Real-time threshold determination of auditory brainstem responses by cross-correlation analysis

ABRpresto: An algorithm for automatic thresholding of the Auditory Brainstem Response using resampled cross-correlation across subaverages

The Auditory Brainstem Response (ABR) is an essential diagnostic indicator of overall health of cochlea, auditory brainstem and midbrain, used extensively in both basic research and clinical studies. A key quantification of the ABR is threshold, the lowest sound level that elicits a response. Because the morphology of ABR waveforms shift with stimulus level and the overall signal-to-noise ratio is low, threshold estimation is not straightforward. Although several algorithmic approaches have been proposed, the current standard practice remains the visual evaluation of ABR waveforms as a function of stimulus level. We developed an algorithm based on the cross-correlation of two independent averages of responses to the same stimulus. For each stimulus level, the individual responses to each tone-pip are randomly split into two groups. The median waveform for each group is calculated, and then the normalized cross-correlation between these median waveforms is obtained. This process is repeated 500 times to obtain a resampled cross-correlation distribution. For each frequency, the mean values of these distributions are computed for each level and fit with a sigmoid or a power law function to estimate the threshold. Algorithmic thresholds demonstrated robust and accurate performance, achieving 92% accuracy within ±10 dB of human-rated thresholds on a large pool of mouse data. This performance was better than that of several published algorithms on the same dataset. This algorithm has now fully replaced the manual estimation of ABR thresholds for our preclinical studies, thereby saving significant time and enhancing objectivity in the process.

Derivative-Guided Dual-Attention Mechanisms in Patch Transformer for Efficient Automated Recognition of Auditory Brainstem Response Latency

Accurate recognition of auditory brainstem response (ABR) wave latencies is essential for clinical practice but remains a subjective and time-consuming process. Existing AI approaches face challenges in generalization, complexity, and semantic sparsity due to single sampling-point analysis. This study introduces the Derivative-Guided Patch Dual-Attention Transformer (Patch-DAT), a novel, lightweight, and generalizable deep learning (DL) model for the automated recognition of latencies for waves I, III, and V. Patch-DAT divides the ABR time series into overlapping patches to aggregate semantic information, better capturing local temporal patterns. Meanwhile, leveraging the fact that ABR waves occur at the zero crossing of the first derivative, Patch-DAT incorporates a first derivative-guided dual-attention mechanism to model global dependencies. Trained and validated on large-scale, diverse datasets from two hospitals, Patch-DAT (with a size of 0.36 MB) achieves accuracies of 92.29% and 98.07% at 0.1 ms and 0.2 ms error scales, respectively, on a held-out test set. It also performs well on an independent dataset with accuracies of 88.50% and 95.14%, demonstrating strong generalization across clinical settings. Ablation studies highlight the contributions of the patching strategy and dual-attention mechanisms. Compared to previous state-of-the-art DL models, Patch-DAT shows superior accuracy and reduced complexity, making it a promising solution for object recognition of ABR latencies. Additionally, we systematically investigate how sample size and data heterogeneity affect model generalization, indicating the importance of large, diverse datasets in training robust DL models. Future work will focus on expanding dataset diversity and improving model interpretability to further improve clinical relevance.

Audiogram Estimation Performance Using Auditory Evoked Potentials and Gaussian Processes

OBJECTIVES

Auditory evoked potentials (AEPs) play an important role in evaluating hearing in infants and others who are unable to participate reliably in behavioral testing. Discriminating the AEP from the much larger background activity, however, can be challenging and time-consuming, especially when several AEP measurements are needed, as is the case for audiogram estimation. This task is usually entrusted to clinicians, who visually inspect the AEP waveforms to determine if a response is present or absent. The drawback is that this introduces a subjective element to the test, compromising quality control of the examination. Various objective methods have therefore been developed to aid clinicians with response detection. In recent work, the authors introduced Gaussian processes (GPs) with active learning for hearing threshold estimation using auditory brainstem responses (ABRs). The GP is attractive for this task, as it can exploit the correlation structure underlying AEP waveforms across different stimulus levels and frequencies, which is often overlooked by conventional detection methods. GPs with active learning previously proved effective for ABR hearing threshold estimation in simulations, but have not yet been evaluated for audiogram estimation in subject data. The present work evaluates GPs with active learning for ABR audiogram estimation in a sample of normal-hearing and hearing-impaired adults. This involves introducing an additional dimension to the GP (i.e., stimulus frequency) along with real-time implementations and active learning rules for automated stimulus selection.

