Comparative Study on the Acoustic Analysis of Voice in Auditory Brainstem Implantees, Cochlear Implantees, and Normal Hearing Children

Abstract

The aim of the study was to compare the acoustic characteristics of voice between Auditory Brainstem Implantees, Cochlear Implantees and normal hearing children. Voice parameters such as fundamental frequency, formant frequencies, perturbation measures, and harmonic to noise ratio were measured in a total of 30 children out of which 10 were Auditory Brainstem Implantees, 10 were Cochlear Implantees and 10 were normal hearing children. Parametric and nonparametric statistics were done to establish the nature of significance between the three groups. Overall deviancies were seen in the implanted group for all acoustic parameters. However abnormal deviations were seen in individuals with Auditory Brainstem Implants indicating the deficit in the feedback loop impacting the voice characteristics. The deviancy in feedback could attribute to the poor performance in ABI and CI. The CI performed comparatively better when compared to the ABI group indicating a slight feedback loop due to the type of Implant. However, there needs to be additional evidence supporting this and there is a need to carry out the same study using a larger sample size and a longitudinal design.

Introduction

Speech is undoubtedly an essential form of communication, and its production must be carefully regulated to convey the required information [1]. For seamless communication, adequate functioning of the auditory feedback system is obligatory and this system oversees and ensures the proper fluidity of speech by monitoring the speech motor movements by detecting and modifying production errors. [2, 3]. The auditory system contains both feedback and feedforward loops, which are integral in maintaining the flow of speech. The feedback loop monitors the deviations and detects errors and the feedforward loop monitors the task execution from the already learned commands [4]. Apart from monitoring speech motor movements, these loops are integral for maintaining normal voice characteristics, the importance of this feedback loop is well demonstrated in individuals with hearing impairment and its significance has rightly been extensively studied and well understood [5, 6]. In the phonation subsystem of speech, the auditory feedback enables the maintenance of a normalized fundamental frequency (F0), and manipulating the feedback has resulted in a shift in F0 and formant frequencies [7, 8].

This auditory loop deviates in individuals with hearing loss [3] and Literature reports that individuals with prelingual hearing loss exhibit variation in voice quality such as hoarseness, breathiness [9], strain, high F0 [10], high variability, monotone, altered speech rate, increased loudness, loudness either too soft or too loud [11], resonance irregularity [9], and instability [9]. These deviations in the voice characteristics also depend on the severity of hearing loss. Individuals with severe to profound hearing loss, present with a marked increase in F 0 with vocal instability. In contrast, in individuals with mild to moderate hearing loss deviances are predominantly perceived in oral resonance [12]. This highlights that impairment in auditory perception can, in turn, affect speech sound production including the voice [3].

The current hearing restoration techniques for hearing loss individuals are the use of hearing aids, Cochlear Implants (CI), and Auditory Brainstem Implants (ABI) [13, 14]. In individuals using CI, the auditory feedback loop tends to develop post-implantation as a a consequence of developing auditory perception and vocal exploration, which facilitates auditory-motor representations and neural connections between concerned cortical areas [15]. Improvement in fundamental frequency (F0), formant frequencies, amplitude, and stability is noticeable after implantation thus, contributing to better perceptual voice characteristics [3, 16]. Before the fourth year of implantation, children receiving CI attain better acoustic control over speech, normalization of fundamental frequency [17], and attaining articulatory control [18]. It is a well-known fact that the auditory feedback loop is impaired in hearing-impaired individuals, and cochlear implantation can help restore function. However, the feedback provided by the still-maturing system may be deemed insufficient to restore the vocal parameters due to the vocal changes caused by deafness. This could be due to several factors associated with deafness such as duration, nature, and onset as well as factors associated with implantation such as unilateral vs bilateral, age at implantation, etc. Investigations suggest that to restore the voice and other parameters of speech in implanted individuals, there is a need for combined auditory verbal and speech therapy post-implantation to bring up near-normal speech [19]. The concept and importance of auditory feedback can be understood by the excellent illustration by Lattenkamp EZ (2020) (Fig. 1) which depicts a feedback loop where there is a need to compare the self-emitted vocalization and internally memorized targets to facilitate normal production [20]. However, in individuals with hearing loss, the perception of self-emitted vocalization and the external target is affected thereby affecting the internally memorized target and affecting the production [21].

Fig. 1.

