The Hearing Aid Fitting Process for Frequency Lowering

Hearing Aid Fitting Process
The Hearing Aid Fitting Process for
Frequency Lowering Amplification
• Rehabilitation Plan – 6 stages
1.
2.
3.
4.
5.
6.
www.tinyURL.com/FLassist
Joshua M. Alexander
Ph.D., CCC-A
Assessment
Planning
Selection
Verification
Orientation
Validation
Pre-Fitting
Benefit
Fitting
Post-Fitting
• Evidence-Based Practice
– “Guidelines for Hearing Aid Fitting for Adults”
www.tinyURL.com/PurdueEAR
• 1998 ASHA Ad Hoc Committee, Am. J. Audiol. 7, 5-13
– American Academy of Audiology Guidelines (2006)
2
-10
Typical High Frequency Hearing Loss
0
10
x
x
20
Hearing Level, dB HL
30
Assessment & Planning
• Type and magnitude of hearing loss
• Candidacy for frequency lowering
• Areas of need and potential benefit
40
50
f
x
z v
j md b
n
ng
e l
u
i a
o r
p
h g
ch
sh
th
s
k
x
60
x
70
x
80
x
90
100
110
3
4
120
250
500
1000
2000
4000
8000
Frequency, Hertz
1
-10
Limits of Conventional Amplification
Typical HA Receiver Response
0
x
0
x
20
Hearing Level, dB HL
40
50
f
x
z v
30
j md b
n
ng
e l
u
i a
o r
p
h g
ch
sh
-5
th
s
Relative Level, dB SPL
10
k
x
60
x
70
x
80
x
90
-10
-15
Triple Whammy
Gain is least where
speech energy is least &
hearing loss is greatest
-20
-25
-30
-35
100
250
500
1000
2500
5000
10,000
Frequency, Hz
110
120
5
250
500
1000
2000
4000
8000
6
Frequency, Hertz
-10
Nonlinear Frequency Compression
0
0
x
10
x
20
Hearing Level, dB HL
30
40
50
x
z v
j md b
n
ng
e l
u
i a
o r
30
p
x
60
h g
ch
sh
70
x
80
f
k
x
7
th
s
40
50
110
1000
2000
4000
8000
p
x
h g
ch
sh
x
f
kk
x
8
120
250
500
1000
2000
th
s
x
90
110
Frequency, Hertz
i a
o r
70
100
500
x
j md b
n
ng
e l
u
60
100
250
x
z v
80
x
90
120
x
20
Hearing Level, dB HL
10
Frequency Transposition
-10
4000
8000
Frequency, Hertz
2
Candidacy
Importance of High Frequencies
High occurrence of fricatives/affricates that carry
significant morphological importance in English
Expected outcomes vary depending on
severity of loss
Developmental delays
Moderately severe or greater loss
Precipitous high-frequency loss
Mild to moderate loss
Fricative/affricate production in toddlers, despite
early amplification (Moeller et al., 2007)
Morpho-syntax in 5-7 year-olds identified late,
especially for the morpheme /s/ (Moeller et al., 2010)
Influenced by the information we hope to
‘recover’ from the speech spectrum
Even NH adults listening in quiet can benefit
(Stelmachowicz et al., 2001)
Different needs/goals for different hearing losses
9
Relative importance increases in challenging listening
conditions and with decreased linguistic context
10
Benefits of Extending Bandwidth
Frequency Lowering Evidence
Increasing bandwidth >5 kHz in the laboratory (no lowering)
1. Improved speech recognition (Stelmachowicz & colleagues; Hornsby &
Alexander (2013)
Ricketts, 2006; Hornsby, 2007)
Seminars in Hearing, 34(2), 86-109.
Review of modern techniques
2. Improved novel word learning by children (Pittman 2008)
3. Improved speech clarity and music quality (Moore & Tan,
McCreery et al. (2012)
2003; Ricketts et al., 2008; Sjolander & Holmberg, 2009; Füllgrabe et al., 2010; Moore et al., 2011)
American Journal of Audiology, 21, 313-328.
Evidence-based review focusing explicitly on children
4. Less effortful listening (Karlsen et al., 2006)
5. Improved localization (Best et al., 2005; Brungart & Simpson, 2009)
Simpson (2009)
Trends in Amplification, 13(2), 87-106.
