Introduction
Since the launch of the first commercial cochlear implantation (CI) model in 1985, the Nucleus 22, significant improvements have been made to both the cochlear implant device itself and the associated technology and surgical techniques. Continuous upgrading of CI devices has resulted in greater convenience and functionality. Additionally, surgical techniques have undergone standardization. Moreover, many studies on postoperative management and rehabilitation have been conducted, benefiting patients with severe to profound hearing loss. In the rehabilitation process, CI mapping, which involves customized programming of a sound processor to meet the individual requirements of its user, plays a crucial role. This personalized process enables the patient to perceive sound with optimal clarity through the electrodes of the cochlear implant array. CI mapping encompasses several factors; threshold and comfort level (T&C level), pulse-width (duration of the pulse), sensitivity, volume, and frequency allocation are representative factors [1,2].
For a high level of speech perception, mapping of acoustic frequency information onto the appropriate cochlear location without frequency mismatch or spectral distortion is necessary. To achieve this, clinical considerations such as the insertion depth of electrode, proximity of the electrode to the spiral ganglion, and the actual length of the cochlea should be taken into account [3]. Even after surgery, adjustments can be made by varying the frequency at the corresponding active electrode site, a process known as frequency allocation.
In the normal cochlea, there is a tonotopic map that transmits specific frequencies to corresponding positions in the spiral ganglion. Frequency allocation in CI devices is employed to mimic the tonotopic characteristics of a normal cochlea, wherein higher frequency sounds are assigned to basal electrodes and lower frequency sounds to apical electrodes. Currently, audiologists modulate frequency allocation to enhance individualized hearing ability in specific situations. Falcón- González, et al. [4] modified the frequency allocation of electrodes based on the fundamental frequency of musical instruments, under the premise that musical sounds and voices are complex tones rather than the pure tones typically used in CI mapping. They reported improved musical sound perception by altering frequency allocation. Subsequently, they also observed enhanced perception of disyllabic words and openset sentences using the same method [5]. These findings indicate that frequency allocation is a parameter worth modifying to enhance sound perception in specific situations.
In Korean, the frequency of vowel and consonant phonemes generally falls within the range of 250–6,500 Hz. While the frequency range of Korean vowels is similar to that of English vowels, Korean consonants occupy a narrower frequency range compared to English consonants, which reach up to 8,000 Hz [6]. As the target language for most CI device default settings is English, the standard frequency range is typically 100–8,000 Hz (Cochlear: 188-7,938 Hz; MED-EL: 70–8,500 Hz).
Considering these differences in the frequency characteristics between Korean and English, we hypothesized that a narrower frequency range allocation in Korean CI users would contribute to enhanced speech perception. Accordingly, we investigated the relationship between narrower frequency allocation and speech perception outcomes in Korean CI users.
Subjects and Methods
Study design
This study was conducted prospectively from January 2020 to October 2023 at a single tertiary hospital. We employed two different frequency allocations: one ranging from 188–7,938 Hz (standard setting), and the other from 188–5,938 Hz (modified setting). The former represents the standard company setting used for all patients, while the latter signifies a modified, narrower frequency range specifically implemented for this study. Patients activated their CI devices approximately 1 month after surgery, depending on their condition. After initial mapping with the standard setting, a speech perception evaluation was performed. The frequency allocation was then adjusted to the modified setting, and following a 2-week adaptation period, speech perception was reassessed.
Participants
Among patients who visited the Korea University Anam Hospital Otorhinolaryngology-Head and Neck Surgery Department for CI mapping, those ≥15 years of age exhibiting relatively high performance as CI users and without cognitive problems affecting communication were included in the study. Individuals who underwent revision surgery were excluded. To reduce heterogeneity between cochlear implant manufacturers, we selected participants using implants from CochlearTM Nucleus® (Macquarie University), and all patients used fewer than 21 electrode channels. Patients were included in the study only after providing informed consent. Both bilateral and unilateral CI users were included, and for bilateral CI users, each ear was evaluated separately and considered as an independent participant for analysis.
A total of nine patients (6 women, 3 men) participated in this study. Five of these used bilateral CIs, resulting in a total of 14 ears enrolled. Three patients were prelingually deaf, while six were postlingually deaf, with heterogeneous etiologies of deafness. The mean age of patients was 26.6±9.9 years (median: 24 years, range: 15–44 years), and the average duration of CI use was 92.1±77.4 months (median: 82 months, range: 0–228 months). The adjustment of comfort (C) and threshold (T) levels was performed for all patients according to their responses (Table 1).
