Novel Psychoacoustic Measure of Spectral-Temporal Processing for Discrete Regions of the Cochlear Array: A Proof-of-Concept Study

Abstract

Introduction

Despite the general success of cochlear implantation (CI), the impact of known pathophysiological processes occurring within the cochlea after surgery on hearing outcomes are poorly understood. Previous studies suggested localised intracochlear variations in the electrode-cochlear interface and neurosensory targets may be contributing to poor outcomes for some CI users. Research has highlighted the need to define optimal cochlear biomarkers and behavioural measures to enhance diagnostic and therapeutic approaches for CI users with poor outcomes. The current proof-of-concept study introduces a novel adaptation of a measure of spectral-temporal processing and aimed to explore its association with electrophysiological measures and speech perception.

Methods

Eighteen adult CI users with 24, 400, 500, and 600 series Cochlear Ltd devices were recruited from a large not-for-profit CI clinic in Australia. Participants completed a “sectional” adaptation of the spectral ripple STRIPES test (secSTRIPES) and bipolar transimpedances (4PI). Original STRIPES stimuli were band-pass filtered for secSTRIPES, targeting discrete high-frequency (basal) and low-frequency (apical) sections of the electrode array. A linear mixed model approach was used to determine if any relationship existed between localised impedance measures and secSTRIPES scores. Simple linear regression was used to investigate within-subject correlations and the association with conventional measures of speech understanding.

Results

The analysed data in this study indicated within-subject sectional variability, with a meaningful difference in secSTRIPES scores for 13 of 18 participants (scores different by >1). Most participants (72%) scored worse on the basal secSTRIPES test. This sectional bias was significant when controlling for age and impedances for 14 CI users with bilateral hearing loss (p = 0.005). There was no statistically significant relationship between 4PI and secSTRIPES scores; however, the results suggested there may be positive trend between localised impedance and spectral-temporal processing. The association of secSTRIPES and speech scores was also nonsignificant; however, positive trends between speech and apical STRIPES scores were observed.

Conclusions

This pilot study, the first to describe the sectional adaptation of the STRIPES test, provides sufficient proof of concept for secSTRIPES as a psychoacoustic measure of listening ability in localised sections of the electrode array. While results must be interpreted cautiously due to limitations of the sample size, preliminary findings reveal intracochlear variability in processing performance which warrants further investigation. Following formal validation, secSTRIPES may be beneficial for future research to explore how sectional biases influence overall processing for CI users and as a cochlear biomarker to guide interventions to improve CI performance.

Plain Language Summary

Cochlear implants (CIs) have had significant positive impact for many people with severe-to-profound hearing loss; however, some users still experience poor hearing outcomes. Previous research suggests that differences in the inner ear environment after surgery could play a role in how electrodes interact with the nerve cells. To improve outcomes, researchers need more precise ways to measure the internal environmental of the inner ear. This pilot study tested a new version of an existing listening test called STRIPES. The adapted test, secSTRIPES, measures specific sections of the CI, high-frequency (basal) and low-frequency (apical), rather than the whole array. Eighteen adult CI users took the secSTRIPES test along with conventional and new electrode impedance measures. Researchers looked for links between these measures and conventional speech test scores. Results showed that most participants (72%) performed worse on the basal section of the secSTRIPES test, and 13 out of 18 had noticeable differences between sections. This bias towards poorer basal performance was statistically significant for users with hearing loss in both ears. New impedance measures correlated better with secSTRIPES than with conventional impedance. While no strong statistical link was found between impedance and secSTRIPES scores or speech understanding, some positive trends appeared, especially between apical scores and speech. This study introduces secSTRIPES as a promising tool to measure listening ability in specific sections of the CI. These early findings suggest that physiology and listening performance varies along the cochlea, which could help guide future research and interventions to improve CI outcomes.

Introduction

Despite the success of the cochlear implant (CI) to restore hearing and communication, it is recognised that outcomes vary and are difficult to predict, with a small portion of recipients experiencing very poor or regressive speech outcomes [1–4]. There has been incremental progress in improving surgical techniques, optimising sound-processing and programming strategies, and refining rehabilitation. However, there is still limited understanding of how the morphology and physiology of the cochlea pre- and postimplantation influences CI function and outcomes. There is now broad recognition within the CI literature of the need for advanced biomarkers and diagnostic techniques to better understand the electrode-neural interface to support improvements in CI design and rehabilitation [5, 6]. Factors such as electrode placement, neural responsiveness, and the foreign body response are potentially modifiable with surgical techniques or pharmacological intervention and are now measurable with emerging electrophysiological and psychoacoustic tools.

