Can You Hear Me Now: A Meta-Analytical Perspective of the Benefits of Frequency-Modulated (FM) Systems for People with Cochlear Implants

Abstract: 

Cochlear implants (CIs) can significantly improve hearing for people with severe-to-profound hearing losses, but they do not restore hearing in noise. Frequency-modulated (FM) systems, however, can help combat the interference of background noise. Three kinds of FM receivers can be used with a CI: (1) a classroom soundfield, (2) a desktop soundfield, or (3) a direct-audio input (DAI). There is no consensus, however, on which type of FM system provides the best performance. Speech-recognition data were extracted from eight studies for FM system conditions in noise: (1) CI only, (2) classroom soundfield, (3) desktop soundfield, and (4) DAI. A meta-analysis was performed to compare the improvements in speech recognition with the different receivers. The best performance was with the DAI system, followed by the desktop soundfield. The classroom soundfield provided little or no benefit. Thus, to provide optimal hearing in noise, DAI systems should be the first type of FM system considered for people with CIs.

Table of Contents: 

    Introduction

    While hearing aids can amplify sounds for people with mild to profound hearing loss, listening becomes more difficult in adverse situations (e.g., a noisy restaurant). As with all hearing losses, amplification from hearing aids will not restore hearing. Cochlear implants (CIs) are an option for some people with severe-to-profound hearing losses who do not show benefit from use of hearing aids. With such a significant hearing loss, most of the hair cells of the cochlea are not functioning, disallowing benefit from hearing aids. The internal device of a CI consists of a magnetic receiver-stimulator that is connected to several electrodes that have been surgically threaded through the cochlea. The external device includes a magnetic headpiece, which attaches to the magnet of the internal implant, and a speech processor, which can be worn on the body or behind the ear. As with people who use hearing aids, people with CIs also have substantial difficulty in noisy situations.

    Although CIs significantly improve hearing for people with severe-to-profound hearing losses, these devices often do not restore normal hearing, especially in noise (Schafer & Thibodeau, 2004). Furthermore, the speech-recognition abilities of the person with hearing loss often decline in adverse conditions, most likely because the minute differences between phonemes are difficult to discriminate. One solution to reduce the interference of background noise is the use of an assistive listening device, such as a frequency-modulated (FM) system. An FM system allows the listener to focus on both the signal from the desired stimulus, which is difficult to hear over the background noise without the aid of the system, as well as other stimuli in the listening environment. FM systems are ideal for numerous environments including educational and work settings, entertainment venues, such as live theatre and concerts, and places of worship. An FM system consists of a microphone into which the speaker talks, a receiver located close to or attached to the CI, and a transmitter that carries the acoustic signal wirelessly from the microphone to the receiver. There are three types of FM receivers that can be used with a CI: a classroom soundfield, which utilizes stationary speakers throughout a room; a desktop soundfield, which uses a small, portable speaker that is compact enough to sit on a desk; and a direct-audio input (DAI), which is wired directly to the CI. Currently there is no comparison available to suggest which type of system provides the greatest benefit for adults and children who use CIs.

    Research Topic

    The purpose of this project is to compare the improvements in speech recognition in noise for people who have CIs when using classroom soundfield, desktop soundfield, and DAI FM systems.

    Literature Review

    In order to examine and compare the benefits of FM systems, the following three sections will discuss the preexisting data about performance with the different types of FM systems available.

    Classroom Soundfield FM Systems

    In many classrooms across America, classroom soundfield systems provide amplification for children who are hearing impaired and normal hearing. Recently, the Acoustical Society of America (ASA) released a position statement regarding amplification in the classroom stating, “Sound amplification should not be routinely employed in typical small mainstream classrooms” (Acoustical Society of America, n.d.). This document mentions that reverberation in a room often hinders the effectiveness of a classroom soundfield system. In addition, the ASA reports only the communication mode from the teacher to student is impacted, while other modes of communication such as the student to teacher and between students are hindered (Acoustical Society of America, n.d.). Despite these findings, classroom soundfield systems are available and frequently used today. These are often recommended for children with CIs because they do not require specialized cords and are easy to troubleshoot.