Review of early techniques
6. Improved spatial unmasking (Hamacher et al., 2005; Moore et al., 2010)
11
12
3
Frequency Lowering Evidence
Effectiveness - Points to Consider
There is not a lot of good evidence
Results might be highly individualized
Double-blinded, randomized control experiments
Many uncontrolled variables within and between
studies, especially how settings are chosen for FL
Perhaps the focus should be on individuals, in addition to
group means — do they form a subpopulation along one
or more predictive factors?
Interactions with audiometric configuration, peripheral
integrity, cognitive status, FL settings, etc. are expected
The evidence so far indicates that, when properly
set, frequency lowering does not do harm
Benefit may not always manifest as improvement in
speech intelligibility, but perhaps, ease of listening
and effort (like noise reduction), production
accuracy, etc.
Estimate that about ½ of research subjects fail to show
a difference in outcomes with FL
Depends on how outcomes are measured
Plural detection in quiet vs. sentence recognition in noise
13
14
Modeling Individual Outcomes
Extrinsic Factors
A. Which sounds?
Usually, most improvement for fricative consonants,
especially /s/ (e.g., Alexander et al., 2012; 2014)
B. Which technique and settings?
Likely depends on the sounds one is trying to restore (A)
and the individual loss (C)
C. Which losses?
As severity increases, so do potential benefits (Glista et
al., 2009; Souza et al., 2013), but unfortunately, so do
the challenges (e.g., Hopkins et al., 2014)
Alexander (2013)
15
16
4
Intrinsic Factors
Potential Side Effects
A. Adults vs. children
While the speech code is relatively ‘scale invariant,’ it is
heavily dependent on frequency
No differences observed when both groups have the same
treatment (McCreery et al., 2014; Brennan et al., 2014)
No currently implemented strategy has as much potential to
change the identity of individual speech sounds
Potential to make speech understanding worse because lowfrequency information has to be altered to accommodate
displaced high-frequency information
B. Age of hearing loss onset
Pre/post linguistic differences and etiology may relate to the
amount of deficit, hence amount of benefit – no data
C. Acclimatization
Maybe some improvement in children over 6 weeks to 6
months (Wolfe et al., 2011; Glista et al., 2012); maybe no
improvement in adults (Hopkins et al., 2014)
Re-coded information must go somewhere
Regions that otherwise would be amplified normally
Concern is not so much fidelity of re-coded information as it is
newly introduced distortion and sound quality
D. Role of cognition
Amount of relearning needed may mediate this factor
(YES: Arehart et al. 2013; NO: Ellis & Munro, 2013)
17
18
Information-Theoretic Model
Potential Pros vs. Cons
(a) Wideband
A successful fitting
requires giving the
patient more than
is taken away
less can equal more
Frequency (kHz)
DO NO HARM
Depends on interaction
b/w hearing loss &
re-coding technique
ee
S
ee
S
(c) NFC
ee
S
1
0
9
8
7
6
5
4
3
2
1
100
200
300
400
500
100
200
300
Time (ms)
Alexander (2013)
19
(b) Low Pass Filtered
400
500
100
200
300
400
500
(Alexander, 20XX)
20
5
Settings vary Potential Pros vs. Cons
Candidacy Summary
Loss ≥ Moderately-severe
Go ahead, but be careful of distorting low frequencies
WARNING! If the hearing aid has FL, it might default to “ON” for
certain audiometric configurations
Loss ≤ Moderate
Do not base hearing aid selection on whether it has FL
There are other important considerations: cost, wireless
connectivity, etc.
If the hearing aid you select has FL, it might be worth
trying (i.e., experimenting)
21
22
(Rallapalli & Alexander, 20XX)
To Fit or Not to Fit
Ultimately, a decision has to be made whether the
potential pros outweigh the cons
Does the patient experience speech perception deficits
with conventional amplification, despite your best efforts
to achieve high-frequency audibility?
Selection
• Choosing the frequency lowering
technology whose electroacoustic
characteristics are most likely to address
the areas of need and result in benefit
If the decision is to fit, there are a few things to ask:
1. How does the technology of choice work?
Fundamental differences between manufacturers: techniques,
terminology, adjustments, etc.