Evidence of an immune response to implantation and subsequent development of fibrotic tissue around the array has been well documented in immunohistochemical imaging of cadaveric temporal bones from implant users [7, 8]. However, characterisation of this response in the living CI recipient has proved challenging. Using in vitro models, Newbold et al. [9] first demonstrated that biphasic electrode impedance increases could be used as a biomarker for protein absorption and fibrotic cell growth surrounding the electrode. These “conventional” impedance measures are routinely used in clinical settings to assess implant function [10]. Newer impedance measures, such as the transimpedance matrix (TIM), can be used to detect adverse events such as electrode tip fold-over and extracochlear electrodes [11]. By utilising an additional bipolar stimulation mode, a novel application of the TIM called four-point impedance (4PI) has been deployed in clinical research. 4PI uses four adjacent electrodes to quantify the resistance between a pair of inner electrodes while stimulating the two outer electrodes. By recording impedance for each set of four consecutive electrodes, the entire array yields 19, 4PI measurements (e.g., electrodes 1–4, 2–5, 3–6, etc.). In contrast, conventional monopolar and common ground impedance (CGI) stimulate and record from each of the 22 electrodes, with CGI using the remaining electrodes as the ground. As such, CGI is highly influenced by the electrode contacts and gross cochlear morphology. Bester et al. [12] demonstrated that unlike CGI, 4PI was sensitive to the tissue environment surrounding the target electrodes. They found elevated 4PI corresponded with suspected presence of red blood cells in the scala tympani, potentially because of surgical trauma to the vascular lateral wall. Increased 4PI also predicted loss of residual hearing and higher rates of postoperative vestibular dysfunction [13, 14]. Needham et al. [15] also established the correlation between 4PI and fibrosis formation using an animal model.

Another source of intra-cochlear variability is the pattern of survival and function of neurosensory targets. Histopathological studies have demonstrated that despite similar preoperative hearing thresholds between two ears of the same individual, there is greater degeneration of spiral ganglion (SGN) cell bodies and dendritic processes in the implanted cochlea when compared with the non-implanted contralateral ear [16, 17]. Posthumous studies of temporal bones have shown that SGN density varies greatly both between individual implanted ears and across different sections of the cochlea within the same individual [18, 19]. These publications have also highlighted that speech outcomes did not correlate with SGN survival. This points to the limitations of using measures of gross anatomy to predict CI outcomes but that measures of localised neural function may provide stronger predictive power for performance [20].

While advanced biomarkers are still under investigation, CI outcomes are predominantly assessed using speech discrimination scores from test material made up of speech sounds common in the CI user’s spoken language [21]. However, several limitations of speech tests have been identified such as the reliance on the recognition of learned phonemic patterns, translation of material across global clinics, and the constraints of ceiling effects [22]. As an alternative, language-independent measures of spectral-temporal processing (e.g., spectral-ripple discrimination tests) overcome many limitations of speech tests. Various spectral-ripple tasks have been developed and are now commonly reported in the CI research literature [23–26]. Examples such as the spectral-temporally modulated ripple test (SMRT) correlate well with speech performance while also avoiding the bias of learning or acclimatisation [27]. And yet, the analysis of the SMRT found that there may be an effect of additional loudness cues in the modulated stimuli, potentially diluting its effectiveness as a measure of spectral and temporal processing only [23].

An alternative spectral ripple test, the Spectro-Temporal Ripple for Investigating Processor EffectivenesS (STRIPES) test, was developed by Archer-Boyd et al. [28]. Designed to assess spectral-temporal processing for CI users, listeners must discriminate between stimuli based on the direction of the frequency trajectories over time. The STRIPES test demonstrated test-retest reliability across manufacturer and implant-type, processing strategies, presentation level and modality and in both research and clinical settings [29–31]. Furthermore, STRIPES scores are significantly positively correlated with clinically used speech-in-noise test scores [32]. While the SMRT also correlated with speech scores, in a separate study, only STRIPES successfully predicted speech-in-noise scores when comparing different mapping parameters for individual CI users [25].