A functional auditory feedback loop of a normal-hearing individual. The sef-emitted vocalization (bak arrow) is compared with the internal memorization target and production is faciliated. By deafening the ear (black lightning) the perception of sef-emiited vocalization is affected and hence internal memorization wouldn’t be present with no perception of external input. This effects the production. Same can be applied to ABI and CI. Permission obtained from Lattenkamp et al. [21]

Changes in fundamental frequency, voice onset time, and vowel duration have been documented in ABI post-neurofibromatosis type 2. Their fundamental frequency, voice onset time, and frequency contours were observed to have shifted post-implantation [20]. The F0 inflection and average SPL were reported to be higher in an individual with ABI, and along with these parameters, reduced speech intelligibility and increased duration of words were observed [22]. The authors highlight that reduced or impaired auditory feedback can impact the vocal parameters and other parameters of speech in individuals with ABI [20, 21]. In an attempt to better understand the auditory feedback system's impact on voice of ABI users, we have conducted a first-of-its-kind investigation by profiling acoustic voice characteristics of ABI users and comparing the same with those of CI users and normal hearing peers. We hypothesize that individuals with ABI will have poorer vocal production quality when compared to individuals with CI followed by individuals with normal hearing individuals. Also, we predict that the acoustic characteristics of voice will be poorer in individuals with ABI followed by individuals with CI than normal hearing individuals.

Methods

Participants

Ninety-seven children in the age group of 12 to 84 months were recruited for the present study. The subjects were age-matched and gender-matched and divided into group 1: Children using auditory brainstem implants (n = 10, mean age = 4.9 years), group 2: Children using cochlear implants (n = 44, mean age = 3.1 years), and group 3: Normal hearing children (n = 43, mean age = 3.68 years).

Selection Criteria

Groups 1 and 2 consisted of children with the chronological age of 36 to 84 months ABI users and CI users, respectively (unilateral) with Opus II speech processor and FS4 sound coding strategy. Individuals who have undergone ABI had inner ear abnormalities. Six of them had absent auditory nerves bilaterally and three of them with a common cavity on both ears and one had a hypoplastic nerve bilaterally. All the children in the ABI group had no response to Electrically Evoked Auditory Brainstem Response (EABR) and were confirmed as candidates for ABI. On the other hand, 44 CI users had normal anatomy of the cochlea and nerve. Children with GDD/other physical/ cognitive disabilities and those with a history of laryngeal pathology were excluded from the study. Only those who received continuous auditory habilitation were recruited for the study. Subjects included in the normal hearing group presented with normal hearing in both ears on screening, no history of any speech, language, and hearing disorders nor any history of neurological impairment.

Procedure

Phase I: Collection of demographic details

Demographic details and history of the child’s deafness, implant age, speech and language intervention, previous use of amplification devices such as hearing aid, and history of laryngeal pathology, if any, were collected from the parents.

Phase II: Voice Recording and analysis

Test Environment

The recording was done in an environment with acceptable ambient noise level measured using Sound Level Meter, Volcraft 322 data logger with electret condenser microphone in an A-weighting network. Noise levels were observed to be ranging approximately from 32.8 to 37.4 dB(A).

Equipment

Data collection and analysis were performed using PRAAT software version 4.4.33. Installed in Acer laptop with Intel processor N3050, Windows 8.1 and recorded using Microphone Xpro-106 Clair.

Recording Procedure

Subjects were instructed and demonstrated to repeat the sustained vowel /a/. Voice samples were mono-recorded using PRAAT software (4.4.33 version) at a sampling rate of 44,100 Hz. The cochlear implant and brainstem implant were activated and set into regular use settings during voice tasks. The subjects were seated in an upright position; the mouth-to-microphone distance was fixed at 5 cm, with a 45° angle placement to reduce aerodynamic noise. All data were collected in real-time with the output screen in full view of the subjects. Commencement of recording was done post-verification of the child’s understanding of the task via practice trials.

Voice Sample

The task consisted of sustaining the low central vowel /a/ at a comfortable pitch and loudness levels for 3–5 s. Three trials were recorded. The mean value across three sustained vowel production trials was then considered for data analysis [18]. Vowel /a/ was chosen as it is easy for young children to produce accurately and consistently [18]. Its steady-state nature allowed for easy extraction of frequency and amplitude measurements.

