The clinical usefulness of aided cortical auditory evoked potentials (CAEPs) remains unclear despite several decades of research. One major contributor to this ambiguity is the wide range of variability across published studies and across individuals within a given study; some results demonstrate expected amplification effects, while others demonstrate limited or no amplification effects. Recent evidence indicates that some of the variability in amplification effects may be explained by distinguishing between experiments that focused on physiological detection of a stimulus versus those that differentiate responses to two audible signals, or physiological discrimination. Herein, we ask if either of these approaches is clinically feasible given the inherent challenges with aided CAEPs. N1 and P2 waves were elicited from 12 noise-masked normal-hearing individuals using hearing-aid-processed 1000-Hz pure tones. Stimulus levels were varied to study the effect of hearing-aid-signal/hearing-aid-noise audibility relative to the noise-masked thresholds. Results demonstrate that clinical use of aided CAEPs may be justified when determining whether audible stimuli are physiologically detectable relative to inaudible signals. However, differentiating aided CAEPs elicited from two suprathreshold stimuli (i.e., physiological discrimination) is problematic and should not be used for clinical decision making until a better understanding of the interaction between hearing-aid-processed stimuli and CAEPs can be established.
The potential clinical benefits of a measure of brain encoding and plasticity in hearing aid users have driven a growing interest in aided cortical auditory evoked potentials (CAEPs). A better understanding of the effects of hearing aids on brain function, and resulting behavior, may improve the current science underlying successful rehabilitation of hearing loss. CAEPs, a type of event-related electroencephalography (i.e., scalp-recorded electrical brain activity) recorded 50–300 ms following stimulus onset, are thought to reflect neural activity in reverberant thalamocortical circuits (for a review see [
Methodological differences contribute to the variable results that exist in the aided CAEP literature. Much of the existing aided CAEP literature can be grouped into two general approaches: (1) a focus on physiological response detection, or (2) emphasis on physiological response discrimination. The physiological detection approach compares the CAEPs from an (inaudible or barely audible) unaided stimulus, to the response obtained from the same stimulus that has been processed by a hearing aid and delivered at a suprathreshold level. In this case, the unaided CAEP is often absent or weak, while the aided CAEP is often present with robust waveform morphology. The absence or presence of a response demonstrates a good correlation with inaudible and audible stimuli, amounting to a physiological correlate of detecting the presence of sound. Historically, CAEPs have successfully been used to estimate behavioral thresholds (approx. within 10 dB of behavioral threshold) of both normal-hearing and hearing-impaired populations [
In contrast to the physiological detection approach, the physiological discrimination approach compares CAEPs from two audible stimuli to determine differences between the waveforms (e.g., unaided versus aided conditions in normal-hearing individuals or two audible aided conditions with varied parameters). Physiological discrimination is measured by specific differences in waveform morphology (i.e., differences in peak latencies and peak amplitudes) between two present waveforms.
The different focus of these two approaches contributes to the variability in the existing literature involving aided CAEPs. Significant changes in waveform morphology (i.e., amplification effects) were often found when individuals/groups were tested in studies that used a physiological detection approach, comparing inaudible to audible conditions (e.g., [
Outline of the experimental conditions used in the studies cited in Figure
Subjects | Experimental design | Source | ||||||
---|---|---|---|---|---|---|---|---|
|
Characteristics | Stimulus | Duration (ms) | ISI (ms) | Signal level | Conditions | ||
Figure |
||||||||
(a) McNeil et al., 2009 [ |
1 | 68 yo male; severe to profound SNHL | /ba/ | 115 | 750* | 60 dB nHL | Unaided and aided | Figure |
(b) Rapin and Graziani, 1967 [ |
1 | 21 mo female; rubella, sedated | 500-Hz tone | Not specified | Not specified | 109 dB re: 0.0002 dynes/cm2 | Unaided and aided | Figure |
(c) Gravel et al., 1989 [ |
1 | 7 mo male; severe to profound SNHL | /da/ | Not specified | Not specified | Not specified | Unaided and aided (aid set to user settings) | Figure |
(d) Korczak et al., 2005 [ |
7 | Adults; severe to profound SNHL | /ba/ and /da/ | 150 | 950 | 80 dB ppeSPL | Unaided and aided (aid set to MCL) | Figure |
Figure |
||||||||
(e) Billings et al., 2007 [ |
13 | Young adults; normal hearing | 1000-Hz tone | 757 | 1910 | 50 dB SPL | Unaided and aided (20 dB gain) | Figure |
(f) Tremblay et al., 2006 [ |
7 | Young adults; normal hearing | /si/ | 655 | 1910 | 64 dB SPL | Unaided and aided (average gain of 19 dB) | Figure |
(g) Billings et al., 2011 [ |
9 | Young adults; normal hearing | 1000-Hz tone | 756 | 1910 | 40 dB SPL | Unaided and aided (gain of 20 dB) | Figure |
(h) Korczak et al., 2005 [ |
4 | Adults; moderate SNHL | /ba/ and /da/ | 150 | 950 | 85 dB HL | Unaided and aided (aid set to MCL) | Figure |
*Reference does not state whether this value refers to onset to onset, or offset to onset. All other interstimulus intervals (ISIs) refer to offset to onset.