METHODS

The GP's accuracy was evaluated using the "hearing threshold estimation error," defined as the difference between the GP-estimated hearing threshold and the behavioral hearing threshold to the same stimuli. Test time was evaluated using the number of preprocessed and artifact-free epochs (i.e., the sample size) required for locating hearing threshold at each frequency. Comparisons were drawn with visual inspection by examiners who followed strict guidelines provided by the British Society of Audiology. Twenty-two normal hearing and nine hearing-impaired adults were tested (one ear per subject). For each subject, the audiogram was estimated three times: once using the GP approach, once using visual inspection by examiners, and once using a standard behavioral hearing test.

RESULTS

The GP's median estimation error was approximately 0 dB hearing level (dB HL), demonstrating an unbiased test performance relative to the behavioral hearing thresholds. The GP additionally reduced test time by approximately 50% relative to the examiners. The hearing thresholds estimated by the examiners were 5 to 15 dB HL higher than the behavioral thresholds, which was consistent with the literature. Further testing is still needed to determine the extent to which these results generalize to the clinic.

CONCLUSIONS

GPs with active learning enable automatic, real-time ABR audiogram estimation with relatively low test time and high accuracy. The GP could be used to automate ABR audiogram estimation or to guide clinicians with this task, who may choose to override the GP's decisions if deemed necessary. Results suggest that GPs hold potential for next-generation ABR hearing threshold and audiogram-seeking devices.

ABRpresto: An algorithm for automatic thresholding of the Auditory Brainstem Response using resampled cross-correlation across subaverages

The auditory brainstem response (ABR) is an essential diagnostic indicator of overall cochlear health, used extensively in both basic research and clinical studies. A key quantification of the ABR is threshold, the lowest sound level that elicits a response. Because the morphology of ABR waveforms shift with stimulus level and the overall signal-to-noise ratio is low, threshold estimation is not straightforward. Although several algorithmic approaches have been proposed, the current standard practice remains the visual evaluation of ABR waveforms as a function of stimulus level. We developed an algorithm based on the cross-correlation of two independent averages of responses to the same stimulus. For each stimulus level, the individual responses to each tone-pip are randomly split into two groups. The median waveform for each group is calculated, and then the normalized cross-correlation between these median waveforms is obtained. This process is repeated 500 times to obtain a resampled cross-correlation distribution. For each frequency, the mean values of these distributions are computed for each level and fit with a sigmoid or a power law function to estimate the threshold. Algorithmic thresholds demonstrated robust and accurate performance, achieving 92% accuracy within ±10 dB of human-rated thresholds on a large pool of mouse data. This performance was better than that of several published algorithms on the same dataset. This algorithm has now fully replaced the manual estimation of ABR thresholds for our preclinical studies, thereby saving significant time and enhancing objectivity in the process.

Rapid hearing threshold assessment with modified auditory brainstem response protocols in dogs

Introduction

Auditory brainstem response (ABR) is the gold standard for hearing testing in dogs. ABR is commonly used in puppies to diagnose congenital sensorineural deafness. Long test times limit the use for a more comprehensive hearing screening in veterinary practice. This study aimed to establish a super-fast hearing screening protocol in dogs.

Methods

Hearing thresholds were routinely measured with a mobile device designed for newborn hearing screening in 90 dogs. We introduced modifications of the ABR protocol, e. g., a binaural test mode, higher stimulus rates, a broadband chirp stimulus, and an algorithm for automatic peak V detection in a stepwise fashion. Hearing thresholds were then measured with fast protocols utilizing either 30 Hz click or 90 Hz broadband chirp stimuli with 80, 60, 40, 30, 20, 10, 0 and -10 dBnHL stimulation intensities. Interrater reliability, agreement between click and chirp hearing thresholds and correlations with clinical characteristics of the dogs were assessed.

Results

Using all innovations, the test time for hearing threshold assessment in both ears was reduced to 1.11 min (mean). The chirp stimulus accentuated both, peak V and the subsequent trough, which are essential features for judgement of the hearing threshold, but preceding peaks were less conspicuous. Interrater reliability and agreement between click and chirp hearing threshold was excellent. Dogs >10 years of age and dogs with abnormal hearing score or otitis score had significantly higher hearing thresholds than younger dogs (p ≤ 0.001) or dogs without abnormalities (p < 0.001).

Conclusion

The results demonstrate that modifications in ABR protocols speed-up test times significantly while the quality of the recordings for hearing threshold assessment is maintained. Modified ABR protocols enable super-fast hearing threshold assessment in veterinary practice.

