Two-dimensional filter to facilitate detection of transient-evoked otoacoustic emissions

Denoising click-evoked otoacoustic emission signals by optimal shrinkage

Click-evoked otoacoustic emissions (CEOAEs) are clinically used as an objective way to infer whether cochlear functions are normal. However, because the sound pressure level of CEOAEs is typically much lower than the background noise, it usually takes hundreds, if not thousands, of repetitions to estimate the signal with sufficient accuracy. In this paper, we propose to improve the signal-to-noise ratio (SNR) of CEOAE signals within limited measurement time by optimal shrinkage (OS) in two different settings: covariance-based optimal shrinkage (cOS) and singular value decomposition-based optimal shrinkage (sOS). By simulation, the cOS consistently enhanced the SNR by 1-2 dB from a baseline method that is based on calculating the median. In real data, however, the cOS cannot enhance the SNR over 1 dB. The sOS achieved a SNR enhancement of 2-3 dB in simulation and demonstrated capability to enhance the SNR in real recordings. In addition, the level of enhancement increases as the baseline SNR decreases. An appealing property of OS is that it produces an estimate of all single trials. This property makes it possible to investigate CEOAE dynamics across a longer period of time when the cochlear conditions are not strictly stationary.

Signal-to-noise ratio improvement of swept-tone-generated transient otoacoustic emissions

In this study, we utilized the swept-tone (ST) deconvolution method for comparing the signal-to-noise ratio (SNR) characteristics of ST otoacoustic emissions (OAE) to conventionally acquired click, or transient-evoked (TE), OAE. We generated a hearing-level equalized (HLeq) ST stimulus based on normative loudness metrics at the different frequencies present in the ST. Due to noise-shaping properties of the ST deconvolution method, we anticipated a theoretical SNR gain of +4.26 dB in STOAE compared to TEOAE acquired under comparable settings. This prediction was confirmed by computer simulation. HLeq STOAE and TEOAE were then acquired from each of the 22 ears that were tested at five stimulation levels from 5 to 45 dB HL, and analyzed responses in terms of their overall SNR. We found that the overall SNR of the HLeq STOAE responses at stimulation levels at or above 15 dB HL was significantly higher than that of TEOAE by an average of +3.6 dB. Importantly, this leads to recording quality and time-saving improvements in clinical hearing screenings.

The Milan Project: a newborn hearing screening programme

AIM

Since 1997 a newborn hearing screening programme has been implemented by the U.O. Neurologia-Neurofisiopatologia and Dipartimento di Neonatologia of the Istituti Clinici di Perfezionamento ICP in Milan for both babies with no risk and those at risk of hearing impairment. This programme was named the Milan Project.

METHODS

The protocol for no-risk babies consisted of three stages: in the first two stages, newborns were tested with transient click-evoked otoacoustic emissions (TEOAE), in the third one with conventional auditory brainstem responses (ABR). The first TEOAE test was performed by 36 h of age, before discharge, the second one after 15-30 d in case of referral, and the third one, by ABR, for those babies who failed the second TEOAE stage. Newborns at audiological risk were submitted to conventional ABR before the third month of corrected age. Some of this latter population was also submitted to the TEOAE test. The entire tested population (no-risk babies and newborns at audiological risk) consisted of 19 777 babies: 19 290 without risk ("no risk") and 487 at risk ("at risk").

RESULTS

During the course of the Milan Project, hearing impairment (ABR threshold equal to or greater than 40 dB nHL) was identified in 63 newborns (19 from the no-risk and 44 from the at-risk population), with a prevalence of 0.32%. Bilateral hearing impairment (BHI) was found in 33 newborns (10 from the no-risk and 23 from the at-risk population), corresponding to 0.17%. Among infants with bilateral hearing impairment, 30.3% had no risk factors. The prevalence of hearing impairment was determined on days 15-30 after birth.

CONCLUSIONS

The results show that the implementation of a hospital-based, universal neonatal hearing screening programme for babies with and without audiological risk is feasible and effective. The effectiveness of the programme has increased as a function of the years since its inception, with a strong decrease in the referral rate. Further improvement is obtained if the TEOAE measurements are repeated in cases of referral scoring before discharge.