Examples of physiological detection (a–d) and physiological discrimination (e–h) approaches from the aided CAEP literature. Results across these studies demonstrate significant amplification effects (unaided versus aided) for physiological detection, but very limited amplification effects for physiological discrimination. All figures were modified from published figures; the appropriate citation is indicated for each panel (see Table
In addition to considering the audibility of the signal, it is equally important to take into account the audibility of the underlying noise and its relationship to the signal (i.e., signal-to-noise ratio or SNR). SNR is a key contributor to both unaided and aided CAEP morphology [
The purposes of this study are to characterize the existing aided CAEP literature relative to two potential clinical approaches, and to determine whether these approaches (i.e., physiological detection and physiological discrimination) demonstrate clinical utility for encoding hearing-aid-processed signals. Specifically, we asked two questions: Will aided CAEPs be different for a near-threshold signal relative to a suprathreshold signal (i.e., physiological detection)? Will aided CAEPs be different for two suprathreshold signals (i.e., physiological discrimination)?
We set out to answer these questions in the aided CAEP domain by recording hearing aid output and eliciting CAEPs with the recorded stimuli.
CAEPs were recorded using hearing-aid-processed stimuli. In addition, a background noise masker was presented to the normal-hearing participants as a means of simulating the audibility factors that are present when an individual with hearing impairment is fit with a hearing aid. Testing noise-masked normal-hearing participants allowed tight control of audibility (i.e., audibility of hearing-aid-processed signals and hearing aid noise) while avoiding hearing-impairment-related confounds associated with recording CAEPs from individuals with hearing impairment.
Twelve individuals (six women and six men) participated in the study (mean age = 22.1 years; SD = 2.47). All participants were right handed with normal hearing from 250 to 8000 Hz (≤20 dB HL), were in good general health, reported no significant history of otologic or neurologic disorders, and denied use of mood or sleep-altering medications. All participants provided informed consent, and all research complied with the regulations of the Portland Veterans Affairs Medical Center Institutional Review Board.
Hearing-aid-processed stimuli were used to elicit CAEPs. Signals were 1000-Hz tones recorded from the output of two hearing aids: Hearing Aid A was a currently available digital hearing aid, Hearing Aid B was the same analogue hearing aid used in previous studies from our laboratory [
Three recordings of hearing aid output. Specific characteristics of the three hearing aid recordings used in this study to elicit aided CAEPs.
Hearing Aid | Gain at 1000 Hz1 | Input2 | Output SNR3 | |
---|---|---|---|---|
Recording 1 | A | 30 dB SPL | 25 dB SPL | 11 dB |
Recording 2 | B | 30 dB SPL | 25 dB SPL | 8 dB |
Recording 3 | B | 10 dB SPL | 45 dB SPL | 23 dB |
1Gain with a 45 dB SPL input (electroacoustically verified).
2Input: input to hearing aid microphones in the sound field.
3Output SNR: difference between hearing-aid-processed signal and noise at 1000 Hz (1/3 octave band).
The goal was to record three signals that varied in signal-to-noise ratio (SNR). As indicated in Table
A schematic depicting the hearing aid signal (i.e., 1000-Hz tone) and noise levels in relation to the background noise masker for each hearing aid condition is shown in Figure
Experimental design. One-third octave band levels at 1000 Hz are shown for the three hearing-aid-processed recordings. Scaling of the recordings resulted in Near
The hearing aid noise spectra obtained from the three 59 dB hearing aid condition presentations are represented in Figure
Frequency spectra of hearing aid noise for each of the three hearing aid conditions. Values are 1/3 octave bands with center frequencies between 200 and 6300 Hz. Hearing aid noise was measured for the 59-dB signal level condition for each recording. The general pattern of noise spectra is similar across conditions with a spectral peak at 1000 Hz, the frequency of the signal. The noise floor of the measurement system is shown with the dashed line (note: the lower limit of the sound level meter was 10.5 dB).
