Objective measures for detecting the auditory brainstem response: comparisons of specificity, sensitivity and detection time

Evaluating Fixed Single-Point Parameters When Applied to Vestibular Evoked Myogenic Potentials: The Effect of Single Point and Signal Window

OBJECTIVES

Several studies have applied a common objective detection algorithm (fixed single point [ Fsp ]) for detection of the vestibular evoked myogenic potential (VEMP). However, fundamental parameters of Fsp , such as establishing the location and duration of a signal window, have not been examined. In addition, Fsp criterion values used for response detection have not been established for cervical VEMPs (cVEMPs) or ocular VEMPs (oVEMPs). The purpose of this article was to investigate the effect of various single points and signal windows on Fsp , as well as determining Fsp criteria to determine response presence for cVEMP and oVEMP in a group of young healthy participants.

DESIGN

Twenty young healthy adults under the age of 30 and with no history of hearing or balance concerns were enrolled in the study protocol. Air-conducted cVEMPs and oVEMPs were evoked using 500 Hz tone bursts at 123 dB pSPL recorded at a fixed electromyography activation of 50 µV for cVEMPs and 35° gaze angle for oVEMPs. Responses were analyzed off-line using visual and objective detection. Fsp was applied to cVEMPs and oVEMPs using a range of single points and signal windows.

RESULTS

Noise variance was lowest for cVEMPs at the latency of P1, and for oVEMPs noise variance was not significantly different across the single-point latencies. On average, extending the length of the signal window lowered the Fsp value in cVEMPs and oVEMPs. An Fsp value of 2.0 was chosen as the criterion cutoff associated with the 95th percentile during no-response conditions using group data for cVEMPs and oVEMPs, respectively. Fsp values for cVEMPs and oVEMPs were not significantly different from each other.

DISCUSSION

This study established single-point latency and time-window parameters for VEMP-related applications of the Fsp detection algorithm. Fsp criteria values were established for cVEMP and oVEMP. Using these parameters, responses were detected in all participants.

Gaussian Processes for Hearing Threshold Estimation Using Auditory Brainstem Responses

The Auditory Brainstem Response (ABR) plays an important role in diagnosing and managing hearing loss, but can be challenging and time-consuming to measure. Test times are especially long when multiple ABR measurements are needed, e.g., when estimating hearing threshold at a range of frequencies. While many detection methods have been developed to reduce ABR test times, the majority were designed to detect the ABR at a single stimulus level and do not consider correlations in ABR waveforms across levels. These correlations hold valuable information, and can be exploited for more efficient hearing threshold estimation. This was achieved in the current work using a Gaussian Process (GP), i.e., a Bayesian approach for non-linear regression. The function to estimate with the GP was the ABR's amplitude across stimulus levels, from which hearing threshold was ultimately inferred. Active learning rules were also designed to automatically adjust the stimulus level and efficiently locate hearing threshold. Simulation results show test time reductions of up to  ∼ 50% for the GP compared to a sequentially applied Hotelling's T2 test, which does not consider correlations across ABR waveforms. A case study was also included to briefly assess the GP approach in ABR data from an adult volunteer.

Vestibular evoked myogenic potential (VEMP) test-retest reliability in adults

BACKGROUND

The technique of measuring ocular vestibular evoked myogenic potentials (oVEMP) in response to Mini-shaker vibration is relatively new, there is a limited normative data to define the presence or absence of a response in the literature.

OBJECTIVE

To determine the test-retest reliability of cervical and ocular VEMPs (cVEMP and oVEMP, respectively) to air-conducted sound (ACS) and bone-conducted vibration (BCV) stimulation and to determine normative ranges for the responses.

METHODS

Twenty normal-hearing individuals (40 ears) and 20 hearing impaired volunteers with normal balance function (40 ears) were examined in this study. ACS cVEMP and BCV oVEMP (using a Mini-shaker) were recorded from both groups to assess the test-retest reliability and to collect normative VEMP data for P1/N1 latencies and amplitudes from 20 normal hearing individuals. To test reliability, VEMP recordings were repeated within the same session.

