• Nenhum resultado encontrado

Development and validation of Finnish speech audiometry in noise

N/A
N/A
Protected

Academic year: 2023

Share "Development and validation of Finnish speech audiometry in noise"

Copied!
224
0
0

Texto

Speech perception in noise tests provide a better measure of functional hearing in noise, and are therefore. The Finnish Matrix Sentence Test (FMST) has been the standard speech perception test in noise in Finland since 2014.

Word-level speech perception tests

The test material was used to determine both SRT and speech intelligibility in silence. In the past, an SRS in silence was considered a valid estimate of a listener's speech perception.

Digits as speech test material

  • The anatomy of a DTT
  • Comparability to other hearing measures
  • DTTs in hearing screening
  • DTTs in monitoring hearing rehabilitation results
  • The future of DTTs

Kaandorp et al., 2015) reported only a moderate correlation (r = 0.56) between a sentence-level speech perception test in noise and the DTT. Broadband versions of DTT have been used successfully for screening for noise-induced HL (Jansen et al., 2013).

Sentence-level speech perception tests in noise

Tests with everyday sentences as test material

Some tests (AzBio (Spahr et al., 2012)) use the same fixed presentation level throughout the test and present results as a percentage of correct responses. A certain level of presentation predisposes to floor or ceiling effects (ie, the test is too difficult or too easy), especially if the same test material is used to assess hearing before and after rehabilitation (Gifford, Shallop, & Peterson, 2008; Dornhoffer et al., 2021).

Table 1. Overview of the speech material of commonly used English speech perception tests in noise
Table 1. Overview of the speech material of commonly used English speech perception tests in noise

Matrix sentence tests

The test-retest difference has been reported to be typically lower than 1 dB SNR (Kollmeier et al., 2015). During the test, the word matrix is ​​not available and the listener has to repeat the words they hear.

The effect of age on testing speech perception in noise

  • Speech perception in noise in children
  • Testing sentence-level speech perception in noise in
  • Speech perception in noise in the elderly
  • Testing sentence-level speech perception in the noise in the

Walker et al., 2019) and Magimairaj and Nagaraj (Magimairaj and Nagaraj, 2018) did not find any relationship between working memory and speech perception in noise. Variation in onset and severity of age-related HL is large (Hoff et al., 2020;.

Table 4. Pediatric versions of commonly used sentence-level speech perception tests in noise and other sentence  level speech perception tests used with children
Table 4. Pediatric versions of commonly used sentence-level speech perception tests in noise and other sentence level speech perception tests used with children

Learning effect related to repeated speech perception testing

Perceptual learning

Auditory perceptual learning can be defined as an enhanced ability to detect, distinguish, or categorize sounds or sound features (Halliday et al., 2012). Some studies of the repeated use of speech intelligibility tests have interpreted the improvements in test results as perceptual learning (Cainer, James, and Rajan, 2008; Bugannim et al., 2019).

Test-retest difference

Procedural learning

Although the test material may be new to the participant, prior familiarity with the psychoacoustic test procedures may affect the amount of procedural learning. The HINT (Nilsson, Soli, & Sullivan, 1994) is one of the most well-translated tests of speech perception in noise, and also has the most evidence of procedural learning.

Table 5. Procedural learning for some word- and sentence-level speech perception tests in noise
Table 5. Procedural learning for some word- and sentence-level speech perception tests in noise

Content learning

The amount of content learning is likely related to exposure to the speech material. Adaptive testing procedures typically converge to the SNR, recognizing 50% of the test material correctly.

Learning effect in the elderly and in children

The pediatric versions of speech perception tests in noise have fewer test lists than the original versions (Wagener and Kollmeier, 2005; the overall aim of this thesis was to expand the Finnish speech audiometry test battery and improve the reliability of the new and current Finnish speech perception tests in noise .

