The prescription method for hearing aids Harvey Dillon, Gitte Keidser, Teresa Ching, Matt Flax, Scott Brewer HEARing CRC National Acoustic Laboratories, Australia Audiology Online June 212 Talk structure (?) Intelligibility in noise Frequency resolution OAEs Dead regions Thresholds NL2 formula Age Gender Experience Binaural Temporal resolution Intelligibility derivation Language Speed Loudness Amplification rationales NAL-NL1 NL1 deficiencies CR Children Severe profound Infants Gain preferences Outcomes nhl ehl RECD The Fitting Process Sing Soc Audiol Professionals 1
Adult Measure hearing thresholds (db HL) Child Measure hearing thresholds (db HL or db SPL) Measure individual RECD (or estimate from age) Enter into manufacturer software (hearing aid auto adjusted to approximate prescription) Enter into manufacturer software (hearing aid auto adjusted to approximate prescription) Verify with real ear measurement Adjust amplification to better match prescription Adjust hearing aid in coupler via computer to better match prescribed coupler gain Prescription goals Prescribe hearing aids to: Make speech intelligible Make loudness comfortable Prescription affected by other things localization, tonal quality, detection of environmental sounds, naturalness. Sing Soc Audiol Professionals 2
Hearing aids amplify, so. How much amplification? Gain 5 db SPL 65 db SPL 8 db SPL Frequency The lolly shop Deriving Sing Soc Audiol Professionals 3
The rationale for NAL proceudres Maximize calculated speech intelligibility, but Keep total loudness less than or equal to normal NAL-NL1 (1999) empirical studies psychoacoustic studies speech intelligibility models Deriving optimal gains -step 1 Speech spectrum & level Loudness model Normal loudness Gain-frequency response Compare Intelligibility achieved Intelligibility model Amplified speech spectrum Loudness model Loudness (hearing impaired) Audiogram The audiograms Inverted hearing loss profiles used Rejection criterion : -3<= G <=6, where G is the slope sum(h(f))/3 <=1, where f is in the set {.5, 1, 2} khz Sing Soc Audiol Professionals 4
The audiograms, continued Deriving optimal gains -step 1 Audiogram 1 Speech level 1 Optimal gain frequency response Audiogram 1 Speech level 2 Optimal gain frequency response Audiogram 1 Speech level 3 Optimal gain frequency response Audiogram 2 Speech level 1 Optimal gain frequency response 2 audiograms x 6 speech levels 12 gain frequency responses, each at 2 frequencies from 125 Hz to 1 khz Overall prescription approach Psychoacoustics Assumptions, rationale Theoretical predictions Adjust Final formula Compare Speech science Empirical observations Sing Soc Audiol Professionals 5
Multi-dimensional equation: A neural network H 25 H 5 H 1 H 2 H 8k SPL G 25 G 5 G 1 G 2 G 8k The two key ingredients 1. A loudness model 2. An intelligibility model Speech spectrum & level Loudness model Normal loudness Gain-frequency response Compare Intelligibility achieved Intelligibility model Amplified speech spectrum Loudness model Loudness (hearing impaired) Audiogram Calculating loudness Loudness model of Moore and Glasberg (24) Allowance for hearing loss External & middle ear Filtering into auditory bands Calculate loudness per band Sum across bands Free field speech level Input to cochlea Excitation level Loudness per band Total loudness Sing Soc Audiol Professionals 6
Speech intelligibility 1/3 octave SPL 3 Audibility:... 5... 16... 17... x x x x Importance:.1.2.3.2 = = = =....5....32....51... Freq =.3 Speech Intelligibility Index Sum SII = A i I i Audibility Importance But intelligibility gets worse if we make speech too loud! Sing Soc Audiol Professionals 7
Speech intelligibility also depends on Level distortion Normal hearing people perform poorer at high speech levels tor Level distortion fact 1 73 14 Speech level (db SPL) SII = A i I i L i Level distortion factor 1 Percen nt Correct 8 6 4 2 The transfer function Sentences Nonsense syllables.2.4.6.8 1 Speech Intelligibility Index (SII) Sing Soc Audiol Professionals 8
