A total of 60 participants who had bilateral moderate sensorineural hearing loss with a flat configuration were involved in the study. Flat configuration was operationally defined as the difference between least and highest air-conduction thresholds being less than 20dB in the range from 0.25 to 8kHz.31 The age range of the participants was from 15 to 65 years. They had speech identification scores (SIS) that was greater than or equal to 75% at 40dB SL (re: speech reception threshold, SRT). The test ear had normal middle ear status as indicated by ‘A’ type tympanogram with middle ear peak pressure ranging from +50daPa to −100daPa, and admittance ranging from 0.5mL to 1.75mL. The auditory brainstem response (ABR) was recorded at two repetition rates of 11.1s and 90.1s at 90dB nHL to ensure that there was no retro cochlear pathology. The latency difference of V peak of ABR was found to be less than 0.8ms between the two repetition rates. All participants were naïve hearing aid users and there was no self-reported history of other otological and neurological problems. The participants were further classified into Good or Poor Hearing aid Performers (GHP or PHP) using the acceptable noise level (ANL).6 Those participants who obtained an ANL score of ≤7 were considered as good hearing aid performers and a score of ≥13 were considered as poor hearing aid performers.6 The demographic data of each participant in clinical group are tabulated in Table 1. The hearing thresholds at each audiometric frequency of the test ear of the good and poor hearing aid performers are depicted in Fig. 1. The study was approved by the All India Institute of Speech and Hearing Ethics Committee for Research in Humans. Informed consent was obtained from each participant.

Figure 1.

Audiograms of good and poor hearing aid performers.

(0.06MB).

Naturally produced consonant vowel (CV) tokens were utilized as target test stimuli. An adult male with normal voice was used to record the CV tokens. The duration of /da/ and /si/ stimuli was 94ms and 301ms, respectively. For /da/, the voice onset time was 18ms, burst duration was 5ms, transition duration was 37ms, and vowel duration was 34ms. For /si/, the fricative duration was 159.3ms, transition duration was 47.1ms and the vowel duration was 94.6ms. Both the stimuli were converted from ‘.wav’ to ‘.avg’ format using wavtoavg m-file of Brainstem tool box. The ‘.avg’ format of both the stimuli were band pass filtered from 30 to 3000Hz using Neuroscan (Scan 2-version 4.4) to know the functional relationship between the acoustic structure of speech and the brain stem response to speech. The ‘stimulus.avg’, waveforms and spectrograms of the two CV tokens are depicted in Fig. 2. Table 2 summarizes the fundamental frequency (F0) and the first two formant frequencies (F1 and F2) of the vowel component of /da/ and /si/ stimuli. The onset to steady state F0, F1 and F2 within the transition duration (37ms) of /da/ stimulus, and frequency components within the transition duration (42ms) of /i/ portion of /si/ stimulus were measured using Praat (version 5.1.29) software.

Figure 2.

(A) and (D) are the spectrograms of /da/ and /si/ stimuli. The dark black solid line in both stimuli indicates the F0, which has a falling pattern. The formant frequencies are represented by white lines. The F1 and F2 of /da/ is flat in pattern. The F1 of /si/ stimulus is falling in pattern and F2 is raising in pattern. (C) and (F) are the waveforms of /da/ and /si/ stimuli. For sound /da/, the voice onset time was 18ms, the burst duration was 5ms, the transition duration was 37ms and vowel duration of 34ms. For the sound /si/, the fricative duration was 159.3ms, transition duration was 47.1ms and vowel duration was 94.6ms.

(0.27MB).

Further, Kannada passage developed by Sairam and Manjula32 was read out in normal vocal effort by a female speaker was recorded using Adobe Audition (version 3) software. This recorded passage was used to determine the acceptable noise level (ANL). A goodness test was performed in order to verify the quality of the recorded Kannada passage, in which ten listeners with normal hearing rated the passage for naturalness.

Digital Behind The Ear (BTE) hearing aid was used to record the output at the ear canal and at the auditory brainstem response from each participant. According to the technical specifications, frequency range of test hearing aid extended from 0.210 to 6.5kHz. The peak full-on gain was 58dB and high-frequency average full-on gain was 49dB. The functioning of the hearing aid was ensured at the beginning of the data collection and repeated periodically during data collection.

