Computers in Biology and Medicine 52 (2014) 66–72

Contents lists available at ScienceDirect

Computers in Biology and Medicine journal homepage: www.elsevier.com/locate/cbm

A user-operated test of suprathreshold acuity in noise for adult hearing screening: The SUN (SPEECH UNDERSTANDING IN NOISE) test Alessia Paglialonga n, Gabriella Tognola, Ferdinando Grandori CNR – National Research Council of Italy, Institute of Electronics, Computer and Telecommunication Engineering IEIIT, Milan , I-20133 Italy1

art ic l e i nf o

a b s t r a c t

Article history: Received 16 April 2014 Accepted 17 June 2014

A novel, user-operated test of suprathreshold acuity in noise for use in adult hearing screening (AHS) was developed. The Speech Understanding in Noise test (SUN) is a speech-in-noise test that makes use of a list of vowel–consonant–vowel (VCV) stimuli in background noise presented in a three-alternative forced choice (3AFC) paradigm by means of a touch sensitive screen. The test is automated, easy-to-use, and provides self-explanatory results (i.e., ‘no hearing difficulties’, or ‘a hearing check would be advisable’, or ‘a hearing check is recommended’). The test was developed from its building blocks (VCVs and speechshaped noise) through two main steps: (i) development of the test list through equalization of the intelligibility of test stimuli across the set and (ii) optimization of the test results through maximization of the test sensitivity and specificity. The test had 82.9% sensitivity and 85.9% specificity compared to conventional pure-tone screening, and 83.8% sensitivity and 83.9% specificity to identify individuals with disabling hearing impairment. Results obtained so far showed that the test could be easily performed by adults and older adults in less than one minute per ear and that its results were not influenced by ambient noise (up to 65 dBA), suggesting that the test might be a viable method for AHS in clinical as well as non-clinical settings. & 2014 Elsevier Ltd. All rights reserved.

Keywords: Adult hearing screening (AHS) Adults Elderly Speech-in-noise Speech recognition Speech understanding Suprathreshold acuity in noise

1. Introduction The development of reliable and quick techniques for adult hearing screening (AHS) is a key technological challenge nowadays. Hearing loss is a common health problem in adults and older adults. The World Health Organization reported it as the first among the 20 leading causes of moderate-to-severe disability in the adult population in 2011 [1]. The demand for viable methods and systems for AHS is growing rapidly as well as the need for readily accessible AHS programs [2,3]. Screening is particularly important because people with age-related hearing loss typically get used to the slow progression of their impairment and seek help very late (or do not seek help at all) with negative repercussions on speech communication, cognitive decline, and quality of life [4–8]. As in the past decades the implementation of newborn hearing screening (NHS) programs was boosted by the availability of

n Correspondence to: Istituto di Elettronica e di Ingegneria dell'Informazione e delle Telecomunicazioni, Consiglio Nazionale delle Ricerche (IEIIT CNR), c/o DEIB (Ed. 21), Politecnico di Milano, Piazza Leonardo da Vinci, 32, I-20133 Milano, Italy. Tel.: þ 39 02 2399 3343; fax: þ 39 02 2399 3367. E-mail address: [email protected] (A. Paglialonga). 1 Formerly with the Institute of Biomedical Engineering ISIB at CNR, Milan I-20133, Italy.

http://dx.doi.org/10.1016/j.compbiomed.2014.06.012 0010-4825/& 2014 Elsevier Ltd. All rights reserved.

portable, easy-to-use equipment providing automated pass/fail decisions in a few minutes, similar dedicated systems might likewise help the initiation and spread of AHS programs in the near future. Nevertheless, methods used in NHS and AHS are substantially different as the former rely upon objective measures (e.g., otoacoustic emissions and evoked potentials [9,10]) whereas the latter make use of subjective measures (typically, threshold detection of tones, suprathreshold speech recognition, or selfadministered questionnaires [11–13]) to identify people with hearing problems and refer them for further audiological evaluation. Increasing evidence is showing that measures of hearing handicap and hearing disability may be better than measures of hearing impairment to identify individuals with hearing problems that could benefit from intervention. This is because some individuals with hearing impairment may not perceive any hearing disability and, vice versa, others with minimal or no hearing impairment may report considerable disability [14,15]. For example, recent studies showed that measures of self-perceived activity limitation are very important determinants in aural rehabilitation and can predict key outcome measures (i.e., help seeking, hearing aid uptake, use, and satisfaction) better than threshold measurements [16,17]. Similarly, as one of the most common complaints of older adults with impaired hearing is just the inability to understand speech in noise [18,19], it is agreed that speech-in-noise tests, which are suprathreshold measures of hearing acuity, may be able to reflect the

A. Paglialonga et al. / Computers in Biology and Medicine 52 (2014) 66–72

real-life communication performance more accurately than conventional threshold measures of hearing sensitivity (detection of pure-tones) [20,21]. Recently, we have developed a fully automated, user-operated test of suprathreshold acuity in noise for AHS: the speech understanding in noise (SUN) test. Prior papers [22–26] documented the main results obtained with the test in a population of more than 6000 subjects in a variety of test protocols and settings (including unchecked background noise). This paper presents in detail the development of this user-operated speech-in-noise screening test through the description of design criteria (Section 2.1), building blocks (Section 2.2), and test development (Section 2.3). The test is currently available in different languages (e.g., English, French, German, Italian, and Brazilian Portuguese) and is currently being developed in novel versions (e.g., Spanish and Mandarin Chinese). As the test was originally developed in the Italian version, only data relevant to this language will be shown here.