In order to simulate hearing-impaired thresholds and to control the audibility of the hearing-aid-processed signal and underlying hearing aid noise, a continuous background noise masker was created in Matlab passing a Gaussian white noise through a series of 1/3 octave band filters with center frequencies from 100 to 5000 Hz. The output of the filters was adjusted to generate a noise masker with a spectrum matching the thresholds of a patient with a moderate, sloping hearing loss. The spectrum of the noise masker was verified with a spectrum analyzer in 1/3 octave bands. In addition, behavioral thresholds at octave and interoctave frequencies from 250 to 8000 Hz were established using ER-3A (Etymotic Research, Inc., Elk Grove Village, IL) insert earphones. The thresholds of four participants were established using 1 dB steps in a 1-up, 2-down procedure while the background masker was played through the audiometer. Mean behavioral thresholds for the four individuals were 11.0, 15.75, 22.25, 24.25, 28.0, 34.75, 44.75, 43.25, and 8.75 dB HL at frequencies of 250, 500, 750, 1000, 2000, 3000, 4000, 6000, and 8000 Hz, respectively.
For each of the 13 conditions, a recorded 50-tone wav file was repeated three times, yielding a total of 150 tone presentations recording over approximately six minutes for each condition. Both the noise masker and hearing aid noise were continuous throughout each block of trials, and all stimuli were presented in the right ear using an ER-3A insert earphone and the Stim2 system (Compumedics Neuroscan, Charlotte, NC). The presentation order of the three hearing aid conditions was randomized across subjects, and the various stimulus levels were randomized within each hearing aid condition to minimize order effects across participants. Two-minute listening breaks were given between each condition, and subjects were offered a longer break after one hour of testing. Acquisition sessions lasted three hours and consisted of consenting, audiometric testing, electrode placement, and CAEP acquisition. Participants were seated comfortably in a double-walled sound attenuating booth, and they were instructed to ignore the auditory stimuli and to watch a closed-captioned movie of their choice.
Evoked potential activity was recorded using an Electro-Cap International, Inc. cap which housed 64 tin electrodes. The ground electrode was located on the forehead and Cz was the reference electrode. Data were rereferenced offline to an average reference. Horizontal and vertical eye movement was monitored with electrodes located inferiorly and at the outer canthi of both eyes. The recording window consisted of a 100-ms prestimulus period and a 700-ms poststimulus time. Using Scan 4.5 (Compumedics Neuroscan, Charlotte, NC), evoked responses were analog bandpass filtered online from 0.15 to 100 Hz (12 dB/octave roll off) and converted using an analog-to-digital sampling rate of 1000 Hz. Trials with eye-blink artifacts were corrected offline using Neuroscan software. This blink reduction procedure calculates the amount of covariation between each evoked potential channel and a vertical eye channel using spatial, singular value decomposition and removes the vertical blink activity from each electrode on a point-by-point basis to the degree that the evoked potential and blink activity covaried [
We fit the linear mixed model representation of the repeated measures analysis of variance (ANOVA). This model has the two-fold advantage of (1) being fit using maximum likelihood so that all observations are included in the analysis and not just observations for subjects with complete data, and (2) taking into account nonsymmetrical variances that may occur across conditions. Main effects of hearing aid condition were tested for two contrasts: Low versus Mid and Mid versus High. The Mid versus High comparison directly tests the physiological discrimination approach while the Low versus Mid comparison verifies that when SNR is changing, the aided CAEP is also likely to change. Where main effects were found, post-hoc comparisons were made for each hearing aid condition.
To test the effectiveness of the physiological detection approach, paired comparisons were completed on rectified area measures of Near
The current study investigated the ability of aided CAEPs to demonstrate amplification effects using both neural physiological response detection and physiological discrimination approaches. Based on a review of literature, we hypothesized that aided CAEPs would show the most robust effect of amplification in neural response detection approaches which correlate with the difference in response to audible versus inaudible stimuli. In contrast, aided CAEP discrimination approaches that reflect the ability of the auditory system to represent differences between audible hearing-aid-processed stimuli were expected to show weak amplification effects. Results, presented below, are organized to address these two primary hypotheses.
This study addressed the effects of amplification on audible and inaudible stimuli by comparing the Near
Grand average (
Area measurements for Near
Hearing Aid A (Recording 1):
Hearing Aid B (Recording 2):
Hearing Aid B (Recording 3):
The discrimination task of the current study measured the ability of aided CAEP measures to reflect differences between two clearly audible stimuli (i.e., Mid and High conditions). To demonstrate the main effect of signal level, butterfly and GFP plots were constructed from grand average responses across subjects and hearing aid recordings for Mid and High conditions (Figure
N1 and P2 peak amplitude and latency data for the Low, Mid, and High conditions are displayed in Figure
Statistical anlalysis. A linear mixed model representation of the repeated measures ANOVA resulted in a main effect of the level contrast with post-hoc comparisons where the main effect was significant.
