Threshold Testing

15 February 2022
10 - 30 mins

What is threshold testing?

The Auditory Brainstem Response (ABR) is an evoked potential that originates at the auditory nerve (Cranial Nerve VIII) and the response is picked up by surface electrodes typically placed at Vertex and Left and Right mastoids. An ABR test is used to assess the auditory system’s function from the cochlea through the brainstem.

The response to auditory broad band or frequency specific stimuli is identified by “peaks” that occur typically between 1 and 15 milliseconds from the stimulus onset. The ABR peaks are measured and marked traditionally as I, II, III, IV, and V. Each peak has an expected latency range to be considered “normal”. Delayed or missing peaks are consistent with abnormal auditory function. The presence or absence of responses can be used to estimate hearing thresholds. An ABR threshold is an electrophysiological threshold that can be used to predict 
the behavioral audiogram. The difference between the two may vary quite a lot, but correction of 20dB at 500Hz, 15dB at 1kHz, 10dB at 2kHz and 5dB at 4kHz are typically applied correction factors.


Threshold recording using 2kHz Tone Burst. Note the large PAM response from the right side caused by the loud stimulus of 80dBnHL. The ABR threshold at 20dB nHL at 2kHz found here would be well within the range of normal hearing - applying a typical correction factor would estimate the behavioral audiogram threshold to be 10dBHL at 2kHz. 


How to improve threshold testing with Eclipse

Challenges arise when evaluating hearing thresholds. They can include:

Five challenges each with their own solution. The first challenge is small amplitudes, to which the CE-Chirp stimulus family is the solution. The second challenge is unknown noise levels, to which residual noise calculation is the solution. The third challenge is residual noise, to which Fmp calculation is the solution. The fourth and fifth challenges are patient state and optimizing test time, respectively, to which the solution is Bayesian weighting.


1. CE-Chirp® stimulus family

We know that as we get closer to threshold, the waveform latencies increase and the amplitudes decrease. This presents a challenge in “peak picking”. A solution is the implementation and use of the CE-Chirp® LS and NB CE-Chirp® LS. The CE-Chirp® LS Stimulus Family compares to the traditional click while the NB CE-Chirp® LS compares to traditional tone burst. Research indicates that the use of the CE-Chirp® LS and NB CE-Chirp® LS particularly at modest stimulation levels results in waveform amplitudes up to double that of traditional stimuli (Elberling & Don, 2008; Ferm et al., 2013). This is achieved by simply accounting for timing issues within the cochlea. If amplitudes can be increased, the ability of the user to quickly and accurately identify waveform peaks near threshold increases. This clearly reduces test time and increases user confidence. Please see example 

ABR waveforms using CE-Chirp vs Click, and displayed side by side. Wave five is much more prominent in the CE-Chirp tracings compared to the Click tracings.

The larger amplitudes of the CE-Chirp® LS when compared to the traditional Click allows for faster and more reliable testing. Similar benefits can be found when substituting traditional Tone Bursts with the NB CE-Chirps® LS.


2. Residual Noise calculation

Noise levels also create a challenge as noise can eliminate the ability to obtain or view the necessary waveform peaks. Lower noise levels increase the ability to identify the waveform peaks and increase the confidence of the presence or absence of a response.

Traditionally, users run a set number of sweeps in an effort to reduce the amount of noise in the recording. However, the number of sweeps may tell us little about the Residual Noise in the tracing. An objective Residual Noise measure should be used instead. Typically, if Residual Noise levels are 40nV or less, the tracing is sufficiently clean to be able to reveal a response if present. Therefore, Interacoustics implemented the Residual Noise calculation. The use of Residual Noise calculation either by monitoring or as a stop criterion greatly 
improves the certainty. The Residual Noise bar placed on the right side of the Fmp graph indicates the Residual Noise level and will turn green with a checkmark when the criterion for residual noise is reached (e.g. 40nV).