 2. Materials and methods

67

0.82 and the within-subject repeatability, as measured by the within-subjects standard deviation on replications, was 0.79 (i.e. lower than 1 score point) with no significant learning effect in repeated measures [22]. Tests based on meaningless VCVs can be more easily adapted to different languages than tests based on complex stimuli (e.g., words with meaning or sentences). Tests based on meaningless VCVs in a multiple-choice task are largely independent on subjects' literacy and education [21,35] and, also, can be performed by non-native listeners who have at least a rough familiarity with spoken and written language in the host country. This is particularly relevant in AHS campaigns because, typically, no prior assessment or selection of subjects is done and, as a significant number of non-native listeners are expected, tests based on words with meaning or sentences might be biased due to limited word understanding and poor vocabulary. Multiple-choice tests can be easily implemented in a useroperated, automated procedure by using, for example, a touch sensitive screen interface to display the alternatives and record the subjects’ responses. User-operated, automated tests based on a small set of alternatives (e.g., three in the SUN test) and simplified interactions (i.e., multiple-choice recognition) are particularly useful to limit anxiety and concern in the tested subjects, as well as the involvement of cognitive functions [36,37] such as short-term memory or reading speed, which are known to influence speech recognition, particularly in older adults [7,27,28].

2.1. Design criteria



Specific design criteria were set to make the test viable to be used for AHS and implemented on a stand-alone, portable, easy-touse device. A core set of relevant attributes were identified for the test, i.e.: (i) reliable as an AHS method; (ii) easy to be performed by the largest possible population; (iii) self-explanatory and convincing; and (iv) short in duration. As the test had to be reliable as an AHS method, it was conceived as a speech-in-noise consonant recognition test. This was because:

As the test had to be self-explanatory and convincing, the test outcomes are given as a clear, easy-to-remember outcome, i.e. as the combination of:

 Adults and older adults typically experience hearing difficulties just with speech recognition in noise, particularly with fast transients and consonants [18,19,27,28].  Speech-in-noise tests are particularly promising for AHS (see, e.g., recently developed speech-in-noise tests for remote, athome self-screening by telephone or internet [29–31]).  Speech-in-noise tests, differently than pure-tone threshold measures, are particularly viable for use in noisy environments such as in typical AHS settings (e.g., hospital wards, waiting rooms, pharmacies, outdoor places, or special rooms in the workplace) because their results are largely independent from ambient conditions [13,21]. This was also true for the SUN test as results in 899 participants tested in high ambient noise (up to 65 dBA) in nonclinical settings (e.g., supermarkets, drugstores, outdoor sites, and in some “Universities of the Third Age”) showed that the test outcomes were not influenced by the noise level in the test room [22,24]. As the test had to be easy to be performed by the largest possible population, it was based on a simple task, i.e.: multiplechoice recognition of short meaningless stimuli. More specifically, meaningless vowel–consonant–vowel (VCV) stimuli (e.g., asa, apa, …) were used, in a three-alternative forced-choice (3AFC) task with ‘maximal opposition’ alternatives (i.e. with the two ‘wrong’ alternatives differing from the spoken stimulus in all the relevant features: voicing, manner, and place of articulation – such as, e.g.: ata, ava, ama). This choice was supported by a number of reasons:

 The test score, i.e., the number of stimuli correctly identified in the test out of the total number of stimuli delivered (e.g., 10 out of 12). This provides the subject with direct feedback on the number of correct and wrong responses, making him/her more conscious of possible difficulties experienced while performing the test.  A screening outcome, i.e. a ‘traffic-light’ color: green, yellow, or red to indicate that the test results are within, slightly below, or well below the normal range, respectively.  A short statement associated with the different outcomes, i.e.: ‘no hearing difficulties’ (green light), ‘a hearing check would be advisable’ (yellow light) or ‘a hearing check is recommended’ (red light). As the test had to be short in duration, it was conceived as a short list of stimuli with pre-determined levels of noise to check quickly whether the subject's speech recognition performance is within, below, or well below the normal hearing range. The exact number of stimuli may vary in the different languages. Typically, 12–18 stimuli are used across the different languages, and test duration is below 1 min per ear, including in very old subjects (mean test time ¼40 s, s.d. ¼10 s; range¼ 27–114 s [22]). It is worth noting that short test duration is important to test a large number of subjects in a short time in AHS applications and, also, to limit possible related unwanted effects, such as concerns, anxiety, or decrease in subjects' arousal state. 2.2. Building blocks

 Tests based on meaningless VCVs have good test–retest reliability and precision [21,32] and can be successfully used when highly accurate measures of speech recognition abilities are needed [33,34]. This was also true for the SUN test as Spearman's rank order correlation between test and retest scores was