Abstract

The text corresponds to the original Accepted Manuscript except for the heading numbers for figures and tables, which have been changed. Also the font and the references have been formatted to match the general format of this thesis.

Introduction

The basic concept of the DTT uses digit triplets (for example 2-4-7, pronounced two-four-seven) as speech material and adaptive up-and-down tracking for the SRT determination. The success of the early versions of the DTT led to the development of similar tests in different languages ​​(Zokoll et al., 2012; Akeroyd et al., 2015).

Material and Methods

  • Recording of the speech material
  • Cutting the speech material and re-synthesizing the
  • Development of the masking noise
  • Participants
  • Optimization
  • Evaluation measurements

The equipment and setup were the same as in the optimization of speech material for the Finnish matrix sentence test (Dietz et al., 2014). At the beginning of the session, each subject performed two training lists in an adaptation procedure.

Results

Results from the optimization measurements

The purpose of the evaluation measurements was to obtain a normative reference function for normal-hearing listeners and to verify that there was no systematic difference in intelligibility between the final test lists. The order of test lists and SNRs used was randomized to minimize the effect of training or fatigue on the results.

Figure 1. SRTs of the individual digits before and after optimization. Error bars indicate standard deviations
Figure 1. SRTs of the individual digits before and after optimization. Error bars indicate standard deviations

Results from the evaluation measurements

The crossed and solid horizontal black lines indicate the mean plus/minus one standard deviation of the adaptively determined SRT. As expected, the digit score resulted in a slightly flatter mean slope of dB for all subjects.

Figure 2. Psychometric functions the Finnish digit triplet test with triplet scoring. Left: gray lines indicate the  psychometric functions of the six different test lists
Figure 2. Psychometric functions the Finnish digit triplet test with triplet scoring. Left: gray lines indicate the psychometric functions of the six different test lists

Discussion

For the Finnish version, all digits 0 to 9 were used because after optimization no large differences in intelligibility could be measured (see Figure 1). For the Finnish tests, however, the difference between the reference SRTs is very small (see Table 8).

Table 7. Comparison of key parameters of the Finnish DTT with other  existing DTTs. The data is summarized for triplet scoring and the
Table 7. Comparison of key parameters of the Finnish DTT with other existing DTTs. The data is summarized for triplet scoring and the

Conclusions

This predictability may contribute to the apparently better comprehensibility (i.e., lower SRT) of the Finnish version of the matrix test and explain the similarity of the reference SRTs for the Finnish DTT and the matrix test. 5 THE FINNISH SIMPLIFIED MATRIX TEST FOR THE ASSESSMENT OF SPEECH INtelligibility IN THE ELDERLY.

Abstract

For older HI listeners, mean SRT50 scores for FINSIMAT and FMST were −4.1 and −3.6 dB SNR, respectively. Conclusion: The FINSIMAT showed comparable characteristics to the FMST and proved feasible for measurements in older HI listeners.

Introduction

Since the most widely used and the only internationally comparable diagnostic speech intelligibility test in noise in the Finnish language is the Finnish matrix sense test (FMST) (Dietz et al.), it was logical to base the new test on the FMST. , the OLKISA was found to be more appropriate than the OLSA when testing normal hearing (NH) children aged 6-8 (Wagener and Kollmeier, 2005) and considered appropriate for HI children in noise and in silence (Ebner, Steffens, and Hellbrück, 2008 ; Weissgerber et al., 2012).

Study 1: The development and evaluation of the Finnish

Materials and Methods

The word matrix of the Finnish matrix sentence test (FMST), adapted from (Dietz et al. 2014). The words in bold are also included in the word matrix of the Finnish Simplified Matrix Sentence Test (FINSIMAT).