Authors of the STRIPES procedure recognised that a listener’s overall STRIPES score may be dominated by their processing ability derived from a discrete frequency region of the electrode array as processing performance would be nonuniform along the array [28]. The flexibility of the language-independent broadband STRIPES stimuli to target specific regions of the electrode array and corresponding tonotopic cochlea presents an opportunity to investigate localised spectral-temporal processing. Modifications of the STRIPES test using sectioned stimuli have not been reported in the literature to date, nor has any comparison been made between advance impedance and STRIPES to better understand the relationship of sectional variation in the electrode-neural interface and processing ability. This study aimed to assess the feasibility of an adapted STRIPES test as a language-independent measure of spectral-temporal processing in localised sections of the cochlea and compare those outcomes with conventional and emerging measures of implant function and performance to provide a preliminary characterisation of the diagnostic relevance of sectional STRIPES.

Methods

Recruitment

Ethics approval was obtained from the Human Research Ethics Office, The University of Western Australia (Crawley, WA, Australia), approval: 2023/ET000135. Participants were recruited from the Ear Science Implant Clinic (Subiaco, WA, Australia). Adult clients with an upcoming review appointment were invited to participate if their clinical record indicated consent to be contacted for research purposes and had a Cochlear (Cochlear Ltd, Sydney, NSW, Australia) Nucleus^® 24, 400, 500, or 600 series implant for at least 3 months. A further requirement was that all 22 electrodes must have been active, as a requirement to complete the STRIPES test. Given the pilot nature of the study, sample size was determined by feasibility. A total of 18 adult individuals completed the research session (female = 8, male = 10). The mean age of participants was 69.4 years (range 35–91 years). Most participants had bilateral hearing loss (n = 15). Three participants had single-sided deafness (SSD) defined as having an implanted ear and normal hearing (3FAHL ≤20 dB HL) in the contralateral ear. A breakdown of the demographic, audiological, and implant profiles of each participant is given in Table 1.

Table 1.

Summary of participant demographic and audiological information

Participants	​	​	BHL	SSD
Total	n = 18	HL configuration	n = 15	n = 3
Female	n = 8
Male	n = 10
Age, years	​	3FAHL – TE (dB HL)	​	​
Mean (SD)	69.4 (13.3)	Mean (SD)	111.5 (11.9)	102.3 (16.6)
Range	35–91	Range	85–120	87–120
Aetiology	​	3FAHL – NTE (dB HL)	​	​
Acquired/progressive	n = 11	Mean (SD)	69.1 (34.2)	7 (5)
Congenital	n = 3	Range	27–120	2–12
Unknowna	n = 4	​	​	​
Duration of implantation	​	​	​	​
Mean (SD)	4.6 (4.6)	​	​	​
Range	0.5–19	​	​	​
Test ear	​	​	​	​
Right	n = 8	​	​	​
Left	n = 10	​	​	​
Implant type	​	​	​	​
24RE series	n = 3	​	​	​
400 series	n = 1	​	​	​
500 series	n = 4	​	​	​
600 series	n = 10	​	​	​

SD, standard deviation; HL, hearing loss; BHL, bilateral hearing loss; SSD, single-sided deafness; dB HL, decibels hearing loss; 3FAHL, three frequency average; TE, test ear; NTE, non-test ear.

^a Unknown: Aetiology not documented in available clinical record.

Research Sessions

Each participant completed one 60-min research session consisting of two tests: secSTRIPES and TIM. All tests were performed on one ear only and administered by a Elitebook laptop (HP Inc., CA, USA). For participants with bilateral implants, tests were performed with the better ear as determined by their most recent CNC word scores.