Acceptable noise level (ANL) evaluates the reaction of the listener to background noise while listening to speech. For the measurement of ANL, the method given by Nabelek et al.5 was adopted. Each study participant was made to sit comfortably on a chair in front of the loudspeaker of the audiometer that was located at 1m distance and 45° Azimuth. To compute the ANL, most comfort level (MCL) and background noise level (BNL) were measured.

The recorded Kannada passage was routed through the auxiliary input to the loudspeaker of the audiometer. The presentation level set at the level of SRT. Gradually, the level was adjusted in 5dB-steps to establish the most comfortable level (MCL) and then in smaller steps size of +1 and −2dB, until the MCL was established reliably. After the MCL was established, speech noise was introduced at 30dB HL. The level of the speech noise was increased in 5dB-steps initially, and then in 2dB-steps, to a point at which the participant was willing to accept the noise without becoming tired or fatigued while listening to and following the words in the story. The maximum level at which he/she could accept or put up with the speech noise without becoming tired was considered as the background noise level (BNL). The level of the speech noise was adjusted until participant was able to ‘put-up-with’ the noise while following the story. The resultant level was the BNL. The ANL quantifies the acceptable level of background noise and is calculated as the difference between MCL (dB HL) and BNL (dB HL).4 Based on the ANLs, each participant was classified as good (ANL of ≤7dB) or poor (ANL of ≥13dB) hearing aid performers.4 The procedure of ANL was repeated twice and the average of the two values was considered as the ANL for each participant.

Each participant was fitted with the digital BTE test hearing aid using a custom made soft shell mould. The hearing aid was programmed using NAL-NL1 prescriptive formula. The real ear measurement was carried out to match the gain of hearing aid with the target gain objectively. Further, the Ling's six speech sounds were presented at 65dB SPL to optimize the hearing aid gain. Through fine tuning option, the gain and the frequency shaping of the hearing aid were optimized for the audibility of Ling's six sounds.

The level of the each signal (stored in personal computer) was varied in the audiometer so that the intensity measured was 65dB SPL in sound level meter. Larson Davis 824 sound level meter (SLM) was positioned at the test ear of the participant. The SLM was set at fast weighting function, and it was ensured that the stimuli /da/ and /si/ were presented at 65dB SPL, based on peak amplitude level read on the SLM. After the calibration of stimulus was ensured, output spectrum at the ear canal was recorded using the probe tube microphone measurement, in both unaided and aided conditions. The probe tube microphone in the ear canal picks up the spectral energies at approximately half-octave bands from 0.25kHz to 8kHz for each speech stimulus. The levels as a function of frequency from 0.25kHz to 8kHz, in octaves, were noted down for each stimulus, in the unaided and aided conditions.

Each participant was seated comfortably in a reclining chair with arm. The electrode sites were cleaned up with skin preparing gel. Disc type silver coated electrodes were placed using conduction gel at the test sites. The FFR was recorded using vertical montage. The non-inverting electrode (+) was placed on the vertex (Cz), the ground electrode was on upper fore head (Fpz) and the inverting electrode (−) was placed on nose. It was ensured that the electrode impedance was less than 5kΩ for each of the electrodes and that the inter-electrode impedance was less than 2kΩ.

Prior to recording, calibration of stimuli was ensured using Larson Davis System 824 SLM. The SLM was positioned at reference point. It is the point where the test ear of the participant would be positioned at the time of testing. The SLM was set at fast weighting function for the measurement. It was ensured that both stimuli /da/ and /si/ were presented at 65dB SPL, based on peak amplitude level read on the SLM.

The loudspeaker of the Auditory Evoked Potential equipment was placed at 45° Azimuth from the participant test ear, located at the calibrated position of 12inch distance. The height of loudspeaker was adjusted to the level of participant test ear. The participant was instructed to ignore the stimulus and to watch a movie that was muted and played through a battery operated laptop computer. He/she was also asked to minimize the eye and head movement.