Observed and Predicted performance correct 14-56 Hz 1 MF MS SF SS 8 6 Percent 4 2 12 24 36 12 24 36 12 24 36 12 24 36 Sensation level (db) Ching, Dillon & Byrne, 1998 Subjects 2 adults with normal hearing 55 adults with sensorineural hearing loss mild to profound Experienced hearing aid users Hearing threshold leve el (db HL) -2 2 4 6 8 1 12 Audiogram 25 5 1 2 4 8 Frequency (Hz) Speech perception Stimuli: Filtered speech CUNY sentences VCV syllables Shaping: POGO prescription p Conditions: Quiet at high and low sensation levels Babble Noise Headphones: Sennheiser HD25 evel (db SPL) Le 8 6 4 2 HP7 HP14 HP28 LP7 LP14 LP28 LP56 25 5 1 2 4 8 Frequency (Hz) Sing Soc Audiol Professionals 9
Audibility and Speech intelligibility H.I. 1. VCV.8 ect Proportion corre 6.6.4.2...2.4.6.8 1. Calculated SII LP7: Q LP7: N LP14: Q LP14: N LP28: Q LP28: N LP56: Q LP56: N HP28: Q HP28: N HP14: Q HP14: N HP7: Q HP7: N Deficit = S ansii Percen nt Correct 1 8 6 4 2 ansii -SII eff SII eff SII ansi Deficit =.6 -.4 =.2.2.4.6.8 1 Speech Intelligibility Index (SII) VCV deficit vs CUNY deficit.25.2.15 R=.77.1 CUNY SIIansi-S SIIeff.5. -.5 -.1 -.15 -.2 -.25 -.15 -.1 -.5..5.1.15.2.25.3 VCV SIIansi-SIIeff Sing Soc Audiol Professionals 1
Intelligibility and audibility 1 m p 3 Sensation level (db) BKB, VCV and CUNY Optimizer results: 3 data sets BKB VCV CUNY Q & N Sing Soc Audiol Professionals 11
Desensitisation for hearing loss ve audibility Effecti 1.9.8.7.6.5 4.4.3.2.1 2 4 6 Sensation level (db) 2 4 6 8 1 12 m parameter 1.8.6.4.2 2 4 6 8 1 12 Hearing threshold (db HL) Psychoacoustics Why measure only pure tone thresholds? Other measurements Outer hair cell function click-evoked otoacoustic emissions Frequency resolution psychophysical tuning curves cochlear dead regions TEN test Cognitive ability Age Sing Soc Audiol Professionals 12
Healthy PTC no dead region Psychophysical tuning curve 11 A29 Masker Level (db SPL) 1 9 8 7 6 Signal Masker 5 5 1 15 2 Masker Frequency (Hz) Poor PTC: Dead region at 4 khz 12 Psychophysical tuning curve SPL) 11 Masker Level (db 1 9 8 1 2 3 4 5 Masker Frequency (Hz) Dead regions RIP NAL-NL1 only allows for hearing loss desensitization on average Sing Soc Audiol Professionals 13
Off-frequency frequency listening: TEN test Basilar membrane e vibration Threshold Equalizing Noise (TEN) Frequency or position Based on Moore (24) TEN and PTC (non) agreement 2 khz TEN: Alive TEN: Dead TEN uncertain PTC: 6 1 Tip in place 1 PTC: Tip shifted 4 3 2 PTC uncertain 1 2 1 Psychoacoustic correlations 4 khz HL Matrix Plot (Prof and Psy 11 March 9.sta 659v*75c) PTC Q1 OAE strength TEN elevation Cognition Age Sing Soc Audiol Professionals 14
Psychoacoustic correlations 2 khz Matrix Plot (Prof and Psy 11 March 9.sta 659v*75c) HL2k PTC2k_Q1 C2k El2k Cognition AgeLim Can we better predict intelligibility if we use psychoacoustic results? Deficit (VCV & CUNY), HL, Q1, OAE, TEN, Cog, & Age: 56 Hz low pass es-siieff (VCV Avg(QH QL N) LP56) SIIdes-SIIeff (CUNY Avg(QH QL N) LP56) L56HL Q1 LP56 COAE L56 EL LP56 Cogni tion AgeLim Sing Soc Audiol Professionals 15
Yes, a little speech deficit increases as frequency selectivity gets broader But not once we fully build HL into the SII prediction Correlations 5 Hz HL 5 PTC5_ OAE TEN Cognit Age Q1 5 5 HL 5 -- -.7.4.56 -.38.26 PTC5_Q1 -.7 -- -.6 -.48.33 -.17 OAE 5.4 -.6 -- -.1 -.5 -.5 TEN 5.56 -.48 -.1 -- -.35.4 Cognition -.38.33 -.5 -.35 -- -.45 Age 26.26 -.17 17 -.5 5 4.4 -.45 45 -- 1 khz HL 1k PTC1k OAE TEN Cognit Age _Q1 1k 1k HL 1k -- -.63 -.36.46 -.38.3 PTC1k_Q1 -.63 --.3 -.1.23 -.17 OAE 1k -.36.3 -- -.18.2 -.2 TEN 1k.46 -.1 -.18 -- -.13.1 Cognition -.38.23.2 -.13 -- -.45 Age 3.3 -.17 -.2 1.1 -.45 -- 2 khz HL 2k PTC2k OAE TEN Cognit Age _Q1 2k 2k HL 2k -- -.79 -.65.53 -.4.35 PTC2k_Q1 -.79 --.49 -.47.36 -.25 OAE 2k -.65.49 -- -.36.35 -.36 TEN 2k.53 -.47 -.36 -- -.38.15 Cognition -.4.36.35 -.38 -- -.42 Age.35 -.25 -.36.15 -.42 -- 4 khz HL 4k PTC4k OAE TEN Cognit Age _Q1 4k 4k HL 4k -- -.84 -.6.58 -.43.41 PTC4k_Q1 -.84 --.5 -.49.33 -.23 OAE 4k -.6.5 -- -.31.44 -.4 TEN 4k.58 -.49 -.31 -- -.45.38 Cognition -.43.33.44 -.45 -- -.45 Age.41 -.23 -.4.38 -.45 -- Correlations Age PTC HL OAE Cognit TEN Sing Soc Audiol Professionals 16