Analysis of Acoustic Parameters

All the following acoustic parameters were analyzed in the middle portion of sustained recorded waveform as it was considered owing to the steady nature (i) Habitual fundamental frequency (F0 in Hz) is defined as the number of times a sound wave produced by the vocal cords repeats during a given period. It is also the number of cycles of opening/closure of the glottis for a modal register. (ii) Intensity (I0 in dB), (iii) Jitter (%)is defined as the parameter of frequency variation from cycle to cycle. (iv) Shimmer (%) is defined as the cycle-to-cycle amplitude variation in the sound. (v) Harmonic to-noise ratio (HNR) is defined as the ratio between periodic components and non-periodic components comprising a segment of voiced Speech. (vi) Formants (F1 & amp; F2 Hz) are defined by the size and shape of the acoustic spaces of the vocal tract (the oral, nasal, and pharyngeal cavities) and the coupling between those spaces. Formant frequency was subjected to statistical analysis without normalization of data.

Phase III: Statistical Analysis

The collected data were tabulated and statistically analysed. Mean, standard deviation (SD), and P values were extracted using Statistical Package for Social Science (SPSS) version 27.0. An array of tests were carried out to establish the nature of significance across the groups. The normality test was carried out using Shapiro–Wilk’s test of normality across the three groups individually and as the data followed the normal distribution pattern (p > 0.05) at a 95% confidence interval, a parametric ANOVA test was carried out to establish the nature of significance. Upon carrying out Levene’s test of equality across the parameters, equal variances were not assumed for the parameters (p < 0.05), and hence Welsch’s ANOVA (Analysis of Variance) test was carried out to establish the nature of significance between the groups. A Games-Howell post hoc test was further done to establish the nature of significance between various pairs.

Results

Out of 97 samples collected, each group of auditory brainstem implantees (ABI), cochlear Implantees (CI), and normal-hearing children contained a sample size of 10, 44, and 43 respectively. A total of 291 voice samples were collected from these 97 children with major voice parameters measured such as habitual fundamental frequency (F0 in Hz), Intensity (I0 in dB), Jitter (%), Shimmer (%), Harmonic to-noise ratio (HNR), a Formants (F1& F2). All these measured parameters were compared between individuals with ABI, CI, and normal hearing.

Fundamental Frequency (F0)

Upon comparing F0 across the groups using the Welsch ANOVA test, results reveal statistically significant different scores (F(2,35.61) = [45.91], p < 0.01) obtained by at least two groups. Games-Howell post hoc multiple comparison tests reveal a presence of statistical significance between all three groups with significance between ABI and CI (p < 0.01, 95% CI = [53.6, 108.2]), between ABI and normal hearing (p < 0.01, 95% CI = [73.31, 126.92]), and between CI and normal hearing (p < 0.01, 95% CI = [− 5.45, 43.88]). Amongst the three groups, ABI individuals had a higher F0 value of 389.51 Hz followed by CI individuals with a value of 308.6 Hz and normal hearing children with a value of 289.39 Hz.

Formant Frequencies (F1& F2)

Upon comparing F1 and F2 across the groups, results revealed a statistical significance between at least two groups on the Welsch ANOVA test where F1 (F(2,37.59) = [125.34], p < 0.01) and F2 (F(2,62.53) = [33.51], p < 0.01), both had at least two groups significant. Games-Howell post hoc multiple comparison tests revealed a presence of statistical significance between all three groups with significance between ABI and CI (p < 0.01, 95% CI = [− 239.38, − 83.42]), between ABI and normal hearing (p < 0.01, 95% CI = [− 421.30, − 306.23]), and between CI and normal hearing (p < 0.01, 95% CI = [− 275.38, − 129.34]) for F1. A similar statistical significance was found for F2 with significance between ABI and CI (p < 0.01, 95% CI = [− 197.15, − 37.83]), between ABI and normal hearing (p < 0.01, 95% CI = [− 328.77, − 175.35]), and between CI and normal hearing (p < 0.01, 95% CI = [− 234.16, − 34.97]). Amongst the three groups, ABI individuals had a lower F1 value of 740.86 Hz, followed by CI individuals with a value of 942.26 Hz followed by normal hearing children with a value of 1144.63 Hz. Similarly for F2, ABI had a lower value of 1463.18 Hz followed by CI with a value of 1580.67 Hz followed by normal hearing children with a value of 1715.24 Hz (Fig. 2).

Fig. 2.