The building blocks of the test are: (i) speech stimuli (VCVs) and (ii) background noise. The set of VCVs is language specific as only consonants relevant to the specific language are included. For example, 14 VCVs were recorded for the Italian version (aba, ada,

68

A. Paglialonga et al. / Computers in Biology and Medicine 52 (2014) 66–72

aga, afa, aka, ala, ama, ana, apa, ara, asa, ata, ava, aza). Of the 16 consonants in the Italian alphabet, two (i.e., h and q) were not used because h does not correspond to any phoneme, and q overlaps with k, which is used much more frequently in the Italian language. VCV utterances were single exemplars recorded by a professional, native-Italian, male speaker who was instructed to pronounce the VCVs with no prosodic accent, with the stress on the first vowel and with constant pitch across the list. Stimuli were recorded in a professional recording studio using a Neumann TLM 103 microphone, a SSL S4000 64-channels mixer, Motu HD 192A/D converters (44.1 kHz, 16 bit) and GENELEC 1025A control room monitor, with the speaker standing in a sound-treated room. The level of VCV recordings was digitally equalized across the set to meet the ‘Equal speech level’ requirement as set in the ISO 82533:1996(E) Standard for Speech Audiometry [38]. The background noise in the SUN test is a steady speechshaped noise, i.e. a wide-band noise generated by filtering a steady-state unmodulated white noise with the reference long term average speech spectrum for the relevant language [39,40].

2.3. Test development The development of the test from its building blocks includes two main steps, i.e.: (i) development of the test list through equalization of the intelligibility of test stimuli across the set and (ii) optimization of the test results through maximization of the test sensitivity and specificity.

2.3.1. Development of the test list The intelligibility of stimuli across the list was equalized by setting the ‘presentation SNR’ (signal-to-noise ratio) for each VCV at or above the reception threshold (RT), which is defined as the SNR where the psychometric curve reaches 79.4% correct responses in a 3AFC task [41]. The RTs were estimated from the psychometric curves of VCVs. The psychometric curves were measured in a group of normal hearing subjects as a function of the SNR. For example, the SNR was varied, in 2 dB steps, from 12 dB to þ6 dB SNR in the Italian language, corresponding to chance performance and 100% recognition, respectively [42–44]. VCVs that exhibit psychometric curves with slope higher than 30%/2 dB were not included in the test list because, as suggested by Strasburger [45], they could lead to unreliable estimates of speech recognition and unstable test results. To measure the psychometric curves, the output level of VCVs was fixed above 100% correct responses in normal hearing subjects (e.g., 60 dB HL [46]) and the output level of noise was varied to set the desired SNR (e.g., from 72 dB HL to 54 dB HL for the Italian version). Noise onset and offset were 500 ms before and 100 ms after the VCV onset and offset, respectively (see, e.g., [47]). Stimuli were presented monaurally through headphones. At each SNR, VCVs were presented twice. The overall test sequence (i.e., 14 VCVs  10 SNRs  2 repetitions) was completely randomized across SNRs and across VCVs. To familiarize participants with the test stimuli and procedure, VCVs were also presented before the test sequence at a comfortable presentation SNR, above 100% recognition (e.g., þ8 dB [42–44]). The psychometric curves were measured in a group of 50 young adults (24 males, 26 females; age range 21-25 years; mean 23 years, s.d. 1 year), otologically normal as in the ISO 7029:2000 Standard [48] (pure-tone thresholds r10 dB HL in the range 0.125-8 kHz). Subjects were tested only in one ear (the ear with better hearing thresholds). Testing was carried out in low ambient noise as in the ISO 8253-1:1989 Standard for Pure Tone

Audiometry [49], with subjects seating in a sound attenuating booth (PRO 30, Puma Srl, Settimo Milanese, Italy).

2.3.2. Optimization of the test results The SUN test results had to be optimized both monaurally and binaurally to target the highest agreement with the reference criteria for pure-tone screening. On a monaural basis, two cut-off thresholds were defined (T1 and T2) to classify the test scores in the three different categories (‘no hearing difficulties’, scoreZT1; ‘a hearing check would be advisable’, T2 oscoreoT1; and ‘a hearing check is recommended’, scorerT2). The cut-off threshold scores T1 and T2 were systematically varied and the test results were compared to the outcomes of pure tone screening, as defined by the following classes:  Class I (pure-tone hearing thresholds r40 dB HL at 1, 2, and 4 kHz), i.e. ears that would pass a pure-tone screening at a level of 40 dB.  Class II (pure-tone hearing thresholds 440 dB HL at 4 kHz and r40 dB HL at 1 and 2 kHz), i.e. ears that might either fail or pass a pure-tone screening at 40 dB, depending if the frequency 4 kHz is included or not in the pass/refer criterion, respectively [50,51].  Class III (pure-tone hearing thresholds 440 dB HL at 1 or 2 kHz), i.e. ears that would fail a pure-tone screening, irrespective of the criterion for high-frequency hearing thresholds [50–52]. On a binaural basis, the test results in the two ears were combined and all the possible pass/fail criteria were analyzed, from the tightest (criterion A) where a ‘pass’ outcome requires ‘no listening difficulties’ (green light) in both ears to the loosest (criterion F) where a ‘fail’ outcome requires ‘a hearing check is recommended’ (red light) in both ears (see inset table in Fig. 3). Two reference pure-tone benchmarks were considered, i.e.: the Ventry and Weinstein [51] criterion for pure-tone screening (puretone hearing thresholds440 dB HL at 1 and 2 kHz monaurally or at either frequency binaurally) and the World Health Organization (WHO) [53] criterion for disabling hearing impairment (hearing threshold level, averaged over 0.5, 1, 2, and 4 kHz, 440 dB HL in the better ear; hearing aids are usually recommended). Testing was carried out in a population of 150 adults and older adults (67 males, 83 females; age range 40–89 years; mean 65 years; s.d. 9 years) with varying degrees of hearing impairment, also including normal hearing. 136 subjects were tested in both ears whereas 14 subjects, who had a hearing aid, were tested only in the unaided ear, for a total of 286 ears tested.

