Operating Instructions
67
Important Considerations When Monitoring
Typical of all DPOAE and TEOAE measurements, the lower limit of measurement is determined by noise,
most of which has a Gaussian distribution. In the absence of an emission, both the signal and reference
channels contain nothing but noise. Just like tossing 10 coins simultaneously where the long-term
average will be five heads, it is not unusual to see seven or eight heads in any one toss. Similarly, an
apparent SNR of 5 dB or even 10 dB can occasionally appear even though the long-term average SNR in
a cavity is 0 dB. In the absence of an emission, a single SNR reading will exceed 5.5 dB one time in six
on the average. This is just as true at high frequencies (8, 10, and 12 kHz) as it is at lower frequencies
(2, 3, 4, 6 kHz). However, the noise floor is slightly higher at frequencies above 6 khz. To safeguard
against mistaking noise for the signal, it is recommended that minimum DPOAE amplitudes be used as
part of any protocol measuring frequencies above 6 khz.
To give an actual example: Figure 2 on the previous page indicates the absence of emissions for an
adult. Figure 3, a subsequent test on the same subject, shows apparent emissions at 8, 10, and 12 kHz,
with displayed SNRs of 7, 5, and 4 dB, respectively. Figure 3 was selected from a large number of tests
on that ear to illustrate that such readings can occur by chance.
There appear to be no published norms for the use of high-frequency DPOAEs for ototoxic monitoring,
but we can determine the significant difference between the averaged test results from one session to
the averaged test results of another session based on known statistical variation. Averaging the results
of several DPOAE test improves the reliability compared to a single test. This is particularly important
when DPOAE test results are used to monitor changes in hearing sensitivity at specific frequencies over
two or more different sessions.
Table 1, below, gives the number of tests required for the comparison between two test sessions.
The critical differences at two different confidence levels (80% and 95%) are shown. An 80%
confidence level is normally adequate for clinical testing while a 95% confidence level is common for
research reporting, where a reduced risk of error is normally required.
The
critical difference
is the difference between two measurements that probably did not occur by
chance. An example: You want to be reasonably (80%) certain that any shift beyond 5 dB will be
statistically significant. How many tests must be averaged in each session to obtain this result? The table
below indicates that the average of 6 tests in each session will give the desired critical difference of 5.1
dB. Since a 5 dB change is normally considered clinically significant, we suggest the use of at least six
tests during each session.
If a more sensitive test is required, more tests must be averaged. For example, averaging 10 tests in each
condition gives a 3.9 dB critical difference at the same 80% confidence level. In contrast, a 12.4 dB
change would be required at the 80% confidence level when comparing a single test from each session.
Alternately, a 50% increase in testing can be used to improve from an 80% to a 95% confidence level
at a given criterion. Example: The average of 4 tests will give a 6.2 dB critical difference at the 80%
confidence level; the average of 6 tests will provide the same 6.2 dB at the 95% confidence level.
Table 1
Number of tests averaged 1
2
3
4
6
10
16
95% C.D.
15.2
10.7
8.8
7.6
6.2
4.8
3.8
80% C.D.
12.4
8.8
7.2
6.2
5.1
3.9
3.1
Содержание Ero Scan Pro
Страница 74: ...1162 0802 Rev E 05 12...