For recording the unaided and the aided FFR, the stimulus /da/ was presented through loud speaker at the presentation level of 65dB SPL to the test ear. The PC-based evoked potential system, Neuroscan 4.4 (Stim 2-version 4.4), controlled the timing of stimulus presentation and delivered an external trigger to the evoked potential recording system, Neuroscan (Scan 2-version 4.4). To allow for a sufficient refractory period within the stimulus sweep, while minimizing the total recording time, an inter-stimulus interval (ISI) of 93ms. from offset to onset of the next stimulus was included for recording FFR to /da/ stimulus. A similar procedure was repeated to record the unaided and aided FFR for /si/ stimulus. However, for recording unaided and aided FFR to /si/ stimulus, an ISI of 113ms was used. The order of stimuli while testing on each participant was counter balanced. The FFR was recorded from 1500 sweeps each in condensation and rarefaction polarities, delivered in a homogenous train using the stimulus presentation software Neuroscan 4.4 (Stim 2-version 4.4).

The FFR recording was initiated once a stable electroencephalogram (EEG) was obtained. The ongoing EEG was converted from analogue-to-digital with the rate of 20,000Hz. The continuous EEG was online band-pass filtered from 30 to 3000Hz with 12dB/octave roll-off. This was stored to disc for offline analysis.

The output of the hearing aid in the ear canal for each stimulus in the unaided and aided conditions were analyzed for spectra. Further, the FFR recorded was analyzed for F0, F0 energy and F1 energy obtained for each stimulus. The continuous EEG data were epoched over a window of 160ms for /da/ stimulus (which included a 30ms pre-stimulus period and a 130ms post-stimulus time). The response for /si/ stimulus was epoched over a window of 360ms (which included a pre-stimulus period of 30ms and a post-stimulus period of 330ms). The epoched waveforms were corrected for baseline. The responses were averaged and filtered off-line from 0.030kHz (high-pass filter, 12dB/octave) to 3kHz (low-pass filter, 12dB/octave). All artefacts exceeding ±35μV were rejected while averaging the response for each averaged response, in rarefaction and condensation polarity. A minimum of 1450 artefact-free epochs was ensured. The averaged waveforms of rarefaction and condensation polarities were added. Further, the added waveforms were created by averaging two trials recorded for each stimulus, in unaided and aided conditions.

For all the participants, the unaided responses were absent for both the stimuli. From FFR recorded for /da/ stimulus in aided condition, the latency of ‘V’ peak was identified by visual inspection. The default MATLAB-code of autocorrelation was utilized, in which a range for latency was specified to obtain F0 in the FFR i.e., from noted ‘V’ peak latency till transition duration. Whereas, the latency of ‘a1’ corresponding to CV boundary30 in the FFR was identified for /si/ stimulus is shown in Fig. 3. The latency of ‘a1’ till transition duration was specified in autocorrelation MATLAB code to obtain F0 in the FFR. Further, F0 energy and F1 energy were determined, using ‘Brainstem Toolbox’ which utilizes the FFT technique (Fig. 4), from the transient response (‘V’ peak for /da/ stimulus; and ‘a1’ for /si/ stimulus) till specified transition duration (37ms for /¿a/ stimulus; and 47.1ms for /si/ stimulus).33

Figure 3.

Transition response of FFR obtained from /si/ stimulus in aided condition.

(0.09MB).

Figure 4.

(A) Response corresponding to stimulus at transition portion of /da/ stimulus; (B) showing grand average spectrum of GHP and PHP sub-group. (B) Fundamental frequency and frequency of first formant in FFR for /da/ stimulus.

(0.2MB).

Spectral energy at frequencies from 0.25 to 8kHz (in octaves) in the unaided and aided conditions, for both the stimuli was analyzed. It was performed to determine representation of energy across frequencies at the ear canal, in good and poor hearing aid performers. The data of spectral energy met the assumption of normal distribution on Kolmogorov–Smirnov normality test (p>0.05) and homogeneity on Levene's test (F<2). The spectral energy (0.5–8kHz in octaves) for both stimuli obtained from both groups, in unaided and aided conditions, was subjected to MANOVA. The result revealed that there was no significant difference between groups in the spectral energy at each octave frequency, in both the unaided and the aided conditions, for /¿a/ and /si/ stimuli. Thus, the data of spectral energy was combined between groups. Descriptive analysis was carried out separately in the unaided and aided conditions. For /¿a/ stimulus (Fig. 5), at extreme low frequency (0.25kHz) and at extreme high frequencies (4kHz and 8kHz) the energy in both unaided and aided conditions is relatively minimal than at other frequencies (0.5kHz, 1kHz and 2kHz). For /si/ stimulus (Fig. 6), at extreme low frequencies (0.25kHz) and at extreme high frequency (8kHz) the energy in both unaided and aided conditions is relatively minimal compared to other frequencies (1kHz, 2kHz and 4kHz).

