`
TEOAE Recording Protocols Revised. Data from adult subjects (paper summary from IJA 42, 2003 )
A reprint of the original article can be downloaded from here
Stavros Hatzopoulos S., Joe Petrucelli J2., Thierry Morlet T3. and Alessandro Martini A 1
1 Dept. of Audiology, University of Ferrara, Italy.
2 Dept. of Mathematical sciences, Worcester Polytechnic, Worcester, Mass., USA
3 Dept of Otolaryngology and Biocommunication, Louisiana State University, New Orleans, LA , USA
1. INTRODUCTION
Traditionally the transient Evoked Otoacoustic Emission (TEOAE) responses can be evoked by two types of train stimuli: (a) by a set of four clicks of equal magnitude (referred to as the linear protocol) or (b) by three clicks of positive polarity followed by a fourth click of an inverse polarity with a relative magnitude of 9.5 dB higher than the corresponding positive clicks (referred to as the non-linear or the derived non-linear protocol). Under the hypothesis that the TEOAE recordings originate from saturated cochlear generators, it is assumed that the nonlinear protocol removes stimulus artifacts of linear nature which can be misinterpreted as TEOAE responses (Kemp et al, 1986; 1990). Despite the lack of sufficient research on the statistical verification of the functional premises of the nonlinear protocol (i.e. how much the induced linear artifacts at stimulus levels > 80 dB SPL are suppressed ) , it is generally accepted that the nonlinear ILO protocol is a practical compromise to maximize the reliability of a TEOAE recording , and this protocol is used to assess of the integrity of the cochlear function of neonatal and adult subjects.
The fact that the non-linear protocol contains the words Ònon-linearÓ has created considerable confusion in the OAE clinical community. Over the years, after the introduction of this methodology by Kemp et al. (1986, 1990), two questionable assumptions have been generated: (1) the non-linear protocol captures the non-linear cochlear responses due to the non-linearity of its stimulation paradigm. This statement is not true because the utilized clicks are linear ; (2) the non-linear protocolÓ assures the capture of the non-linear TEOAE responses, because the high intensity click stimuli the clicks stimulatesaturate the cochlea. The TEOAEs are of non-linear origin due to the nonlinear operation of the cochlear amplifier and mainly of the outer hair cells of the organ of Corti (Kim, 1986; Zwicker, 1986; Leeuw and Dreschler 1998; Thorton et al, 2001; Zinn et al, 2001), but not all segments of the cochlear amplifier saturate at the same stimulus intensity (Patuzzi, 1987; Shera and Guinnan, 1999). For example, the regions related to TEOAE frequencies < 1.0 kHz saturate even with mid-intensity stimuli of approximately 60 dB SPL (Grandori et al, 1994) and in many occasions (neonatal subjects and children) even the use of high intensity stimuli ( > 80 dB SPL) do not fully saturate the cochlea . In this context, it should be clarified that both linear and nonlinear TEOAE protocols can efficiently capture the nonlinear responses (TEOAEs) of the cochlear amplifier.
This issue of ÒOAE capturing efficiencyÓ has been an interesting argument for the signal processing OAE community during the last 10 years (Lutman, 1993; Grandori and Ravazzani, 1993; Grandori et al, 1994; Berlin et al, 1995; Hatzopoulos et al, 1999; 2000b; Tognola et al, 2001; Von Specht et al, 2001). Recently the impetus from the diffusion of the TEOAE neonatal screening technology has set the proper grounds for further TEOAE protocol re-design and re-evaluation. This goal is also one of the majormain objectives of the European Concerted Action on Otoacoustic Emissions (AHEAD-II, 2002), aiming the optimization of the current TEOAE recording methods (i.e. an improvement of the signal to noise ratio ÒS/NÓ referred to as the quality of the TEOAE recording). The optimized protocols are expected to offer improvements in two areas: (1) anan improvement in the amplitude and the S/N ratio of the TEOAE response; and (2) a decrease of the TEOAE acquisition time, a requirement imposed by the neonatal hearing screening programs.
A way to minimize the TEOAE time recording requirements is based on the fact that the characteristics of the final TEOAE response depend on the number of averaged sweeps (or acquisition frames), which for an adult subject vary from 100 to 260 (260?). In this context, it has been proposed to drive the cochlea with a faster stimulus rate, in order to capture TEOAE responses in less time. This is the basis of the Maximum Length Sequence (MLS) protocol using linear clicks and stimulation rates up to 5 kHz (Thornton et al, 1993, 2001). This protocol has not been implemented commercially yet, because it requires significant changes in the current TEOAE instrumentation. An alternate approach(like the Quickscreen protocol?) (used in the popular neonatal protocol QuickScreen) is to decrease the value of the inter-stimulus interval. The latter by default is approximately 20 ms and corresponds to a stimulation rate of 50 stimuli per second. In this context, the use of a shorter 12.5(12.5?) ms TEOAE response-window increases the stimulation rate from 50 to 80 reps / s. The processing of information derived from smaller TEOAE response windows, as the one mentioned above, offers an additional advantage. By shortening the response window the average noise level, which is usually higher at the latter segments of the TEOAE response, decreases and the resultingresulting waveform presents higher signal to noise ratios (Kemp et al, 1994; Hatzopoulos et al, 1995; Whitehead et al, 1995, Fitzgerald and Prieve, 1997). Despite these two positive aspects, the TEOAE-window processing proposals suffer from the same disadvantages of the traditional nonlinear recordings (low S/N ratio), which is is caused by the effect of the forth click in the stimulus train (Kemp et al, 1986).
