Upload
ngohuong
View
219
Download
0
Embed Size (px)
Citation preview
For internal use only / Copyright © Siemens AG 2006. All rights reserved.For internal use only / Copyright © Siemens AG 2006. All rights reserved.
Possible Pitfalls of InSitu Measurements
Page 2 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Contents
Contents
Impact of hearing instrument setting / Connexx model 7
Impact of test setup and protocol 11
Impact of individual anatomy 19
Modifications and fixes 24
Verification of the modified Connexx model 26
Page 3 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Background
Deviations between NAL-NL1 target and REM gain during InSituverification of First Fit were reported.
Bad shape of curve between 1.5 and 3kHz“2kHz dip”Peaks at higher frequencies
Observed especially in UK and AustraliaObserved for various hearing instruments Observed for various test systems
Individual characteristics
Page 4 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Background
A project was started to
Investigate the reproducibility of the reported observations If applicable: find reasons for the deviationsIf applicable: perform fix and validate the fix
The actual presentation gives an overview about the investigations and actions performed within this project.
Page 5 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Agenda of actual presentation
Outcomes of investigations (reproducible impact factors)Connexx modelTest setupIndividual anatomy
Actions based on these findingsValidationFurther Recommendations
Questions and discussion at the end of each section
Page 6 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Investigated issues
Impact of hearing instrument setting / Connexx modelInternal clocking / Connexx ModelInput compression (AGCi)Output limiting (AGCo)
Impact of test setup and protocol StimulusSound source positionProbe tube PositionRoom acousticsMeasurement system
Impact of individual anatomyInter-individual variation of OEG and RECDHI tube lengthVolume of ear canal Diffractions of sound signal caused by head and shoulder
Page 7 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of hearing instrument setting / Connexxmodel
Wireless applications and standard on the ASIC make a 5% over-clocking of the chip necessary. However, all correction curves of the electro-acoustic elements stored in Connexx 5.3 are still based on the original clocking.The effect on the gain curve differs depending on the shape of the curve and is hardly predictable. In general, stronger effects can be expected for moderately sloping curves. Amount: about 2dB less gain for f<5kHz.
(Acuris S, First Fit Cluster3, pure tone stimulus)
LE65
-10
0
10
20
30
40
50
60
100 1000 10000Frequency (Hz)
gain
/ dB
UPLCxx
Deviation at LE65
-10
0
10
100 1000 10000
Frequency (Hz)
UPL
-Cxx
/ dB
Page 8 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of hearing instrument setting / Connexxmodel
An impact of sub-optimal AGCisettings (level meter) can be observed across the whole frequency range. Amount: up to 3dB @ 2kHz
(Acuris S, First Fit Cluster3, LTASS stimulus)
LE65
-10
0
10
20
30
40
50
60
100 1000 10000Frequency (Hz)
gain
/ dB
UPLCxx
Deviation at LE65
-20
-10
0
10
20
100 1000 10000
Frequency (Hz)
UPL
-Cxx
/ dB
Page 9 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of hearing instrument setting / Connexxmodel
The impact of AGCo (intended behavior) may cause large deviations from the Connexxsimulation of LE90 curve, if modulated noise is used as stimulus.Direct impact cannot be excluded for modulated stimuli and higher amplification levels.In fact, a reduction of about 6dB was observed for this study for 15dB linear gain with AGCo limiting while using the Aurins stimulus. When output liming was set to PC, correct gain curves were observed .(Acuris S, FOG curve for IEC-2 noise)
LE90
-10
0
10
20
30
40
50
60
100 1000 10000Frequency (Hz)
gain
/ dB
UPLCxx
Deviation at LE90
-10
0
10
100 1000 10000
Frequency (Hz)
UPL
-Cxx
/ dB
Page 10 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of hearing instrument setting
Impact of Connexx model and chip clocking
5% over-clocking was not accounted for in the Connexx model=> mis-shaping of curve especially between 1 and 2kHz, about 2dB
less gain for f<5kHz
AGCi parameters / level meters were sub-optimal=> up to 3dB less gain than expected
AGCo may influence measurement in some cases=> up to 6dB less gain was observed
Page 11 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of test setup and protocol
If two different stimuli are used for measurement, gain may be falsely estimatedby Unity1 using the 16bit software.Example: If Aurins is used for the measurement of insertion response and pink noise for the measurement of open ear response, the gain calculated by Unity1 will be over-estimated in the frequency range around 1kHz.