    Several studies have researched the benefit of wall-mounted, classroom soundfield amplification for people with CIs (Anderson, Goldstein, Colodzin, & Iglehart, 2005; Crandell, Holmes, Flexter, & Payne, 1998; Iglehart, 2004). Two groups of researchers examined classroom soundfield in relation to other forms of FM-system amplification in the presence of noise (Anderson, et al., 2005; Iglehart, 2004). Iglehart (2004) evaluated the benefit of classroom soundfield systems in a room with ideal acoustic characteristics and another with poor acoustics. He evaluated word recognition of 14 children, ages 6.4 to 16.1 years old, while using the no FM system, the classroom FM, and a desktop FM system. The classroom system provided an overall average of 12.4% benefit over the no-FM system condition in the acoustically poor classroom and 9.8% benefit for the classroom with good acoustics. Anderson, et al. studied six children with an average age of 11.45 years old in four hearing conditions: a no-FM system, a classroom soundfield system, a desktop soundfield system, and a DAI system. At a +10 dB signal-to-noise ratio (SNR), the use of the classroom FM system resulted in a 12.5% decrease in sentence recognition in noise when compared with using a CI with no FM system.

    Crandell, et al. (1998) examined the speech recognition in multibabble background noise of ten children and eight adults with CIs, with and without a soundfield FM system. Using a total of eight conditions for children and adults, the soundfield system use yielded slightly greater gains when compared to the CI alone; however, these gains were not statistically significant. For the adults, there was a 5.7% increase in the pattern perception and a 3.7% increase in spondee words, and for children there was a 1.2% increase in spondee words and a 0.5% increase in monosyllables. In all other circumstances, the soundfield system condition did not yield any improvement in speech recognition (Crandell, et al., 1998). According to results in these studies, there seem to be limited improvements in speech recognition when using soundfield FM systems in noise, ranging from 0.5% to 12.5%.

    Desktop Soundfield FM Systems

    Three studies measured the benefit of portable desktop soundfield systems for people using CIs (Anderson, et al., 2005; Schafer & Thibodeau, 2003, 2004). In addition to classroom soundfield, Anderson, et al. tested desktop soundfield systems and reported a significant increase of 10% in sentence recognition in noise for the desktop soundfield FM system over the no-FM condition. In both of the Schafer and Thibodeau studies (2003, 2004), sentence recognition in noise was evaluated for children and adults with CIs using three different FM systems: desktop, body worn, and DAI. Schafer and Thibodeau (2003, 2004) investigated speech perception of children with CIs with the average age of 8.9 years old and studied adults with CIs from age 20 to 58 years old, respectively. The desktop system provided a speech perception in noise benefit of 25.2% for children and 9% for adults over the no-FM system condition.

    Direct-Audio Input (DAI) FM Systems

    Seven studies have examined the advantage of personal DAI FM systems compared to performance with a CI alone (Aaron, Sonnevelt, Arcaroli, & Holstad, 2003; Anderson, et al., 2005; Catlett & Brown, 2003; Davies, Yellon, & Purdy, 2001; Schafer & Thibodeau, 2003, 2004; Thibodeau, Schafer, Overson, Whalen, & Sullivan, 2005). All studies compared speech perception with their CI device alone in a noisy environment to their perception with a DAI FM system. Speech perception in a +5 dB SNR with the DAI receiver improved ranging from 29.8% to 56.3% over the CI alone (Aaron, et al., 2003; Catlett & Brown, 2003; Thibodeau, et al. 2005). Davies, et al. (2003) measured accuracy in 0 SNR and -3 SNR situations and found an increase of 0.8% and 9.1%, respectively, over the CI alone. Overall, personal FM systems positively impacted speech perception when compared to a CI alone by providing an increase in speech perception ranging from 9.1% to 56.3%, and many researchers reported significant improvements. In fact, many studies show that DAI is superior to classroom or desktop soundfield systems (Anderson, et al., 2005; Schafer & Thibodeau, 2003, 2004).

    Research Methodology

    To examine the relative benefit of the three types of FM systems, a meta-analysis was conducted. A meta-analysis provides a means to compare the weighted-average data from within each FM system condition (i.e., classroom soundfield, desktop soundfield, DAI). In addition, the meta-analysis will combine all of the FM conditions to illustrate whether or not FM systems, in general, provide a significant benefit.

