How to interpret aided cortical waveforms

This video explores how to interpret the results of an aided cortical waveform, or a set of waveforms. The information covered in this video is also crucial to understanding how to perform the test, when to stop testing and when to repeat a measurement.

You can read the full transcript below.

Morphology of cortical waveforms

Once an aided cortical waveform, or set of waveforms, has been recorded, it is important to be aware of how to interpret the results obtained.

The morphology of cortical waveforms has a high degree of variability between infants, as you can see from these waveforms here.

There is a much less clearly defined response morphology than in adults, and the literature shows that there is a large variability in the response morphology between subjects.

This is largely due to the immature cortex in this age group.

In fact, the cortex does not fully mature until teenage years.

In addition to this, infants and young children are more prone to being more active and producing more movement and vocalizations than adults during awake tests such as cortical testing.

This leads to greater amounts of noise entering the waveforms than we find in adult recordings, and this can make it challenging to accurately detect the presence of a response with confidence in these younger patients.

Detection methods

This is something that is highlighted in the British Society of Audiology Cortical Evoked Response Audiometry Recommended Procedure, which emphasizes the value of having an objective response detection method to help interpret infant cortical waveforms.

An interesting paper by Carter et al. in 2010 concluded that the inclusion of an automatic objective response detection algorithm increased the clinical utility of cortical testing by making the test more accessible and easier to interpret [1].

This study found that clinicians' confidence was much higher as a result of having this automatic response detection method, particularly for those clinicians with less experience.

In auditory evoked potentials, detection algorithms have well-established for some time.

Fsp was first developed in the late 1980s, which then evolved into Fmp in the 1990s.

And now we have the latest evolution of this method, called Fmpi.

Fsp and Fmp essentially calculate a signal-to-noise ratio based on the statistical variance of the measured average overall waveform in relation to the estimated residual noise of the waveform.

Fsp estimated the residual noise based on a single point of data on the waveform.

Fmp extended this to making use of multiple points of data on the waveform, 5 to be precise.

Fmpi also calculates the signal-to-noise ratio, but now all of the measured data points on the waveform are included in the calculation, which in the case of an aided cortical measurement, comes to 250 data points.

And furthermore, the individual listener's brain activity, their EEG, is taken into consideration, which doesn't happen in Fsp or Fmp, where a conservative estimation of the EEG is used.

The end result is that with Fmpi the detection of a cortical response can be established with fewer numbers of sweeps and thus less time than Fsp or Fmp in the majority of patients.

Here you can see the Fmpi graph developing.

The percentage value appears after the first 10 sweeps have been gathered, and updates every 2 sweeps thereafter.

Here we can see the response confidence has been set to 95%, which means the bar will turn green and a green tick will be displayed once the Fmpi value reaches 95% or greater.

It is also possible to adjust this target value to 99% in the software.

The Fmpi line is in purple in this instance as this waveform is being recorded in the aided binaural condition.

If the aided right condition is selected the line will be red and it will be blue when testing aided left.

The line will be black when the unaided condition is selected.

We can also see a steadily decreasing black line which displays the residual noise value found within the waveform as the recording progresses.

And here we can see that an Fmpi value of over 99% has been recorded in just 66 sweeps.

Residual noise

The residual noise value of a waveform can be used to inform when to stop testing, when it looks likely that no response is present and going to be recordable.

The Interacoustics Research Unit has explored the datasets available to understand the residual noise values in aided cortical responses in more detail.

A good starting point was the Ladies in the Van study [2].

In this study, 160 sweeps were tested per stimulus.

So a logical extrapolation is that we are likely to see whether a response is present or not at this point after 160 sweeps.

The data analyzed by the Interacoustics Research Unit showed that after 160 sweeps, the residual noise value for adults was on average 0.88 microvolts and 1.28 microvolts for children.

These values could be used as stopping criteria to help determine when there is no recordable response present in the waveform.

It is very important to highlight that these are provisional values based on internal unpublished data and these values may change in the future as further data emerges.

These provisional values have not been implemented into the default protocols as stopping criteria until there is more published data available.

For now these are suggested values and individual end users can set this up within their protocols themselves if they wish to use these values.

Waveform labeling

Within the software you will also find options to label the waveforms.

CR for clear response, RA for response absent, and INC for inconclusive.

These labels can be selected by the clinician based on their judgement of the waveforms and the Fmpi value should be used to help inform this decision.

A clear response label should be applied when there is an Fmpi value of greater than 95%.

If this Fmpi value has been achieved, the residual noise value is inconsequential.

There should be a reasonably acceptable morphology present in the waveform and anything unusual should be explored further and taken into account.