RESULTS

The test-retest reliability for all the cVEMP parameters showed excellent reliability whereas oVEMP parameters showed between fair and excellent reliability depending on the parameter tested. Normative data for VEMP P1/N1 latencies and amplitudes were established.

CONCLUSIONS

Normative data and test-retest reliability for BCV oVEMP using the Mini-shaker at 100 Hz were established in our study for the first time in the literature. Responses appear reliable.

Detecting cortical responses to continuous running speech using EEG data from only one channel

OBJECTIVE

To explore the detection of cortical responses to continuous speech using a single EEG channel. Particularly, to compare detection rates and times using a cross-correlation approach and parameters extracted from the temporal response function (TRF).

DESIGN

EEG from 32-channels were recorded whilst presenting 25-min continuous English speech. Detection parameters were cross-correlation between speech and EEG (XCOR), peak value and power of the TRF filter (TRF-peak and TRF-power), and correlation between predicted TRF and true EEG (TRF-COR). A bootstrap analysis was used to determine response statistical significance. Different electrode configurations were compared: Using single channels Cz or Fz, or selecting channels with the highest correlation value.

STUDY SAMPLE

Seventeen native English-speaking subjects with mild-to-moderate hearing loss.

RESULTS

Significant cortical responses were detected from all subjects at Fz channel with XCOR and TRF-COR. Lower detection time was seen for XCOR (mean = 4.8 min) over TRF parameters (best TRF-COR, mean = 6.4 min), with significant time differences from XCOR to TRF-peak and TRF-power. Analysing multiple EEG channels and testing channels with the highest correlation between envelope and EEG reduced detection sensitivity compared to Fz alone.

CONCLUSIONS

Cortical responses to continuous speech can be detected from a single channel with recording times that may be suitable for clinical application.

Modified T2 Statistics for Improved Detection of Aided Cortical Auditory Evoked Potentials in Hearing-Impaired Infants

The cortical auditory evoked potential (CAEP) is a change in neural activity in response to sound, and is of interest for audiological assessment of infants, especially those who use hearing aids. Within this population, CAEP waveforms are known to vary substantially across individuals, which makes detecting the CAEP through visual inspection a challenging task. It also means that some of the best automated CAEP detection methods used in adults are probably not suitable for this population. This study therefore evaluates and optimizes the performance of new and existing methods for aided (i.e., the stimuli are presented through subjects' hearing aid(s)) CAEP detection in infants with hearing loss. Methods include the conventional Hotellings T2 test, various modified q-sample statistics, and two novel variants of T2 statistics, which were designed to exploit the correlation structure underlying the data. Various additional methods from the literature were also evaluated, including the previously best-performing methods for adult CAEP detection. Data for the assessment consisted of aided CAEPs recorded from 59 infant hearing aid users with mild to profound bilateral hearing loss, and simulated signals. The highest test sensitivities were observed for the modified T2 statistics, followed by the modified q-sample statistics, and lastly by the conventional Hotelling's T2 test, which showed low detection rates for ensemble sizes <80 epochs. The high test sensitivities at small ensemble sizes observed for the modified T2 and q-sample statistics are especially relevant for infant testing, as the time available for data collection tends to be limited in this population.

Objective hearing threshold identification from auditory brainstem response measurements using supervised and self-supervised approaches

Hearing loss is a major health problem and psychological burden in humans. Mouse models offer a possibility to elucidate genes involved in the underlying developmental and pathophysiological mechanisms of hearing impairment. To this end, large-scale mouse phenotyping programs include auditory phenotyping of single-gene knockout mouse lines. Using the auditory brainstem response (ABR) procedure, the German Mouse Clinic and similar facilities worldwide have produced large, uniform data sets of averaged ABR raw data of mutant and wildtype mice. In the course of standard ABR analysis, hearing thresholds are assessed visually by trained staff from series of signal curves of increasing sound pressure level. This is time-consuming and prone to be biased by the reader as well as the graphical display quality and scale.In an attempt to reduce workload and improve quality and reproducibility, we developed and compared two methods for automated hearing threshold identification from averaged ABR raw data: a supervised approach involving two combined neural networks trained on human-generated labels and a self-supervised approach, which exploits the signal power spectrum and combines random forest sound level estimation with a piece-wise curve fitting algorithm for threshold finding.We show that both models work well and are suitable for fast, reliable, and unbiased hearing threshold detection and quality control. In a high-throughput mouse phenotyping environment, both methods perform well as part of an automated end-to-end screening pipeline to detect candidate genes for hearing involvement. Code for both models as well as data used for this work are freely available.