Multiple regression including HL causes: correlations between age and PTC / OAE / TEN to disappear correlations between cognition and PTC / OAE / TEN to disappear PTC Age HL OAE Cognition TEN Likely intermediate effects Cognition Mechanical? PTC Age Cardiovascular Noise Stria OHC OAE TEN IHC HL Why are hearing thresholds so useful? Frequency selectivity Hearing thresholds Temporal resolution Central auditory processing Speech Perception proficiency Age Other Cognitive ability Sing Soc Audiol Professionals 17
Factors affecting prescription National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Gain; 187 adults, medium input level National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Source: Keidser et al., 28 Sing Soc Audiol Professionals 18
National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Adjustments to prescription to allow for experience Gain adjustm ment 12 1 8 New 6 Experienced 4 2-2 2 4 6 8 1 12-4 -6-8 Hearing threshold (db HL) National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Adjustments to prescription to allow for experience Gain adjustm ment 12 New 1 8 Experienced 6 Experienced - 4 New 2-2 2 4 6 8 1 12-4 -6-8 Hearing threshold (db HL) National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Gain preference over time N = 11 Source: Keidser et al. (submitted) Sing Soc Audiol Professionals 19
National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Gain; adults, low and high input levels ation from eferred at 65 db Preferred gain devi NAL-NL1 re gain pre db SPL in d 1.5 1.5 -.5-1 -1.5-2 Smeds et al. 26 Zakis et al. 27 5 8 Input level in db SPL Desired gain Output level Gain and compression; adults vs children Children, NL2 NAL-NL1 Adults, NL2 Age dependent gain Input level National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Adults congenital or acquired? eviation (db) Preferred gain de from NAL-RP -2-4 -6-8 -1-12 -14 Congenital (N=15) Acquired (N=28) LFA HFA Sing Soc Audiol Professionals 2
Compression Limiting compression for severe/profound hearing loss (Fast compression) Source: Keidser et al., 27 The maximum C NL2 prescribes different compression for slow and fast compressors for those with high degrees of hearing loss creating sound value TM 3. 2.5 Ca 2. 1.5 1. H 5 2 1 6 4 log2 f 125 National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Compression ratio preferences: severe and profound hearing loss LF band Average 1/CR in L 1.1 1..9.8.7.6.5.4.3 1:1 1.8:1 3:1.2 45 5 55 6 65 7 75 8 85 9 Average HTL in LF band (db HL) Source: Keidser, Dillon, Dyrlund, Carter, and Hartley (27) Bilateral loudness correction Gain preference by aid configuration (N = 187) Bilateral loudness correction too large supported by newer data (e.g. Whilby, 26; Epstein & Florentine, 29) Sing Soc Audiol Professionals 21
New bilateral loudness correction riation (db) Gain va 7 6 5 4 3 2 1 2 4 6 8 1 12 Input level (db) National Acoustic Laboratories, Sydney, Australia Keidser and Dillon Effect of language Gain at each frequency depends on importance of each frequency Low frequencies more important in tonal languages Two versions of Tonal languages Non-tonal languages Tonal versus non-tonal language tion gain in db Insert 4 35 NAL NL1 3 Male, exp, non tonal 25 Male, exp, tonal 2 15 1 5 1 1 1 Frequency in Hz Sing Soc Audiol Professionals 22