Conditions |
Main effect | Hearing aid A (Recording 1) | Hearing aid B (Recording 2) | Hearing aid B (Recording 3) | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
|
|
|
|
|
|
|
|
|
|
|
| |
Low to Mid level | ||||||||||||
N1 Latency (ms) | 9.36 | 3,97 | <0.0001 | 13.78 | 1,97 | 0.0003 | 4.03 | 1,97 | 0.0474 | 10.25 | 1,97 | 0.0018 |
P2 Latency (ms) | 17.53 | 3,96 | <0.0001 | 30.52 | 1,96 | <0.0001 | 0.47 | 1,96 | 0.4955 | 21.6 | 1,96 | <0.0001 |
N1 Amplitude ( |
6.92 | 3,97 | 0.0003 | 0.03 | 1,97 | 0.868 | 4.18 | 1,97 | 0.0435 | 16.54 | 1,97 | <0.0001 |
P2 Amplitude ( |
3.16 | 3,96 | 0.0282 | 2.84 | 1,96 | 0.0954 | 0.67 | 1,96 | 0.4134 | 5.96 | 1,96 | 0.0164 |
Mid to High level | ||||||||||||
N1 Latency (ms) | 0.64 | 3,97 | 0.5894 | — | — | — | — | — | — | — | — | — |
P2 Latency (ms) | 0.38 | 3,97 | 0.7698 | — | — | — | — | — | — | — | — | — |
N1 Amplitude ( |
0.31 | 3,97 | 0.8163 | — | — | — | — | — | — | — | — | — |
P2 Amplitude ( |
2.07 | 3,96 | 0.109 | — | — | — | — | — | — | — | — | — |
Mean latency and amplitude measures for Low, Mid, and High conditions as a function of hearing aid recording (error bars: standard error of the mean). Generally, a change from Low to Mid conditions results in decreases in latency and increases in amplitude, and a change from Mid to High results in minimal change in latency and amplitude.
Results from previous literature indicate that CAEP responses are more sensitive to changes in SNR than to changes in absolute signal level [
This visual impression is confirmed by the significant main effect of Low to Mid level conditions on N1 and P2 amplitudes and latencies, which is not found in the comparison between Mid to High level conditions (Table
The purpose of this study was to help clarify some of the variability that is seen in decades of aided CAEP research. This variability is shown in Figure
The physiological detection approach appears to be a reasonable use of aided CAEPs because these measures are sensitive to differences in detectability of an inaudible or barely audible signal and a suprathreshold signal. Our results, and the results of other past studies, demonstrate robust amplification effects when taking a detection approach [
Cz-electrode waveforms for two representative individuals across the three hearing aid recordings. For both participants, the Near
In contrast, these data and examples from the literature [
It is important to consider subject factors as well. The audibility of a broadband stimulus and the underlying noise will vary depending on the hearing configuration of the individual being tested. The participants in this study were young normal-hearing individuals, and a noise masker was used to simulate thresholds that were comparable with a typical sloping hearing loss. However, even with tightly controlled audibility and a pure tone stimulus, variability across participants was found. Figure
Individual N1 latency and amplitude values demonstrating the physiological discrimination approach. Low, Mid, and High conditions are shown for Hearing Aid B (Recording 3). The general trends, consistent with Figure
It should be noted that the method of scaling used in the design of this study to modify signal level is different than clinical hearing aid gain adjustments in that modification of gain can lead to a wide range of acoustic modifications to the signal. While SNR has been shown to remain similar across gain settings in some hearing aids [
As mentioned above, the problems related to the physiological discrimination approach likely result from a combination of subject and stimulus factors. SNR and onset modification are two stimulus characteristics whose importance has been demonstrated in the literature (e.g., [
The results of this study seem to indicate that the contribution of stimulus factors can be minimized when the physiological detection approach is used. Specific signal-processing modifications made by the hearing aid are less important when the comparison response waveform is absent. In contrast, subtle acoustic changes (e.g., modification of SNR or onset characteristics) are essential when comparing two audible signals in a physiological discrimination approach. To characterize acoustic changes, it is necessary to complete in-the-canal recordings of hearing-aid-processed signals. Only then can measures of the important signal modifications be made and related to the resulting aided CAEPs.
Two approaches for using aided CAEPs, physiological detection and physiological discrimination, were tested to determine the clinical usefulness of each. Results are in agreement with an analysis of the literature (see Figure
Central auditory system
Signal-to-noise ratio
Standard deviation
Global field power
Cortical auditory evoked potentials
Analysis of variance.
The authors wish to thank Samuel Gordon, Roger Ellingson, and Garnett McMillan for assistance with experiment setup and data analysis. The authors also thank Oticon (Oticon, Inc., Somerset, NJ) for the use of a hearing aid. This work was supported by the National Institutes of Health through the National Institute on Deafness and Other Communication Disorders (R03-DC10914) and a Veterans Affairs Rehabilitation Research and Development Center of Excellence Grant (C4844C).