Objective Detection of the Speech Frequency Following Response (sFFR): A Comparison of Two Methods

Speech frequency following responses (sFFRs) are increasingly used in translational auditory research. Statistically-based automated sFFR detection could aid response identification and provide a basis for stopping rules when recording responses in clinical and/or research applications. In this brief report, sFFRs were measured from 18 normal hearing adult listeners in quiet and speech-shaped noise. Two statistically-based automated response detection methods, the F-test and Hotelling’s T² (HT²) test, were compared based on detection accuracy and test time. Similar detection accuracy across statistical tests and conditions was observed, although the HT² test time was less variable. These findings suggest that automated sFFR detection is robust for responses recorded in quiet and speech-shaped noise using either the F-test or HT² test. Future studies evaluating test performance with different stimuli and maskers are warranted to determine if the interchangeability of test performance extends to these conditions.

Auditory Brainstem Response Detection Using Machine Learning: A Comparison With Statistical Detection Methods

OBJECTIVES

The primary objective of this study was to train and test machine learning algorithms to be able to detect accurately whether EEG data contains an auditory brainstem response (ABR) or not and recommend suitable machine learning methods. In addition, the performance of the best machine learning algorithm was compared with that of prominent statistical detection methods.

DESIGN

Four machine learning algorithms were trained and evaluated using nested k-fold cross-validation: a random forest, a convolutional long short-term memory network, a stacked ensemble, and a multilayer perceptron. The best method was evaluated on a separate test set and compared with conventional detection methods: Fsp, Fmp, q-sample uniform scores test, and Hotelling's T2 test. The models were trained and tested on simulated data that were generated based on recorded ABRs collected from 12 normal-hearing participants and no-stimulus EEG data from 15 participants. Simulation allowed the ground truth of the data ("response present" or "response absent") to be known.

RESULTS

The sensitivity of the best machine learning algorithm, a stacked ensemble, was significantly greater than that of the conventional detection methods evaluated. The stacked ensemble, evaluated using a bootstrap approach, consistently achieved a high and stable level of specificity across ensemble sizes.

CONCLUSIONS

The stacked ensemble model presented was more effective than conventional statistical ABR detection methods and the alternative machine learning approaches tested. The stacked ensemble detection method may have potential both in automated ABR screening devices as well as in evoked potential software, assisting clinicians in making decisions regarding a patient's ABR threshold. Further assessment of the model's generalizability using a large cohort of subject recorded data, including participants of different ages and hearing status, is a recommended next step.

Controlling test specificity for auditory evoked response detection using a frequency domain bootstrap

BACKGROUND

Statistical detection methods are routinely used to automate auditory evoked response (AER) detection and assist clinicians with AER measurements. However, many of these methods are built around statistical assumptions that can be violated for AER data, potentially resulting in reduced or unpredictable test performances. This study explores a frequency domain bootstrap (FDB) and some FDB modifications to preserve test performance in serially correlated non-stationary data.

METHOD

The FDB aims to generate many surrogate recordings, all with similar serial correlation as the original recording being analysed. Analysing the surrogates with the detection method then gives a distribution of values that can be used for inference. A potential limitation of the conventional FDB is the assumption of stationary data with a smooth power spectral density (PSD) function, which is addressed through two modifications.

COMPARISONS WITH EXISTING METHODS

The FDB was compared to a conventional parametric approach and two modified FDB approaches that aim to account for heteroskedasticity and non-smooth PSD functions. Hotelling's T²(HT2) test applied to auditory brainstem responses was the test case.