The Fmp value (red curve) has at this point of testing passed the 99% response confidence criteria, and the Residual Noise level (black curve) has not yet reached the 40nV residual noise level suitable for quality recordings around threshold


3. Fmp calculation and Response Confidence

The Fmp is an indication of the Response Confidence of the recorded response. While looking at the waveform being recorded, a confidence level is calculated, and provided as a percentage of statistical certainty of a response being present in the recorded waveform. This statistical analysis assists the clinician in that it can reduce test time since it relies on statistical information and not solely on the experience of the user. The experienced user can encompass the Fmp values when evaluating response presence or absence. Often the Fmp will detect strong responses sooner than confident eyeballing can reach the same conclusion. On the other hand, Fmp may not identify smaller responses close to threshold, in which case eyeballing is to be applied, being the golden standard in response detection at threshold. When the response confidence reaches the set criteria (e.g. 99%) the bar 
indicating the Response Confidence turns green with a checkmark indicating that the Response Confidence criterion is met.


4. Bayesian weighting

In the best testing scenario that one should always try to obtain, the patient will be very quiet or sleeping and the EEG at a constant and low level throughout the data acquisition. This is not always the case and Bayesian Weighting is of great help in such test situations. It is an averaging technique that weighs each sweep individually giving more “weight” or importance to quiet sweeps and less “weight” to sweeps with more noise. This is different from traditional averaging since traditional averaging simply accepts or rejects each sweep and not weighing them individually.

The use of Bayesian Weighting will assist the clinician in situations that are not ideal testing conditions, and will not alter the response waveform’s morphology.


A few additional practical hints

The Rearrange Curves and Group Curves functions in the upper tool bar ensure easy manual positioning of waveforms.
The A&B tool in the upper tool bar allows waveform reproducibility to be evaluated without the need to run repeated waveforms.


Important highlights

  • Use eyeballed (or Fmp based) response identification or Noise Level as stop criteria, not a fixed number of sweeps
  • Use Bayesian Weighting for lowest residual noise
  • Use CE-Chirp® LS rather than click for larger response amplitudes
  • Neither Click nor CE-Chirp® LS is sufficient information on which to fit a hearing aid
  • Use NB CE-Chirps® LS rather than Tone Bursts for larger response amplitudes, when doing the frequency specific testing needed for hearing aid fitting
  • Consider typically using a relatively fast rate, as the time to get a low noise level rather than wave morphology is of the essence in threshold testing. If in doubt then lower the rate
  • Spend your time close to threshold rather than trying to obtain overly beautiful responses at loud intensities
  • Tough filtering (e.g. Low Pass display filtering at 1500Hz) may help eyeballing responses
  • Consider using the A&B buffers as your reproducibility tool, rather than spending time actually running double sweeps
  • Remember Fmp is typically not capable of picking out response presence close to threshold, but may be a valuable tool confirming response quickly well above threshold
  • A typical good threshold practice requires waveforms around threshold to have a residual noise level of 40nV or less, and requires eyeballing of a repeated responses (or similar A&B) at threshold and no visible response just below threshold



Don, M. & Elberling, C. (1996). Use of quantitative measures of auditory brain-stem response peak amplitude and residual background noise in the decision to stop averaging. J. Acoust. Soc. Am., 99(1).

Elberling, C. & Don, M. (1984). Quality Estimation of averaged auditory brainstem responses. Scand Audiol. (13) 187-197.

Elberling, C., & Don, M. (2008). Auditory brainstem responses to a chirp stimulus designed from derived-band latencies in normal-hearing subjects. J. Acoust. Soc. Am. (124) 3022-3037.

Elberling, C. & Wahlgreen (1985). Estimation of auditory brainstem response, ABR, by means of Bayesian interference. Scand. Audiol (14) 89-96.

Ferm, I., Lightfoot, G. & Stevens (2013). J. Comparison of ABR response amplitude, test time, and estimation of hearing threshold using frequency specific chirp and tone pip stimuli in newborns. International Journal of Audiology, (52) 419-423



Popular Academy Training

Interacoustics - hearing and balance diagnosis and rehabilitation
Copyright © Interacoustics A/S. All rights reserved.