The figure demonstrates the descriptive statistics (mean) across all three groups for frequency parameters which include the fundamental frequency and both the format frequencies

Intensity (I0)

Upon comparing I0 across the groups, results reveal a statistical significance between at least two groups on the Welsch ANOVA test (F(2,40.93) = [8.01], p < 0.01). Games-Howell post hoc multiple comparison tests reveal a presence of statistical significance between ABI and CI (p < 0.01, 95% CI = [2.31, 10.47]) and between CI and normal hearing (p < 0.01, 95% CI = [− 9.42, − 1.25]). Amongst the three groups, ABI individuals had a higher I0 value of 72.98 dB followed by normal hearing individuals with a value of 71.92 dB followed by CI individuals with a value of 66.58 dB.

Perturbation Measures

Comparing the perturbation measures such as jitter and shimmer across the groups revealed a statistical significance between at least two groups on the Welsch ANOVA where jitter (F(2,29.07) = [9.93], p < 0.01) and shimmer (F(2,28.95) = [6.51], p < 0.01), both had at least two groups significant. Concerning jitter, Games-Howell post hoc multiple comparison tests revealed a presence of statistical significance between ABI and normal hearing (p < 0.01, 95% CI = [0.21, 0.84]) with ABI having a higher value of 1.1% when compared to normal hearing with a value of 0.56%. No statistical significance was obtained between ABI- CI and CI—normal hearing for both the perturbation parameters of jitter and shimmer. On the other hand, using Games-Howell posthoc tests for shimmer, a statistical significance was obtained between ABI and normal hearing (p < 0.01, 95% CI = [0.78, 4.81]. The ABI group obtained a higher mean value of 6.71 dB when compared to the 3.91 dB documented in the normal hearing group. No statistical significance was obtained between ABI—CI and CI-normal hearing groups (Fig. 3).

Fig. 3.

The figure demonstrates the descriptive statistics (mean) across all three groups for perturbation measures which include both jitter and shimmer

Harmonic to Noise Ratio (HNR)

Upon comparing HNR across groups, a statistical significance was obtained between at least two groups on the Welsch ANOVA (F(2,25.66) = [4.54], p = 0.02). On carrying the Games-Howell post hoc test, a significant difference was obtained between ABI and normal hearing (p = 0.03, 95% CI = [1.12, 4.84]) with ABI having a higher mean value of 17.9 when compared to normal hearing with a mean value of 16.04.

Discussion

The present study investigated children's voice characteristics implanted with ABI and compared them with children using CI and normal-hearing children. The outcomes of ABI were observed to be poorer and more challenging than CI [23]. However, with improved surgical techniques in ABI, there have been significant overall outcomes for children [13]. While ABI provides hearing benefits to patients, the outcomes challenge our understanding of how the brain processes neural patterns of auditory information. The neural pattern of activation produced by an ABI is highly unnatural, yet some patients achieve high speech understanding levels; however, they might not show similar production skills [23]. The present study studied voice parameters such as F0, I0, jitter, shimmer, HNR, F1, and F2 in ABI, CI, and normal hearing groups. The mean pitch indicated by higher mean F0 values was found in CI and ABI users when compared to normal-hearing children. These findings are supported by a study conducted by Coelho et al. [3]. In a study comparing patients with acoustic neuroma implanted with ABI and hearing aid users, F0 and VOT values were noted to have increased significantly, much similar to our findings [21]. Karin Lundin [24] suggested that anatomical deviation in the tonotopic organization of the cochlear and the cochlear nucleus results in the varied performance of CI and ABI users. The findings demonstrate that children with ABI can take a longer time to “learn to hear” when compared to children with CI. This could attribute to higher fundamental frequencies in ABI users.

In our study, formant frequencies were decreased in children using ABI and CI when compared to controls in all age groups. These findings suggest that perception through implant can alter speech perception and thus change a patient`s production or performance.

Formants (F1 & F2) have been reported not being significantly deviant for children undergoing cochlear implantation before four years of age and significantly deviant for children who undergo surgery after four years of age, as the former attain better acoustic control over speech, normalize their fundamental frequency and formants [25].

In our study, the mean Intensity (Io) was almost similar between the groups within normal limits. However, an earlier study by Liker et al. [25] reported contradictory findings for children with CI where the mean intensity was observed to be comparatively higher [25]. Alike to Holler and Campisi [26] stated that loudness could be normalized in individuals with CIs if the child has been exposed to a longer duration of auditory feedback [26].