Principal component analysis as a method to facilitate fast detection of transient-evoked otoacoustic emissions

Transient-evoked otoacoustic emissions (TEOAE) are acoustic signals coming from the inner ear (outer hair cells of the cochlea) after acoustic stimulation by clicks. They can be used to investigate the status of the peripheral hearing system. Some of their potential applications (e.g., their use as a tool in newborn hearing screening programs) are deeply related to the duration of each recording session. This duration can be strongly reduced by applying a principal component analysis approach to a set of TEOAE recorded from the same ear at different stimulus levels averaging only a few sweeps (a maximum of 100 versus the classical 260). The PCA approach is shown to be able to enhance the signal-to-noise ratio and, in turn, to allow a correct detection of the responses. Results of the application of this approach in comparison with responses recorded from the same subjects with the classical technique will be shown.

"Linear" and "derived" otoacoustic emissions in newborns: a comparative study

OBJECTIVE

To investigate the effects of a specific aspect of the acquisition procedure, the averaging technique, on the evaluation of click-evoked otoacoustic emissions (CEOAEs) in newborns.

DESIGN

CEOAEs were recorded by an Otodynamic ILO88 system from 89 full-term newborns at the third day after delivery. For each ear and in the same test session, CEOAEs were evoked by 75 to 85 dB pSPL acoustic clicks and averaged according to two different modes: the "linear" (classic average) and the "derived" mode, which allows the cancellation of linear behaving components (such as acoustic artifacts). All examined ears had a normal auditory function as assessed by conventional ABR between the ages of 2 and 4 mo. CEOAEs obtained by both averaging techniques were compared on the basis of several quantitative parameters: the waveform similarity; the levels of signal and noise and the inter-test reproducibility of the broadband response and of four different frequency bands centered at 1.6, 2.4, 3.2, and 4 kHz; the amplitude as a function of time; the test time. To eliminate the contribution of the stimulus artifact, linear CEOAEs were windowed 6 to 20 msec, whereas derived emissions were windowed using the default ILO88 window (2.5 to 20 msec). Additionally, CEOAEs were classified as "pass" or "fail" accordingly to screening criteria used in the daily clinical practice.

RESULTS

Linear and derived emissions had very similar wave shapes and no time shifts during the first 12.5 msec. On the contrary, clear differences in the waveforms and time shifts were observed at longer latencies. The use of both averaging techniques resulted in identical CEOAE levels for both the broadband response and for the first two tested frequencies. For the last two frequencies, emission levels were lower when averaged with the linear technique owing to the use of the time window 6 to 20 msec, which reduces the amplitudes of high-frequency components. The residual noise in derived traces is 6 dB higher than that from linear traces. Also, derived CEOAEs had a lower inter-test reproducibility in both the broadband compound emission and in the four frequency bands examined here. The greatest difference in reproducibility was observed at the lowest band (1.2 to 2 kHz). Scoring of emissions was influenced by the averaging technique: 14% CEOAEs obtained with linear averaging and scored as passes were classified as fails when averaged with the derived mode. Moreover, if a CEOAE was scored as pass when using the derived technique, it also was scored as pass when using linear averaging. The increased number of false positives most likely was due to the higher noise floor/lower signal to noise ratio (SNR) of CEOAEs obtained with the derived technique.

CONCLUSIONS

In the tested newborns and at the levels of stimulation used in this study, the emissions obtained with the derived technique were noisier than those obtained with the linear technique, this being intrinsically due to the type of averaging. Therefore, screening criteria based on the evaluation of the SNR (or similar parameters) could be influenced by the type of averaging used during the acquisition.