Sectional Adaptation of the Spectral Ripple STRIPES

A customised implementation of the STRIPES test, based on original code developed and shared by the Cambridge Hearing Group (Cambridge University, UK), was executed within MATLAB^® 2022a (MathWorks, MA, USA). The interactive STRIPES program generated sinusoidal sweep stimuli in an automated, adaptive-staircase (two-up, one-down) procedure, and participants recorded their responses via an on-screen interface. To present the STRIPES stimuli to the listener, a Cochlear Mini Microphone 2+ (Cochlear Ltd, Sydney, NSW, Australia) was connected to the laptop via a Linsoul Kiwi Ears Allegro Mini DAC (32 bit/384 kHz; USB-C to 3.5 mm audio jack). The Mini Microphone 2+ streamed to a research-designated Cochlear Nucleus^® CP1000 sound processor (Sydney, NSW, Australia), preprogrammed with the participant’s most recent clinical map. All optional speech-enhancing features were deactivated, and audio mix was selected as accessory only. Listening comfort and audibility was verified by demonstrating an up and down STRIPES stimulus, followed by a short familiarisation phase of the three-interval two-alternative forced choice task. In this study, the STRIPES implementation described by the original authors in Goehring et al. [25] was modified to a secSTRIPES test targeting discrete sections of the array. Each participant completed two runs of the STRIPES task using the custom-designed band-passed basal and apical, and secSTRIPES stimuli that stimulated only electrodes 2–10 and 14–22, respectively. Frequency specificity of the secSTRIPES stimuli was verified by the electrodogram output of the ACE processing strategy simulator [33]. The order of basal and apical tests was alternated for each participant. Spectral ripple density (RD) and listener response (correct/incorrect) were recorded by the STRIPES program for each trial, and trials continued until 10 reversals of the adaptive procedure were completed. The final STRIPES score (average RD of the last four reversals) was then generated for each stimulus type, with secSTRIPES scores reported as equivalent RD based on 5-octave bandwidth of the broadband STRIPES stimulus [28].

Impedance Measurements

Two impedance measures were collected. A TIM was measured by CustomSound^® EP 6.0 software (Cochlear Ltd, Sydney, NSW, Australia) with the bipolar TIM research add-on and an unassigned, research-designated Cochlear Nucleus^® CP900 sound processor (Cochlear Ltd, Sydney, NSW, Australia). The voltage per unit of stimulating current at each electrode (in kΩ) induced by a cathodic-anodic biphasic pulse (using CustomSound^® default settings) was automatically recorded across the 22-by-22 electrode matrix [34] at the 25-μS time point. The bipolar TIM measurements were extracted from Custom Sound EP 6.0. Conventional impedances were extracted from most recent clinical record of Custom Sound. CGI for electrodes corresponding to the secSTRIPES stimuli (basal: electrodes 2–10; apical: electrodes 14–22) were averaged to calculate the sectional impedance for each participant.

Statistical Analysis

Data analysis was completed using R Statistical Software (2024.12.1 Build 563). Sectional impedance and STRIPES data were compared by a nonparametric test (Mann-Whitney U test). Shapiro-Wilk tests were used to determine normality of the data. STRIPES scores were normally distribution in the apical section but deviated from normal distribution in the basal section (p = 0.001 and p = 0.198, respectively). A linear mixed effects model was used to further analyse any within-subject relationship between impedance measures and secSTRIPES scores. The model was repeated using both measures of impedance as the independent variable. Estimated marginal means were derived from the mixed model to visualise predicted values of STRIPES across impedance, while controlling for section. The relationship between STRIPES performance and conventional speech perception scores was also examined using simple linear regression. To support interpretation of the models’ and to project appropriate sample sizes for future studies, a simulation-based sensitivity analysis was conducted for the fixed effect of TIM using the fitted linear mixed effects model and linear regression.

