Signal Transparency of Remote Microphone Technology in Pediatric Bone Conduction Device Users

Bone conduction hearing devices (BCDs) are a common treatment modality for children with chronic and/or permanent conductive hearing loss (CHL). Significant gains in audibility and improved speech perception with BCDs have been long demonstrated in this population [Papsin et al., 1997; Bejar-Solar et al., 2000; Priwin et al., 2007; Verhagen et al., 2008]. While these benefits have been well studied, less is known about hearing performance in children in noisy environments. Competing noise in the environment negatively affects speech intelligibility, particularly when listening at a distance [Crandell and Smaldino, 2000; Wolfe et al., 2013a]. This is especially true for children with hearing loss (HL) in classroom environments where the signal-to-noise ratio (SNR) is often poor [AAA, 2011; ANSI/ASA, 2010; Neuman et al., 2010; Wroblewski et al., 2012]. Speech signals decrease in overall intensity with increasing distance, leading to reduced speech perception [Noh and Park, 2012]. Access to critical speech information is negatively affected, even at small distances typical of classroom environments [Leavitt and Flexer, 1991], and compounded by the presence of competing noise. Further, previous studies have demonstrated that listening abilities are compromised in classroom environments even in mild and unilateral HL [Hawkins, 1984; Ruscetta et al., 2005; Noh and Park, 2012] and aided listening [Hawkins, 1984; Smaldino and Crandell, 2000; Lewis et al., 2004; Mondelli et al., 2019; Snapp et al., 2020].

The most common method to reduce the negative impacts of competing noise in classroom environments for children with HL is with the use of a personal remote microphone (RM) system. In personal RM systems, the talker wears a microphone that transmits the signal wirelessly to the listener who wears a receiver coupled to their hearing device. Direct transmission of the signal to the device worn by the hearing-impaired individual increases the SNR, thereby improving speech intelligibility in adverse listening environments [Wolfe et al., 2013b, 2015c; Thibodeau, 2014]. Various RM systems and methods exist for signal delivery from an external microphone to hearing instrument. Coupling of the devices, applied gain, and transmission method/frequency can vary [Wolfe et al., 2015c]. It has been suggested that digital adaptive RM technology (_DARM_T) such as the Phonak Roger™ system which transmits the intended speaker’s voice directly to an integrated receiver in a hearing device using frequency hopping on a 2.4 GHz bandwidth should be recommended for all persons with HL who struggle to hear in complex environments such as classrooms [Wolfe et al., 2015]. Thibodeau [2014] reported improvements with _DARM_T of up to 54% compared to traditional frequency modulation and 35% over dynamic RM technology. Direct wireless transmission to an integrated receiver has been shown to be highly beneficial at improving the SNR in both cochlear implant [Wolfe et al., 2013a, 2015b, 2015c, 2016] and hearing aid (HA) users [Hawkins, 1984; Lewis et al., 2004; Thibodeau, 2014]. However, there is limited evidence of the potential added benefits of RM technology in BCD users [Snapp et al., 2020]. Snapp et al. [2020] investigated hearing in noise performance in pediatric BCD users with unilateral CHL using a Cochlear™ Baha^® processor with and without wireless audio streaming from a Mini Microphone 2+ (MM2+) digital RM system (Cochlear Bone-Anchored Solutions, Sweden) direct to the Baha^® processor. Results demonstrated hearing in noise deficits for both soft and moderate speech inputs that are significantly improved with the MM2+ personal RM [Snapp et al., 2020]. While promising, personal RM systems for BCD users only allow for a 1:1 talker to listener configuration, meaning the microphone worn by the speaker is only able to be transmitted to a single device (i.e., user). _DARM_T such as the Phonak Roger^™ can stream to multiple hearing devices at once. This combined with the reported benefits of _DARM_T [Thibodeau, 2014], makes the Phonak Roger^™ an optimal and widely used resource for classroom-based listening [Wolfe et al., 2015c; Wolfe et al., 2015]. To benefit from classroom-based _DARM_T, children with BCDs must rely on indirect transmission of the acoustic signal via an intermediate device. In such cases, the transmitting microphone system links to an intermediate device (i.e., MM2+) that then transmits the signal wirelessly to the BCD processor. Such a setup should ensure an acoustically transparent, non-interfering delivery of the signal to the sound processor. There are no studies on the effectiveness of using a relay method of signal delivery using RM technology to improve speech intelligibility in adverse listening environments in BCD users. The aim of this study was to assess speech intelligibility in adverse listening conditions when using a BCD and compare performance outcomes for direct audio streaming using a personal RM system (Cochlear^™ MM2+) to performance when using the relay method where _DARM_T (Phonak Roger^™) transmits the signal to the personal RM (Cochlear^™ MM2+) which then wirelessly sends the signal the BCD processor.