Otoacoustic emissions and improved pass/fail separation using wavelet analysis and time windowing

A new method is presented for the purpose of improving pass/fail separation during transient evoked otoacoustic emission (TEOAE) hearing screening. The method combines signal decomposition in scales using the discrete wavelet transform, non-linear denoising and scale-dependent time windowing. The cross-correlation coefficient between two subaveraged, processed TEOAE signals is used as a pass/fail criterion and assessed in relation to the pure-tone, mean hearing level. The performance is presented in terms of receiver operating characteristics for a database of 5,214 individuals. The results show that the specificity improves from 68% to 83% at a sensitivity of 90% when compared with the conventional wave reproducibility parameter.

Frequency and temporal analysis of contralateral acoustic stimulation on evoked otoacoustic emissions in humans

Previous studies have shown that the effect of contralateral acoustic stimulation (CAS) on ipsilateral evoked otoacoustic emissions (EOAE) depends somewhat upon the spectrum of the eliciting stimulus. The latency of the EOAE, however, is itself frequency-dependent. Consequently, two general ways of analyzing the effects of CAS may be considered: by frequency band or by temporal segment. In this study, we analyzed the effects of CAS both ways in the same subjects, essentially simultaneously. The frequency analysis of the EOAE derived from the wavelet transform (WT). The WT is known to provide a robust approach to the analysis of non-stationary signals and was anticipated to avoid possible time-frequency confounds of the cochlear mechanical system. For comparison, a more basic analysis - using a temporal moving window - was employed. The results largely support earlier findings and confirm that in humans the greatest suppression of EOAEs by CAS is obtained for lower frequency and/or longer latency EOAE components. Despite expectations for the WT analysis, the more basic, temporal, analysis tended to yield the clearer results.

Frequency-specific information from click evoked otoacoustic emissions in noise-induced hearing loss

Click evoked otoacoustic emissions (CEOAEs), pure-tone audiograms (PTAs), and Bekesy sweep frequency audiograms were recorded from 15 ears of 11 subjects with noise-induced hearing loss. For all ears, hearing threshold levels > or = 30 dB HL were found at the high frequencies. The aims of the study were to examine whether the decomposition of CEOAEs into narrow band components could identify hearing loss in a frequency-specific manner and to what extent audiometric thresholds could be predicted. CEOAEs were parcelled into 0.5-kHz-wide components by means of the wavelet transform. Reproducibility of CEOAE components was compared with audiometric threshold at corresponding frequencies. A general trend of low reproducibility for increasing audiometric thresholds was found. A reproducibility value of 60 per cent was found to best separate normal and elevated thresholds. The presence of a CEOAE component at a given frequency was always associated with audiometric thresholds < or = 20-25 dB HL. On the other hand, the absence of a component was equally associated either with normal or abnormal hearing levels. Large inter-subject variability was observed. A weak linear relationship was found between reproducibility and audiometric thresholds at corresponding frequencies, indicating that analysis of narrow band CEOAE components is valuable for separating normal from hearing-impaired ears but cannot replace the audiogram.

Evaluation of click evoked otoacoustic emissions in newborns: effects of time-windowing

The objective of this study was to investigate the effects of time-windowing on click evoked otoacoustic emissions (CEOAEs); 466 CEOAEs from full-term babies were considered. Data were acquired according to the default ILO88 response window (2.5-20 ms). Because CEOAEs are time-varying signals, each emission was analysed by means of the wavelet transform (WT), a technique which allows a simultaneous representation of the time and frequency features of a signal. By means of the WT it was possible to extract the temporal pattern of the elementary frequency components of the compound emissions. The effects of time-windowing were evaluated on each single emission component: rms and correlation values were computed from six 2.5-ms long time windows in the 2.5-17.5 ms range. Results indicated that both rms and correlation were not constant with time but reached a maximum in specific time-windows, depending on the frequency of the component. For low-frequency components, the correlation typically had a maximum in the interval 10-12.5 ms, whereas high-frequency components had a maximum around 5 ms. Above 12.5 ms, the correlation was greatly decreased for all frequency components. As a result, the comparison between the performance of the default ILO88 window (2.5-20 ms) and the window 2.5-12.5 ms showed that for all frequencies in the 1.5-6 kHz range there was a statistically significant improvement in the correlation. No improvement was observed for frequencies below 1.5 kHz because of their contamination by the residual background noise.