Table 9. The word matrix of the Finnish matrix sentence test (FMST),  adapted from (Dietz et al
Table 9. The word matrix of the Finnish matrix sentence test (FMST), adapted from (Dietz et al

Results

The speech intelligibility functions of the Finnish Simplified Matrix Sense Test (FINSIMAT) are derived from the evaluation measures in Study 1 with normal hearing young adults. The list-specific speech intelligibility functions of the Finnish Simplified Matrix Sentence Test (FINSIMAT) are derived from the evaluation measures in Study 1 with normal-hearing young adults.

Figure 3. The speech intelligibility functions of the Finnish simplified matrix  sentence test (FINSIMAT) derived from the evaluation measurements in  Study 1 with normal-hearing young adults
Figure 3. The speech intelligibility functions of the Finnish simplified matrix sentence test (FINSIMAT) derived from the evaluation measurements in Study 1 with normal-hearing young adults

Study 2: Evaluation of the FINSIMAT in the Elderly

Materials and Methods

Independent sample t-test was used to assess differences between ears, and paired sample t-test was used to assess the possible differences between the ear-specific FINSIMAT and FMST results. Pearson's correlation coefficient was used to assess the correlation between the FMST and FINSIMAT results.

Figure 7. Results of the pure-tone air-conduction audiometry for the  elderly hearing-impaired participants in Study 2
Figure 7. Results of the pure-tone air-conduction audiometry for the elderly hearing-impaired participants in Study 2

Results

The correlation between the two tests was good (r2 = 0.78, p < 0.001) as depicted in Figure 8, where the results for both tests are shown for each participant. Correlation between Finnish Matrix Sentence Test (FMST) and Finnish Simplified Matrix Sentence Test (FINSIMAT) results in elderly hearing-impaired patients in Study 2.

Figure 8. Correlation between the Finnish matrix sentence test (FMST) and  Finnish simplified matrix sentence test (FINSIMAT) results in the elderly  hearing-impaired patients in Study 2
Figure 8. Correlation between the Finnish matrix sentence test (FMST) and Finnish simplified matrix sentence test (FINSIMAT) results in the elderly hearing-impaired patients in Study 2

Discussion

Most of the improvements to the SRT80 rating were made during the first two adaptive measurements, and after these measurements. However, the first measurement for SRT50 estimation was preceded by four training measurements with the FINSIMAT test material.

Table 11. The reference speech reception threshold (SRT 50 ) estimates  (across listeners) and the slopes of the intelligibility functions (across lists)  for adults for the Finnish matrix sentence test (FMST, Dietz et al
Table 11. The reference speech reception threshold (SRT 50 ) estimates (across listeners) and the slopes of the intelligibility functions (across lists) for adults for the Finnish matrix sentence test (FMST, Dietz et al

Conclusions

6 THE LONG-TERM LEARNING EFFECT ASSOCIATED WITH THE REPEATED USE OF THE FINNISH MATRIX TEST AND THE FINNISH NUMBER. This chapter is fully derived from the Accepted Manuscript of an article published by Taylor & Francis in International Journal of Audiology at available at: https://doi.org The text corresponds to the original Accepted Manuscript except for the heading numbers for figures and tables, after which have been changed.

Abstract

Conclusions: Repeated use of the FMST led to significant learning-related improvements, but improvements appeared to level off by the third test session. For the FDTT, overall improvements were smaller, but there was a significant difference between the first and consecutive FDTT measurements between test sessions.

Introduction

Content learning refers to improvement through familiarity with the test material and benefiting from memory of parts of the test material from previous test sessions. A training effect, attributed to both content and procedural learning, has been reported for different versions of the test.

Materials and methods

  • Participants
  • Apparatus
  • Procedure
  • Data Analysis

In the test, the numbers 0-9 are combined into quasi-random triads (i.e., 2-6-4, pronounced two-six-four) and presented against background noise with a preceding announcement phrase "Numerot..." ("The numbers… ”). All measurements were performed in a sound field chamber in the S0N0 condition, i.e. the signal and noise were presented by the same loudspeaker at 0° azimuth in the horizontal plane.