2. How much of the lowered information is audible?
3. Can the patient use the lowered information?
23
24
6
Techniques in Today’s Hearing Aids
Techniques in Today’s Hearing Aids
Dynamic Linear Frequency Compression (DLFC)
AVR Sonovation: “Digital Frequency Compression” (DFC) – Introduced in
1991 (body), 1998/2000 (BTE/ITE), 2004 (DSP)
Inaudible
Frequency Transposition (FT)
Widex: “Audibility Extender” (AE) – Introduced in 2006
Nonlinear Frequency Compression (NFC)
Phonak: “SoundRecover” (SR) – Introduced in 2008
Unitron: same as Phonak – Adopted 2012
Siemens: “Frequency Compression” (FCo) – Introduced in 2012
ReSound: “Sound Shaper”– Introduced in 2014
Frequency
Frequency
Frequency
Frequency
Nonlinear
Frequency
Compression
Spectral
Feature
Translation
Linear
Frequency
Transposition
Frequency
Composition
Spectral Envelope Warping
Starkey/Microtech: “Spectral iQ” – Introduced in 2011
Time
Bernafon
Time
Widex
Time
Starkey /
Microtech
Frequency Composition
Time
Phonak/Unitron,
Siemens, ReSound
25
Bernafon: Introduced in 2012
26
Linear Frequency Transposition
Dynamic Linear Freq. Compression
Filter
10k
9k
8k
7k
6k
Input
5k
4k
1768 Hz
3536 Hz
target
source
3k
2k
Start Freq.
1k
Start frequency is start of inaudibility
Basic AE: dominant peak shifted down by 1/2 (octave),
source region begins ½ octave below start frequency
Expanded AE: dominant peak shifted down by 1/3, source
region begins ½ octave above frequency start; peak shift
from basic AE remains
Adapted from: Widex (95020880001 #01)
ch il d r en l i ke s trawberrie /z/
6k
5k
Output
4k
3k
2k
1k
27
28
7
Nonlinear Frequency Compression
Linear Frequency Transposition
10k
Input
9k
7k
*
6k
*
*
5k
*
4k
Input
Actual start
3k
2k
1k
ch il d r en li ke strawberrie/z/ ee SH
6k
5k
Target
30
25
25
20
Output
3k
10
0
Source content
mixes with content
in target region
2k
1k
Source Region
≈ 4500 Hz wide
15
5
?
4k
Output
30
(fend)
(fmin)
Input BW
Start
Relative Level, dB
*
8k
Relative Level, dB
6 kHz
‘start’
Source
20
15
4000
Frequency, Hz
6000 8000
Target Region
≈ 1800 Hz wide
10
5
2000
CR=2:1
0
fmax
fmin
2000
4000
6000 8000
Frequency, Hz
30
29
Side Effects - Revisited
NFC for Moderately Severe Loss
Formant alteration, vowel reduction with low start freq.
6k
5k
Source
4k
3k
*
2k
*
*
* *
*
Target
1k
ch
31
i l d
r
en
l i ke s t r aw b err ie
Flattened formant transitions
/z/
32
8
NFC Differs between Manufacturers
NFC for Mild-Moderate Loss
What does “nonlinear” mean anyway???
9k
Source
8k
From the start frequency, the input-output
frequency relationship is defined by the
compression ratio (CR)
