Noise Health Home 

ARTICLES
[Download PDF]
Year : 2005  |  Volume : 7  |  Issue : 29  |  Page : 12--23

The benefit method: Fitting hearing aids in noise

I Svard1, KE Spens2, L Back1, BH Ahlner3, ML Barrenas4,  
1 BFM Support AB, Sweden
2 Department of Speech, Music and Hearing, Royal Institute of Technology, Stockholm, Sweden
3 Department of Audiology, Karolinska Hospital, Stockholm, Sweden
4 Department of Paediatrics, Queen Silvia Children's Hospital, Göteborg, Sweden

Correspondence Address:
M L Barrenas
Institute of Health of Women and Children, Department of Paediatrics, Queen Silvia Children's Hospital, SE 41685 Göteborg
Sweden

Abstract

The most common complaint among individuals with hearing impairment is the inability to follow a conversation when several people are talking simultaneously, a noisy listening situation which is completely different from the quiet surrounding of the conventional pure tone audiometry used as basis for the hearing aid settings. The purpose of this report was to present important characteristics of the BeneFit Method (BFM), a procedure that fits the hearing aid under simulated conditions of competing speech and also a clinical pilot evaluation study comparing the BFM to the NAL-R recommendations and also to the Logic procedure, a GN resound proprietary fitting algorithm representing a modern digital hearing aid fitting procedure. Speech recognition scores in noise (SRSN) using monosyllabic words presented under different background noise levels were evaluated on 21 randomly selected subjects with hearing impairment. The subjects were fitted with the same type of hearing aid Danalogic 163D according to the BFM procedure as well as the logic procedure, the latter developed and recommended by the manufacturer. A comparison of the SRSN when using the subjects' current hearing aid fitted according to the NAL-R procedure was also made. Only the BFM procedure provided a significant SRSN improvement compared to the unaided condition (P< 0.01) in a signal/speech-noise level of 75/65 dB corresponding to a normal cocktail party condition. Moreover, patients performed significantly higher SRSN when fitted according to the BFM, than when fitted according the Logic or NAL-R procedures. The BFM procedure, which is based on individual and functional detection of hearing thresholds in noise levels corresponding to a cocktail party condition, can improve SRSN significantly. Hearing aids should be fitted under conditions similar to those when the hearing disability is perceived the most, i.e., in an environment with background noise.



How to cite this article:
Svard I, Spens K E, Back L, Ahlner B H, Barrenas M L. The benefit method: Fitting hearing aids in noise.Noise Health 2005;7:12-23


How to cite this URL:
Svard I, Spens K E, Back L, Ahlner B H, Barrenas M L. The benefit method: Fitting hearing aids in noise. Noise Health [serial online] 2005 [cited 2022 May 22 ];7:12-23
Available from: https://www.noiseandhealth.org/text.asp?2005/7/29/12/31874


Full Text

 Introduction



Despite considerable hearing aid technology improvements, such as digitalization, a wide range of adjustment options and tailored fitting procedures especially designed to take advantage of the characteristics of a particular hearing aid, the user's satisfaction was unchanged in year 2000 compared to 1990. [1],[2] Still, the most common complaint is the inability to follow a conversation when several people are talking simultaneously, a noisy listening situation completely different from the quiet surrounding of the conventional pure tone audiometry used as basis for the hearing aid settings. So far, no current hearing aid fitting method is based on the assessment of detection thresholds in noise at normal and loud speech levels. [3] Moreover, when compared on coupler (IEC-136, 1961), the prescribed gain could differ by as much as ± 10 dB between prescriptive methods. [4] Finally, since the pure-tone audiogram only reflects a reduced sensitivity to sound, other important aspects associated with impaired hearing, such as distortion, masking and non-linearities might not be taken into account. For example, speech intelligibility in noise might depend on increased sensitivity to cochlear distortion. [5] The BeneFit Method (BFM), a procedure where the hearing aid is fitted under simulated conditions of competing speech, was developed bearing these concerns in mind.