    Prior to starting this project, literature relating to CIs and FM systems was collected from peer-reviewed journals and poster presentations. Though unpublished, data from poster presentations at national and international conventions were collected because of the limited number of published articles available. In order to be used in this meta-analysis, previous studies had to meet the following five criteria: testing done in Standard English, testing of speech recognition in noise, testing of no-FM and FM conditions, providing mean percent-correct and standard deviations for the data, and a fixed intensity speech and noise stimuli at a +5 to +10 dB SNR. From the original 11 sources found, 8 studies met the required criteria. Wood, et al. (2005) and Schafer and Thibodeau (2006) were excluded because adaptive testing techniques were used in their studies. Davies, et al. (2001) was eliminated because only SNRs of 0 and +3 dB were utilized.

    Data, including sample size, mean percent-correct performance, and standard deviation, were extracted for all listening conditions: no FM system in noise (CI alone), classroom soundfield FM system, desktop soundfield FM system, and DAI FM system. In addition to mean, standard deviation, and sample size, other important characteristics of these studies are explained in Table 1. All of the subjects in the studies were implanted monaurally, and most participants were several years post-implant, though not all studies mentioned an exact duration of CI use. Each condition comparing no-FM to FM-system scores was treated as an independent experiment to increase sample size, giving a total of 29 different experiments. Ten were of classroom FM systems, 4 utilized desktop FM systems, and 15 used DAI FM systems.

    Differences between no-FM and FM-system conditions were calculated for each individual experiment using The Number Cruncher Statistical System (2004) software program. The program was also used to calculate weighted average, percent-correct difference scores, and 95% confidence intervals (CI95) for each type of system. Using a random effects model, a meta-analysis was performed and a chi-square value was computed to test the null hypothesis (no FM benefit). Results of the main analysis allowed us to determine if FM systems provide a significant benefit over no-FM conditions. Furthermore, the results show if one type of FM system provides better gains in speech recognition than the others.

    After the primary analysis, post hoc meta-analyses were conducted to explore how three other factors contributed to better speech recognition in noise with certain types of FM systems. These factors included the age of the listener (e.g., child, adult), the type of background noise presented (e.g., speech noise, babble noise), and the type of internal CI used by the listeners.

    Results

    Main Analysis

    Degrees of freedom, weighted-average differences, CI95, chi-square, and probability level values were calculated for each of the three FM conditions, and the values for the main analysis are in Table 2. According to the weighted-average differences, the average benefits with the three FM systems were: 3.7% (CI95 ± 7.3) for the classroom soundfield system, 21.6% (CI95 ± 7.4) for the desktop soundfield system, and 36.8% (CI95 ± 7.0) for the DAI system. The data in the forest plot in Figure 1 allows for the comparison of benefit across different systems by illustrating the weighted-average difference and CI95 for each type of system.

    The main meta-analysis confirmed significant benefit for the desktop soundfield (Χ2 [4, 26] = 47.13, p < 0.000) and the DAI system (Χ2[15, 26] = 564.53, p < 0.000) relative to the no-FM conditions in noise. However, the classroom soundfield did not provide significant benefit (Χ2 [7, 26] = 12.81, p = 0.0770). As shown in Figure 1, significant individual experiments and significant weighted-average differences for the types of FM systems did not overlap with the 0% mean difference line (no FM benefit). The confidence intervals for the three types of systems do not overlap, suggesting significantly different weighted-averages across the devices (p < .05). The best performance was with the DAI system, followed by the desktop soundfield system. The classroom soundfield provided little to no benefit in any experiments.

    Post Hoc Findings

    Three additional meta-analyses were conducted to examine any effects of the age of the listener, the type of background noise used, and the type of internal CI used by the listeners. Degrees of freedom, weighted-average differences, CI95, chi-square, and probability level values of these experiments can be found in Table 3.