RESULTS

When using conventional HT2, false-positive rates deviated significantly from the nominal alpha-levels due to serial correlation. The false-positive rates of the modified FDB were consistently closer to the nominal alpha-levels, especially when data was strongly heteroskedastic or the underlying PSD function was not smooth due to e.g. power lines noise.

CONCLUSION

The FDB and its modifications provide accurate, recording-dependent approximations of null distributions, and an improved control of false-positive rates relative to parametric inference for auditory brainstem response detection.

Efficient Detection of Cortical Auditory Evoked Potentials in Adults Using Bootstrapped Methods

BACKGROUND

Statistical detection methods are useful tools for assisting clinicians with cortical auditory evoked potential (CAEP) detection, and can help improve the overall efficiency and reliability of the test. However, many of these detection methods rely on parametric distributions when evaluating test significance, and thus make various assumptions regarding the electroencephalogram (EEG) data. When these assumptions are violated, reduced test sensitivities and/or increased or decreased false-positive rates can be expected. As an alternative to the parametric approach, test significance can be evaluated using a bootstrap, which does not require some of the aforementioned assumptions. Bootstrapping also permits a large amount of freedom when choosing or designing the statistical test for response detection, as the distributions underlying the test statistic no longer need to be known prior to the test.

OBJECTIVES

To improve the reliability and efficiency of CAEP-related applications by improving the specificity and sensitivity of objective CAEP detection methods.

DESIGN

The methods included in the assessment were Hotelling's T2 test, the Fmp, four modified q-sample statistics, and various template-based detection methods (calculated between the ensemble coherent average and some predefined template), including the correlation coefficient, covariance, and dynamic time-warping (DTW). The assessment was carried out using both simulations and a CAEP threshold series collected from 23 adults with normal hearing.

RESULTS

The most sensitive method was DTW, evaluated using the bootstrap, with maximum increases in test sensitivity (relative to the conventional Hotelling's T2 test) of up to 30%. An important factor underlying the performance of DTW is that the template adopted for the analysis correlates well with the subjects' CAEP.

CONCLUSION

When subjects' CAEP morphology is approximately known before the test, then the DTW algorithm provides a highly sensitive method for CAEP detection.

The global Beta test for hidden periodicities in signals and its extensions to multivariate systems

BACKGROUND AND OBJECTIVE

There are many phenomena that lead to changes in the power spectrum of a given signal, and their detection has been a challenge that has received considerable attention over the years. Objective Response Detection (ORD) techniques are a set of tools that perform automated tests for such a task, allowing thus to automatically track changes in the spectrum. The performance of these detectors is affected by the signal-to-noise ratio (SNR) of the recorded signal as well as the length of the available data. The Global F Test (GFT) is a promising detector that can be used to test whether there is a statistically significant difference between the spectrum before and during an event. In fact, this detector has proved useful in the detection of event-related desynchronization/synchronization (ERD/ERS), where only amplitude, but not the phase, changes are locked to the stimulus. In order to improve the statistical power of the GFT (for the same length of recording), multiple channels recorded simultaneously can be included. This concept is called Multivariate Response Detection. The aim of the current work is to extend the GFT to the multivariate (multichannel) case.

METHODS

Firstly, the single channel normalization of the GFT is presented as a new ORD detector - the global Beta test (GBT). After that, three multivariate extensions of this new test are derived. The critical values used in the detection of spectral changes are obtained by using theoretical distributions, and where this is intractable, by means of Monte Carlo simulations. The probability of detection (PD) of each technique was estimated using simulation and was used in order to compare the detectors performance. A practical example with the electroencephalogram (EEG) from 10 volunteers under intermittent photic stimulation was also provided.

RESULTS

The statistics under both the null and alternative hypothesis could be obtained for all detectors. Simulated results for PD demonstrate the strong potential of the proposed method and the performances in EEG data are always improved with increasing number of signals.

CONCLUSION

If more than one signal is available, then the multivariate extensions may provide significant benefit compared to the original GFT.