2.3.3. Equipment A PC connected to the RCA analog input of a clinical audiometer (Amplaid 177þ with TDH49 headphones) was used to deliver the stimuli. The PC was equipped with RME HDSP9632 soundcard (24 bit resolution, sampling frequency 44.1 kHz) compliant with the EN 60645-2:1997 Standard for equipment for speech audiometry [54]. The soundcard output was calibrated by using a 1 kHz tone as in the ‘Equal speech level’ requirement in the ISO 82531:1996 Standard for speech audiometry [38]. A touch sensitive screen (resistive LCD Viper 10.4″, resolution 800  600 pixels, brightness 350 cd/m2, contrast ratio 250:1) was used to display the written alternatives, record the subjects' responses, and display the test score. The actual size of the complete three-letter VCVs displayed on the screen was 6 cm (width)  3 cm (height). An ad hoc software and user interface were implemented in MATLAB (R2007b, v. 7.5.0.342, MathWorks™) to perform experiments. Testing was carried out in

A. Paglialonga et al. / Computers in Biology and Medicine 52 (2014) 66–72

a quiet room as in the ISO 8253-1:1989 Standard for Pure Tone Audiometry [49].

3. Results and discussions 3.1. Development of the test list Fig. 1 summarizes data from the first experiment. The upper panel shows the psychometric curves of eleven out of the 14 VCVs recorded (aba, ara, and aza were excluded according to Ref. [45]).

69

A large variability of RTs was observed across the set, from values lower than  12 dB (e.g., asa) to about þ 2 dB (e.g., ata and aga), further supporting the need to equalize test stimuli. Based on the RTs, three subsets of VCVs were identified. Their pooled psychometric curves are shown in the center panel in Fig. 1:  Subset I: asa, ala, ama (RT r  8 dB, continuous black line);  Subset II: afa, ada, ana, apa (  8 oRT r  2 dB, dashed black line);  Subset III: ata, aka, aga, ava (  2 oRT r þ2 dB, continuous grey line). The presentation SNRs were set at or above the RT for each subset and for each VCV in the subset (i.e.,  8 dB SNR for subset I,  2 dB SNR for subset II, and þ2 dB SNR for subset III). The bottom panel in Fig. 1 shows the psychometric curves as a function of the ‘SNR offset’, i.e. the difference between the actual SNR and the presentation SNR. As a reference, SNR offset¼0 dB identifies the presentation SNR for each VCV. This figure shows that:  At SNR offset ¼0 dB percent recognition was equal to 97%, 94%, and 92% for subsets I, II, and III, respectively, indicating that the intelligibility of VCVs was equalized if stimuli were delivered at the presentation SNR.  The psychometric curves largely overlapped, particularly for SNR offset 4  2 dB, indicating that the intelligibility of VCV was equalized across the whole range of SNRs. The test list was thus developed as the set of VCVs delivered at the presentation SNR in a decreasing-SNR fashion (e.g., from þ2 to  8 dB SNR). At the last presentation SNR (  8 dB) the stimulus asa was presented twice, for a total of 12 stimuli (chance performance ¼4/12). Moreover, as a preliminary ‘training’ phase, a short list of VCVs is also presented (but not scored) before the actual test list at þ8 dB SNR (above 100% recognition). 3.2. Optimization of the test results

Fig. 1. Development of the test list. Top panel: psychometric curves of the 14 VCVs as a function of the SNR. Center panel: psychometric curves of the three VCV subsets: subset I (RTr -8 dB, continuous black line), subset II (  8o RT r  2 dB, dashed black line), and subset III (  2 oRT r þ 2 dB, continuous grey line). Bottom panel: psychometric curves of the three VCV subsets as a function of the ‘SNR offset’, i.e., the difference between the actual SNR and the presentation SNR.

