−0.35 (−0.61 to −0.12)	0.02 (−0.33 to 0.35)	0.15 (−0.20 to 0.46)

^a Sex was categorized as female or male, age is shown in years, onset of deafness is categorized as prelingual or postlingual, manufacturer is categorized by Advanced Bionics, Cochlear, or MED-EL, array type is categorized as precurved or straight, scalar location is categorized as full scala tympani insertion or translocated insertion, mean modiolar distance is shown in mm, angular insertion depth is shown in degrees, length of cochlear implant use is shown in months, and datalogging is shown in hours per day.

To determine what factors help the prediction of cochlear implant–aided music perception scores, we used linear mixed-effects models with stepwise regression to model the most meaningful factors associated with each of the UW-CAMP subtests. Participant was included as a random effect in each model. We used the MASS package to create the models, which selects the model with the lowest Akaike information criterion.³⁷ Separate models were run for each subtest, including pitch, melody, and timbre, as shown in Table 3. Any variables that exceeded moderate correlation (>0.50) with another factor or did not pass the multicollinearity test (variance inflation factor >10) were removed from the models (sex and electrode array). Unfortunately, we were missing data for the datalogging factor, as not all participants’ devices recorded this information at the time of the study. Missing data were handled using listwise deletion.

Table 3. Model Specifications for Predicting Cochlear Implant–Aided Music Perception^a.

Characteristic	Estimate (SE)	df	t Value (95% CI)
Pitch
(Intercept)	4.82 st (1.29)	29	3.73 (2.4 to 7.3)
Scalar location (ST vs ST-SV/SV)	2.50 st (0.63)	29	3.95 (1.3 to 3.7)
Mean modiolar distance, mm	−1.54 st (0.90)	29	−1.71 (−3.2 to 0.16)
Length of cochlear implant use, mo	0.01 st (0.01)	29	1.63 (−0.002 to 0.03)
Datalogging, h	−0.15 (0.08)	29	−1.85 (−0.3 to 0.003)
Melody, %
(Intercept)	68.66 (24.85)	27	2.76 (23.2 to 114.1)
Age, y	−0.69 (0.28)	27	−2.42 (−1.2 to −0.2)
Onset of deafness (prelingual vs postlingual)	32.43 (11.05)	27	2.94 (12.2 to 52.6)
Scalar location (ST vs ST-SV/SV)	−9.94 (6.56)	27	−1.52 (−21.9 to 2.1)
Mean modiolar distance, mm	26.60 (9.95)	27	2.67 (8.4 to 44.8)
Angular insertion depth, °	−0.06 (0.04)	27	−1.43 (−0.1 to 0.02)
Datalogging, h	−1.38 (0.91)	27	−1.52 (−3.0 to 0.3)
Timbre, %
(Intercept)	87.10 (17.23)	26	5.05 (56.1 to 118.1)
Age, y	−0.65 (0.21)	26	−3.06 (−1.0 to −0.3)
Onset of deafness (prelingual vs postlingual)	13.69 (8.60)	26	1.59 (−1.8 to 29.2)
Manufacturer (A vs C)	11.21 (6.67)	26	1.68 (0.8 to 23.2)
Manufacturer (A vs M)	20.75 (10.31)	26	2.01 (2.2 to 39.3)
Scalar location (ST vs ST-SV/SV)	−13.50 (4.78)	26	−2.82 (−22.1 to −4.9)
Mean modiolar distance, mm	20.33 (7.55)	26	2.69 (6.8 to 33.9)
Angular insertion depth, °	−0.08 (0.037)	26	−2.18 (−0.2 to −0.01)

Abbreviations: A, Advanced Bionics; C, Cochlear; M, MED-EL; st, semitones; ST, scala tympani insertion; SV, scala vestibuli.

^a For pitch perception, only a full ST insertion was associated with lower (ie, better) pitch perception thresholds. For melody recognition, only younger age, postlingual onset of deafness, and a higher mean modiolar distance was associated with better melody recognition. For timbre recognition, only younger age, a full ST insertion, a higher mean modiolar distance, and a shallower insertion depth was associated with better timbre recognition.

Experiment 2

Figure 1A displays median thresholds in semitones for pitch discrimination at each of the 3 base frequencies on the UW-CAMP. Error bars represent the 95% CIs. A lower threshold represents better pitch discrimination abilities and thus a lower threshold (or just noticeable difference). Scores obtained at baseline are represented by dark blue bars and scores following 2 to 4 weeks with IGCIP are represented by light blue bars. Median (IQR) thresholds for the 3 base frequencies were 2.9 (3.4), 4.4 (4.8), and 3.9 (4.6) semitones at baseline and 2.8 (3.0), 3.1 (4.4), and 3.0 (3.8) semitones following IGCIP.

Figure 1. Music Perception With the Baseline and Image-Guided Cochlear Implant Programming (IGCIP) Maps.