Development and validation of the protocol for the evaluation of voice in patients with hearing impairment (PEV-SHI)

Introduction

The voice of individuals with hearing impairment has been widely described, and can be compromised in all levels of the phonatory system.

Objective

To develop and validate an instrument for evaluating the voice of this population.

Methods

The instrument underwent the validation steps suggested by the Scientific Advisory Committee of the Medical Outcomes Trust. The study sample consisted of seventy-eight Brazilian people with cochlear implants (experimental group) and 78 individuals with normal hearing (control group), divided in groups by age range — children from 3 to 5 years; children from 6 to 10 years and adults from 18 to 46 years. The study sample participated in a voice recording of the sustained vowel /a/, connected speech and spontaneous conversation, in which three voice specialists rated using the proposed instrument. It consists of visual-analog scales of suprasegmental aspects, respiratory-phonatory coordination, resonance, phonation, additional parameters and general vocal perception.

Results

Evaluation by an expert committee and a pilot test established content validity. Reliability measures showed excellent test-retest reproducibility for the majority of the parameters. Analysis with the ROC curve showed that perceptual evaluation with the sustained vowel did not strongly differentiate individuals with cochlear implants from those with normal hearing, and the parameter “speech rate” did not differentiate the groups at all. For the connected speech and spontaneous conversation, the majority of the parameters differentiated the experimental group from the control group with an area under the curve ≥0.7. The cutoff values with maximum specificity and sensitivity were 30.5 for mild, 49.0 for moderate and 69.5 for intense deviation.

Conclusions

The protocol for the evaluation of voice in subjects with hearing impairment, PEV-SHI, is a reliable and useful tool for assessing the particularities of the voice of individuals with hearing impairment treated with cochlear implants and can be used in research and clinical settings to standardize evaluation and facilitate information exchange among services.

A group sequential test for ABR detection

Objective: To detect the auditory brainstem response (ABR) automatically using an innovative sequentially applied Hotelling's T ² test, with the overall goal of optimising test time whilst controlling the false-positive rate (FPR). Design: The stage-wise critical decision boundaries for accepting or rejecting the null hypothesis were found using a new approach called the Convolutional Group Sequential Test (CGST). Specificity, sensitivity, and test time were evaluated using simulations and subject recorded data. Study sample: Data consists of click-evoked ABR threshold series from 12 normal hearing adults, and recordings of EEG background activity from 17 normal hearing adults. Results: Reductions in mean test time of up to 40-45% were observed for the sequential test, relative to a conventional "single shot" test where the statistical test is applied to the data just once. To obtain these results, it will occasionally be necessary to run the test to a higher number of stimuli, i.e. the maximum test time needs to be increased. Conclusions: The CGST can be used to control the specificity of a sequentially applied ABR detection method. Doing so can reduce test time, relative to the "single shot" test, when considered across a cohort of test subjects.

The Convolutional Group Sequential Test: Reducing Test Time for Evoked Potentials

When using a statistical test for automatically detecting evoked potentials, the number of stimuli presented to the subject (the sample size for the statistical test) should be specified at the outset. For evoked response detection, this may be inefficient, i.e., because the signal-to-noise ratio (SNR) of the response is not known in advance, the user would usually err on the cautious side and use a relatively high number of stimuli to ensure adequate statistical power. A more efficient approach is to apply the statistical test repeatedly to the accumulating data over time, as this allows the test to be stopped early for the high SNR responses (thus reducing test time), or later for the low SNR responses. The caveat is that the critical decision boundaries for rejecting the null hypothesis need to be adjusted if the intended type-I error rate is to be obtained. This study presents an intuitive and flexible method for controlling the type-I error rate for sequentially applied statistical tests. The method is built around the discrete convolution of truncated probability density functions, which allows the null distribution for the test statistic to be constructed at each stage of the sequential analysis. Because the null distribution remains tractable, the procedure for finding the stage-wise critical decision boundaries is greatly simplified. The method also permits data-driven adaptations (using data from previous stages) to both the sample size and the statistical test, which offers new opportunities to speed up testing for evoked response detection.