Ethics statement
This study protocol was reviewed and approved by the Institutional Review Board of Korea University Anam Hospital (approval number 2020AN0509). Written informed consent was obtained from all participants or their legal guardians prior to inclusion in the study.
Audiometry
All patients underwent aided pure tone audiometry (PTA) to assess their hearing thresholds for conversational frequencies. The results were calculated using the weighted four-frequency average ([0.5 kHz + 2×1 kHz + 2×2 kHz + 4 kHz]/6), for 0.5, 1, 2, and 4 kHz thresholds in decibels hearing level (dB HL).
To evaluate speech perception, tests were conducted after a 2-week adaptation period following each cochlear implant mapping change. Ling’s six-sound test, which encompasses individual phonemes designed to address low-, middle-, and high-frequency sounds commonly encountered in continuous speech, was used [7]. Ling’s six sounds consist of /a/, /i/, /u/, /sh/, /s/, and /m/. This test measured both detection and identification in a closed-set setting.
Additionally, a speech perception test for adults was conducted using the Korean consonants and vowels imitation test to evaluate open-set speech perception after CI. The test includes a total of nine vowels comprising monophthongs (/a, eo, o, eu, u, i, e/) and diphthongs (/ae, oe/), as well as 18 consonants (/d, dd, g, gg, r/l, n, m, b, bb, j, jj, s, ss, ch, k, t, p, h/) [8].
Furthermore, an assessment of speech performance included open-set recognition of monosyllabic and disyllabic words, along with a test for everyday sentence repetition. The Korean Speech Audiometry test for adults was used for evaluation, consisting of disyllabic word lists for speech recognition threshold testing, monosyllabic word lists for word recognition score testing, and sentence lists for sentence recognition score testing [9].
Statistical analysis
Data were analyzed using linear mixed-effects models (LMM) to account for repeated measures and nested data (ears within subjects). Frequency allocation condition was specified as a fixed effect, and subject was included as a random effect to account for inter-subject variability. Separate LMMs were fitted for each speech perception outcome using the MixedLM function from the Statsmodels library (version 0.14.0) in Python (version 3.11). Model parameters were estimated using restricted maximum likelihood, and statistical significance was determined at a two-tailed threshold of p<0.05.
Results
The mean threshold measured by aided PTA was 28.6±5.3 dB HL for the standard setting (188–7,938 Hz) and 30.6±6.1 dB HL for the modified setting (188–5,938 Hz). The difference in mean thresholds between the two conditions was not statistically significant (p=0.376). With respect to the Ling’s six-sound test, no statistically significant difference was observed between the two conditions, as all participants demonstrated 100% detection and identification of the sounds used in the Ling’s six-sound test.
Speech perception scores were significantly higher with the modified setting for vowels (estimate=-5.59, p=0.032), consonant perception (estimate=-7.35, p=0.035), and monosyllable perception (estimate=-11.11, p=0.022). Disyllable performance showed a trend (estimate=-9.3, p=0.056), but was not statistically significant. Sentence perception showed no meaningful difference (estimate=-3.43, p=0.367) (Table 2). Individual data were visualized in Fig. 1, including lines connecting individual performance before and after allocation change. Participants who were within 6 months of cochlear implant activation were highlighted in red in Fig. 1 to account for possible early-stage auditory adaptation effects, which may influence speech perception performance independently of frequency allocation.
Discussion
Listening is a complex process that involves more than just sound sensation. It encompasses various factors, such as sound perception, pitch sensation, and speech discrimination. Because cochlear implants process sound signals differently compared to normal hearing, speech and hearing outcomes are not yet equivalent to those of normal hearing. Due to technological developments over the past few decades, the detection of sound itself through CI has reached a satisfactory level. Therefore, we are now at a stage where optimizing parameters to enhance speech recognition, improve listening comfort, and provide a more natural music perception experience similar to that of individuals with normal hearing is crucial.
