This tip discusses why you need to be careful when analyzing otoacoustic emissions (OAEs) using the traditional signal to noise ratio method and which criterion to use instead to improve confidence in analysis.
DPOAEs are traditionally interpreted as being present when the difference between the OAE and the noise level is significant, i.e., when you obtain a signal-to-noise ratio (SNR) of 6 dB. Caution is needed for two reasons:
Reason 1: Noise level is an average level
Firstly, the noise level is an average level. While the OAE level is relatively stable, noise fluctuates.
If fluctuations around the measured frequency are large, a signal to noise ratio of 6 dB does not imply that the OAE appears significantly outside the noise spectrum. The OAE can actually be an accidental peak of the noise.
Reason 2: A 6 dB SNR may not be the best criterion
Secondly, the often-used criterion of a 6 dB SNR finds its origin in newborn hearing screening.
Test protocols in this area require high sensitivity values of around 99.9%. As more than one frequency, usually 3 out of 4, is required for a pass, a 6 dB SNR at 3 of the 4 frequencies provides certainty that the pass result reflects the measurement of real OAEs.
As the number of frequencies required for a pass is reduced to - let's say 2 out of 4 - a greater SNR for the remaining frequencies is required in order to maintain an overall test sensitivity of 99.9%.
When evaluating each frequency individually, as required in diagnostic evaluations, a simple 6 dB SNR cutoff may not be sufficient or reliable enough to be certain that the DP point reflects a true OAE response from the cochlea.
Use the reliability criterion instead
This is where use of a reliability setting alongside the SNR can provide more certainty that the detected OAE is not just part of the noise. For example, a reliability setting of 98% for an individual frequency equates to a chance that a single response is actually just noise, 1 out of every 20 times, whereas using a 99.99% reliability setting reduces the chance to only 1 out of 1000.
In those few cases, where longer measurement times are needed, it allows you to make a diagnostic conclusion that can be trusted.