The functional state of the BFM is now reached, bringing about good results at several Hearing Aid Centres in terms of satisfied hearing aid users and improved speech recognition score in noise (SRSN). The purpose of this report is to present important characteristics of the BFM and also a clinical pilot evaluation study comparing the BFM to the NAL-R recommendations[6] and the Logic procedure (a GN resound proprietary fitting algorithm representing a modern digital hearing aid fitting procedure).

 The BFM



According to the BFM, the main key to optimize speech intelligibility under competing speech conditions is to find a hearing aid setting that offers a good balance between the gain in the high and low frequency ranges. Therefore, not only is the BFM based upon individually assessed measurements of the signal detection threshold in noise (SDTN). It also focuses at optimizing listening under adverse conditions (cocktail noise) by bringing the aided SDTNs at several noise levels and frequencies into the range of that for people with normal hearing [Table 1]. Since an expanded high frequency range improves both listening comfort and speech audibility [7],[8] and especially in background noise, [9] it is assumed that detection of speech sounds up to 6 kHz is optimized, when the aided SDTNs at noise levels ranging from 45 dB SPL and up to 75 dB SPL (corresponding to quiet and loud competing speech, respectively) are within that for the normative SDTNs. The selection of SDTN assessments is considered a feasible simulation of the most common adverse listening conditions. Accordingly, no prescriptive formula based on the hearing thresholds can be provided for the BFM. For all other present fitting procedures, on the other hand, a certain algorithm is used when setting the gain in the available frequency bands, which depends on the characteristics of that particular hearing aid. Usually, such settings give a too low amplification of the high frequencies relative to the low and middle frequencies. This constitutes a risk that the hearing aid user will not receive the high frequency information, which might be drowned by distortion products created from the lower frequency ranges of the cochlea. These differences between conventional methods and the BFM are probably crucial.

Test stimuli, speech-weighted noise and calibration

Computer-generated sequences ad modum Victoreen [10] [Figure 1] were chosen as test stimuli, because the Victoreen signal is easily recognized in SWN and it shares certain characteristics with speech. The resemblance between stimulus and the building blocks in human speech with regard to onset and duration probably constitutes an important prerequisite to good perception of similar speech elements, since both the Victoreen stimulus and a glottal pulse from a typical vowel sound are built up of a fast click onset and a tonal decay as shown in [Figure 1],[Figure 2]. Vowels and other voiced sounds are repetitive glottal pulses or clicks, which activate the resonance in the mouth cavity as illustrated by an "a" and "k" in [Figure 1]. In the high-frequency range, the Victoreen stimuli resemble the single click found in stops such as the p, t, k and b, d, g sounds. This similarity becomes evident, when looking at the time domain or listening to the stimuli. The broad spectral information of the fast click onset (as illustrated by the click during the first millisecond in [Figure 2]) is very important and should be perceived as precise as possible. This is a task for all intact sensory cells in the cochlea.

To simulate masking competing speech when assessing both unaided and aided SDTNs and SRSN, the BFM uses a speech weighted noise (SWN). Its spectral configuration resembles that of an average long-term speech spectrum. [11],[12] In [Figure 3], the noise spectrum is shown as long-term root-mean-square (rms) 1/3-octave band levels of 35, 45, 55, 65 and 75 dB SPL, respectively. When presented at a fixed overall level of 45 dB SPL, the SWN corresponds to quiet competing speech. At the 55, 65 and 75 dB SPL it corresponds to subdued, normal and loud speech masking levels, respectively. All levels correspond to the assessed values obtained in the position between the subjects' ears. A computerized audiometer of the fixed Bιkιsy type (PortaRem, Digital 2000, Rastronics, Copenhagen, Denmark) generates both the test signal and the SWN, which both are routed from the audiometer to the respective signal and noise loudspeakers [Figure 4]. The repetition frequency of the stimulus is 3 Hz. Within the stimulus, each period has 10% lower amplitude than the preceding one meaning that its effective duration is dependent on the frequency [Figure 1], left. Due to the short decay of the Victoreen stimulus when compared to the sinusoidal stimuli, the standing wave problem is small.

All measurements were carried out in a soundproof booth having a reverberation time of [13] Moreover, when measuring in a sound field it is easy to separate the ears with a 45°-approach angle from the signal loudspeaker. [14] A headrest, small enough to avoid any influence from baffle effects, is used to ensure correct positioning of the subject's head.

