Download ENGLISH for ENGINEERS This is an example of a clinical proposal

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Earplug wikipedia , lookup

Auditory system wikipedia , lookup

Noise-induced hearing loss wikipedia , lookup

Soundscape ecology wikipedia , lookup

Audiology and hearing health professionals in developed and developing countries wikipedia , lookup

Sound from ultrasound wikipedia , lookup

Sensorineural hearing loss wikipedia , lookup

Transcript
ENGLISH for ENGINEERS
This is an example of a clinical proposal written by an MSc student. The student was
given the following scenario:
You are based in an Audiology department that carries out standard hearing tests
including pure tone audiometry, speech audiometry in quiet and speech audiometry in
noise. The department has standard audiometric equipment, including audiometers
with CD players for speech audiometry. Your head of department has heard that
measurement of frequency resolution may be a useful additional test and has asked you
to research the issue and make a proposal to the department for a test that must be
completed in a maximum of 30 minutes for both ears.
Executive summary
1. Frequency resolution ability is the capability of the ear to discriminate a tone at one
frequency from a tone, or sound, at a different frequency. It is important for speech
discrimination in noise.
2. Frequency resolution ability can be derived from the psychophysical tuning curve (PTC).
3. This proposal describes a clinical test which uses a masking method, based on Lutman
and Wood (1985), for measuring one PTC in both ears within 30 minutes. Characteristic
parameters of the PTC are identified. However, there is no reference data available to
relate these parameters directly to a subjects’ frequency resolution ability. This data has
to (and can) be derived before the test can be employed reliably in the clinic.
4. The test can be performed with the existing audiometric equipment in the department.
5. The test should be used for subjects with hearing aids, or subjects with normal hearing,
who complain about problems with speech discrimination in noise.
6. Knowledge about a subjects’ frequency resolution might be useful in the clinic for
counselling but there is no consensus about its usefulness in fitting hearing aids.
7. Taking into account the previous point and the absence of reference data, it is advisable
to start with a pilot study in order to get a better feeling if PTCs, and consequently a
clinical test for frequency resolution, is useful.
1. Introduction
1.1
Frequency resolution and the auditory filter
Frequency resolution ability can be interpreted as the filtering capability of the auditory
system and is important for speech recognition. For example a reduction in this (spectral)
ability makes it difficult to distinguish speech elements like formants in vowels. It can be
studied by the shape of the underlying filter – called the auditory filter – that is centred at the
frequency component of interest. Different methods can be used to estimate the shape of this
filter. In general these methods use a pure tone signal (mostly just above hearing threshold)
and they are based on the assumption that a subject can detect the tone if the signal-to-noise
ratio (SNR) within the auditory filter, exceeds a critical threshold. The reduction of the SNR
is the reason that problems arise in a noisy environment. Noise can be interpreted as all
sounds different from the signal pure tone. Most methods use the principle of changing the
level and/or frequency content of the noise in order to obtain the shape of the filter. The noise
added on purpose in the method is called ‘the masker’ and the differences between methods
are mainly related to the maskers applied. The auditory filter is nonlinear and its shape
depends on the method and signal level used, e.g. the filter broadens with increasing signal
level (Moore, 2012; p. 78).
A psychophysical (or psychoacoustical) tuning curve (PTC) is closely related to the auditory
filter1 and it shows the frequency sensitivity of the ear as a function of frequency for a pure
tone of a chosen frequency, see Figure 1.
Two characteristics of the PTC are very important for frequency resolution:
1. The location of the dip of the V-shaped curve on the frequency axis. This is the most
sensitive frequency and it will be called the characteristic frequency (CF). This frequency
should correspond to the signal frequency SF. If this is not the case the subject has no
access to the auditory filter tuned to the signal frequency and he uses the auditory filter
tuned to CF. This phenomenon is called ‘off-frequency listing’ and it is an indication for a
cochlear dead region (e.g. Dillon, 2012; p. 298). Frequency resolution around the signal
frequency will be poor because the subject cannot distinguish between the frequencies SF
and CF. The difference DF = SF – CF is an appropriate parameter to describe the
correctness of the peak location. Its value should be low for good frequency resolution.
2. The sharpness of the curve. This can be expressed with the so-called Q10dB factor (Q for
quality; Moore, 2012, p. 28) which corresponds to a relative bandwidth and is calculated
as the bandwidth 10 dB above the threshold at the tip, divided by the frequency at the tip
(CF). A sharp curve (high Q10dB) indicates a good frequency resolution because the