Comparison of Aurins and Pink Noise Spectra
0102030405060708090
100
100 1000 10000
Frequency / Hz
SPL
/ dB
Pink NoiseAurins
Page 12 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of test setup and protocol
If AURINS noise is used for the IG measurement, the measured curves must not be compared to the target curves displayed in the Unity, because the latter are NAL targets which are, by definition, based on an LTASS input. The correct targets are shown in Connexx and yield about 6dB less gain for Aurins between 1 and 2.5kHz (Cluster3, typical age related hearing loss).
Comparison of Aurins and LTASS Target, Prisma2 Pro
-200
-100
0
100
200
300
400
100 1000 10000
Frequency
dB *
1/10
Aurins
LTASS
Page 13 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of test setup and protocol
The position of the reference microphone yields a different sound field compared to the field reference point. If the input level is regulated with respect to the reference microphone, less input than expected is presented to the hearing instrument
effective sound field deviation
-20
-10
0
10
20
100 1000 10000
Frequency (Hz)
Lo /
dBsp
l
diff Kemar 0°
diff Kemar 30°
diff Kemar 45°diff Kemar 60°
settings:MP1, Centra S
deviation causedby equalizationbelow Kemar pinna
Page 14 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of test setup and protocol
The sound source position at 45deg combined with the different sound field cause a deviation from the standard REUG as assumed by Connexx. This may cause a peak / dip pattern in the aided response between 2.5 and 6kHz.
(Acuris S, First Fit Cluster3, pure tone, KEMAR
measurement)
Deviation of REUG at 45deg from exprected value
-15
-10
-5
0
5
10
15
100 1000 10000f / Hz
dB
Page 15 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of test setup and protocol
The position of the reference microphone yields a different sound field compared to the field reference point. In addition, the MLE for 45° sound direction differs from the 0° MLE. This leads to a different input to the instrument and, as a consequence, to a different gain curve than expected.
effective HI input deviation (45°)
-20
-10
0
10
20
100 1000 10000
Frequency (Hz)
Lo /
dBsp
l
insitu HI input deviation
settings:MP1, Centra S
1) equalization below pinna !2) Kemar 45°
Page 16 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Summed effects of setup and Connexx model
If the impacts of AGCi, clocking, and test setup are regarded isolated from each other, they are all within the tolerance range. However, if they are combined, the total effect rises up to about 6dB.
-10
-8
-6
-4
-2
0
2
4
6
8
10
100.00 1000.00 10000.00
Frequency
dB
total difftest setupAGCiclocking
Page 17 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Example case
The stimulus spectrum was not equalized during the measurement. This caused a prominent dip at 2.5 kHz during the InSitu measurement which cannot be attributed to the instrument. Rather, this is a measurement artifact caused by diffractions at the head and the shoulders of the patient. Expected gain curve: about 20dB flat from 250Hz to 6kHz.
Artis P, subject RS
-20
0
20
40
60
80
100
100 1000 10000
Frequency / Hz
Leve
l / d
B
REIG REAR Aurins
stimulus at ref mic
aided response(InSitu)
insertion gain(InSitu)
Page 18 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of test setup and protocol
If two different stimuli (Pink Noise and Aurins) are used for measurement, this may result into an over-estimation of gain at 1kHz with old Unity1 software
The position of reference microphone below the ear may cause dip at 1.8kHz of about 3dB
The sound source position at 45° causes peaky shape for f>2kHz
NAL-NL1 is based on LTASS and does not support the AURINS stimulus, instead Connexx makes a calculation based on the LTASS target. Unity is unable to do this and always shows the LTASS targets.
For some stimuli (incl. AURINS) only the overall level is equalized during the measurement, but not the spectrum shape. For these stimuli, diffraction at head and shoulders may cause dip at 2.5kHz of 10dB. The amount and shaping of the dip are determined by the individual anatomy of the patient.
Page 19 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of individual anatomyLength of HI tube
For HI tubes longer than standard, a more peaky shape of the gain curve can be expected for f>1kHz.
Impact of HI tube length at given ear canal volume
-10
-8
-6
-4
-2
0
2
4
6
8
10
100 1000 10000Frequency (Hz)
dB
length 33mm
length 35mm
length 37mm
length 39mm
length 43mm
settings:
CIELO PFOG - 20dBLE70
audio input:Zwislocki (var. length)
Parameter:real ear tube length
Page 20 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of individual anatomyVolume of ear canal
Less overall gain is measured for identical HI settings in case of larger ear canal volumes (i.e. large ear canal diameter).