Means and standard error measures in the baseline (dark blue bars) and IGCIP (light blue bars) conditions for pitch discrimination thresholds in semitones (A) and melody and timbre recognition in percentage correct (B). Error bars represent 95% CIs.

Figure 1B displays median performance for melody and timbre recognition at baseline (dark blue bars) and following 2 to 4 weeks with IGCIP (light blue bars). Error bars represent the IQR. The horizontal lines display chance performance for the closed-set melody (8%) and timbre (12.5%) measures. Median (IQR) scores for melody and timbre recognition were 15.3% (11.1%) and 37.5% (24.0%) at baseline and 16.7 (13.2) and 29.2 (16.7) following IGCIP.

A series of Wilcoxon tests were completed and found a difference between the 2 conditions for the melody subtest (W =  91; 95% CI, −9.7 to −0.01) but not the 262-Hz (W = 364; 95% CI, −0.4 to 1.8), 330-Hz (W = 380; 95% CI, −0.08 to 1.7), or 392-Hz frequency discrimination (W = 273; 95% CI, −1.0 to 0.9), nor the timbre subtest (W = 306; 95% CI, −4.2 to 8.3).

Because the UW-CAMP is an assessment of primarily low-frequency stimuli (<500 Hz), the deactivation plans of the 34 participants were analyzed, and participants were stratified by whether or not they had an electrode deactivated in the region the frequencies tested for pitch discrimination. Of the 34 participants, 24 (71%) had an electrode deactivated in the region where the 3 base frequencies were represented. Figure 2 displays pitch discrimination thresholds with IGCIP as a function of thresholds at baseline. The diagonal solid line in each panel displays equivalent performance for each point. The diagonal dashed lines on either side of the equivalence line represent ± 1 semitone (or half step). For 262 Hz, 9 of the 24 participants (38%) who had an electrode deactivated in this region exhibited more than 1 semitone improvement in threshold, whereas just 1 of the 10 participants (10%) who did not have an electrode deactivated in this region displayed this magnitude of benefit. For 330 Hz, 7 of 24 participants (29%) with a deactivated electrode in this region exhibited more than 1 semitone improvement in threshold, and 3 of 10 (30%) without a deactivated electrode in this region displayed such benefit. For 392 Hz, 4 of 24 participants (17%) with a deactivated electrode in this region exhibited more than 1 semitone improvement in threshold, and just 1 of 10 (10%) without a deactivated electrode in this region displayed this magnitude of benefit.

Figure 2. Association of Apical Electrode Deactivations With Image-Guided Cochlear Implant Programming (IGCIP) Benefit for Pitch Perception.

Individual University of Washington Clinical Assessment of Music subtest scores at baseline and with IGCIP. The line of identity is shown by the dotted line, and the dashed lines represent the 95% CI for each subtest. Circles represent participants with at least 1 electrode among electrodes 1 to 3 deactivated in their IGCIP map. Upside down triangles represent participants with electrodes 1 to 3 active in their IGCIP map.

Discussion

Experiment 1

Electrode positioning factors, such as electrode type, scalar location, electrode-to-modiolus distance, and angular insertion depth, are associated with speech outcomes.^{27,28,29,30,31,32,33} However, it is important to investigate how these electrode positioning factors affect music perception to understand which characteristics are associated with better outcomes. Therefore, experiment 1 modeled the variance in music perception scores as a function of electrode positioning and demographic factors.

None of the factors investigated in this study predicted performance in all 3 UW-CAMP subtests, suggesting that these subtests measure unique domains of music perception. However, a full ST insertion and a higher mean modiolar distance were associated with better melody and timbre recognition. Additionally, a shallower insertion depth was also associated with better timbre recognition. A full ST insertion was consistent with our hypothesis and previous results with speech perception.^29,34 A higher mean modiolar distance contradicted our assumption that closer modiolar proximity would be associated with better music perception. However, it did corroborate our previous work, showing that a higher mean modiolar distance is associated with better subjective musical sound quality ratings.³⁸ In that study, this association was only significant for precurved arrays, which typically have shallower insertion depths than straight arrays. From these data, we postulated that a continuous signal, such as music, may benefit from an array positioned closer to the lateral wall to stimulate wider bands of neurons. In contrast, discrete signals, such as speech, may benefit from an array closer to the modiolus to stimulate discrete neurons with less channel interaction.