Results

Electrodograms demonstrated effective band-pass filtered secSTRIPES that selectively stimulated basal or apical sections of the Nucleus^® array (Fig. 1). The intention was to design secSTRIPES so that the electrode specificity of the apical and basal stimuli was mutually exclusive. However, for both up (target) and down (reference) stimuli, there was some overlapping stimulation on electrode 11 and, to a much lesser extent, electrode 12. Up and down secSTRIPES demonstrated symmetrical stimulation patterns in the basal section. Asymmetric temporal resolution was observed in the most distal electrodes of the apical secSTRIPES response. secSTRIPES scores were obtained for all 18 participants, presented in Figure 2. Scores ranged from 1.65 to 10.10 RD in the basal section and 2.50–8.50 RD in the apical section. Basal and apical scores differed by >1 RD for 13 out of 18 listeners, which indicates a meaningful variation in spectral-temporal processing between sections of the array [29]. Of those 13 participants, 9 had poorer secSTRIPES scores in the basal section, and overall, 13 out of 18 participants had poorer scores in the basal section. Within-subject linear regression analysis indicated a moderate, positive correlation between basal and apical secSTRIPES scores that was just above the conventional threshold for significance (r² = 0.215, p = 0.053). Scores were subsequently analysed by separating participants with SSD from those with bilateral hearing (BHL: 3FAHL in the better ear >20 dB HL). It was suspected that the participants with SSD would have differences in underlying aetiology and SGN function which may influence secSTRIPES performance. secSTRIPES scores for both the basal and apical sections were better for the SSD group (n = 3) compared to the BHL group (n = 15), as shown in Figure 3. The difference in the median secSTRIPES scores between the SSD and BHL groups was statistically significant using the Mann-Whitney U nonparametric test (basal section: p = 0.044; apical section: p = 0.044). This difference should be considered exploratory, however, and may reflect individual differences rather than an inherent difference in performance in this small subgroup. Between-subject comparison of BHL participants indicated that there was a significant difference between the distribution of the secSTRIPES scores, with poorer median scores in the basal section compared with the apical section: median score (basal) = 2.90 RD, interquartile range (basal) = 1.48 RD, median score (apical) = 4.5 RD, interquartile range (apical) = 1.93 RD, p = 0.05, n = 15. When the within-subject correlation of secSTRIPES scores was repeated without data from SSD participants, no association was present between apical and basal scores (r² = 0.005, p = 0.796).

Fig. 1.

Electrodograms demonstrating band-pass secSTRIPES stimuli processed by the ACE processing strategy simulator (Swanson [33]). Plots on the left represent the target “upward” STRIPES stimulus; plots on the right represent the reference “down” STRIPES stimulus. For two different ripple densities (2 RD and 8 RD), the plots display the response derived from the basal and apical secSTRIPES stimuli (filtered to target electrodes 1–10 and 14–22, respectively). Both basal and apical secSTRIPES activated electrode 11 and, to a much lesser extent, electrode 12. However, sectional specificity of the novel stimuli is evident.

Fig. 2.

secSTRIPES scores for all participants (n = 18). Darker grey indicates scores using stimuli targeting the apical electrodes and light grey indicates scores form the basal electrodes. Participants with SSD had a 3FAHL in the better ear of <20 dB HL. All other participants had bilateral hearing loss. Participants marked with + had basal and apical scores different by >1 RD. RD, ripple density; SSD, single-sided deafness.

Fig. 3.

Distribution of secSTRIPES scores for two hearing loss groups. 3FAHL in better ear ≤25 dB HL. Black lines indicate statistically significance difference in distribution of secSTRIPES scores between groups. RD, ripple density; BHL, bilateral hearing loss; SSD, singled-sided deafness.

CGI and 4PI was collected from all but one participant, whose CI map indicated an open circuit at electrode 2, preventing the bipolar TIM recording. CGI and 4PI measurements from each of the remaining 17 participants are presented in Figure 4, along with the sample mean across all electrodes. Sectional impedances for each participant were calculated based on the average impedance across the apical and basal electrodes corresponding with the secSTRIPES stimuli. The median sectional impedances for each measurement were: basal CGI = 4,937 Ω; apical CGI = 3,931 Ω; basal 4PI = 404 Ω; and apical 4PI = 238 Ω. For both measurement techniques, the average sectional impedance was higher in the basal section compared with apical (4PI: p = 0.023, CGI: p = 0.001). Impedance data from one subject with elevated basal 4PI are indicated by the dashed line in both Figure 5a,b. The corresponding CGI values for the same subject’s basal electrodes were not elevated and showed no exceptional patterns. Excluding this participant from the dataset did not alter the overall distribution, which continued to show a bias towards higher basal impedance, with the difference remaining statistically significant (p (CGI) = 0.003; p (4PI) = 0.044; n = 16). Data from the individual with elevated 4PI was retained in subsequent analysis described below.

Fig. 4.

Distribution of the averaged common ground impedance (CGI) and four-point impedance (4PI) for each section for all participants where a bipolar TIM was measurable (17 of 18). Black bars indicate comparisons with statistical significance on Mann-Whitney U test. 4PI (basal ∼ apical): p = 0.023). CGI (basal ∼ apical): p = 0.001.

Fig. 5.

Common ground impedance (CGI; a) and four-point impedance (4PI; b) measures across the full electrode array, basal to apical. Electrodes ordered from base to apex. Light grey lines = individual data; black line = average across the sample. Dashed line represents the same individual participant across a, b.