system can distinguish between a tone at SF and a tone with a slightly different frequency.
A broad PTC will give a low Q10dB and it indicates poor frequency resolution.
1.2
Importance of measuring frequency resolution in the clinic
Knowledge about frequency resolution of a subject might be useful in clinic for:
1. A better understanding of speech recognition difficulties of the subject in noise, which is
important for counselling. For example subjects with an early sensorineural (noiseinduced) hearing loss might complain but their PTA shows no, or only a small, increase in
pure tone thresholds (for references see Larsby and Arlinger, 1998).
2. Detection of the presence of a cochlear dead region (side effect). Although other tests
(such as TEN-HL, see Moore et al., 2004) are faster and specially developed for this
1
A PTC shows a level in dB SPL (so a reference level) and not a gain in dB as is usual for a filter.
purpose, they are not routinely done in the clinic. The proposed test can show the
necessity to investigate (the extent of) the cochlear dead region with different tests.
There is no consensus in the literature about its usefulness in fitting hearing aids (Dillon,
2012; p. 301, Summers, 2004). A subject with serious loss of frequency resolution (broad
PTC), will still have difficulties with frequency resolution because the aid cannot change the
shape of the PTC (without introducing distortion for other auditory filters).
2. Method
2.1 Motivation of the method proposed
The filter shape, and in particular its bandwidth, is studied in several methods (Moore, 2012;
chapter 3). Most methods need some special equipment or do not give a complete PTC which
is needed to calculate DF and Q10dB. For example in methods based on broadband noise with a
spectral notch at SF (e.g. see Larsby and Arlinger, 1998) the width of the notch is varied.
Another disadvantage of using broadband noise is the need for high masker levels for hearing
impaired subject which can be uncomfortable. Fast methods (5-7 minutes) for obtaining a
PTC have been brought under attention again (e.g. Malika et al. 2009). They use an
automatically sweeping masker across the frequency axis and a Békésky threshold tracking
scheme. However, this appealing method needs a PC with sound card and special software.
An improved method to measure frequency resolution is proposed that is an extension of the
method of Lutman and Wood (1985), and does not need investment in new equipment. It
consists of measuring the relevant part of a PTC with a (standard) masking narrowband noise
centred at different frequencies. Lutman and Wood (1985) calculated a measure for sharpness
of the PTC at the signal frequency by considering three points on the PTC (one at the signal
frequency, one at higher frequency FMAX and one at lower frequency FMIN). However, with
three points the frequency shift DF cannot be calculated and Q10dB seems to be a more general
measure than their measure for defining the sharpness of a filter. For these reasons we
propose to measure more points on the PTC such that DF and Q10dB can be calculated. Only
one PTC for each ear at SF = 2000 Hz (i.e. centre of speech relevant frequencies) is calculated
in order to limit the time of testing.
2.2
Stimuli
The signal tone used in the test is a pulsed pure tone of 2000 Hz at 10 dB SL (duration 170
ms, rise/fall 50 ms). Pulsation is used to make the task less difficult for the subject, see
Lutman and Wood (1985). The maskers used in the test are narrowband noise centred at seven
different frequencies ranging from FMIN = 1250 Hz to FMAX = 2500 Hz, including the
signal frequency SF = 2000 Hz. The other frequencies are logarithmically equally spaced
between these frequencies, see Figure 1. The design of the narrowband noise masker and the
extreme frequencies (FMIN and FMAX) are taken from Lutman and Wood (1985) such that
the masker level would not be undesirable loud.
2.3
Procedure
The test consists of two stages: (1) measuring a PTC in each ear and (2) processing the results
in the computer to calculate the parameters DF and Q10dB2. The procedure and instructions to
measure the PTC (in each ear) are similar to these for measuring thresholds with PTA (see
British Society of Audiology, 2011). The main points are:
1. The audiometric data (PTA) is assumed to be available, including the 2000 Hz AC.
2. The threshold for the pulsed signal tone (no noise masker) is measured using the standard
protocol (5 dB steps up/10 dB steps down): starting at the level found in (1).
3. The modulated signal is set at 10 dB above the level found in (2), which equals 10 dB SL.
4. The order of the 7 frequencies (1250 Hz,... , 2500 Hz ) is chosen at random.
5. The level of masking required to just mask the signal tone (both channels to the same ear!)
is measured for each frequencies in (4) with an inverse protocol (10 dB steps up/5 dB
steps down) because the masker level, and not the signal level, is changed during the
procedure (Lutman and Wood, 1985); start level as (2). The patient should be instructed to
“ignore the noise and listen for the tone, and always indicate if the noise is too loud”.
The time needed for the test is 30 minutes: 25 minutes for the (2 + 2x7 = 16) measurements3 5
minutes for putting the results in the computer and the interpretation.
2.4
Equipment
The test proposed can be performed with the existing audiometric equipment in the