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.

Spatial patterns of noise-induced inner hair cell ribbon loss in the mouse mid-cochlea

In the mammalian cochlea, moderate acoustic overexposure leads to loss of ribbon-type synapse between the inner hair cell (IHC) and its postsynaptic spiral ganglion neuron (SGN), causing a reduced dynamic range of hearing but not a permanent threshold elevation. A prevailing view is that such ribbon loss (known as synaptopathy) selectively impacts the low-spontaneous-rate and high-threshold SGN fibers contacting predominantly the modiolar IHC face. However, the spatial pattern of synaptopathy remains scarcely characterized in the most sensitive mid-cochlear region, where two morphological subtypes of IHC with distinct ribbon size gradients coexist. Here, we used volume electron microscopy to investigate noise exposure-related changes in the mouse IHCs with and without ribbon loss. Our quantifications reveal that IHC subtypes differ in the worst-hit area of synaptopathy. Moreover, we show relative enrichment of mitochondria in the surviving SGN terminals, providing key experimental evidence for the long-proposed role of SGN-terminal mitochondria in synaptic vulnerability.

Subtle alterations of vestibulomotor functioning in conductive hearing loss

Introduction

Conductive hearing loss (CHL) attenuates the ability to transmit air conducted sounds to the ear. In humans, severe hearing loss is often accompanied by alterations to other neural systems, such as the vestibular system; however, the inter-relations are not well understood. The overall goal of this study was to assess vestibular-related functioning proxies in a rat CHL model.

Methods

Male Sprague-Dawley rats (N=134, 250g, 2months old) were used in a CHL model which produced a >20dB threshold shift induced by tympanic membrane puncture. Auditory brainstem response (ABRs) recordings were used to determine threshold depth at different times before and after CHL. ABR threshold depths were assessed both manually and by an automated ABR machine learning algorithm. Vestibular-related functioning proxy assessment was performed using the rotarod, balance beam, elevator vertical motion (EVM) and Ferris-wheel rotation (FWR) assays.

Results

The Pre-CHL (control) threshold depth was 27.92dB±11.58dB compared to the Post-CHL threshold depth of 50.69dB±13.98dB (mean±SD) across the frequencies tested. The automated ABR machine learning algorithm determined the following threshold depths: Pre-CHL=24.3dB, Post-CHL same day=56dB, Post-CHL 7 days=41.16dB, and Post-CHL 1 month=32.5dB across the frequencies assessed (1, 2, 4, 8, 16, and 32kHz). Rotarod assessment of motor function was not significantly different between pre and post-CHL (~1week) rats for time duration (sec) or speed (RPM), albeit the former had a small effect size difference. Balance beam time to transverse was significantly longer for post-CHL rats, likely indicating a change in motor coordination. Further, failure to cross was only noted for CHL rats. The defection count was significantly reduced for CHL rats compared to control rats following FWR, but not EVM. The total distance traveled during open-field examination after EVM was significantly different between control and CHL rats, but not for FWR. The EVM is associated with linear acceleration (acting in the vertical plane: up-down) stimulating the saccule, while the FWR is associated with angular acceleration (centrifugal rotation about a circular axis) stimulating both otolith organs and semicircular canals; therefore, the difference in results could reflect the specific vestibular-organ functional role.

Discussion

Less movement (EVM) and increase time to transverse (balance beam) may be associated with anxiety and alterations to defecation patterns (FWR) may result from autonomic disturbances due to the impact of hearing loss. In this regard, vestibulomotor deficits resulting in changes in balance and motion could be attributed to comodulation of auditory and vestibular functioning. Future studies should manipulate vestibular functioning directly in rats with CHL.

Bats experience age-related hearing loss (presbycusis)

Hearing loss is a hallmark of aging, typically initially affecting the higher frequencies. In echolocating bats, the ability to discern high frequencies is essential. However, nothing is known about age-related hearing loss in bats, and they are often assumed to be immune to it. We tested the hearing of 47 wild Egyptian fruit bats by recording their auditory brainstem response and cochlear microphonics, and we also assessed the cochlear histology in four of these bats. We used the bats' DNA methylation profile to evaluate their age and found that bats exhibit age-related hearing loss, with more prominent deterioration at the higher frequencies. The rate of the deterioration was ∼1 dB per year, comparable to the hearing loss observed in humans. Assessing the noise in the fruit bat roost revealed that these bats are exposed to continuous immense noise-mostly of social vocalizations-supporting the assumption that bats might be partially resistant to loud noise. Thus, in contrast to previous assumptions, our results suggest that bats constitute a model animal for the study of age-related hearing loss.