Further analysis was conducted using a linear mixed model approach. Fixed effects were the participant age at the time of the test session, electrode array section, and sectional impedance, with participant ID as the random effect. Participants with SSD were excluded from the analysis due to the previously described differences in STRIPES performance. The initial model (illustrated in Fig. 6a) included the remaining participants who had obtained both CGI and 4PI measures (n = 14), using 4PI as the predictor variable. While the model indicated a trend towards a positive relationship between 4PI and STRIPES, the observation was not significant (p = 0.098). The model confirmed basal STRIPES scores were significantly lower than apical scores (p = 0.005) at a sample level. Despite validity of the elevated 4PI measure obtained from one participant, the analysis was repeated to test the sensitivity of the model to the outlier. The model indicated a weaker relationship between both 4PI and STRIPES scores without the outlier. When the elevated 4PI was included, a 1.08-RD increase in STRIPES score for every 1 kΩ increase in average sectional 4PI was observed (i.e., effect of sectional 4PI on secSTRIPES: +1.08 RD/kΩ [95% CI −0.10 to +2.27; p = 0.086], equivalent to +0.52 RD per 1 standard deviation [SD] of 4PI). A 0.95-RD increase in STRIPES score was observed when outlier was excluded. Evidence of poorer STRIPES in the basal section persisted in both models (Fig. 6a, c).

Fig. 6.

Linear mixed effect modelling (a, c, e) for sectional impedance and STRIPES data. The estimated marginal mean is also plotted, which demonstrated the predicted STRIPES score while controlling for age (b, d, f). a–d Sectional 4PI was used as the predictor variable, whereas e–f modelled sectional CGI. a, b, e, f Elevated basal 4PI data point (n = 14). c, d Excluded outlier (n = 13). 4PI, four-point impedance; CGI, common ground impedance; RD, ripple density.

The model was finally repeated with the CGI measure as a fixed effect (Fig. 6e). The alternative model indicated CGI was a much poorer predictor of STRIPES score, which increased by 0.22 RD per 1-kΩ increase in CGI, which was not significant (p = 0.51; n = 14). Figure 6b, d, f illustrates the estimated marginal means from the linear mixed effects model, showing STRIPES score as a function of impedance, while controlling for section. The results reflected the trends observed in the mixed model that were not statistically significant. Sensitivity analysis demonstrated that the pilot design (n = 14) had 80% power to detect an effect of ≥1.76 RD/kΩ. Using a prespecified smallest effect size of interest of 1 RD per SD of 4PI, sample-size projections suggested that approximately 11 participants would be sufficient to detect a clinically meaningful effect with 80% power. However, the effect observed in the current pilot study was smaller than this predefined threshold [35].

Simple linear regression models of within-subject trends in the STRIPES results and speech testing data are displayed in Figure 7. There were no significant trends observed between STRIPES performance and phoneme or word scores from the CNC word test in quiet in this sample (n = 15). When retaining data from the individual with elevated 4PI results, apical STRIPES and word scores showed some indication of a positive relationship, approaching significance (r² = 0.23, p = 0.073; Fig. 7a). Analysis also indicated no substantive association between basal STRIPES scores and word scores. For the association between CNC word scores and apical secSTRIPES performance, a sample size projection indicated that approximately 32 participants would be required to confirm this relationship with 80% power at α = 0.05.

Fig. 7.

Within-subject correlation of secSTRIPES scores and CNC word and phoneme scores, including the participant with elevated basal 4PIs (n = 15). a, c Speech scores and apical STRIPES were positively correlated, with a p value approaching significance, whereas b, d show no trend between speech scores and basal STRIPES scores in this sample size. RD, ripple density.

Discussion

Spectro-Temporal Ripple for Investigating Processor EffectivenesS

This is the first publication to describe a specific sectional adaptation of the STRIPES procedure using filtered stimuli to test discrete halves of a CI electrode array. All 18 participants successfully completed the secSTRIPES session. The range of secSTRIPES scores were similar to those presented in previous studies such as Archer-Boyd et al. [30] or Van Groesen et al. [36], and no individual score exceeded the estimated limit for spectral aliasing (2.5 RPO or 12.5 RD) [37].