RECD in infants (own mold; HA2) 35 3 25 REC CD_4k 2 15 1 5-5.6.8 2. 4. 6. 8. 2. 6. 4. 8. 2. 4. Age (months) Summary New features in Different gain-frequency response shape and higher compression ratios Different compression ratios for fast and slow compressors (severe/profound hearing loss) Gender dependent gain Age dependent gain Gain adaptation for new hearing aid users Language dependent gain (tonal vs nontonal) Examples of prescriptions Sing Soc Audiol Professionals 23
Example audiogram: moderate sloping Frequency (Hz) HL) Hearing threshold (db 125 25 5 1k 2k 4k 8k 2 4 6 8 1 12 n (db) Insertion Gain 4 35 3 25 2 15 1 5 5 db 65 db 8 db 1 1 1 Frequency (Hz) Example audiogram: flat 6 Frequency (Hz) HL) Hearing threshold (db 125 25 5 1k 2k 4k 8k 2 4 6 8 1 12 n (db) Insertion Gain 4 35 3 25 2 15 1 5 5 db 65 db 8 db 1 1 1 Frequency (Hz) Example audiogram: steeply sloping Frequency (Hz) HL) Hearing threshold (db 125 25 5 1k 2k 4k 8k 2 4 6 8 1 12 n (db) Insertion Gain 4 35 3 25 2 15 1 5 5 db 65 db 8 db 1 1 1 Frequency (Hz) Sing Soc Audiol Professionals 24
Example audiogram: extreme ski-slope Frequency (Hz) HL) Hearing threshold (db 125 25 5 1k 2k 4k 8k 2 4 6 8 1 12 n (db) Insertion Gain 35 3 5 db 25 2 15 1 5 65 db 8 db 1 1 1 Frequency (Hz) Example audiogram: reverse sloping Frequency (Hz) HL) Hearing threshold (db 125 25 5 1k 2k 4k 8k 2 4 6 8 1 12 n (db) Insertion Gain 35 3 25 5 db 2 15 1 5 65 db 8 db 1 1 1 Frequency (Hz) Johnson & Dillon, JAAA, 211 Figure 5. Specific loudness for A-1 through A-4 based on a 65 db international long-term average speech spectrum input in quiet. Overall loudness is shown to the right of the prescriptive method in the legends (NH designates normal hearing). Sing Soc Audiol Professionals 25
Johnson & Dillon, JAAA, 211 Figure 5. Specific loudness for A-1 through A-4 based on a 65 db international long-term average speech spectrum input in quiet. Overall loudness is shown to the right of the prescriptive method in the legends (NH designates normal hearing). Figure 7. Overall loudness of each prescriptive method averaged across the five sensorineural hearing losses (A-1 through A-5) based on a 65 db international long-term average speech spectrum input in quiet. Johnson & Dillon, JAAA, 211 Figure 8. Average Speech Intelligibility Index (SII) value for speech in quiet across the five sensorineural hearing losses for each prescriptive method using both the ANSI S3.5-1997 and the National Acoustic Laboratories SII methods. Also shown is the SII transformed value into a predicted speech recognition score (% correct) for the Connected Speech Test (Cox et al, 1987) using the transfer function of Humes (22). Johnson & Dillon, JAAA, 211 Sing Soc Audiol Professionals 26
We measure several times, NAL & DSL groups have equal language outcomes T Ching, NAL, CRC HEAR Relationships between variables in Age Vent Tube RECD Comp speed Language Bi-uni Gender Experience REDD WBCT BWC UCT N I/O CR RECD REUG Depth Aid type Transducer AC' BC' AC ABG BC REIG RESR SSPL 2cc REAG CG ESCD MLE ESG Blue = User i/p Grey = internal variable Red = effect of saturation Dash-dot = alternatives Green = stored data Limiting type SSPL ES Taking the pressure off prescription The trainable hearing aid Sing Soc Audiol Professionals 27
Aid user adjusts settings... Trainable Aid Dillon et al (26) Zakis et al (27) Process repeats for other sounds Trainable Aid After training... Trainable Aid...preferred settings are automatically applied Sing Soc Audiol Professionals 28
Training gain, CR, CT. 3 25 Gain (db) 2 15 1 5 Gain CR CT 3 4 5 6 7 8 9 Input level (db SPL) A challenge for the profession is to devise fitting procedures that are scientifically defensible and the challenge for the individual audiologist is to choose the best procedures from whatever are available Denis Byrne, 1998 Acknowledgements www.hearingcrc.org www.nal.gov.au This research was financially supported by the HEARing CRC established and supported under the Australian Government s Cooperative Research Centres Program creating sound value TM Sing Soc Audiol Professionals 29