Nine pediatric patients aged 5–17 years old with unilateral or bilateral CHL were prospectively enrolled for the study. Unilateral CHL was defined as an air-bone gap of >30 dB in the treatment ear, and normal hearing (NH) in the contralateral ear as determined by a pure tone average of <25 dB HL at 500–4,000 Hz. A control group consisting of twelve adults with NH (mean age 28 years, range 23–41) participated in the study. All adult controls had NH thresholds bilaterally as determined by air conduction hearing thresholds of <20 dB HL across the independent standard audiometric test frequencies of 250–8,000 Hz, and no prior history of hearing impairment. All participants were primary English speakers. Participants were excluded if they did not meet the above-described inclusion criteria. All participants and their guardians (where applicable) consented to participate in the study, and all study procedures were approved by the University of Miami Institutional Review Board.

This study was designed as a prospective, within-subject repeated-measures experiment. Each subject served as their own control. Recorded commercially available Pediatric AzBio sentences [Spahr et al., 2014] were used for the speech in noise stimuli in the presence of a 10-talker babble. All sentences were presented in the sound field in a 4 m × 4 m × 2 m acoustically treated sound booth with speech presented from the front at 0° azimuth and the noise signal presented from four separate single-cone loudspeakers located at 30°, 135°, 225°, and 330° azimuth. The noise was maintained constant during each sentence list. Stimuli were generated in MATLAB (Mathworks, Natick, MA) and processed through a TDT RX8 real-time multichannel processor (Tucker Davis Technologies, Alachua, FL) and Crown Audio CT-8150 8-channel amplifiers. The output was delivered via Avantone MixCube single-cone speakers (Avantone Pro, New York).

Participants were tested at three SNRs (0 dB, +5 dB, and +10 dB) with noise fixed at 70 dB, A-weighted (dBA) to approximate the average noise level experienced in an occupied classroom [Wang and Brill, 2021]. Pediatric participants were tested in the BCD only, direct (BCD + RM), and relay (BCD + RM + _DARM_T) listening configurations. The participant’s everyday hearing configuration was maintained for BCD listening. The experimental conditions are presented in Figure 1. In total, pediatric participants underwent nine listening conditions: 1) BCD_-10_SNR, 2) BCD_0_SNR, 3) BCD_+5_SNR, 4) BCD+RM _-10_SNR, 5) BCD+RM _0_SNR, 6) BCD+RM _+5_SNR, 7) BCD+RM+ _DARM_T _-10_SNR, 8) BCD+RM+ _DARM_T _0_SNR, and 9) BCD+RM+ _DARM_T _+5_SNR.

For the NH control group, HL was simulated using deeply inserted bilateral ear plugs (BP) seated flush with the subject’s tragus. Controls were tested at −10 dB SNR for all listening configurations. Controls were also assessed in HA listening configuration as a control experiment to compare direct streaming of the _DARM_T to compatible device (Fig. 1d, e), to the direct BCD listening configuration (BCD + RM). All controls used a unilateral HA (UHA) in the right ear with contralateral plug (UHA_CP). The HA configuration was tested last so as not to disrupt ear plug placement and ensure consistent attenuation across conditions. In total, controls underwent six listening conditions: 1) BP, 2) BP + BCD, 3) BP+BCD+RM, 4) BP+BCD+RM+ _DARM_T, 5) UHA_CP, and 6) UHA_CP + _DARM_T.