Results

FMST Results

The average washout effect between sessions was assessed by comparing M1 scores with M4 scores from the previous test session. FMST scores from the last test session were compared with the number of words participants correctly recalled from the FMST word matrix.

Table 13. The mean results with standard errors (SE) for the Finnish matrix  sentence (FMST)
Table 13. The mean results with standard errors (SE) for the Finnish matrix sentence (FMST)

FDTT Results

However, the amount of statistically significant differences quickly decreased, and between the last three test sessions (S3, S4, and S5), a pairwise comparison of SRT scores detected only three statistically significant differences (S3M1-S3M4, S3M1-S4M4, and S4M1). -S4M4). All statistically significant differences for the last three test sessions were detected only when compared to the first adaptive measurement of the test session, which is considered a training measurement according to the current recommendation for clinical use of the FMST.

Table 14. The mean results with standard errors (SE) for the Finnish digit  triplet test (FDTT)
Table 14. The mean results with standard errors (SE) for the Finnish digit triplet test (FDTT)

Discussion

  • Within-Session Learning Effect for the FMST
  • The Long-Term Learning Effect for the FMST
  • Learning Effect for the FDTT
  • Procedural Learning
  • Content Learning
  • Comparison of the long-term learning effect between
  • Learning effect for hearing-impaired patients
  • Limitations

In both studies, the results of the last test measurements (the mean of the 5th and 6th test measurements) of the first and the last (5th) test session were compared, and in both studies the improvements were greater than in our data ( Table 13) . However, even with part of the procedural learning eliminated, the learning effect for FMST was both clinically and statistically significant and greater than the learning effect for FDTT.

Conclusions

American Speech-Language-Hearing Association, pp. 2014) 'The development and evaluation of the Finnish Matrix sentence test for speech intelligibility assessment', Acta Oto-Laryngologica, 134(7), pp. Neurotology: Official Publication of the American Otological Society, American Neurotology Society [and] European Academy of Otology and Neurotology, 37(2), pp. 2017) 'Prevalence of age-related hearing loss,.

General discussion

Development of the Finnish digit triplet test (Study I)

Development of the FDTT started in 2014, and between then and writing this, interest in DTTs and the available research data on them has exploded (for a recent review, see (Van den Borre et al., 2021). current multicultural world , speech audiometry must accommodate the different language skills of the listeners.

Development of the Finnish simplified matrix sentence test

Evaluation of long-term learning effects (Study III)

We assumed that the learning effect we observed probably represented maximal learning, and that the learning effects could be significantly smaller in a clinically representative patient population. Although the learning effect persisted throughout the study sessions, the improvement leveled off, indicating that the learning effects for MST are limited.

Study limitations and future perspectives

2011) 'Assessment of auditory nonlinearity for listeners with different hearing losses using temporal masking and categorical loudness scaling', Hearing Research, 280(1), pp. 2016) 'The influence of lexical access and vocabulary knowledge on measures of speech recognition in noise', International Journal of Audiology, 55(3), pp. New York, NY: Springer, pp. 2020) 'Cognitive and Linguistic Contributions to Masked Speech Recognition in Children', Journal of Speech, Language, and Hearing Research, 63(10), pp. 2017) 'Individual differences in language and working memory influence children's speech.

Imagem

Table 1. Overview of the speech material of commonly used English speech perception tests in noise
Table 1. Continued. 1)  The material includes 3 additional list pairs that are equal in intelligibility, if used in pair 2)  15 test lists released as a commercial CD; 8 separate lists used in the MST 3) Quick Speech in Noise (QuickSIN); 4)Minimum speech t
48  Table 2. Overview of test practices of commonly used English speech perception tests in noise
Table 2. Continued. SNR, signal-to-noise ratio; dB SNR, decibel signal-to-noise ratio; SRT, speech recognition threshold
+7

Referências

Documentos relacionados

– Models noise as a sum of various small noise sources, which is indeed what. happens