Figure 5.

Mean and standard deviation of intensity of /si/ stimulus as a function of frequency in unaided and aided conditions.

(0.09MB).

Figure 6.

Mean and standard deviation of intensity of /¿a/ stimulus as a function of frequency in unaided and aided conditions.

(0.09MB).

The F0, F0 energy and F1 energy of FFR between the groups for each stimulus met the assumption of normal distribution on Kolmogorov–Smirnov normality test (p>0.05) and homogeneity on Levene's test (F<2) was also performed. Hence, an independent samples t-test was conducted on each data of FFR between GHP (n=34) and PHP (n=24) groups. From the mean value of F0 of FFR (Fig. 7), it can be inferred that the F0 was represented better in GHP than in PHP, for each stimulus. Further, the F0 of FFR was compared between GHP and PHP using independent samples test. The result showed that there was a significant better F0 encoding in GHP than PHP for /da/ stimulus (t=3.41, p=0.001) and /si/ stimulus (t=2.84, p=0.006).

Figure 7.

Mean, standard deviation and p-value of independent samples t-test on F0 of FFR for each stimulus, in GHP and PHP groups.

(0.06MB).

In addition, the F0 energy and the F1 energy of FFR was compared between GHP and PHP groups to each stimulus. From Fig. 8, it was noted that the mean and standard deviation of F0 energy and F1 energy of FFR to each stimulus were higher in GHP than in PHP. Further, to know if there was any significant difference between GHP and PHP in the mean the F0 energy and the F1 energy of FFR for each stimulus, independent samples t-test was performed. The result revealed a significant higher F0 energy in GHP than in PHP for /da/ stimulus (t=6.80, p=0.000) and /si/ stimulus (t=6.20, p=0.000). Further, a significant higher F1 energy was observed in GHP than PHP for /da/ (t=3.11, p=0.002) and /si/ stimulus (t=5.20, p=0.000).

Figure 8.

Mean and standard deviation of F0 energy and F1 energy for each stimulus in GHP and PHP groups.

(0.11MB).

In the unaided condition for /da/ and /si/ stimuli, energy measured at 2kHz was relatively larger than at other frequencies (in octave). Further, there was a decline in energy after 4kHz i.e., an approximately 10dB per octave for /da/ stimulus and 12dB per octave for /si/ stimulus (Figs. 5 and 6). This pattern of energy representation as a function of frequencies, for both the stimuli, could be because of frequency response of microphone used in recording the target test stimuli. In the aided condition for /da/ stimulus and /si/ stimulus, the energy measured was relatively less at the two extreme cut off frequencies, which is at low frequency below 0.25kHz and high frequencies above 4kHz. Thus, at extreme frequencies, the mean energy in both unaided and aided conditions was less. At low frequency, reduced energy could be because of less gain in that frequency region provided by the prescriptive formula.34 Additionally, low energy noted in the low frequency region of /da/ and /si/ stimulus could also be because of frequency response of the hearing aid. The low frequency cut-off of the frequency response of the test hearing aid was 0.21kHz. At high frequencies i.e., above 4kHz, the energy reduced approximately at the rate of 10dB per octave for /da/ and 14dB per octave for /si/ stimulus. This could be the frequency response of /da/ and /si/ stimulus had energy till 4kHz as noted from unaided condition. Yet another reason could be though the frequency response of hearing aid had 0.216 to 6.5kHz, energy after 4kHz gradually reduced per octave. Thus, remarkable energy was noted in the frequency range from 0.5kHz to 4kHz. It can be inferred that there is a relatively high amplification in the mid-frequency region of the hearing aid than other two extreme cut-off frequencies (low and high). Informally, participants were instructed to repeat the syllables which were randomly presented for three times. In unaided condition, the participants were unable to identify the CV tokens as the presentation level was 65dB SPL, which failed to reach audibility range. However, in aided condition, all the participants consistently identified syllables. Further, on spectral analysis, it was noted that the amplitude of aided burst spectrum of /da/ was similar to the unprocessed burst spectrum amplitude of /da/. It was also observed that amplitude spectrum of fricative /si/ was similar to the unprocessed fricative spectrum amplitude of /si/. This infers that hearing aid preserves inherent speech cues at the ear canal.