Per American Academy of Audiology Remote Microphone Guidelines [American Academy of Audiology, 2008], the transmitting microphones were positioned 6 inches away from the center of the cone of the loudspeaker presenting the speech, and stimulus and noise levels were verified at the location of the RM and at ear level of the listener with signals calibrated individually to ensure the designated SNR as measured at the location of the listener.

Before the experiment started, the participant was familiarized to the task by completing a practice list with speech presented at 70 dBA and noise at 50 dBA to ensure understanding of the task and ability to complete the experiment. One sentence list of 10 sentences was presented for each experimental condition, totaling 9 lists for pediatric participants and 6 lists for adult controls. SNR configuration and sentence list were randomized across participants to control for order effects. Participants were instructed to repeat the sentence or to repeat as many words as possible in the event they only heard a portion of the sentence. They were advised that the difficulty of the task would vary randomly and were provided opportunities for breaks between each trial. The experimenter wore a headset wired to a microphone worn by the participant to ensure accurate recording of responses in the presence of the background noise.

A Baha^® 5 sound processor on a Softband (Cochlear Bone-Anchored Solutions, Sweden) was used for all BCD conditions. The BCD was programmed using in-situ bone conduction thresholds prior to experimental testing. For the RM conditions, the BCD was wirelessly coupled to the MM2+. The MM2+ allows for direct wireless audio streaming of the audio signal to the Baha^® 5 sound processor without the need for an intermediary device using near-field magnetic induction [Wolfe et al., 2015b].

A Phonak Sky™ B90 SP behind-the-ear HA was used for the HA listening configurations (UHA_CP, and UHA_CP + _DARM_T). The UHA was configured with a standard ear hook and a universal insert ear tip coupled to standard size 13 tubing. The HA was programmed and electroacoustically verified for a moderate (flat 45 dB) HL.

In the UHA_CP + _DARM_T condition, the UHA’s battery door was replaced with a Phonak _DARM_T 18 integrated receiver, which was then paired to the Roger Pen™ transmitting microphone to allow direct audio streaming to the UHA. In all the BCD + _DARM_T conditions, a Roger X^™ (02) (Phonak, Sonova AG, Stäfa, Switzerland) receiver was connected using the 3-pin euro plug on the base of the MM2+ (Fig. 1c), which allows hearing instruments (i.e., BCDs) that do not have an integrated Roger^™ receivers to be compatible with the Roger Pen^™ [Phonak, 2019]. The Roger Pen^™ was set to lanyard mode. In this mode, an adaptive beamformer yielding a directional microphone characteristic is applied. Default gain was maintained with an accessory mixing ratio of 1:1.

All devices were electroacoustically verified using the Audioscan^® Verifit2 device. Maximum output of the devices was verified prior to ensure an 85 dB SPL pure tone sweep was comfortable and did not exceed uncomfortable loudness levels. Controls were asked to report on perceived comfort and tolerance when listening with both the BCD and UHA. Both the RM and _DARM_T systems were objectively measured for transparency (as per manufacturer guidelines, transparency was measured with the Roger™ in verification mode) and the Audioscan Verefit^® skull simulator was used for BCD verification [AAA, 2011]. Volume controls were deactivated and were not manipulated by the participant or experimenter. Prior to each test condition for each participant, the transmitting microphone connection was verified via listening check.

A repeated-measures analysis of variance (ANOVA) was used with two within-subject factors (listening configuration and SNR level) to assess performance across the listening conditions to determine the effect of listening configuration on speech understanding. Post hoc multiple comparison analysis was run with a 0.05 significance level to assess for differences within the experimental conditions. Pairwise comparisons with Bonferroni corrections for multiple comparisons were used to further examine significant main effects. All analyses were performed using SPSS^® software (version 26.0; New York: IBM Corp^®).