The top panel in Fig. 2 shows the percentage of ears in Class I with scoreZT1 (black marks) and the percentage of ears in Classes II and III with score oT1 (white marks) as a function of T1. The highest agreement between test results and pure tone screening results was observed with T1 ¼9, which provided a sensitivity of 75% and specificity of 80% to identify ears that might fail a puretone screening (Classes II and III). The bottom panel in Fig. 2 shows, once that T1 was set, the percentage of ears in Class III with score rT2 (white marks) and the percentage of ears in Class II with T2 oscoreoT1 (black marks) as a function of T2. Accordingly, the cut-off score T2 was set at 6 as this provided about 70% correct sub-classification between Classes II and III and a sensitivity of 67% and specificity of 88% to identify ears in Class III. It is to note that, as the proposed method is test of suprathreshold acuity in noise, a mismatch with the results of threshold pure-tone detection in quiet is largely expected [19,20,27]. This is in line with findings reported in the literature for other speech-in-noise tests. In general, the agreement between speech-in-noise testing and pure-tone testing widely varies as it depends on the speech material, noise, and protocols that are used, but typically the correlation between speech recognition and pure-tone thresholds is below 0.7 [55,56]. Fig. 3 shows the ROC curves of the SUN test obtained from the 136 subjects tested binaurally as a function of the binaural pass/fail criterion (as reported in the inset table) for the two reference pure-tone benchmarks here considered. The figure shows that the third criterion (C), where a ‘fail’ outcome requires ‘a hearing check is recommended’ (red light) in at least one ear, provided the best

70

A. Paglialonga et al. / Computers in Biology and Medicine 52 (2014) 66–72

100 90

C

80

D

70

Sensitivity (%)

A

B

E

60 50

F

40 30 20 10 0

0

10

20

30

Pass

Fail

A

GG

GY, YY, GR, YR, RR

B

GG, GY

YY, GR, YR, RR

C

GG, GY, YY

GR, YR, RR

D

GG, GY, GR

YY, YR, RR

E

GG, GY, YY, GR

YR, RR

F

GG, GY, YY, GR, YR

RR

40

50

60

70

80

90

100

1-Specificity (%) Fig. 3. ROC curves of the SUN test as a function of the binaural pass/fail criterion (G¼ green, Y¼yellow, R¼ red) for the two reference pure-tone benchmarks: the Ventry and Weinstein criterion for pure-tone screening (dashed black line) and the WHO criterion for disabling hearing impairment (continuous grey line).

Fig. 2. Optimization of the test results. Upper panel: percentage of ears in Class I with scoreZ T1 (black marks) and percentage of ears in classes II and III with scoreo T1 (white marks) as a function of T1. Bottom panel: percentage of ears in Class III with scorer T2 (white marks) and percentage of ears in Class II with T2 o scoreo T1 (black marks) as a function of T2, (with T1 ¼ 9).

the set and (ii) optimization of the test results through maximization of the test sensitivity and specificity. The development steps here shown are general and, as such, they could be used to develop similar speech-in-noise screening tests in different languages. Some minor modifications may be introduced to account for some peculiar characteristic of the different languages. For example, VC or CV stimuli could be used in some languages in place of VCVs, if relevant to the language structure and use. Results obtained so far showed that the test could identify people with hearing impairment with good sensitivity and specificity (around 83–86%) and that, due to short test time, high test-retest repeatability, reliability in high ambient noise, it shows promise as a tool for AHS in clinical and non-clinical settings. Such a test could be easily implemented in a portable, stand-alone system that might also, for example, incorporate additional screening methods (e.g., automatic pure-tone audiometry, other speech-in-noise tests, or screening questionnaires). Future studies would be essential to define the necessary specifications and requirements for such a system for the implementation of different measures on the same platform and, also, to assess the feasibility of an integrated, multiple-test device for use in AHS in different test settings.

Summary trade-off: 82.9% sensitivity and 85.9% specificity compared to the Ventry and Weinstein criterion for pure-tone screening and 83.8% sensitivity and 83.9% specificity compared to the WHO criterion for disabling hearing impairment.

4. Conclusions This paper presented the development of the SUN test, a useroperated speech-in-noise screening test for AHS, developed according to a core set of design criteria, i.e.: (i) reliable as an AHS method; (ii) easy to be performed by the largest possible population; (iii) self-explanatory and convincing; and (iv) short in duration. This paper showed that, starting from the languagespecific ‘building blocks’ (i.e., test stimuli and noise), the test could be developed through two general steps: (i) development of the test list through equalization of the intelligibility of test stimuli across

This paper described the development of the speech understanding in noise (SUN) test, a novel, user-operated test for AHS. The SUN test is a suprathreshold speech-in-noise screening test made up of short list of VCVs in background speech-shaped noise presented in a 3AFC task by means of a touch-screen. The test score is computed as the number of VCVs correctly identified and the test outcome is given as: ‘no hearing difficulties’ (green light) or ‘a hearing check would be advisable’ (yellow light) or ‘a hearing check is recommended’ (red light). This paper described in detail the design criteria (i.e.: (i) reliable as an AHS method; (ii) easy to be performed by the largest possible population; (iii) self-explanatory and convincing; and (iv) short in duration), the building blocks (i.e., (i) test stimuli and (ii) speech-shaped noise), and the main steps for test development (i.e., (i) development of the test list, and (ii) optimization of the test results).

A. Paglialonga et al. / Computers in Biology and Medicine 52 (2014) 66–72

Conflict of interest statement None declared.