    The first post hoc meta-analysis, shown in Figure 2, was conducted to explore the relationship between age (e.g., child or adult) and the amount benefit from two types of FM systems, classroom soundfield and DAI. A total of four conditions were examined: adult use of classroom soundfield, child use of classroom soundfield, adult use of DAI, and child use of DAI. With the classroom soundfield, both age-groups provided limited or no benefit, as the weighted-average differences were -0.7% (CI95 ± 27.7) for the adult group and 3.0% (CI95 ± 9.0) for the child group. The meta-analysis illustrates the classroom soundfield for the adult category is not significant (Χ2 [2, 22] = 0.1335, p = 0.9354), whereas the classroom soundfield for the child category is significant (Χ2 [5, 22] = 12.67, p = 0.0267). For the DAI conditions, the weighted-average differences showed a benefit of 33.2% (CI95 ± 18.4) for the adult and 37.1% (CI95 ± 7.5) for the child. The DAI results were significant for the adult category (Χ2 [2, 22] = 12.63, p = 0.0018) and child category (Χ2 [13, 22] = 0.0267, p < 0.0001). The confidence intervals for the DAI conditions overlapped, suggesting there are no effects between age and DAI FM system use (p > .05). Similarly, the confidence intervals for the classroom soundfield conditions overlapped on the forest plot, implying there is no effect for age and the use of a classroom soundfield FM system (p > .05). Even though the classroom soundfield CI95 for the child crosses the 0% benefit line on the forest plot, the pvalue (p = 0.0267) is significant. However, while statistically significant, a simple 3% average benefit is not clinically significant.

    The meta-analysis illustrated in Figure 3 explored the relationship between the type of FM system (desktop soundfield and DAI) and the internal portion of the CI. The Nucleus22 (N22) and Nucleus24 (N24) models from Cochlear Corporation served as the internal devices, giving a total of four conditions. The results for the desktop soundfield were 24.1% (CI95 ± 10.6) for the N22 internal device and 26.8% (CI95 ± 17.1) for the N24. The DAI results were 33.1% (CI95 ± 3.1) for N22 and 42.3% (CI95 ± 5.1) for N24. The meta-analysis verified these benefits as significant for the desktop soundfield used with the N22 (Χ2 [3, 22] = 20.51, p = 0.0001), desktop soundfield with the N24 (Χ2 [3, 22] = 37.08, p< 0.0001), DAI FM system with N22 (Χ2 [5, 22] = 456.14, p < 0.0001), and DAI FM system with the N24 (Χ2 [11, 22] = 414.34, p < 0.0001). Although the CI95 overlapped for the N22 and N24 internal implant in the desktop soundfield FM-system condition (p > .05), the confidence intervals for the type of internal implant did not overlap for the DAI system (p < .05), suggesting the N24 device had greater gains than the N22.

    The last meta-analysis, found in Figure 4, investigated the effect of the type of background noise (e.g., babble, speech noise) when testing speech recognition in desktop soundfield and DAI FM-system conditions. The results for the desktop soundfield were 18.8% (CI95 ± 14.9) for the babble noise and 21.2% (CI95 ± 13.7) for the speech noise, while results for the DAI system were 41.2% (CI95 ± 9.6) for the babble noise and 33.2% (CI95 ± 9.4) for the speech noise. The meta-analysis proved values to be significant for the desktop soundfield with babble (Χ2 [2, 19] = 31.71, p < 0.0001), desktop soundfield with speech noise (Χ2 [6, 19] = 362.69, p < 0.0001), DAI system with babble (Χ2 [2, 19] = 15.89, p = 0.0004), and DAI system with speech noise (Χ2 [9, 19] = 201.84, p < 0.0001). All of the CI95 values overlapped for desktop soundfield and DAI systems (p > .05), suggesting there are no effects of the type of background noise on the amount of benefit gained from FM system use.

    Discussion and Clinical Implications

    Although 14% and 50% of the individual experiments within the classroom and desktop soundfield FM-system categories, respectively, provided significant benefit over the no-FM condition, DAI FM systems provided the greatest benefit, as 80% of all the individual DAI experiments showed significant FM-system benefit. When each category of FM system was weighted and averaged, both the desktop and DAI FM systems showed significant improvements in speech recognition while the classroom showed no benefit. The confidence intervals for the desktop and DAI systems did not overlap, illustrating the DAI FM system provides a greater benefit than the desktop soundfield.