The participants ranged in age from 6 to 17 years (M = 12, SD = 3.5). There were 5 males and 4 females. Seven participants had unilateral and two had bilateral CHL. Participants had limited prior exposure to RM technology with no participants reporting regular use of RM technology. The mean air conduction pure tone average of the test ears was 64 dB ± 9 dB with an air-bone gap of 50 dB ± 6 dB. These population values are considered standard criteria for BCD users for the CHL indication. A summary of speech intelligibility scores across all conditions is presented in Table 1. For analysis, percent correct scores were converted to RAUs [Studebaker, 1985].

A repeated-measures ANOVA was calculated to determine whether there were any statistically significant differences between device use on speech performance in the −10 SNR condition. Analyses revealed significant differences between listening configurations (F(4,8) = 155.14, p < 0.001, η_p² = 0.99). Post hoc comparisons indicated that the speech intelligibility score on the Pediatric AzBio for the BCD+RM condition (mean RAU = 92.12, SD = 14.44) significantly improved from the BCD alone condition (mean RAU = −17.93, SD = 2.78; p < 0.001), and from the BCD+RM+ _DARM_T condition (mean RAU = −15.78, SD = 9.77; p < 0.001). However, no significant improvement in performance was observed with the BCD+RM+ _DARM_T from the BCD alone condition (Fig. 2). When the _DARM_T was coupled to the HA, however, a significant improvement in intelligibility is observed, with RAU by 70.9 from the HA alone condition (mean RAU = 22.45, SD = 22.5) to the HA + _DARM_T condition (mean RAU = 93.4, SD = 16.24). There was no significant difference in speech intelligibility when listening with the HA + _DARM_T from listening with the BCD+RM (mean RAU = 92.12, SD = 14.44).

Figure 3 shows the speech in noise performance for the pediatric CHL participants at +5 dB, 0 dB, and −10 dB SNR for the device, direct, and relay configurations (Fig. 1a–c, respectively). A two-factor repeated-measures ANOVA was calculated to determine whether there were any statistically significant differences between device configuration and SNR condition. Analyses revealed there was a main effect for device use on speech performance (F(2,7) = 45.44, p < 0.001, η_p² = 0.850). Specifically, there was a significant difference between the three listening configurations, with the BCD + RM conditions having significantly higher speech intelligibility compared to both the BCD alone and relay configurations (p < 0.001).

There was a significant main effect of noise (SNR condition) on performance (F(2,7) = 92.17, p < 0.001, η_p² = 0.920.) Specifically, there was a significant difference in performance between +5 dB SNR and −10 dB SNR, with performance scores at −10 dB SNR being significantly worse (p < 0.001). Also, there was a significant difference between the 0 dB SNR and −10 dB SNR conditions, with performance scores in the −10 SNR condition being significantly worse (p < 0.001; Fig. 3).

There was also a significant interaction between device configuration and noise SNR condition (F(4,5) = 16.58, p < 0.001, η_p² = 0.675). As can be seen in Figure 3, performance decreases for all device configurations at the poorest SNR, but the BCD and the relay configuration did not significantly differ from one another. The differences, however, are much greater compared to the direct configuration, suggesting that those relying on a BCD alone or the relay configuration are at a far greater disadvantage in adverse listening situations than those using the direct RM configuration.

Figure 4 shows the speech in noise performance for the control subjects compared the pediatric CHL participants when listening at the most challenging SNR, −10 dB. There was no significant difference between groups for any of the configurations.

Figure 5 shows an illustration of the electroacoustic verification of acoustic transparency for the three listening configurations. The BCD processor meets prescriptive targets across the frequency range (solid line). Output in the direct (BCD + RM) configuration is close to target and within +/− 2 dB of the processor output indicating this method is able to achieve transparency (dashed line) [AAA, 2011]. However, responses demonstrate significant reduction in output for the relay configuration (BCD + RM + _DARM_T) (dotted line). The output of the BCD alone configuration was 65 dB for a 65 dB speech input, whereas the output for the relay configuration was 10–20 dB below the desired signal.