7k
6k
Target
5k
Higher CRs = greater reduction of source bandwidth
The relationship on a Hz scale is
4k
3k
1. Nonlinear (but linear on a log scale) – Phonak/Unitron
2. Linear – ReSound
CR is not compatible
3. Indeterminate (by me!) – Siemens across manufacturers!
2k
1k
ch
i l d
r
en
l i ke s t r aw b err ie
33
/z/
34
Spectral Feature Translation
Spectral Feature Translation
Spectral Envelope Warping
Spectral Envelope Warping
After
BEFORE
35
36
9
Spectral Feature Translation
Frequency Composition
10k
9k
8k
7k
6k
5k
4k
3k
2k
1k
ch il d r en l i ke s trawberrie /z/
6k
5k
4k
3k
2k
1k
(Angelo, Alexander, et al., 2015)
37
38
Frequency Composition
Th e
d
o
g
s
l ee
ps in a
b
a
Summary of Strategies
s
k e
t
Activation
Technique
Input Dependent
Compression
AVR
(linear @ 0 Hz)
Transposition
Starkey
(feature lowering)
Compression +
Transposition
39
(Angelo, Alexander, et al., 2015)
Always Active
Phonak/Unitron
Siemens
ReSound
(nonlinear @ start)
Widex
(peak lowering)
Bernafon
(frequency
composition)
40
10
Frequency Lowering Taxonomy
Frequency Lowering Taxonomy
Technique
Potential side effects
Compression (AVR, Phonak, Siemens, ReSound, Bernafon)
Transposition (Widex, Starkey, Bernafon)
Masking (Widex, Starkey, Bernafon)
Formant alteration (AVR, Widex, Phonak, Siemens, ReSound)
Activation
Input dependent (AVR, Starkey)
Always active (Phonak, Siemens, ReSound, Widex, Bernafon)
Bandwidth (not an issue if use probe mics)
Source region preserved (Starkey, Bernafon)
Source region rolled off (AVR, Phonak, Siemens, ReSound, Widex)
Additional gain for lowered signal
No (Phonak, Siemens, ReSound)
Yes (AVR, Widex, Starkey, Bernafon)
41
Methods for preserving low frequencies
1. How: maintain harmonics (Widex, Starkey, Bernafon)
2. Where: > 1.5 kHz only (Phonak, Siemens, Bernafon)
> 2.5 kHz only (ReSound)
3. When: ‘feature detector’ (AVR, Starkey)
42
To Fit or Not to Fit
Ultimately, a decision has to be made whether the
potential pros outweigh the cons
Does the patient experience speech perception deficits
with conventional amplification, despite your best efforts
to achieve high-frequency audibility?
Verification
• Probe microphone measurements to
guide programming of different frequency
lowering techniques for individuals
If the decision is to fit, there are a few things to ask:
1. How does the technology of choice work?
Fundamental differences between manufacturers: techniques,
terminology, adjustments, etc.
2. How much of the lowered information is audible?
3. Can the patient use the lowered information?
43
44
11
Importance of Probe Mic Measures
Audiology Practices
(2014), Vol. 6(4)
With the possibility of
side effects causing
‘harm,’ if you plan to fit a
hearing aid with FL, you
must know what you are
delivering to the patient
www.tinyURL.com/FLassist
http://web.ics.purdue.edu/~alexan14/
Publications_files/FL_ProbeMic.pdf
46
45
Primary Goals for Probe Mic Measures
1. The audible bandwidth after FL is activated should
not be less than it was before it was activated
FL should not unnecessarily restrict the audible bandwidth
Protocol for Fitting FL Hearing Aids
1. Deactivate frequency lowering and fit the hearing aid
to targets using probe mic as for a conventional aid
2. Find the maximum audible frequency, MAF
The highest frequency at which output exceeds threshold on
the SPL-o-gram (Speechmap)
2. The lowered information should be audible
3. The ‘weakest’ FL setting should be used to
accomplish your objective
3. Activate FL and position the lowered speech in the
audible bandwidth (MAF) while not reducing it further
o Most of the target region should be audible
o Avoid too much lowering, which will unnecessarily restrict the
bandwidth you had to start with and reduce intelligibility
o Frequency Lowering Fitting Assistants:
www.tinyURL.com/FLassist
47
48
12
3. Activate NFC, adjust settings
1. Deactivate NFC and Fit to Targets
2. Find the maximum audible frequency
www.tinyURL.com/FLassist
NFC off
Optimal Setting
Too Strong Setting
Under-utilized
Audible BW
49
50
3. Verify Bandwidth of Chosen Setting
Special Caveat with Transposition
With frequency compression, the levels of the
input speech spectrum can be inferred by from the
amplified output and the Fitting Assistants
However, because transposition mixes the signals,
you cannot tell what the contribution of each is to
the overall amplified levels
FL off
FL on
51
While the Fitting Assistants can indicate whether the
transposed speech has been moved to a region where
aided audibility is possible, they cannot specifically verify
that the transposed signal is audible
52
13
Special Verifit Test Signals
Special Verifit Test Signals
(www.audioscan.com/resources/audionotes/audionote2rev3lowrez.pdf)
53
54
More on AudiologyOnline.com
20Q: The Highs and Lows of Frequency Lowering
Amplification (Article # 11772) by J. Alexander
Orientation
20Q: The Ins and Outs of Frequency Lowering
Amplification (Article # 11863) by S. Scollie
• Counsel on use of the technology and
foster realistic expectations
20Q: Frequency Lowering - The Whole Shebang (Article #
11913) by G. Mueller, J. Alexander, & S. Scollie
Validation
Frequency Compression - Understanding the Clinical
Application (Webinar # 23078) by S. Scollie
• Evaluation impact of the technology
on areas of need and overall benefit
55
www.tinyURL.com/FLassist
56
14