 SDTN



The SDTNs indicate the minimum level that the stimulus must attain in order to be detected. It is designated by the dB SPL peak equivalent value and is determined for the 0.5, 1, 2, 4, 6 and 8 kHz frequency. The measured values apply to a point in the middle of the booth at a 1 m distance from the signal loudspeaker, which correspond to an ordinary conversation distance. This point also corresponds to the centre of the head of the hearing aid wearer and not to the eardrum of the assessed individual.

In order to define normative reference values of the SDTNs at all SWN levels and frequencies, SDTNs were determined for 240 normal hearing subjects. The normative SDTN35, SDTN45, SDTN55, SDTN65 and SDTN75 means for each frequency [Table 1] were then defined as zero for the corresponding level. All assessed SDTN values will refer to the corresponding normative reference, i.e. the SDTN is set at zero dB if it equals the normative SDTN value. It should be noted that when the SWN is increased by 10 dB(C), then the mean SDTN threshold is also increased by about 10 dB (SPL peak eq) except for the lowest level and highest frequency [Figure 3] and [Table 1].

 SRSN



The SRSN-test is performed monaurally without ear protectors using lists of 50 monosyllabic phonetically balanced-words [15] in a SWN at a fixed speech-to-noise ratio of 75/65 dB, a condition simulating listening to speech masked by a speech noise level of normal group conversation. The level of the speech signal is defined as the equivalent of the rms of 1 kHz sinusoid with the same level as the average dB peak equivalent levels of the speech. The level of the SWN is measured in dB(C). According to Thornton and Raffin [16] binomial confidence intervals cannot be used to determine whether two individual scores are significantly different or not. Therefore, an individual improvement better than 10% when comparing the unaided and aided SRSN (S/N=75/65 dB) is regarded as a significant improvement of the speech perception ability under adverse listening conditions. This 10% limit was based on a SRSN test-retest (n=30; sd for the difference = 5.09; t0,05 = 2,042; variation limits = 5.09 x 2,042 = 10%, i.e. 5 words).

Test-retest-reliability of normative SDTNs and SRSN

A test-retest study was carried out on 30 subjects with hearing impairments ranging from slight to severe. The average SDTN difference between tests was 0.3 dB, the inter-individual standard deviation (SD) of the difference was 1.7 dB and the intra-individual SD was 1.2 dB. If a 90% confidence interval (CI) is chosen, a change larger than 3.0 dB between two tests will be regarded as a significant difference (90% CI SDTN difference = 0.3 ± 2.7 dB). This means that even a worsening by 3 dB in SDTN should result in some kind of action from the audiologist, such as a new setting of the hearing aid or the selection of another aid. If a 95% confidence interval is chosen, then a 4 dB change in SDTN should render the audiologist to act.

3 SD are considered to cover the normal variation (i.e. about ± 10 dB, [Table 1]) and a ± 3 dB range to cover measurement errors. Thus, a ± 13 dB range around the normative SDTN value when set at zero is used when deciding whether an individual SDTN should be regarded as within the range of people who have normal hearing or not (see BFM procedure). The average SRSN difference between the tests was 0.9%, the inter-individual SD of the difference 5.1% and the intra-individual SD was 3.6%. If a 95% CI is chosen, a difference larger than 10% dB between two tests will be regarded as a significant difference (95% CI = ± (3.6 * 1.96 * o2).

 The BFM procedure



The STDN assessments take place monaurally in a sound field. An ear protector is used in order to avoid interference from the non-tested ear. The aided STDNs are assessed in SWN at the standard frequencies 0.5-8 kHz, using the fixed frequency Bιkιsy technique. On the screen, each SDTN value is presented as the stimulus level above the corresponding normative SDTN value (reference). This means that the threshold of each masking noise level is zero if it equals the reference value. The audiologist follows the test procedure on the screen and manages the situation with possible out-liers. The effect of each new hearing aid setting is continuously monitored in order to ensure that the outcome converges positively i.e., that the functional gain makes the aided SDTN for all frequency bands and noise levels move into the corresponding normative SDTN range. Compression is used if necessary.