The results of this study demonstrate that significant gains in speech intelligibility are consistently observed for direct audio streaming configurations and are confirmed in adult controls with both BCD and HA listening. Conversely, coupling of the adaptive digital _DARM_T to the personal RM negatively affects signal transparency, and no hearing in noise improvements is observed. Behavioral findings are supported by objective verification of the signal transparency between the RM and the BCD (Fig. 5).

Various RM approaches and methods exist; the most common is _DARM_T, which is widely used for assisting hearing-impaired listeners in the classroom [Wolfe et al., 2015]. Previous studies on the benefits of RM technology in HA and cochlear implant users have demonstrated significant benefits for listening in noise when using the Phonak Roger™ _DARM_T [Wolfe et al., 2015a, 2015c]. Despite this, benefit of such systems in BCD users has been largely overlooked. Our results support previous findings [Snapp et al., 2020] demonstrating significant gains in speech understanding in noise in pediatric BCD users when using direct wireless audio streaming via the Cochlear^™ MM2+ personal RM system coupled to a Baha^® processor. In classroom environments, however, children using BCDs often have to rely on indirect transmission of the acoustic signal for commonly used classroom-based RM listening (e.g., “_DARM_T” or digital adaptive microphone).

The data in this study reveal poor acoustic signal transparency using a relay configuration where the _DARM_T does not directly deliver the signal to an integrated receiver in the hearing device but has to rely on an intermediate device to relay the signal to the sound processor. Listeners demonstrated no improvement in speech intelligibility when listening in the relay configuration. Conversely, the direct configuration provided significant gains in speech intelligibility, even at negative SNRs. In the most challenging listening situation (−10 dB SNR), participants demonstrated effectively no speech understanding abilities despite being aided with either a BCD or HA. This is an important consideration, as children with hearing devices may be subject to misperceptions that their hearing device alone will sufficiently improve the SNR in adverse classroom environments. An SNR of +15 dB is recommended for children with HL, yet classroom SNRs typically vary from −7 to +4 [Crandell and Smaldino, 2000]. When directly coupled to the RM, pediatric BCD users improved on average 72% and controls by 88% (Table 1). Control experiments comparing performance using _DARM_T in a direct configuration to a HA with an integrated receiver also resulted in large differences (Fig. 2), demonstrating acoustic transparency was not a result of the RM used, rather the configuration. Children with unilateral CHL are more likely to have delayed intervention and experience delays in speech and language development [Banga et al., 2013; Jensen et al., 2013]. It is possible that such delays may contribute to overall outcomes; however, experiments in controls with NH verified that the relay configuration did not significantly improve performance regardless of SNR. While there was a small but significant difference in pediatric BCD users compared to adult controls with NH in the BCD+RM direct configuration (Fig. 4), this could be attributed to the recognized increase in SNR required by children to obtain similar performance in noisy environments as adults [Boothroyd, 1997].

Coupling of a hearing device to an external microphone should ensure a transparent, non-interfering delivery of the signal to the sound processor. In order to achieve transparency, the acoustic signal from the hearing device and the signal from the RM should be within ±2 dB across the frequency range to verify that the frequency and amplitude characteristics of the signal are not altered [AAA, 2011]. Objectively measuring performance of an amplification system, in terms of sound output and amplitude, is an imperative part of gauging benefit from assistive listening devices. Without achieving transparency, the gain of the signal is either reduced or overamplified, which can compromise access to critical speech signals.