The alternative solution to a nonlinear TEOAE protocol is based on trains of click stimuli having the same polarity. Such protocols have been used before the commercial introduction of the ILO nonlinear method, but TEOAE recordings evoked by such schemes were prone to stimulus induced artifacts, generated by reflections of the acoustic energy from the walls of the acoustic meatus and the tympanic membrane (Kemp et al,1986; Johnsen et al, 1988). For the recordings evoked by a nonlinear protocol, the stimulus artifact is cancelled out by subtracting properly scaled TEOAE responses evoked by different intensity click stimuli of the opposite polarity of the opposite polarity. The stimulus artifact of the linear response can be suppressed , by applying a window function that zeros the initial portion of the response which is corrupted by the artifact. In physical terms, the linear artifact is not cancelled-out but its contribution to the TEOAE response is severely attenuated by the weighting-effect of the window function on the TEOAE response. TEOAE recordings evoked by a linear protocol are expected to present higher S/N ratios, because the stimulus train lacks the differencing (subtracting) action of the fourth click, which reduces the signal strength and increases the high frequency noise. A possible drawback of a windowed linear protocol might be the exclusion of an initial segment of the TEOAE response, which is assumed to contain unique higher frequency components. This assumption is not valid according to a number of studies (Cheng 1993,1995; Hatzopoulos et al, 2000b), which have evaluated the structure of the TEOAE recordings with time-time-frequency (TF) techniques. Data from the TF analyses have demonstrated that the high TEOAE frequency components ( £ 5 kHz) are still detectable at TEOAE latencies greater than 4.0 ms.
The present study was designed to provide a new approach of TEOAE data acquisition for emerging TEOAE clinical applications. The objectives of the study were the following: (1) the identification of the limits of the window function, which can be applied on the linear TEOAE responses, preserving the frequency content of the data and suppressing as much as possible stimulus artifacts; (2) the definition of any significant differences between the means of the linear and nonlinear TEOAE responses; (3) theThe definition of scoring criteria regarding normal TEOAE responses, which can be extrapolated for use in other clinical applications (contralateral stimulation, ototoxicity monitoring , carriers of other genetic syndromes etc). The last objective was sought because despite the longer history of applying TEOAEs to assess the hearing status of adult subjects, there are very few references in the literature (Gorga et al, 1993; Prieve et al, 1993) indicating which criteria to use for the TEOAE hearing evaluation.
To proceed with the project objectives, we have first estimated the extent of the stimulus artifact in the linear recordings (Results: section 1). Based on this information we estimated a new window, whose efficiency was evaluated from TEOAE recording simulations in 2cc and 5cc cavities and ears with severe sensorineural hearing losses (Results: section 2) . After the post-processing of the data, we compared the nonlinear and linear responses in terms of various clinical parameters of interest (Results: section 3). Finally we estimated scoring criteria (minimal normative responses) for the linear recordings (Results: section 4). We have postulated that a post-windowed linear TEOAE response will be characterized by lower noise and higher S/N ratios than the corresponding nonlinear response , a feature which is very useful to the successful application of discriminant models to a population of probable carriers. To simplify the terminology through-out the text, the data evoked by a linear / nonlinear protocol are called linear or nonlinear recordings respectively.
2. Materials and Methods
Subjects
Forty-two42 healthy adults (age 26 ± 3.2 years) participated in the study. The hearing-normality of each subject was assessed with otoscopy,otoscopy, pure tone audiometry (thresholds better or equal to 20 dB HL at 0.5 - 4.0 kHz), and tympanometry . All subjects had a normal medical history and none was under any particular medication. Otoacoustic emissions were recorded from the best ear and for the cases where both left and right hearing thresholds were similar an ear wasrandomly randomly selected. TEOAE recordings (from both linear and nonlinear protocols) were also acquired from 4 patients, with severe sensorineural losses (SNHL), showing mean threshold levels higher than 60 dB HL at 2.0 and 4.0 kHz. The data from these recordings were not analyzed statistically, but were used as validators of the duration of the stimulus artifact in the linear recordings.
Recordings
The recording sessions were conducted in an acoustically isolated room using the ILO-292 apparatus (software version 5.60). The linear recordings were elicited by clicks of a 72 dB SPL (-12 dB ILO) and the nonlinear recordings by clicks of 84 dB SPL (0 dB ILO). Each recording was the average of 260 sweeps. The level of acceptable noise was set to be < 3.4 mPa (or approximately 44.6 dB SPL). For all recordings the default ILO window ( 2.5 to 19.5 ms) was used.