The assessments should continue until the normative SDTN values are reached in the high frequency range at the high noise level also. Then the target is reached. In order not to lose detection in other frequency ranges, all previous results must be checked, too. This is essential, because due to probable masking effects, there is often interference between the threshold gain settings at different frequencies. These paths are, with necessity, individual and could never rely on a prescriptive fitting formula based on hearing thresholds in a silent environment. The amount of efforts necessary to reach the target also gives further diagnostic information on how well the subject will recognize speech with the hearing aid.

After some three months of hearing aid use, the unaided and aided SRSN are assessed in order to verify the performance after adaptation, a time span necessary for the learning of how to decode the new sound of the hearing aid. [17] The comparison between the unaided and aided SRSN (S/N=75/65 dB) is also used as a quality assurance estimate. Under adverse listening conditions, an individual speech perception improvement is regarded as significant when better than 10%. After three months of adaptation, most of our patients have accepted the new sound (see pilot study below) and often the increase in clarity is appreciated. More important however is that the subject learns to benefit from the now available new information carried by the expanded bandwidth and dynamic range, which most patients have not been able to detect for many years. Easier tasks, such as listening to close and distant sources with no background noise, are also checked.

 The Clinical Pilot Evaluation Study



Subjects

The study sample comprised randomly selected cases from the waiting list at the Department of Audiology, Karolinska Hospital, who during standard audiological evaluation procedures were regarded to be in need of a hearing aid and who seven years previously were fitted with a Danavox Aura according to the NAL-R procedure. From this group all available subjects born 1920 or later were chosen. The study group comprised 13 men and 8 women (age range 69-84 years, mean 78 years). [Figure 5] illustrates mean pure-tone thresholds and standard deviations.

Experiment procedures, study design and statistics

Each subject attended a 2 hr appointment twice. The same audiologist pursued all measurements at both sessions and the equipment had the same calibration for all corresponding measurements. At the first visit, the ear canals were checked and the pure tone audiometry performed. Then, unaided and aided SDTN65, SDTN75 and SRSN (65/55 and 75/65) were measured for the old Danavox/Aura hearing aid, which had been fitted according to the NAL-R procedure 7 years earlier. When assessing the SRSN performances with the Danavox Aura, the subjects controlled the volume wheel themselves. The subjects were then fitted monaurally according to the Logic procedure with the digital six-channel BTE hearing aid Danavox/Danalogic163D hearing aid. "Logic" is a proprietary prescriptive fitting procedure, based on the conventional pure tone audiogram and developed by the manufacturer Danavox by revising their older S2000R rule for their hearing aid Danalogic 163D. Aided SDTN45, SDTN65 and SDTN75 were assessed as well as unaided and aided SRSN as above. All ear moulds had a Bakke-horn as tube connection, a 2.5 mm sound channel with a horn shaped opening. The insert had a ventilation channel 0.8-1 mm in diameter.

At the second visit, the subjects were again fitted with the same Danalogic 163D hearing aid, however now according to the BFM. The assessments of SDTN and SRSN were repeated as described above. All SRSN values for the Danalogic/163D were obtained without any adaptation. The non-parametric Wilcoxon signed-rank test was used for comparisons of SRSN between fitting procedures and also the aided versus unaided SRSN. Correlation analyses were performed according to Pearson.

In order to explore the perceived sound quality when using BFM, 60 patients were asked to fill in a questionnaire with 2 queries having a 5 response format ranging between "no" (1) and "much" (5) discomfort about the amount of discomfort perceived when first fitted according to the BFM and after 6 months of usage.

 Results



In [Figure 6], the individual aided SRSN results are presented as a function of the unaided SRSN for the masking condition corresponding to normal speech noise (S/N=75/65 dB).