Volume electron microscopy reveals age-related circuit remodeling in the auditory brainstem

The medial nucleus of the trapezoid body (MNTB) is an integral component of the auditory brainstem circuitry involved in sound localization. The giant presynaptic nerve terminal with multiple active zones, the calyx of Held (CH), is a hallmark of this nucleus, which mediates fast and synchronized glutamatergic synaptic transmission. To delineate how these synaptic structures adapt to reduced auditory afferents due to aging, we acquired and reconstructed circuitry-level volumes of mouse MNTB at different ages (3 weeks, 6, 18, and 24 months) using serial block-face electron microscopy. We used C57BL/6J, the most widely inbred mouse strain used for transgenic lines, which displays a type of age-related hearing loss. We found that MNTB neurons reduce in density with age. Surprisingly we observed an average of approximately 10% of poly-innervated MNTB neurons along the mouse lifespan, with prevalence in the low frequency region. Moreover, a tonotopy-dependent heterogeneity in CH morphology was observed in young but not in older mice. In conclusion, our data support the notion that age-related hearing impairments can be in part a direct consequence of several structural alterations and circuit remodeling in the brainstem.

Objectification of intracochlear electrocochleography using machine learning

Introduction

Electrocochleography (ECochG) measures inner ear potentials in response to acoustic stimulation. In patients with cochlear implant (CI), the technique is increasingly used to monitor residual inner ear function. So far, when analyzing ECochG potentials, the visual assessment has been the gold standard. However, visual assessment requires a high level of experience to interpret the signals. Furthermore, expert-dependent assessment leads to inconsistency and a lack of reproducibility. The aim of this study was to automate and objectify the analysis of cochlear microphonic (CM) signals in ECochG recordings.

Methods

Prospective cohort study including 41 implanted ears with residual hearing. We measured ECochG potentials at four different electrodes and only at stable electrode positions (after full insertion or postoperatively). When stimulating acoustically, depending on the individual residual hearing, we used three different intensity levels of pure tones (i.e., supra-, near-, and sub-threshold stimulation; 250–2,000 Hz). Our aim was to obtain ECochG potentials with differing SNRs. To objectify the detection of CM signals, we compared three different methods: correlation analysis, Hotelling's T² test, and deep learning. We benchmarked these methods against the visual analysis of three ECochG experts.

Results

For the visual analysis of ECochG recordings, the Fleiss' kappa value demonstrated a substantial to almost perfect agreement among the three examiners. We used the labels as ground truth to train our objectification methods. Thereby, the deep learning algorithm performed best (area under curve = 0.97, accuracy = 0.92), closely followed by Hotelling's T² test. The correlation method slightly underperformed due to its susceptibility to noise interference.

Conclusions

Objectification of ECochG signals is possible with the presented methods. Deep learning and Hotelling's T² methods achieved excellent discrimination performance. Objective automatic analysis of CM signals enables standardized, fast, accurate, and examiner-independent evaluation of ECochG measurements.

A New Adaptive GCC Method and Its Application to Slug Flow Velocity Measurement in Small Channels

In this work, an adaptive generalized cross-correlation (AGCC) method is proposed that focuses on the problem of the conventional cross-correlation method not effectively realizing the time delay estimation of signals with strong periodicity. With the proposed method, the periodicity of signals is judged and the center frequencies of the strongly periodical components are determined through the spectral analysis of the input signals. Band-stop filters that are used to suppress the strongly periodical components are designed and the mutual power spectral density of the input signals that is processed by the band-stop filters is calculated. Then, the cross-correlation function that is processed is the inverse Fourier transform of the mutual power spectral density. Finally, the time delay is estimated by seeking the peak position of the processed cross-correlation function. Simulation experiments and practical velocity measurement experiments were carried out to verify the effectiveness of the proposed AGCC method. The experimental results showed that the new AGCC method could effectively realize the time delay estimation of signals with strong periodicity. In the simulation experiments, the new method could realize the effective time delay estimation of signals with strong periodicity when the energy ratio of the strongly periodical component to the aperiodic component was under 150. Meanwhile, the cross-correlation method and other generalized cross-correlation methods fail in time delay estimation when the energy ratio is higher than 30. In the practical experiments, the velocity measurement of slug flow with strong periodicity was implemented in small channels with inner diameters of 2.0 mm, 2.5 mm and 3.0 mm. With the proposed method, the relative errors of the velocity measurement were less than 4.50%.