Impact of ear canal volume
-10
-8
-6
-4
-2
0
2
4
6
8
10
100 1000 10000Frequency (Hz)
dB
volume 0%
volume +20%
volume +40%
settings:
CIELO PFOG - 20dBLE70
audio input:Zw islocki (35mm tubing)
Parameter:ear canal volume
Page 21 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of individual anatomyIndividual OEG and RECD: one example
The target gain curve in this figure was 20dB flat from 250 Hz to 4kHz. The statistical OEG and RECD curves lead to deviations below 900Hz (leakage effects) and a dip between 1 and 2 kHzImporting individual OEGs or RECDs and using tight ear molds led to a satisfactory result in this case.
Acuris
-20
-10
0
10
20
30
40
100 1000 10000
Frequency
dB
stat. curvesind. OEGind. RECDind. OEG & RECD
Page 22 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of individual anatomyAverage individual RECD
Measurements of individual RECDs with 50 subjects at an individual research site yielded an average RECD very much comparable to the standard RECD that is actually used in Connexx (HA2).
-20
-10
0
10
20
30
40
100 1000 10000
Frequency
dB
Median RECD in OL, Unity2UK RECDHA2 RECD
Page 23 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Impact of individual anatomy
Individual factors
Longer HI tube may cause dip at 2kHz
Larger ear canal volumes may cause less overall gain
Diffraction of stimulus at patient’s shoulder may cause dip at 2.5kHz
Individual REUG and RECD may have unpredictable influence which may be compensated easily by importing individual curves. However, the standard RECD used by Connexx fits very good to the average individual RECD.
Page 24 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Modifications and fixesConnexx 5.4
Compensation of clock trimming: included in Connexx 5.4
Optimisation of AGCi level meter: included in Connexx 5.4
Activate PC instead of AGCo during test setting for InSitu and coupler measurements: included in Connexx 5.4
Adjustment to modified sound field in case of InSitu fitting (reference mic and sound source positions): Connexx 5.4 allows this to be optimised via country specific SiFit.
Page 25 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Modifications and fixesUnity
Future versions of Unity software will be able to import the Connexx Aurinstarget curve. Until then
Compare to the Connexx targets to be met, not the Unity targets (on-top modus does not work!) or
Use Icra1 stimulus for InSitu measurements
Page 26 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Verification of the modified Connexx model Coupler measurements (UPL)
typical age related hearing lossmodulated speech-like stimulus (IEC2)Prisma2 Pro
Page 27 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Verification of the modified Connexx model Coupler measurements (UPL)
typical age related hearing lossmodulated speech-like stimulus (IEC2)Prisma2 Pro
Page 28 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Verification of the modified Connexx model First Fit Accuracy of Prisma2 Pro
Median REM insertion gain of 6 subjects
Median REM Gain Curves
-20
-10
0
10
20
30
40
100 1000 10000
Frequency
Inse
rtion
Gai
n (d
B)
LTASS Target
Prisma2 Pro
Page 29 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Verification of the modified Connexx model First Fit Accuracy of Prisma2 Pro P
Median REM insertion gain of 6 subjects
Median REM Gain Curves
-20
-10
0
10
20
30
40
100 1000 10000
Frequency
Inse
rtion
Gai
n (d
B)
LTASS Target
Prisma2 Pro P
Page 30 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Verification of the modified Connexx model First Fit Accuracy of Prisma2 Pro P
Median REM insertion gain of 6 subjects
Median REM Gain Curves
-20
-10
0
10
20
30
40
100 1000 10000
Frequency
Inse
rtion
Gai
n (d
B)
Prisma2 Pro DSP
LTASS Target
Page 31 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Summary
Deviations between target gain and REM gain as reported in the past are the result of various effects. Both instrument settings and measurement artefacts play a role.
Several impacts of HI setting / Connexx model were corrected for in Connexx 5.4.
Individual corrections that are needed due to test setup can be implemented in country specific SiFit (e.g. NHSFit).
The impact of very individual anatomy cannot be excluded. In some cases, it may be useful to import individual curves to Connexx in order to enhance the quality of the fitting. However, this does not always work with Unity1using the 16bit software => switch to 32bit software recommended
Page 32 Siemens Audiologische Technik, ErlangenDr. Heike Heuermann
Recommendations
Hearing instrument settingsSwitch to test settings or switch off all adaptive parameters manually before REM measurement
Test setupUse the same stimulus for the measurement of unaided and aided responsesUse LTASS or ICRA instead of AURINSIf other stimuli are used: do not use the NAL-NL1 target displayed in Unity, but compare with target displayed in the fitting software (import of Connexx targets into Unity will be offered soon)Unity32 or Unity1 with 32bit software is recommended
Individual factorsFor patients with longer HI tubes, 2kHz dips should be expected For patients with large ear canal, less overall gain should be expectedFor patients with very unusual REURs, import OEG and/or RECD into Connexx(applicable approach for Unity2)