Observations above the diagonal line indicate benefit of the hearing aid. In some cases the hearing aid was a disadvantage under this masking condition (observations below the vertical diagonal). Only the BFM gave a significant benefit, i.e. significantly better aided than unaided SRSN ( P P Two individual examples

Subjects 1 and 19 were chosen to exemplify that, despite having similar conventional pure tone hearing thresholds [Figure 12], the results may end up in sub optimizations and that an individual assessment was one way to assure a high quality fitting. When fitted by the Logic procedure [Figure 13], the SDTN75 results deviated substantially between subjects as well as from the norm-SDTN75. Arrows at the 4, 6 and 8 kHz frequencies indicate that true thresholds were not reached and were hence replaced by the max output values of the equipment [Table 2]. When fitted according to the BFM [Figure 14], the SDTN75 assessments were very similar to or within the normal range (framed area) for both patients.

When comparing the SRSN assessed in subject 1, the logic procedure improved the SRSN by only 1% under the low (S/N=65/55 dB) and 2% under the high (S/N=75/65 dB) masking level condition compared to unaided SRSN. The corresponding improvement for the BFM procedure was 8% and 6%, respectively. For subject 19, the Logic fit improved the SRSN by as much as 30% in the low level condition, but only 8% the high level condition. Using the BFM, the SRSN was improved by 14% in the S/N=65/55 dB condition and by 13% for the high level S/N=75/65 dB condition. Note that the SRSN results corresponded well with the SDTN75 data in [Figure 13],[Figure 14], which were below the normative SDTN range for the Logic fit (framed areas), but within the norm-SDTN range for most frequencies when the HA was fitted according to the BFM procedure.

The assessments of subjects 1 and 19 also illustrated how different the SDTN characteristics can be for two hearing-impaired subjects with similar conventional audiograms. This is something we see quite often. One reason may be individual differences in internal cochlear masking. The comparison between the two subjects also elucidated the importance to pursue the fitting to include also high noise levels and not to be misled to conclude the fitting procedure after obtaining good SDTN values at levels corresponding only to subdued or normal levels of competing speech.

Another indication that patients prefer a fitting procedure based upon psycho-acoustical detection thresholds was reported by Lennart, [18] who found the BFM to be superior to the NAL-Non-Linear, version l procedures. The subjective reason for choosing the BFM procedure was in most cases an unacceptable high gain when fitted according to the NAL procedure. Accordingly, a too high gain at lower levels may reduce the speech reception ability when the level is raised, as was the case for subject 19 when fitted according to the logic procedure. However, the SDTN values for subject 19 when fitted with the BFM [Figure 14] were all within the normative SDTN range (0-13 dB) and the overall quality of listening comfort was better than that of the Logic procedure [Figure 13]. Usually, subjects reaching the normative SDTN values will almost directly after the fitting procedure arrive at a reasonable good listening comfort also under quite adverse acoustical conditions. It was also shown from the questionnaire data [Figure 11] that most patients will adapt to high frequency amplification, even though some patients at first may find the sound somewhat sharp.

In our opinion, the conventional fitting procedures are too much focused on the immediate listening comfort in a silent environment instead of the speech intelligibility in noise, which requires the clarity achieved by proper high frequency amplification. It may even be that hearing impaired subjects will not get improved speech reception ability in noise unless the aided SDTN values are detected within the normative SDTN range under corresponding noise conditions. This possibility is supported by the correlation between SRSN and SDTNs at different frequencies for the 55 and 65 dB noise conditions [Table 3], indicating that the higher the level and frequency, the stronger the correlation. This in turn is a further indication that the information carried by the high frequency range at high noise levels is important.

 Conclusions



The present pilot study indicated that the BFM procedure, even without a subsequent adaptation period, arrived at a significantly better hearing ability under adverse conditions than did the manufacturer's procedure Logic and the old conventional NAL-R procedures. The specific correlates to the optimal settings of the BFM procedure are not yet fully understood, but it is suggested that the main reason for the good results was that the BFM offered a better balance between the high frequency gain and that of the low and mid frequencies by using an individual "close-to-reality" procedure based on psycho acoustically assessed thresholds in noise. Moreover, BFM is designed to establish aided detection thresholds in more "natural" masking noise for more "natural" stimuli, which are within the detection range of that for people with normal hearing for the speech frequencies at speech levels varying from subdued to loud. Finally, the Victoreen stimulus is regarded as more "natural" compared to conventional sinusoidal stimuli in the sense that it has a structure resembling that of the building stones of speech rather than having. Both the time domain and the spectral domain of the hearing ability will then be addressed in a more "natural" way. The SDTN values correlated well with the respective SRSN values at loud speech levels, which mean that it is not necessary to confirm the settings by a time consuming SRSN assessment [Table 3].