References [1] World Health Organization and World Bank. World report on disability. WHO press, Geneve, Switzerland, 2011, vol. 295, p. 298. Available from: 〈http:// whqlibdoc.who.int/publications/2011/9789240685215_eng.pdf〉 (last accessed 30.06.14). [2] A. Davis, P. Smith, M. Ferguson, D. Stephens, I. Gianopoulos, Acceptability, benefit and costs of early screening for hearing disability: a study of potential screening tests and models, Health Technol. Assess. 11 (2007) 1–294. [3] F. Grandori, M. Parazzini, G. Tognola, A. Paglialonga, Hearing Screening in Adults and Elderly is Gaining Momentum, in: L. Hickson (Ed.), Hearing Care for Adults: The Challenge of Aging, Phonak press, Chicago, 2009, pp. 191–202. [4] A.E. Morris, M.E. Lutman, A.J. Cook, D. Turner, An economic evaluation of screening 60- to 70-year-old adults for hearing loss, J. Public Health (Oxf.) 35 (2013) 139–146. [5] S. Arlinger, Negative consequences of uncorrected hearing loss - a review, Int. J. Audiol. 42 (Suppl. 2) (2003) (2S17–2S20). [6] J. Nachtegaal, J.H. Smit, C. Smits, P.D. Bezemer, J.H. van Beek, J.M. Festen, S.E. Kramer, The association between hearing status and psychosocial health before the age of 70 years: results from an internet-based national survey on hearing, Ear Hear. 30 (2009) 302–312. [7] F.R. Lin, K. Yaffe, J. Xia, Q.L. Xue, T.B. Harris, E. Purchase-Helzner, S. Satterfield, H.N. Ayonayon, L. Ferrucci, E.M. Simonsick, Hearing loss and cognitive decline in older adults, JAMA Intern. Med. 173 (2013) 293–299. [8] A. Davis, P. Smith, Adult hearing screening: health policy issues – what happens next? Am. J. Audiol. 22 (2013) 167–170. [9] D.T. Kemp, Otoacoustic emissions, their origin in cochlear function, and use, Br. Med. Bull. 63 (2002) 223–224. [10] H.D. Nelson, C. Bougatsos, P. Nygren, Universal newborn hearing screening: systematic review to update the 2001 US Preventive Services Task Force Recommendation, Pediatrics 122 (1) (2008) e266–e276. [11] B. Yueh, N. Shapiro, C.H. MacLean, P.G. Shekelle, Screening and management of adult hearing loss in primary care: scientific review, J. Am. Med. Assoc. 289 (2003) 1976–1985. [12] B.E. Weinstein, Validity of a screening protocol identifying elderly people with hearing problems, Am. Speech Hear. Assoc. 28 (1986) 41–45. [13] M.A. Zokoll, K.C. Wagener, T. Brand, M. Buschermöhle, B. Kollmeier, Internationally comparable screening tests for listening in noise in several European languages: the German digit triplet test as an optimization prototype, Int. J. Audiol. 51 (2012) 697–707. [14] R.L. Schow, Considerations in selecting and validating an adult/elderly hearing screening protocol, Ear Hear. 12 (1991) 337–348. [15] J. Kiessling, M.K. Pichora-Fuller, S. Gatehouse, D. Stephens, S. Arlinger, T. Chisolm, A.C. Davis, N.P. Erber, L. Hickson, A. Holmes, U. Rosenhall, H. von Wedel, Candidature for and delivery of audiological services: special needs of older people, Int. J. Audiol. 42 (Suppl. 2) (2003) (2S92–2S101). [16] L.V. Knudsen, M. Oberg, C. Nielsen, G. Naylor, S.E. Kramer, Factors influencing help seeking, hearing aid uptake, hearing aid use and satisfaction with hearing aids: a review of the literature, Trends Amplif. 14 (2010) 127–154. [17] A. Laplante-Lévesque, L. Hickson, L. Worrall, Predictors of rehabilitation intervention decisions in adults with acquired hearing impairment, J. Speech Lang. Hear. Res. 54 (2011) 1385–1399. [18] L.E. Humes, J.R. Dubno, S. Gordon-Salant, J.J. Lister, A.T. Cacace, K.J. Cruickshanks, G.A. Gates, R.H. Wilson, A. Wingfield, Central presbycusis: a review and evaluation of the evidence, J. Am. Acad. Audiol. 23 (2012) 635–666. [19] L.E. Humes, Understanding the speech-understanding problems of older adults, Am. J. Audiol. 22 (2013) 303–305. [20] M.C. Killion, P.A. Niquette, What can the pure-tone audiogram tell us about a patient's SNR loss? Hear. J. 53 (2000) 46–53. [21] M.E. Lutman, Speech Tests in Quiet and Noise as a Measure of Auditory Processing, in: M. Martin (Ed.), Speech Audiometry, 2nd ed.,Whurr Publishers Ltd., London, 1997, pp. 63–73. [22] A. Paglialonga, F. Grandori, G. Tognola, Using the SUN test (Speech Understanding in Noise) for adult hearing screening, Am. J. Audiol. 22 (2013) 171–174. [23] F. Grandori, G. Tognola, A. Paglialonga, A novel approach for adult hearing screening, Audiol. Neurotol. 17 (Suppl. 1) (2012) S12–S14. [24] A. Paglialonga, G. Tognola, F. Grandori, SUN-test (Speech Understanding in Noise): a method for hearing disability screening, Audiol. Res. 1 (2011) 49–54. [25] A. Paglialonga, G. Tognola, F. Grandori, Pilot initiatives of adult hearing screening in Italy, Audiol. Res. 1 (2011) 64–68. [26] N. Vaez, L. Desgualdo-Pereira, A. Paglialonga, Development of a test of suprathreshold acuity in noise in Brazilian Portuguese: a new method for hearing screening and surveillance, Biomed. Res. Int., 2014 (Special Issue “Preventing Hearing Loss and Restoring Hearing: A New Outlook”), (in press). [27] L.E. Humes, J.R. Dubno, Factors Affecting Speech Understanding in Older Adults, in: S. Gordon-Salant, R.D. Frisina, A.N. Popper, R.R. Fay (Eds.), The Aging Auditory System, Springer, New York, 2010, pp. 211–257.