The FFR in both unaided and aided conditions were obtained from all the participants. In the unaided condition, the brainstem responses were absent, as the stimuli (/da/ and /si/) were presented at 65dB SPL, which failed to reach audibility. In the aided condition, the F0 representation in the FFR to each stimulus (/da/ and /si/) remained robust and similar to that of the unprocessed filtered raw stimulus. This indicated that preserved spectral content from hearing aid is relayed to the auditory brainstem level. For /da/ stimulus, the mean F0 of FFR was higher in GHP (133.46Hz) than in PHP (128.84), such that the difference was found to be significantly different. This was true for F0 of FFR for /si/ stimulus between GHP (134.42Hz) and PHP (130.84Hz). Further, the F0 of the aided stimulus of /da/ was 134.95Hz and that for /si/ was 144.74Hz. The difference in F0 (in Hz), between encoding of F0 at brainstem level and F0 of aided test stimulus was 1Hz in GHP and 6Hz in PHP for /da/ stimulus. Similarly, the difference noted was 10Hz in GHP and 14Hz in PHP for /si/ stimulus.

The mean difference between the GHP and PHP in the encoding of F0 was 5Hz for /da/ and 4Hz for /si/ stimulus. Though this difference was significant in the encoding of F0 between GHP and PHP for both stimuli, this may not bring a change in speaker identity. This is because, according to Iles35 a change of up to ±25Hz in the F0 will not bring about a change in speaker identity. The finding of the study is in accordance with the research report by Horii36 who reported that a difference of greater than 25Hz in the F0 between the same two stimuli does not cause difference in speaker identity. Additionally, the intra-subject variability of F0 in a normal vocal effort ranged between ±9.6Hz.37 Thus, it can be inferred that the mean F0 of FFR to /da/ and /si/ stimuli was neurally well represented in GHP than PHP, and that both the groups were able to recognize the identity of the speaker.

Further, it was noted that the F0 energy and the F1 energy of FFR to each stimulus were significantly higher in GHP than PHP. The higher energies of F0 and F1 in GHP might be due to stronger efferent fibres that inhibit other harmonics that do not correspond to fundamental frequency and formant frequencies. This is in accordance with the research reports by Ashmore38 and Knight.39 To be more specific, central afferent mechanism is stronger in the group of GHP such that neurons at inferior colliculus fire precisely to the harmonics corresponding to F0 and F1. In addition, the efferent mechanism might be stronger such that the efferent fibres inhibit the other harmonics which do not correspond to the fundamental frequency and formant frequencies, thereby fine tuning the auditory input. The excitatory and inhibitory mechanisms of neurons of the underlying neural generator of the inferior colliculus in GHP fire more or less precisely to the corresponding F0 and F1 components of the stimulus. The inference of the present study supports the findings reported by Krishnan.40 He demonstrated that efferent auditory pathway suppresses energies adjacent to the harmonics corresponding to the F0 and the F1 of FFR. Along with an active afferent pathway, the afferent auditory nerve generates the electrical activity more precisely corresponding to the F0 and the F1 of the stimulus. This involves the release of neurotransmitter, thereby reducing the trans-membrane threshold and increase in neural firing. In poor hearing aid performers, though similar physiological activity was present, probably a lack of precision in neural activity due to less sensitive afferent and weak efferent auditory pathway, might have failed to provide higher energy at harmonics corresponding to F0 and F1 of each stimulus. Thus, it is can be inferred from the present study that subtle physiological variations might be present at the inferior colliculus of the auditory pathway in the poor hearing aid performers with reference to that in good hearing aid performers.