The current sample had a 52% translocation rate, which is much higher than the general cochlear implant population.^{27,28,29,30,39} Prior work has shown that translocation is much more common in precurved arrays and limits channel independence.^29,39 Therefore, it could be that the precurved array recipients, who typically have lower modiolar distances, are limited in their channel independence due to a higher than normal translocation rate. This could have resulted in poorer performance than straight electrode recipients, who typically have higher modiolar distances. The unusually high translocation rate in this sample may also help explain why shallower insertion depths were associated with better timbre recognition, as deeper insertions are often associated with higher rates of translocation, at least in precurved arrays, that could compromise spectral resolution.³⁹ Along with this, some straight electrode arrays included in this study, such as the FLEXsoft and FLEX28, afford much larger interelectrode spacing (2.1-2.4 mm) and potentially greater channel independence than shorter translocated precurved arrays, such as the CI632 (Cochlear; 0.7 mm).^39,40 To investigate this further, we conducted post hoc Pearson correlations to compare modiolar distance and performance, separating participants by electrode type and scalar location. However, the 95% CIs for these correlations included 0, likely due at least partially to the current sample being underpowered for these analyses. While this study examined electrode positioning factors across manufacturers, there were manufacturer-specific variations in electrode design and stimulation parameters that could have affected outcomes independently.

Among the demographic factors, younger age was associated with better melody and timbre recognition. This finding replicated previous work that showed that younger age was associated with better spectral resolution, which can improve outcomes on spectral-dependent tasks, such as speech understanding in noise and timbre perception.^41,42,43,44

For melody recognition, postlingual onset of deafness was associated with better performance. This association was unsurprising given that the melody subtest relies on previous experience with the familiar melodies. However, participants with prelingual onset of deafness likely did not have the same amount of prior experience with these melodies as those participants with a postlingual onset to help them perform the task. In general, the identified predictors of scalar location, age, and onset of deafness were similar to those previously found for speech outcomes³⁴; however, we were surprised to see that mean modiolar distance and angular insertion depth affected music and speech differently. Given these observed differences, we wanted to examine the perceptual consequences for music of manipulating channel interaction via IGCIP.

Experiment 2

There are known differences in channel independence and the effect of electrode positioning factors between speech and music.^23,24,25 Therefore, experiment 2 investigated the association of IGCIP with music perception. In support of our hypothesis, participants demonstrated improvements with IGCIP for melody perception. This suggests that a meaningful reduction in channel interaction was achieved with IGCIP. IGCIP also trended toward yielding better performance for pitch and timbre perception, but this was not meaningful at the group level. When apical electrodes were deactivated in the IGCIP map, participants were more likely to demonstrate improvements in pitch perception because the UW-CAMP pitch subtest stimuli were in this apical frequency region. These findings also corroborated previous work, showing that CT-based electrode selection can improve pitch perception.²⁶ However, future work should investigate whether pitch perception can be improved with IGCIP for higher frequencies (>500 Hz).

Generally, these results suggest that selectively deactivating electrodes thought to be causing substantial stimulation overlap may help improve music perception outcomes and outweigh the risk of reducing channel independence by lowering the total number of active electrodes. However, future work with a larger sample size should investigate the association of electrode positioning and demographic characteristics with IGCIP benefits for music perception to understand which patient profiles could benefit most from this intervention.

Limitations

This study had several limitations. First, the current sample had an unusually high translocation rate (52%) compared with the general cochlear implant population, which may have affected the associations observed between electrode positioning factors and music perception outcomes. Second, while this study examined electrode positioning factors across manufacturers, there were manufacturer-specific variations in electrode design and signal coding that could have affected outcomes independently. Future work should examine these associations within manufacturer-specific cohorts to better control for these variables, although this would require substantially larger sample sizes to achieve adequate statistical power. Additionally, we did not directly assess spectral resolution or auditory neural health in the current study, which could have provided insights into the mechanisms underlying the observed associations between electrode positioning and music perception outcomes. Finally, the relatively small sample size in experiment 2 (n = 34) may have limited our ability to detect smaller but potentially meaningful associations of IGCIP with pitch and timbre perception.

Conclusions

The results of this study suggest that electrode positioning factors have different associations with music and speech. Patients who were younger with postlingual onset of deafness and had a device with a full ST insertion, higher modiolar distance, and shallower angular insertion depth demonstrated better music perception abilities. Patients also demonstrated benefits for music perception with the IGCIP programming strategy. These findings have clinical implications for device selection and patient counseling. For patients who prioritize speech communication, such as elderly individuals with substantial communication difficulties, clinicians should continue to prioritize established factors that optimize speech outcomes (namely closer modiolar proximity for precurved arrays and deeper insertion depths for straight arrays). However, for patients with specific musical goals, straight electrode arrays might be considered, which typically have larger interelectrode spacing and higher modiolar distances. For most patients who desire speech and music perception, counseling should emphasize the primary goal of optimizing speech communication while setting realistic expectations about music perception limitations with current cochlear implant technology. Future work will focus on increasing access to IGCIP, collecting larger samples with estimates of peripheral (ie, auditory neural health) and central (ie, top-down processing) factors, integrating this strategy into clinical protocols, and developing electrode arrays and programming strategies that can optimize speech and music perception outcomes.

Supplement 1.

eTable

Supplement 2.

Data sharing statement