It is possible that differences in the threshold for voice activation for the MM2+ versus the Roger^™ resulted in the signal level during speech in noise experiments activating the MM2+ when it was used as the RM, but not when the Roger Pen^™ was used. However, electroacoustic measures of transparency confirmed attenuation of the signal when a relay configuration is used. Phonak Roger^™_DARM_T is not directly compatible with the BCD. The signal is significantly attenuated when relayed to the processor through an intermediate device. This may be a result of an impedance mismatch between MM2+ and Roger X^™. In 2003, Platz suggested that if the microphone impedance differed than that of the _DARM_T system, it may account for variations in the _DARM_T signal strength which can vary by a substantial amount of gain. While this is often overcome by the use of advanced integrated systems which are directly fed into a high-impedance amplifier (integrated Roger™ receivers, etc.), the issue may still be relevant for devices that must connect in parallel to the _DARM_T in order to deliver the signal (such as the configuration of the BCD + Roger™ through means of the MM2+). Similar findings were observed by Sousa et al. [2019] when comparing signal transparency of cochlear implants using the Roger™ system compared to MM2+. Their findings suggest that signal transparency was automatically achieved in 50 out of 51 subjects using the MM2+; however, the Roger™ system only achieved automatic transparency on 28 processors and required gain adjustments and modifications by the speaker to achieve transparency in the other 23 subjects. Although in this study individual gain adjustments to overcome transparency issues were not performed, we would expect that doing so would result in similar findings as that presented in CI users. Further, the difference in device (BCD vs. cochlear implant) should not result in differences in terms of efficiency of signal transmission as in both our study and that of Sousa et al. [2019] reductions in performance were isolated to the relay method of transmission.

While the present study did not endeavor to overcome loss in signal transparency, it is indeed possible that losses in gain when relying on a relay configuration may be overcome through gain adjustments. Although, this should not be assumed, and verification of performance for a given RM configuration must be conducted to estimate benefit for classroom-based listening. This can prove challenging as children are often fit with hearing assistive technology in the school system. It is essential that such strategies are recognized by clinicians and effectively communicated to patients and educational support teams. Additionally, this study did not explore the basis for the mismatch observed in various RM technologies. Future work toward verifying the contributing factors and improvements in achieving signal transparency for classroom-based RM technology is warranted. In our control studies, the Roger™ receiver configuration for the HA experiments (integrated) differed from the BCD conditions (Roger™ X). We do not expect this alone significantly contributed to differences in outcomes between groups or between the control BCD and HA listening conditions, as others have shown no statistically significant effect of type of CI Roger™ receiver on speech recognition [Wesarg et al., 2020].

Lastly, the BCDs used in this study were limited to a single manufacturer. The compatibility with _DARM_T to other BCD systems remains unclear and should be the subject of future investigations.

This study investigated speech intelligibility outcomes in pediatric BCD users under adverse listening conditions with direct versus indirect RM streaming. Pediatric BCD users experience significant difficulties hearing in noise that persists even when aided. The use and benefit of different types of RM technology to overcome poor SNRs in unfavorable listening environments have been well studied in individuals with HAs and cochlear implants; however, performance studies in BCD users are entirely lacking. Unlike in cochlear implant and HA users, BCD users are unable to direct connect to commonly used classroom-based hearing assistive technology. The relay configuration where signals are indirectly streamed from _DARM_T to a personal RM to the processor (relay method) negatively affects signal transparency, and no hearing in noise improvements is observed. Significant gains in speech intelligibility are consistently observed for direct streaming methods. Behavioral findings are supported by objective verification of the signal transparency between the RM and the BCD.

This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki. All participants and their guardians (where applicable) consented to participate in the study, and all study procedures were approved by the University of Miami Institutional Review Board (Reference number: 20190308) and written informed consent was obtained from each participant and their parent/legal guardian for pediatric participants.

This study was funded through Cochlear Corporation. The authors have no other conflicts of interest to declare.

This study was funded through Cochlear Corporation.

Chrisanda Sanchez contributed to writing the original draft and analysis of the manuscript. Kari Morgenstein contributed to the conception and design of the study and participated in data collection. Hillary Snapp contributed to the study design, performed data collection, final analysis, final review, and completion of the manuscript.

The data that support the findings of this study are openly available on Clinicaltrials.gov at https://www.clinicaltrials.gov, reference number [NCT 04147611].