Reiss, et al. [10] emphasized that speech recognition could be achieved by the correct allocation of each electrode in the cochlea and the precise positioning of stimulation on each electrode. However, their study revealed that, in the case of a “wrong” cochlear placement, people have the capacity to adapt to spectrally shifted, mismatched speech. In their research, it appeared that the perception of pitch was predominantly influenced by the implant map rather than the cochlear location. This suggests that the brain might adjust to spectral disparities by reconfiguring pitch mapping. These findings imply that even if frequency allocation is altered, given adequate time, patients can adapt to different allocations. Therefore, audiologists can modify frequency allocation according to specific situations and purposes. Jethanamest, et al. [11] pointed out that a mismatch between the tonotopic array of the cochlea and the electrode array can lead to distortion, which should be addressed by individually altering the frequency allocation table. In their study, patients using a self-selected frequency allocation table showed enhanced speech perception, as well as improved sound clarity and quality. These studies indicate that we can alter frequency allocation based on specific purposes, and there is sufficient evidence supporting this approach due to the brain’s plasticity.
In our experiment, both settings resulted in mild hearing loss in the aided-PTA test, with no statistical difference between them. This indicates that frequency allocation modification does not significantly affect basic sound detection thresholds. All participants showed 100% performance in the Ling’s six-sound test for both settings, suggesting that the modified setting maintains essential sound detection capabilities comparable to the standard setting.
When comparing phoneme-level speech perception, both vowel and consonant tests showed statistically significant improvements under the narrower frequency allocation. This suggests that refining the upper frequency limit (from 7,938 Hz to 5,938 Hz) not only preserves perception of low-frequency vowel sounds but may even enhance it. Vowels primarily convey acoustic energy in the lower-to-mid frequencies (approximately 400–2,000 Hz), while consonants typically rely on higher-frequency spectral cues (often above 2,000 Hz) [12]. The modified setting appears to have optimized spectral resolution within the frequency range most relevant to Korean phonemes, improving clarity for both vowel and consonant perception.
In word-level tests, statistically significant improvement was observed for monosyllables, with a clear trend toward improvement for disyllables. These outcomes may reflect the relative perceptual complexity of the tasks: monosyllabic words provide fewer linguistic cues and thus are more acoustically demanding than disyllables or sentences, making improvements more detectable. Although sentence-level perception did not show significant change, this may be due to contextual redundancy in sentences that can compensate for subtle acoustic variations [13,14]. These findings reinforce that narrowing the frequency allocation range can meaningfully enhance CI user’s ability to recognize individual speech sounds, especially in languages like Korean where phonemic energy is concentrated in a narrower spectrum.
Our study had several limitations. First, we examined a relatively small sample size of 14 ears and exclusively focused on individuals with a high level of speech perception, making it challenging to generalize the findings to all CI users. Second, we included both prelingual and postlingual deaf patients, introducing heterogeneity in the etiology of hearing loss. Since the memory of normal hearing could influence speech performance, analyzing these two conditions separately would provide more precise insights. Third, our study design included three participants in the early period of CI use (<6 months). As speech perception abilities typically improve rapidly during the first year of CI use, this could potentially confound our results. Fourth, we conducted testing after a 2-week adaptation period, which may not have been sufficient time for neural plasticity and brain adaptation to fully develop. A longer adaptation period would likely yield more definitive results. Furthermore, in our study design, we initially used the standard setting and then switched to the modified setting, with patients aware of the change. This awareness may have introduced expectation bias. Moreover, speech performance could potentially improve during the initial period of CI use regardless of frequency allocation changes. Therefore, better performance with the modified setting might not solely be due to the frequency allocation difference but could partly reflect the natural adaptation process to CI use. To address these limitations, a patient-blinded randomized study with a crossover design and longer adaptation periods would be beneficial.
Despite these limitations, this is the first study to analyze the phonemic characteristics of Korean language and adjust frequency allocation accordingly, examining its impact on speech perception. These results underscore the importance of adjusting CIs according to the specific characteristics of each language and provide crucial insights for optimizing CI mapping strategies for Korean speakers. Given that personalized CI mapping can enhance performance, we believe that modifying the standard frequency setting according to language-specific characteristics represents a readily implementable approach to improve outcomes. Along with other efforts to enhance performance through individualized CI electrode insertion [15], adjusting post-surgery mapping according to individual needs and the linguistic characteristics of the user’s native language could further personalize and optimize CI outcomes.
In conclusion, Korean phonemes have different frequency characteristics compared to those of English, primarily occupying the range of 250–6,500 Hz, which is narrower than that typical of English phonemes. Given the limited number of electrodes in CI devices, finding the most effective way to utilize these electrodes is crucial. Our findings demonstrate that using a narrower frequency allocation (188–5,938 Hz) in Korean CI users, which better aligns with the frequency characteristics of Korean speech sounds. This approach represents a simple yet effective strategy for enhancing speech perception outcomes in Korean-speaking CI recipients.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