The evaluation of the BFM procedure and the comparisons with other procedures has indicated results favouring the BFM. The importance of high frequency information was clearly indicated, putting high demands that modern "good" hearing aids must convey enough high frequency gain. Achieving a successful low-level assessment only is too easy and not enough. Also, the BFM assessments at high levels of masking noise are important considering the current demands on quality assurance. In Sweden, the introduction of the BFM procedure has already contributed to an improved quality of audiological rehabilitation. Further evaluation and refinement of the BFM procedure is already in progress and hopefully enough data will be collected for modelling a computerized BFM fitting strategy.

References

1Hjortzby B, Edιn M. H φr vi bδttre med ny teknik? Audionytt, ISSN: 0347-6308, (Swedish) 2002;4:21-2.
2Rosenhall U, Arlinger S, Brorsson B, Lagerbring C, Leijon A, Scherstιn T. Hearing aids for adults - benefits and costs. The Swedish Council on Technology Assessment in Health Care (SBU). Report 164. 2003. ISBN: 91-87890-85-2. Available from: http:/www.sbu.se.
3Kiessling J. Hearing aid fitting procedures - State-of-the-art and current issues. Paper presented at the 4 th European Conference on Audiology - EFAS: Finland; 1999.
4Smeds K, Leijon A. Threshold-based fitting methods for non-linear (WDRC) hearing instruments - comparison of acoustical characteristics. Scand Audiol 2001;30:213-22.
5Houtgast T. Psychoacoustics and speech recognition of the hearing impaired. In: Proceedings of the European Conference on Audiology. Noordwijkerhout: the Netherlands; 1995. p. 165-9.
6Byrne D, Dillon H. The National Acoustic Laboratories' (NAL) new procedure for selecting the gain and frequency response of a hearing aid. Ear Hear 1986;7:257-65.
7Hornsby BW, Ricketts TA. The effects of hearing loss on the contribution of high- and low-frequency speech information to speech understanding. J Acoust Soc Am 2003;113:1706-17.
8Mackersie CL, Crocker TL, David RA. Limiting high-frequency hearing aid gain in listeners with and without suspected cochlear dead regions. J Am Acad Audiol 2004;15:498-507.
9Barrenδs ML, Wikstr φm I. The influence of hearing and age on speech recognition scores in noise in audiological patients and in the general population. Ear Heart 2000;21:569-77.
10Victoreen JA. Equal loudness pressures determined with a decaying oscillatory waveform. J Acoust Soc Am 1974;55:309-12.
11Tarnoczy T. Das durchschnittliche Energi-Spektrum der Sprache (für sechs Sprachen). Acoustica 1971;24:57-74.
12Cox RM, Moore JN. Composite speech spectrum for hearing aid gain prescriptions. J Speech Hear Res 1988;31:102-7.
13Killion MC, Revit LJ. Insertion gain repeatability versus loudspeaker location. Ear Heart 1987;8:68S-73S.
14Nordlund B, Fritzell B. The influence of azimuth on speech signals. Acta Otolaryngol (Stockh) 1963;56:632-42.
15Lidιn G. Speech audiometry. An experimental and clinical study with Swedish language material. Speech audiometry. Acta Otolaryngol Suppl (Stockh) 1954;114.
16Thornton AR, Raffin MJ. Speech-discrimination scores modeled as a binomial variable. J Speech Hear Res 1978;21:507-18.
17Gatehouse S. The time course and magnitude of perceptual acclimatization to frequency responses: Evidence from monaural fitting of hearing aids. J Acoust Soc Am 1992;92:1358-68.
18Lennart I. Utvδrdering av programmeringsmetoder f φr h φrapparater. Essay in Swedish at the Medical Faculty, Lund University, Dept of Logopedics and Phoniatrics. 2001. Available from: http://www.logopedi.lu.se/audmag0il.htm.