department. Insert earphones are used to minimize the risk of cross-hearing. The test uses:
1. The audiometer with CD-player. Signal tone on channel 1 (derived from main oscillator
within audiometer, see Lutman and Wood, 1985) and the CD containing the masker
signals to channel 2. The CD has to be purchased or it can be produced in the clinic4.
2
The definition of both parameters is given in Section 1.1. The script to fit a curve to the data points
from which the parameters can be calculated automatically is beyond the scope of this proposal.
3
Lutman and Wood (1985) made 2 + 2x3 = 8 measurements in 10-15 minutes.
2. The laptop/PC in the clinic to calculate the parameters (with e.g. an Excel-script).
Calibration and daily checks conform an audiometer, see British Society of Audiology (2011).
2.1
Subjects
The test is appropriate for two groups of subjects who complain about problems with speech
discrimination in noise5: (1) subject without a hearing loss (see example in Section 1.2 for
explanation), and (2) subjects with a hearing aid, when using their hearing aid.
3. Results
3.1
Reference data
The author has not found reference data concerning the proposed measures (DF and Q10dB) in
the literature. Lutman and Wood (1985) have derived a nominal value for the sharpness of the
PTC for normal hearing people but their measure is different from Q10dB.
Reference data has to be collected before the test can be used in the clinic. The reference data
consists of the minimal and maximal values of the parameters DF and Q10dB for people
without speech recognition problems in noise. These values vary probably with hearing loss
because the PTC will be broader with signal level (see Section 1.1). Consequently PTCs have
to be measured for each degree of hearing loss, for example with 6 groups; normal hearing to
severe hearing loss according to the standard categories (British Society of Audiology, 2011).
This has to be done for subjects with and without speech recognition problems in noise. If
each subgroup consists of 10 subjects, this means a total of 2 x 6 x 10 = 120 subjects.
3.2
Action to be taken with results obtained from the test
All parameters as measured for the subject are compared with the reference values
(thresholds) for their degree of hearing loss, see Figure 2. This has to be done for both ears.
Note: the results of the test are only valid for the best ear if there is a risk of cross-hearing
(this can be deduced from the subjects’ PTA)6. This should be mentioned in the test results.
4
It would be interesting to know if the test can be done with the default maskers on the audiometer.
5
It would be desirable to use a measure to describe the extent of the problem. This could be coming
from a general questionnaire or a speech recognition test.
6
An additional audiometer might be used to mask the non-test ear (as in PTA).
4. Discussion and conclusion
This proposal is about testing the subjects’ frequency selectivity ability by using PTCs. The
test proposed is easy to implement because it does not need investment in additional
equipment and the procedure is similar to PTA such that no training is required. However, the
proposed test lacks reference (or normative) data and a serious amount of time is needed to
get this data. Consequently, a test for measuring frequency resolution ability is at the moment
more in a research stage than in a clinical applicable stage.
However, it would be worth the effort to start a small pilot study in order to get a better
insight if PTCs are useful in the clinic (Section 1.2) and if the proposed parameters to describe
the PTC (Section 1.1) correlated with the speech recognition problem in a noisy environment,
as experienced by the subjects.
5. References
British Society of Audiology (2011) Recommended Procedure. Pure-tone air-conduction and
bone-conduction threshold audiometry with and without masking. BSA, Reading.
Dillon, H. (2012) Hearing Aids.2nd edition, Thieme, Stuttgart.
Larsby, B. and Arlinger, S. (1998) A method for evaluating temporal, spectral and combined
temporal-spectral resolution of hearing. Scandinavian Audiology, 27, 3-27.
Lutman, M.E. and Wood, E.J. (1985) A simple clinical measure of frequency resolution.
British Journal of Audiology, 19, 1-8.
Malika, A.N., Munro, K.J. and Baker, R.J. (2009) Fast method for psychophysical tuning
curve measurement in school-age children. Int. Journal of Audiology, 48, 546-553.
Moore, B.J.C. (2012) An Introduction to the Psychiology of Hearing. 6th edition, Emarald,
Bingley.
Moore, B.C.J., Glasberg, B.R. and Stone, M.A. (2004) New version of the TEN test with
calibrations in dB HL. Ear & Hearing, 25, 478-487.
Summers, V. (2004) Do tests for cochlear dead regions provide important information for
fitting hearing aids? (L) Journal of the Acoustical Society of America, 115, 1420-1423.
Figure 1. Examples of PTCs measured with a narrowband noise masker centred at 7 different
frequencies (circles) for a pure tone signal of SF = 2000 Hz at 10 dB SL (red star). (A) Typical PTC
for a normal hearing subject. Red arrow illustrates the measure of sharpness Q10dB. (B) Typical PTC
for a subject using ‘off-frequency listening’. Red arrow illustrates the measure of frequency shift DF.
Figure 2. Flow chart for each ear with the test results (parameters DF and Q10dB). Values for
threshold1,2,3 depend on hearing loss and have to be deduced from reference data (to be determined).