Audibility and speech perception in quiet are better with nonlinear frequency compression (NLFC) in adults and children who have severe to profound sensorineural hearing loss (SNHL), previous studies have demonstrated (Int J Audiol 2009;48:632-644; Int J Audiol 2005;44:281-292>; J Am Acad Audiol 2010;21:618-628; Int J Audiol 2011;50:396-404). Results for speech recognition in noise are mixed.
However, to date there have been no published studies evaluating these algorithms in people with asymmetric high-frequency hearing loss or helping clinicians determine the most appropriate settings for NLFC in this population.
We embarked on a study to address this gap, and identified benefits of using nonlinear frequency compression in those with asymmetric high-frequency hearing loss.
Study participants had a sloping, moderately severe to severe SNHL that was significantly asymmetric, defined as at least a 15-dB difference in audiometric thresholds at three of four high frequencies (2,000, 3,000, 4,000, and 6,000 Hz). Candidates were excluded if they showed signs of a conductive hearing loss or retrocochlear disorder.
All 28 participants had previous experience with hearing aids, but none used frequency-lowering technology. They ranged in age from 31.5 to 87.8 (mean, 67.6) at assessment interval one. Nineteen of the participants were male, and nine were female.
Mean audiometric data for right versus left ears and better versus worse ears are provided in figure 1. There was no significant difference between mean right-ear and left-ear thresholds at any frequency. Mean better-ear and worse-ear thresholds differed significantly at every frequency tested (p <.001).
All participants were fit bilaterally with Phonak Nios V micro-sized behind-the-ear (BTE) or Naida IX SP BTE hearing aids, depending on audiometric thresholds. During the study period, features of the Naida IX SP hearing aids that are unavailable in the Nios V aids were deactivated to make the hearing aids as equivalent as possible.
Hearing aid fittings were conducted at the Hearts for Hearing Foundation clinic; all study measures were completed at the University of Oklahoma Health Sciences Center's Communication and Aural Rehabilitation Research Laboratory.
HEARING AID FITTING
Unaided air-conduction audiometric thresholds were obtained with insert earphones coupled to ER-3A foam eartips. Next, the output of the hearing aids was matched (+/- 5 dB) to the Desired Sensation Level (DSL) v5.0 prescriptive fitting targets for adults using the Audioscan Verifit Real-Ear/Hearing Aid Analyzer standard speech signal presented at 55, 65, and 75 dB SPL.
Probe-microphone measures also were employed to ensure that the maximum output of the hearing aid did not exceed the DSL v5.0 adult target for a 90-dB swept pure tone. The aforementioned measures were conducted with NLFC disabled.
Next, NLFC parameters were programmed in each participant's hearing aids with iPFG version 2.5. Two fitting profiles were created for each participant in order to switch between NLFC programs over the course of the study. These profiles were identical in all parameters except for the NLFC crossover frequency and compression ratio.
The first program, hereafter referred to as the binaural NLFC program, used NLFC settings automatically created by the iPFG version 2.5 fitting software. These settings, which are derived from a proprietary algorithm, apply the same NLFC parameters to both hearing aids based upon the recommended setting for the better ear (see Eur Arch Otorhinolaryngol 2010;267:1045-1053 for further information). Presumably, this is the default NLFC profile when audiologists use the fitting software for adults with asymmetric SNHL.
The second fitting profile, hereafter referred to as the monaural NLFC program, was created based on software-recommended crossover frequency and compression ratio for NLFC in each ear individually.
All study participants were assessed at three four-week intervals. To minimize order effects and potential long-term acclimatization, we employed a two-period crossover design. Baseline assessment was conducted following a four-week trial period of study hearing aid use with NLFC disabled.
After this assessment, one of two NLFC programs was enabled in all participants’ study hearing aids (assessment 1). Assessment 2 was conducted after participants used the first of two randomly assigned NLFC programs for four weeks, and assessment 3 was conducted after four weeks with the remaining NLFC program.
The assessments, which were performed in the binaural condition, included:
* The phoneme perception test (PPT), which features six nonsense tokens (/asa/, /asha/, /afa/, /ada/, /aka/, and /ata/; Audiology Online 2009; The Benefits of Nonlinear Frequency Compression for People with Mild Hearing Loss). An additional stimulus—asa6k—was created to directly assess perception of the medial phoneme /s/ centered at 6,000 Hz.
* The Modified Rhyme Test (MRT) for phoneme discrimination of initial and final consonant sounds (J Sp Hear Res 1968;11:536-552).
* The Bamford–Kowal–Bench Speech-in-Noise test (BKB-SIN; Brit J Audiol 1979;13:108-112).
* The Speech, Spatial, and Qualities of Hearing Scale (SSQ) version 5.6 to evaluate participants’ subjective impressions of the study hearing aids (Int J Audiol 2004;43:85-99).
To minimize bias due to hypothesis guessing, participants were blinded to the nature of nonlinear frequency compression and the setting (off vs. binaural vs. monaural) during each trial period. With the exception of the BKB-SIN test, study measure responses were recorded by the participant (MRT, SSQ) or by test software (PPT), removing the potential for investigator subjectivity in scoring.