71

[28] S. Gordon-Salant, Hearing loss and aging: new research findings and clinical implications, J. Rehabil. Res. Dev. 42 (Suppl. 2) (2005) S9–S24. [29] M.C. Leensen, W.A. Dreschler, The applicability of a speech-in-noise screening test in occupational hearing conservation, Int. J. Audiol. 52 (2013) 455–465. [30] C.S. Watson, G.R. Kidd, J.D. Miller, C. Smits, L.E. Humes, Telephone screening tests for functionally impaired hearing: current use in seven countries and development of a US version, J. Am. Acad. Audiol. 23 (2012) 757–767. [31] C. Smits, T.S. Kapteyn, T. Houtgast, Development and validation of an automatic speech-in-noise screening test by telephone, Int. J. Audiol. 43 (2004) 15–28. [32] J.R. Dubno, D.D. Dirks, Evaluation of hearing-impaired listeners using a Nonsense Syllable Test. I. Test reliability, J. Speech Hear. Res. 25 (1982) 135–141. [33] P.C. Loizou, O. Poroy, M. Dorman, The effect of parametric variations of cochlear implant processors on speech understanding, J. Acoust. Soc. Am. 108 (2000) 790–802. [34] D.A. Vickers, B.C. Moore, T. Baer, Effects of low-pass filtering on the intelligibility of speech in quiet for people with and without dead regions at high frequencies, J. Acoust. Soc. Am. 110 (2001) 1164–1175. [35] M. Cooke, M.L.G. Lecumberri, O. Scharenborg, W.A. Van Dommelen, Languageindependent processing in speech perception: identification of English intervocalic consonants by speakers of eight European languages, Speech Commun. 52 (2010) 954–967. [36] M.A. Akeroyd, Are individual differences in speech reception related to individual differences in cognitive ability? A survey of twenty experimental studies with normal and hearing-impaired adults, Int. J. Audiol. 47 (Suppl. 2) (2008) S53–S71. [37] J.C. van Rooij, R. Plomp, Auditive and cognitive factors in speech perception by elderly listeners, Acta Oto-laryngol 476 (Suppl.) (1990) S177–S181. [38] Acoustics – Audiometric test methods – Part 3: Speech audiometry, ISO Standard 8253-3:1996(E). [39] F. Cutugno, S. Prosser, M. Turrini (Eds.), Audiometria Vocale, vol. I, GN ReSound, Montegrotto, Italy, 2000. [40] D. Byrne, H. Dillon, K. Tran, S. Arlinger, K. Wilbraham, R. Cox, et al., An international comparison of long-term average speech spectra, J. Acoust. Soc. Am. 96 (1994) 2108–2120. [41] M.R. Leek, Adaptive procedures in psychophysical research, Percept. Psychophys. 63 (2001) 1279–1292. [42] P. Assmann, Q. Summerfield, The Perception of Speech Under Adverse Conditions, in: S. Greenberg, W. Ainsworth, W. Popper, R. Fay (Eds.), Speech Processing in the Auditory System, Springer, New York, 2004, pp. 231–308. [43] I.J. Hirsh, E.G. Reynolds, M. Joseph, Intelligibility of different speech materials, J. Acoust. Soc. Am. 26 (1954) 530–537. [44] S. Prosser, M. Turrini, E. Arslan, Effects of different noises on speech discrimination by the elderly, Acta Oto-laryngol. 476 (Suppl.) (1991) S136–S142. [45] H. Strasburger, Converting between measures of slope of the psychometric function, Percept. Psychophys. 63 (2001) 1348–1355. [46] F. Cutugno, S. Prosser, M. Turrini, Speech Audiometry [Audiometria vocale], GN ReSound, Montegrotto, Italy, 2010. [47] N.J. Versfeld., L. Daalder, J.M. Festen, T. Houtgast, Method for the selection of sentence materials for efficient measurement of the speech reception threshold, J. Acoust. Soc. Am. 107 (2000) 1671–1684. [48] Acoustics – Statistical distribution of hearing thresholds as a function of age, ISO Standard 7029:2000. [49] Acoustics – Audiometric test methods – Part 1: Basic pure tone air and bone conduction threshold audiometry. ISO Standard 8253-1:1989. [50] T. Frank, D.R. Petersen, Accuracy of a 40 dB HL Audioscope and audiometer screening for adults, Ear Hear. 8 (1987) 180–183. [51] I.M. Ventry, B.E. Weinstein, Identification of elderly people with hearing problems, Am. Speech Hear. Assoc. 25 (1983) 37–42. [52] W.S. McBride, C.D. Mulrow, C. Aguilar, M.R. Tuley, Methods for screening for hearing loss in older adults, Am. J. Med. Sci. 307 (1994) 40–42. [53] World Health Organization, Grades of hearing impairment, Available from: 〈http://www.who.int/pbd/deafness/hearing_impairment_grades/en/〉 (last accessed 30.06.14). [54] Audiometers – Part 2: Equipment for speech audiometry. EN Standard 606452:1997. [55] P.M. Zurek, L.A. Delhorne, Consonant reception in noise by listeners with mild and moderate sensorineural hearing impairment, J. Acoust. Soc. Am. 82 (1987) 1548–1559. [56] R.H. Wilson, R. McArdle, Speech signals used to evaluate functional status of the auditory system, J. Rehabil. Res. Dev. 42 (2005) 79–94. Alessia Paglialonga received the M.Sc. and Ph.D. degrees in Biomedical Engineering from the Polytechnic of Milan (Italy) in 2005 and 2009, respectively. She is Research Scientist at the Italian National Research Council, Institute of Electronics, Computer and Telecommunication Engineering (IEIIT CNR), formerly with the Institute of Biomedical Engineering (ISIB CNR), Milan, Italy, where she has been a Research Scientist since 2011. Her current research interests include the definition and validation of methods and techniques to assess the auditory functionality and in the optimization of technologies for hearing rehabilitation. Her work focused on adaptive and fixedlevels speech and speech-in-noise tests, psychoacoustics and signal processing for biomedical applications, advanced multiresolution algorithms for the analysis of speech and otoacoustic emissions, modeling of the neural coding in the peripheral and central auditory system for application to cochlear implants and therapeutic