The primary finding of the pilot study was the sectional variation observed between high-frequency and low-frequency secSTRIPES scores for individual listeners. The STRIPES test has previously been shown to be sensitive to changes in both spectral and temporal processing performance in CI users [28] and therefore the current study provides preliminary evidence that spectral-temporal processing is nonuniform between the two sections of the electrode array. Broadband STRIPES scores may be dominated by performance in a discrete region of the cochlea. The current study provides evidence of a sectional bias in processing performance and supports continued investigation regarding which section of the array has most influence on the overall score. It remains unclear whether regions of poor processing disproportionately affect overall performance or whether the score reflects the best-performing regions. At a group level, the analysis demonstrated high-frequency (basal) secSTRIPES scores were predictably poorer than low-frequency (apical) scores. This result may reflect a tendency for individuals with hearing loss to have more pronounced, chronic degeneration of neurosensory targets in the basal region of the cochlea relative to the apical region.

The study demonstrated band-pass filtered secSTRIPES stimuli activated the target electrodes as spectral density was adjusted. The temporal smearing observed on the most apical electrodes is presumed to be the result of the ACE processing strategy and may introduce an additional acoustic cue for the apical secSTRIPES trials. And yet, a comparison of broadband STRIPES performance across different manufacturers found STRIPES scores remained consistent regardless of processing strategies [30]. In earlier work, the STRIPES authors also noted that the electrodogram does not represent exact patterns of excitation given the influence of current spread that varies along the array [29]. Further research incorporating measures of current spread and comparing secSTRIPES performance across different processing strategies is warranted. This may help determine if the apparent advantage in apical secSTRIPES stimuli should be attributed to an unintended acoustic cue rather than true spectral-temporal processing ability.

Speech Understanding

The preliminary data demonstrated a positive, nonsignificant trend between secSTRIPES and speech scores. This is in line with previous work by van Groesen et al. [32] who reported a significant, moderate positive relationship between STRIPES and speech understanding. Bysouth-Young et al. [31] found a similar trend; however, the authors acknowledged there was no statistical significance, potentially due to the small sample size (n = 7). These two studies are not directly comparable to the results published here. Both compared STRIPES with speech-in-noise scores, while the current study analysed words-in-quiet. van Groesen et al. [32] also mentioned their cohort had high word scores (mean = 75%) and that the relationship between speech and STRIPES may be weaker when assessing CI users with poorer speech understanding such as in the current study (mean word score = 45%). This may partly explain differences in the reported strength of the speech-STRIPES relationship. Nonetheless, these complementary results provide support for STRIPES as a valid measure of the spectral-temporal processing relevant to understanding speech. There is a recognition that language-independent auditory tests may be a better alternative to conventional speech tests in evaluating incremental changes in hearing performance [38]. secSTRIPES may offer the potential for customised spectral-temporal discrimination tasks as an alternative to selecting one of a variety of phoneme, word, and sentence test material available.

Impedances

The current study also demonstrated localised impedances were generally higher in the basal region, which is consistent with the pattern of postoperative 4PI reported by Razmovski et al. [13]. Overall, poorer STRIPES and higher 4PI were observed in the basal half of the cochlea which aligns with previous research that linked excessively high basal impedance and poor CI performance [39]. However, the linear mixed model analysis also indicated a small positive association between secSTRIPES scores and 4PI. Despite the absence of statistical significance, this novel finding may suggest that when controlling for section, spectral-temporal processing may improve in regions of the array where fibrotic tissue is presumed to be present [8]. The observed effect was smaller than the effect size that this pilot study was able to detect reliably, suggesting that the nonsignificant finding may reflect the limited size of the study rather than the absence of an association. Furthermore, the extent to which the 4PI acts as a composite marker or a measure dominated by localised tissue morphology requires further interrogation. Electrode position may influence 4PI as seen in comparisons between perimodiolar and lateral wall electrodes [14]. However, Leblans et al. [40] did not observe an association between patterns of electrode-modiolus distance (determined by computed tomography scans) and 4PI. Rather, 4PI displayed patterns consistent with the component of impedance which increased only 6–12 months postimplantation, presumably as fibrotic tissue or osteogenesis occurred. Within the current pilot study, which did not incorporate implant type, electrode position, or time since implantation, the relative contribution of fibrosis versus other factors that contribute to current steering were not examined.