    Whereas age of the listener and background noise had no effect on FM-system use, the type of internal implant used with the DAI FM systems made a significant difference. Although both the N22 and N24 used in conjunction with the DAI FM system provided a benefit for the person with a CI, on average, users of the N24 internal device received a greater benefit than those with the N22 internal implant. This suggests newer technology may provide greater benefits with DAI FM systems. The meta-analysis, however, did not show any effect between desktop soundfield systems and the type of internal implant.

    Conclusion

    DAI systems provide optimal benefit to the CI user when compared to all other FM systems. They provide an average benefit of 36.8% improvement in speech recognition over the no-FM conditions. Following the DAI system was the desktop soundfield system with an improvement of 21.6%. Despite the fact that desktop soundfield systems do provide some significant benefit to the CI user, DAI systems should be the first type of FM system considered for adults and children with CIs in order to receive the greatest gain in speech recognition as possible. Furthermore, classroom soundfield FM systems, despite being cost effective, should not be utilized because an increase of only 3.7% in speech recognition is provided. In reality, a mere 4% benefit would not be substantial for a child inside a classroom with a myriad of background noise from classmates, classroom equipment, and reverberation within the room.

    References

    • Aaron, R., Sonneveldt, V., Arcaroli, J., & Holstad, B. (2003, November). Optimizing microphone sensitivity settings of pediatric Nucleus 24 cochlear implant patients using a Phonak MicroLink CI+ FM system. Poster presented at ACCESS – Achieving Clear Communication Employing Sound Solutions, Chicago, IL.
    • Acoustical Society of America, (n.d.). Position on the Use of Sound Amplification in the Classroom. Retrieved October 12, 2006, from http://asa.aip.org/amplification.pdf.
    • Anderson, K., Goldstein, H., Colodzin, L., & Iglehart, F. (2005). Benefit of S/N enhancing devices to speech perception of children listening in a typical classroom with hearing aids or a cochlear implant. Journal of Educational Audiology, 12, 14–28.
    • Catlett, D., & Brown, C. J. (2003, November). Optimal audio mix settings for pediatric Clarion cochlear implant patients using a Phonak MicroLink CI-S FM system. Poster presented at ACCESS – Achieving Clear Communication Employing Sound Solutions, Chicago, IL.
    • Crandell, C., Holmes, A., Flexter, C., & Payne, M. (1998). Effects of soundfield FM amplification on the speech recognition of listeners with cochlear implants. Journal of Educational Audiology, 6, 21–27.
    • Davies, M. G., Yellon, L., & Purdy, S. C. (2001). Speech-in-noise perception of children using cochlear implants and FM systems. The Australian and New Zealand Journal of Audiology, 23(1), 52–62.
    • Hintze, J. (2004). Number Cruncher Statistical System (NCSS) [computer software]. Kaysville, UT: NCSS Statistical Software.
    • Iglehart, F. (2004). Speech perception by students with cochlear implants using sound-field systems in classrooms. American Journal of Audiology, 13(1), 62–72.
    • Schafer, E. C., & Thibodeau, L. M. (2003). Speech-recognition performances of children using cochlear implants and FM systems. Journal of Educational Audiology, 11, 15–26.
    • Schafer, E. C., & Thibodeau, L. M. (2004). Speech-recognition abilities of adults using cochlear implants with FM systems. Journal of the American Academy of Audiology, 15, 678–691.
    • Schafer, E. C., & Thibodeau, L. M. (2006). Speech recognition in noise in children with cochlear implants while listening in bilateral, bimodal, and FM-system arrangements. American Journal of Audiology, 15(2), 114–126.
    • Thibodeau, L., Schafer, E., Overson, G., Whalen, H., & Sullivan, J. (2005, March). Clinical evaluation of the benefits provided by FM systems directly connected to cochlear implants. Poster presented at the 10th Symposium on Cochlear Implants in Children, Dallas, TX.
    • Wood, E. J., Flynn, S. L., & Greenham, P. (2005). The benefit of using an FM radio aid over distance and in noise, with the Nucleus ESPrit 3G speech processor. Poster presented at the 10th Symposium on Cochlear Implants in Children, Dallas, TX.