BETTER SPEECH RECOGNITION
Phoneme perception test data were analyzed using a two-way repeated measures analysis of variance (RM ANOVA) for the main effects of token and NLFC profile on threshold (see figure 2). Testing revealed a significant effect of token (F[5,135] = 9.12, p <.001) and NLFC (F[2,54] = 4.63, p <.05). The interaction between token and NLFC was not significant (p =.47).
Post-hoc NLFC profile comparisons using Tukey's honestly significant difference (HSD) test revealed significantly lower (better) thresholds when participants used NLFC in the binaural profile compared with NLFC disabled for the /asa6k/ token (t = 2.21, p <.05) and the /afa/ token (t = 2.23, p <.05). While there appeared to be an overall trend toward lower thresholds using NLFC compared with NLFC disabled, the remaining four tokens did not show a statistically significant NLFC advantage.
In the two-way RM ANOVA for the Modified Rhyme Test (see figure 3), a significant effect of azimuth was seen (F[1,27] = 16.47, p <.001), but no significant effects were detected for the NLFC profile (F[2,54] = 1.19, p =.31) or interaction between the azimuth and profile (p =.45).
Post-hoc testing using Tukey's HSD revealed significantly higher (better) scores when the MRT stimuli were presented from the front (0 degrees) compared with presentation from the poorer ear side (90 or 270 degrees) in all NLFC conditions: off (t = 2.22, p <.05), binaural (t = 3.45, p <.01), and monaural (t = 3.08, p <.01).
The dB signal-to-noise ratio for 50-percent correct performance on the Bamford-Kowal-Bench Speech-in-Noise test was analyzed using a one-way RM ANOVA for the main effect of NLFC profile (see figure 4). Analysis revealed no significant difference among thresholds across NLFC conditions (F(2,54) = 2.88, p =.06).
Finally, scores on the three subscales of the Speech, Spatial, and Qualities of Hearing Scale version 5.6 were compared using a two-way RM ANOVA for the main effects of NLFC and subscale (see figure 5). There was a significant effect of NLFC (F[2,54] = 6.72, p <.01) and subscale (F[2,54] = 15.09, p <.001); the interaction term was not significant (F[4,108] = 1.44, p =.23).
Post-hoc testing using Tukey's HSD revealed significantly higher (better) scores on the Speech subscale with NLFC enabled in both the binaural (t = 2.99, p <.01) and monaural (t = 2.67, p <.05) profiles compared with scores obtained after the trial period with NLFC disabled. Significantly higher scores were also demonstrated on the Qualities subscale with NLFC enabled in both the binaural (t = 2.32, p <.05) and monaural (t = 2.69, p <.05) profiles compared with NLFC disabled. No significant difference was seen among scores on the Spatial subscale.
Overall SSQ scores were significantly higher with NLFC enabled in both the binaural (t = 2.64, p <.05) and monaural (t = 3.13, p <.01) profiles compared with NLFC disabled. No significant differences were observed between scores obtained in the NLFC binaural and monaural profiles for any subscale or the total SSQ score.
For adults with asymmetrical high-frequency hearing loss, nonlinear frequency compression improved the recognition of high-frequency speech phonemes in quiet, as indicated by better scores on the phoneme perception test, and led to self-reported gains in sound quality and speech understanding in quiet. These benefits were seen with NLFC set to the manufacturer's default and based on the audiogram.
Although there was a trend toward better speech recognition in noise with NLFC use, the difference was not significant. No performance differences were observed when NLFC was set to the better ear threshold versus when it was set to the thresholds of each separate ear.
It is important to note that Jace Wolfe, PhD, et al showed an improvement in performance measured after six months of nonlinear frequency compression use compared with effects after six weeks of use in a group of children with high-frequency hearing loss (Int J Audiol 2011;50:396-404). It is possible that the performance of the adults in this study had not reached asymptotic levels when measured after just four to six weeks of NLFC use.
Additionally, it is important to note that the NLFC settings for this study were based on manufacturer defaults rather than optimized for each subject's needs using probe microphone measures and subjective assessment, as described by Dr. Wolfe and colleagues (J Am Acad Audiol 2010;21:618-628; Int J Audiol 2011;50:396-404).
Participants in the current study also had hearing losses that were characterized by more steeply precipitous slopes than the hearing losses reported in the two previous studies by Dr. Wolfe et al. As a result, it may be even more likely that individual optimization was required to enhance NLFC benefits in this investigation.
Additional research is needed to evaluate the optimization of nonlinear frequency compression parameters in adults with asymmetric hearing loss and to determine if this optimization results in differences in performance when NLFC settings are based on the better ear versus each ear separately.
A grant from Phonak covered the expenses to conduct the study. The authors thank Michael Boretzki, PhD of Phonak, who served as a consultant for the study.
References on Tap
Access the hyperlinks (shown in gray) in this article and throughout the issue by reading it on HJ's free iPad app:bit.ly/AppHearingJ.