The between and within-subject variation in spectral-temporal processing suggests that a processing strategy which gives equal weightage to all active electrodes may not always result in optimal speech outcomes. Van Groesen et al. [36] questioned the sensitivity of the STRIPES measure when assessing the benefit of alternative processing techniques such a focused stimulation and current steering. The current results, however, raise the possibility that the presence of elevated 4PI and presumed fibrosis may explain why relationships between targeted stimulation and improved performance were not observed. The assessment of CI performance may be improved with sectional measures of spectral-temporal processing incorporated and biomarkers of sectional cochlear morphology. This approach avoids the limitations of global measures which may be disproportionately influenced by poorly performing regions. Furthermore, the study highlights the need for a more nuanced understanding of how localised fibrosis influences CI outcomes in the context of future use of drug-eluting arrays to inhibit fibrotic tissue generation [41].

Limitations and Future Directions

The current authors acknowledge several limitations of the pilot study design and interpretation of the preliminary data. The study did not include multiple trials of each secSTRIPES test, and the secSTRIPES was not repeated for test-retest reliability. The overlap of secSTRIPES stimuli on electrodes 11 and 12 does not negate the feasibility to test sectional spectral-temporal processing, it may slightly reduce the independence of the scores. Further versions of secSTRIPES should attempt to improve the electrode specificity. To test the sensitivity of the linear mixed model, the analysis was repeated while excluding data form the individual with elevated basal 4PI. In this result, the association between 4PI and STRIPES was weaker. However, the elevated 4PI values from this individual were not indicative of an anomaly; their conventional impedance measures fell within the acceptable range for cochlear implant function, and there was no evidence of electrode malfunction in the basal region [42]. Consequently, the authors considered it appropriate to retain this participant’s data in the overall analysis.

The interpretation and applications of secSTRIPES requires further investigation beyond the preliminary results presented here. Improvements to the methodology require validation of the secSTRIPES reliability as well as inclusion of speech-in-noise scores, electrode type, and position and for greater interrogation of any stimulus-related factors that may be influencing secSTRIPES scores, namely sectional overlap and temporal smearing in the apical section. A broader study, with a sample size upwards of 32 may also determine which secSTRIPES score best predicts the broadband result and the relationship between secSTRIPES, impedance, and current spread.

Conclusion

The current secSTRIPES pilot suggests a potential relationship between secSTRIPES and speech scores as well as secSTRIPES and measures of the electrode-cochlea interface. There is evidence of sectional patterns of both cochlear implant function and intra-cochlear morphology. These results should be interpreted within the context of a small sample and exploratory design but contribute to the existing literature focussed on understanding cochlear biomarkers and encourages a broader, adequately powered study to validate secSTRIPES as a research and clinical tool.

Acknowledgments

The authors gratefully acknowledge the participants who generously contributed their time and experiences to this study.

Statement of Ethics

This study was performed in accordance with the Declaration of Helsinki. This human study was approved by the University of Western Australia Human Research Ethics Committee – approval: 2023/ET000135. All adult participants provided written informed consent to participate in this study.

Conflict of Interest Statement

Ear Science Institute Australia, which supports the research positions of all authors (E.C., B.H., and C.B.), has received funding from Cochlear Ltd for research projects unrelated to the present study. The authors have not personally received any financial support from this funding.

Funding Sources

This study was supported by the Ear Science Institute Australia that provided institutional support for the employment of the secondary researcher (E.C.). Cochlear Ltd provided research processors for the data collection. Neither organisation had a role in data analysis, interpretation of results, manuscript conception, planning or writing, or the decision to submit the manuscript for publication.

Author Contributions

E.C.: data analysis, interpretation of research data, manuscript writing and editing, and critical revision. B.H.: acquisition of research data, manuscript editing, and critical revision. C.B.: conception and design of the project, acquisition of research data, data analysis, interpretation of research data, manuscript editing, and critical revision.

Funding Statement

This study was supported by the Ear Science Institute Australia that provided institutional support for the employment of the secondary researcher (E.C.). Cochlear Ltd provided research processors for the data collection. Neither organisation had a role in data analysis, interpretation of results, manuscript conception, planning or writing, or the decision to submit the manuscript for publication.

Data Availability Statement

Data are not publicly available due to the inclusion of sensitive clinical information. De-identified summary data may be available from the Ear Science Institute Australia via Dr Christo Bester upon reasonable request.