It is possible to compare the A and B buffers to look for confirmation of repeatable morphology.

If testing a patient with a cochlear implant, it is important to ensure there is no artifact present in the waveform, as this can be misidentified by the Fmpi detection algorithm as a response.

When considering whether to label a waveform as a response absent, the British Society of Audiology's recommended procedure for cortical testing highlights some of the criteria that should be met.

Referring to the objective response detection algorithm should be used to confirm the lack of response, and the residual noise of the overall average waveform should be low enough to confirm that there is no possible response hidden within a high level of noise.

This article also emphasizes the importance of having confidence that a response is genuinely absent rather than simply not identifiable.

To categorize a waveform as a response absent, the Fmpi value should be below 95% and the residual noise value should be sufficiently low for the patient demographic being tested.

When considering a response absent scenario, it is important to be mindful of the possibility of false non-responses.

That is, when no response is detected by the system, but the patient is in fact able to hear the stimulus.

The Ladies in the Van study explored this in some detail, identifying that when a non-response was detected, it was more likely that a response would then be detected upon repetition of the test.

The British Society of Audiology also highlights that absent or non-responses should be interpreted with caution if they have not been repeated.

Therefore, best practice is to repeat any waveforms when the Fmpi value does not reach 95 and the residual noise is low.

This provides a further benefit of allowing you to add the two recorded waveforms together, which can further reduce the residual noise and improve the signal averaging.

In a response-absent waveform, we should also expect to see no discernible waveform morphology.

The A and B buffers can be analyzed to support this aspect of the waveform interpretation.

The inconclusive label should be used when the Fmpi value does not reach 95%, but the residual noise value is excessively high.

This is the scenario whereby a response may be present, but hidden by the high level of residual noise, and therefore there is insufficient confidence that a response is not in fact present.

The ideal scenario here is to continue testing in order to reduce the residual noise further, in order to establish either a clear response or response absent instead of an inconclusive waveform.

It is also important to remember that a single trace which meets the response absent Fmpi and residual noise values is in fact inconclusive if the test has not been repeated.

It is the repetition that confirms the response absent.

Here we can see a summary of these values as they relate to children.

It is important to remember the residual noise values are provisional and liable to change as further data emerges in the future.

And it is important to remember there are different residual noise values for adults compared with children.

Here, we can see clear response labels being added to the individual waveforms for those which have all achieved an Fmpi value of over 95%.

We can also see a response absent label being added to the waveform where the Fmpi was below 95%.

In this case, the same stimulus and intensity level has been repeated and the two waveforms added together to provide additional confirmation and confidence that no response is present in the recorded waveforms.

Lastly, we can see a comment being added to an individual waveform, which can help contextualize the decision and add information as to the quality of the test conditions.

Report tab

On the Report tab, we can see every waveform recorded, and for each of these, the Fmpi and residual noise values are shown, alongside the label assigned by the clinician and any comments are also included here.

Here is an example of a waveform labeled inconclusive.

The Fmpi value has not reached 95% but the residual noise value is high, indicating the possibility that a response is present but obscured by the noise.

Clinical decision making

You may encounter the scenario where the residual noise has reached a low value but the Fmpi has not reached 95% but is very close to that target.

There are two options available for how to manage this situation.

You can continue testing, which will reduce the residual noise further, and improve the signal averaging, which increases the likelihood of reaching the Fmpi target of 95%.

Or, you can stop testing the current waveform, run a repeat of the same waveform, and then add the two recorded waveforms together.

This provides a combined Fmpi value, which benefits from both sets of waveform recordings, providing a greater amount of data and therefore a more reliable end result value.

This will also reduce the residual noise value further, providing more confidence in your interpretation of the waveform.

References

[1] Carter, L., Golding, M., Dillon, H., & Seymour, J. (2010). The detection of infant cortical auditory evoked potentials (CAEPs) using statistical and visual detection techniques. Journal of the American Academy of Audiology, 21(5), 347–356.

[2] Visram, A. S., Stone, M. A., Purdy, S. C., Bell, S. L., Brooks, J., Bruce, I. A., Chesnaye, M. A., Dillon, H., Harte, J. M., Hudson, C. L., Laugesen, S., Morgan, R. E., O'Driscoll, M., Roberts, S. A., Roughley, A. J., Simpson, D., & Munro, K. J. (2023). Aided Cortical Auditory Evoked Potentials in Infants With Frequency-Specific Synthetic Speech Stimuli: Sensitivity, Repeatability, and Feasibility. Ear and hearing, 44(5), 1157–1172.