    Table 1: Description of Studies

    Author (Year) N Age Range (years) CI Company CI Processor Stimulus/Noise FM System Receiver Conditions
    Aaron et al. (2003) 12 4 to 12 CC SPrint Sentence/ Multibabble P MicroLink DAI
    Anderson et al. (2005) 6 7 to 13 CC 3G, SPrint, ESPrit Sentence/ Real Classroom PE 900R Vocalight, LES Desktop Sound Pack, P MicroLink C, D, DAI
    Catle et al. (2003) 4 7 to 18 AB Clarion 1.2, CII Sentence/ Speech P MicroLink DAI
    Crandell et al. (1998) 18 7 to 80 CC Spectra 22 Word/ Multibabble Audio Enhancement Omni Deluxe C
    Iglehart (2004) 14 6 to 14 CC, AB ESPrit, SPrint, S-Series, Clarion 1.2, Spectra Word/ Multibabble PE 210 SF System, PE Toteable C, D
    Schafer, Thibodeau (2003) 10 7 to 12 CC, M ESPrit 22, 3G, Tempo Sentence/ Speech PE Toteable, PE Easy Listener, AVR Sonovocation Logicom-CI, P MicroLink D, DAI
    Schafer, Thibodeau (2004) 8 20 to 58 CC Spectra, SPrint, ESPrit22, ESPrit 24 Sentence/ Speech PE Toteable, PE Easy Listener, AVR Sonovocation Logicom-CI D, DAI
    Thibodeau et al. (2005) 8 5 to 15 CC, AB 3G, S-Series, Platinum, SPrint Word/ Speech P Campus S, Comtek PR216 DAI

    Note: CI=cochlear implant; CC=Cochlear Corporation; AB=Advanced Bionics; M=MED-EL; P=Phonak; PE=Ponic Ear; SF=soundfield
    DAI=direct-audio input; C=classroom soundfield; D=desktop soundfield; SNR=signal-to-noise ratio

    Table 2: Main Analysis

    Study DF µ±CL95 X2 p
    Classroom SF 7 3.66 ± 7.29 12.81 0.0770
    Desktop SF 4 21.56 ± 7.39 47.13 <0.0001
    DAI 15 36.78 ± 7.0 564.53 <0.0001

    Note: DF=degrees offreedom; µ/CI95=average difference with 95% confidence intervals; X2=Chi-square value; SF=soundfield; DAI=direct-adio input; p=probability level

    Table 3: Post Hoc Analysis

    Study DF µ±CL95 X2 p
    Age
    A Class SF
    C Class SF
    A DAI
    C DAI

    2
    5
    2
    13

    -0.74±27.71
    3.01±8.96
    33.20±18.36
    37.08±7.52

    0.1335
    12.67
    12.63
    551.9
    0.9354
    0.0267
    0.0018
    <0.0001
    Internal
    N22 Desktop
    N24 Desktop
    N22 DAI
    N24 DAI

    3
    3
    5
    11

    24.09±10.55
    26.77±17.12
    33.10±3.05
    42.28±5.11

    20.51
    37.08
    456.14
    414.34

    0.0001
    <0.0001
    <0.0001
    <0.0001
    Type of Noise
    B Desktop
    S Desktop
    B DAI
    S DAI

    2
    2
    6
    9

    18.76±14.89
    21.17±13.68
    41.16±9.62
    33.24±9.44

    31.71
    15.89
    362.69
    201.84

    <0.0001
    0.0004
    <0.0001
    <0.0001

    Note: DF=degrees of freedom; µ/CI95=average difference with 95% confidence intervals; X2=Chi-square value; SF=soundfield; DAI=direct-audio input; p=probability level

    Figure 1: Main Analysis Forrest Plot

    Figure 1: Main Analysis Forrest Plot

    Figure 2: Effect of Age Forrest Plot

    Figure 2: Effect of Age Forrest Plot

    Figure 3: Internal Processor Forrest Plot

    Figure 3: Internal Processor Forrest Plot

    Figure 4: Background Noise Forrest Plot

    Figure 4: Background Noise Forrest Plot