In our study, the shimmer was found to be affected in both the implanted groups. This finding is in agreement with the results of Srividya, Premalatha, and Gore [19]. where they reported significant differences between age and gender-matched normal-hearing childrenand children using a CI in shimmer (dB) and shimmer (%) [19]. Shimmer indicates the level of perturbation determining vocal consistency. For individuals with hearing implants, the deviancy in shimmer may be due to the deficit in phonation and neuromuscular control consequent to compromised auditory feedback [21].

The HNR values in CI individuals were comparatively lower than those of normal-hearing individuals, which is in agreement with the findings of Jafari et al. [27], the periodicity of the voice signal is impaired in children with CI as they have an atypical voice quality due to over-aspiration and spectral noise [24]. However, the fact that the HNR readings are lower than those of ABI warrants further investigation.

Based on our findings, in most of the parameters, the CI users had more inadequate results than normal hearing participants, but ABI users performed even lower than CI children. These findings could be attributed to feedback control systems even poorer than CI users [24]. The same can be attributed to ABI users with functional limitations when compared to CI users. Even though ABI restores audibility, its performance was significantly different from CI and normal hearing individuals.

The voice characteristics in ABI and CI were significantly different from those of normal-hearing individuals. One primary reason for abnormal voice deviancy in ABI is the tonotopic organization of the cochlear nucleus. The low-frequency fibres terminate towards the inferior portion of the Dorsal Cochlear Nucleus and the anterior portion of the.

The ventral Cochlear Nucleus and the high-frequency fibres extend towards the superior portion of the Dorsal Cochlear Nucleus and the anterior portion of the Ventral Cochlear Nucleus. This complex tonotopic arrangement makes it cumbersome for the electrodes to stimulate all the regions based on incoming signals [28]. Most of the time, there is only partial coverage of electrodes in ABIs than CI, which has a complete electrode coverage as per the tonotopic nature of the cochlea. The other reason is that the coding strategies used in ABI are CI-based strategies. These coding strategies formulated based on cochlear nerve conduction properties may not be the best fit for cochlear nucleus stimulation [29]. These factors result in low processing of complex signals and lead to auditory feedback deprivation, thus majorly affecting all speech-voice parameters.

The rehabilitation process often includes only auditory habilitation and speech and language therapy, the importance of voice in projecting emotions is often negotiated in children using implants. The voice therapy for ABI users is similar in nature to cochlear implant users as there is a lack of auditory feedback and a feedforward loop. Similar to that of speech-language development in ABI users, voice improvement can also show slow progress.

The study aimed at comparing the acoustic features of voice between ABI, CI, and normal-hearing children. Although the comparison was established, the study could have been carriedout with a larger sample size for the three groups. Future direction would recommend researchers to research with larger samples and varied voice production tasks. A correlation between perceptual and acoustic analysis could have been carried out to yield more valid results. The groups could have been further classified based on their age at implantation, and the effect of age at implantation of voice in individuals with ABI CI could have been explored.

Conclusion

The ABI and CI users perform poorly in terms of voice than normal hearing peers due to a deficient auditory feedback system. The acoustic parameters such as fundamental frequency, intensity, jitter, shimmer, harmonics to noise ratio, first formant frequency, and second formant frequency were significantly different between ABI and the normal hearing group.

The voice characteristics of CI users were observed to be less deviant when compared to the ABI group. However, this conclusion requires further supportive studies. There is also a need to carry out studies analyzing children's speech production abilities with ABI as it will provide insights into the precise influence of ABI. A longitudinal design is warranted to understand vocal development better. Furthermore, studies have to check the influence of changes in programming parameters on the voice characteristics of ABI users has to be accounted.

Authors Contribution

AU: Data collection and manuscript, SR: Data Collection and Manuscript, JLS: Supervision and Manuscript, SD: Co-Supervision and Manuscript.

Funding

None.

Data Availability

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.

Declarations

Conflict of Interest

The authors have no conflict of interest.

Ethical approval

Before the study’s initiation, due approval was obtained from the department and the research committee of MERF Institute of Speech and Hearing (p) Ltd. The paper was presented to the ethics committee and was exempted from presentation due to the usage of current CSE practice and Diagnostic Procedure. Prior to the collection of data, written informed consent was obtained from the participants' parent/legal guardian/next of kin to participate in the study.