72

A. Paglialonga et al. / Computers in Biology and Medicine 52 (2014) 66–72

electrical stimulation, objective diagnosis and treatment of tinnitus. She has served as scientific co-organizer of the HEAL (Hearing Across the Lifespan) Conference in 2014, the AHS (Adult Hearing Screening) Conferences on adult hearing in 2010 and 2012, and the NHS (Newborn Hearing Screening) Conference on infant hearing in 2012. She has authored 30 papers in peer-reviewed journals and more than 50 communications in international conferences.

Gabriella Tognola received the Ph.D. degree in Biomedical Engineering from the Polytechnic of Milan (Italy) in 1999. In 1993, she joined the Centre of System Theory, now Institute of Biomedical Engineering of the Italian National research Council (ISIB CNR) in Milan. Currently, she is Senior Researcher at ISIB CNR. Her list of publications includes more than 60 papers in peer-reviewed journals, 10 chapters in International books, and more than 100 contributions published in Conference Proceedings. Prof. Tognola primary research interests are in processing of audiological signals, including multi-resolution analysis (time-frequency) and higher-order analysis of otoacoustic emissions, speech processing and design of smart systems for hearing amplification and hearing screening, development and application of models in psychoacoustic for peripheral and central auditory processing, models for perceptual evaluation of clear speech, study of non-linear cochlear mechanisms.

Ferdinando Grandori was born in 1946. He received the Laurea degree in Electronic Engineering from the Polytechnic of Milan (Italy) in 1970. From 1996 until 2001, he was the Director of the Center of Biomedical Engineering at the Polytechnic of Milan and from 2001 to 2013, he served as Director of the Institute of Biomedical Engineering of the Italian National research Council (ISIB CNR) in Milan. He holds faculty appointments at the Polytechnic of Milan (Bioengineering of Neurosensory Systems, Faculty of Engineering).

He has been working since the mid 70s on the development on methods, techniques and protocols for hearing screening and audiological assessment. His main research interests are: electrical evoked potentials and their applications to diagnosis and monitoring; electrical and magnetic fields produced by motor and sensory organs; magnetic stimulation of central and peripheral neural structures; effects of electromagnetic fields; laser-based technologies for 3-D reconstruction of surfaces for medical and non-medical applications; pattern recognition of antropometric data; biomedical informatics; response analysis, interpretation of otoacoustic emissions; protocols and standards for newborn hearing screening, early hearing diagnosis and intervention. Dr. Grandori coordinated a number of multi-year international studies and coordination projects financed by the European Commission that gave shape to research and practical applications in the field of newborn hearing screening and, more recently, of adult hearing screening. He promoted the European Consensus Development Conference on Neonatal Hearing Screening (May 1998), the NHS Conferences on Infant Hearing, the AHS Conferences on Adult Hearing Care and Screening, and the HEAL (Hearing Across the Lifespan) Conference in 2014. He has served as Editor of books and Proceedings, and Guest Editor of about one dozen of Special Issues of international Journals. He is author of 85 papers in peer-reviewed Journals and more than 300 communications at International Conferences. He gave invited lectures at more than 80 meetings in 24 countries.