Post-Processing of Linear Recordings
Prior to the comparison of the linear and nonlinear data sets the linear recordings were post-processed by a filtering and a windowing routine. According to a previous study (Hatzopoulos et al, 2000b) the frequency content of TEOAEs for frequencies below 900 Hz is very low. Since these frequencies are often associated with the stimulus artifact, a band-pass filter attenuating frequencies below 830 Hz and above 4800 Hz was applied to data. The digital filter is incorporated in the ILO software. For the windowing of linear recordings, a 3.8 -13.8 ms window function was used with a rise and fall-time of 0.64 ms. The low limit of the window function was defined after the results in section 1. The upper limit of the window function (13.8 ms) was defined according to the results of a previous study on adult subjects (Hatzopoulos et al, 20000b). The window function was defined by the ILO software and it is similar to the cosine tapered window used in the default nonlinear protocol.
Post-Processing of Nonlinear Recordings
For the statistical comparisons of the data, we generated two data sets from the available nonlinear recordings. In the first set, called nonlinear-default (coded as D), the data were used as recorded by the ILO-292 windowed with the ILO default window function (applied from 2.5 to 19.5 ms). For the second set of data (coded as N) , the recordings were post-windowed with the same window function applied to the linear recordings ( 3.8 to 13.8 ms). Both nonlinear datasets were filtered with the same band-pass filter used for the linear recordings.
Statistical Methods
We modeled the difference between the linear and nonlinear response values (TEOAE amplitude) over time using the following repeated measures model:.
D(ij)=mu+T(i)+E(ij), where D(ij) = is the amplitude difference between the linear and nonlinear measurement for subject j at time i.
mu = is the overall amplitude mean,
T(i) = is the effect of time i, and
E(ij) = is the error associated with response D(ij).
For each of the nine parameters described above, we used the following repeated measures model to evaluate differencesdifferences in means among the three TEOAE protocols: Y(ij)=mu + P(i) + E(ij), where
Y(ij) =is the response of subject j to protocol i (linear, nonlinear, or nonlinear-default),
mu =is the overall mean
P(i) = is the effect of protocol i
E(ij) = is the random error associated with the observation.
The SAS procedure Òproc mixedÓ was used to fit each model. As with the previous model, several within-subject covariance structures were considered, but for each parameter the compound symmetric covariance structure gave the best fit as determined by the AIC and BIC measures (see the Appendix for details). To evaluate significant differences in response means for the different protocols, each overall F test significant at the 0.05 level was followed by Tukey-Kramer simultaneous pair-wise comparisons, having an experiment- wise 0.05 significance level.
For the calculation of the scoring criteria we used a free-distribution approach because the TEOAE variables were not normally distributed. The scoring criteria provided us with a minimum estimate of normal performance, which is the lower tolerance bound of the estimated tolerance interval (for every tested variable). To obtain scoring pass/fail criteria, we calculated one-sided distribution-free tolerance intervals. These intervals ensure that for a user-specified confidence, ÒMÓ, and a user-specified population proportion, p, we can be ÓMÓbeÓ MÓ percent confident the computed interval will contain at least a percentage ÒpÓ of measurements for the entire population. We present these intervals for M=90% and 95% and values of p between 89% to 94%. Additional details on the free distribution method can be found in previous publications (Hatzopoulos et al, 1999; Hatzopoulos et al, 2000b).
3. RESULTS
Section 1: Presence of Stimulus artifact in the Linear Recordings
The results from the repeated measuress model indicated that there are no statistically significant differences between the mean linear and the nonlinear responses evaluated in the TEOAE recording segment from 3.2 to 5.2 ms . Five subjects presented a number of outliers but when these were removed and the data were re-fit the results were substantially the same. These findings verify the visual inspection of the linear data-set and a representative case is shown in Figure 1. The similarity of the nonlinear and linear trace-contours demonstrates that in the initial part of the TEOAE response the differences between the linear and nonlinear recording are minimal. Following the findings of the repeated measures model, the value of 3.8 ms was selected to represent the lower limit of the post-processing window function. This choice was followed in order to preserve the highest possible TEOAE signal bandwidth The upper limit of the window function was set to 13.8 ms according to previous data (Hatzopoulos et al, 20000b) indicating that a windowed response limited to this upper value would contain more than 90% of the original energy.
Figure1: The trace contours (TEOAE amplitudes) depict the linear and nonlinear TEOAE responses from subject Bar_Lin_R. To compare the recordings we have used the ÒcompareÓ feature of the ILO software. To reveal the details of the response in the first 8 ms we have expanded the time scale using the ÒExpand ResponseÓ option from the View menu of the ILO software. The contours of the two recordings are very similar and at a time = 5ms the traces overlap completely. The arrows indicate the trace of the linear response.
Section 2: TEOAE simulations and data from ears with severe hearing losses.
The artifact-suppression efficiency of the window defined in the previous section was tested on simulated TEOAEs, in 2cc and 5cc cavities, and on ears presenting severe sensorineural hearing losses. The post-processing of the data suppressed completely the induced artifact. Twenty simulated responses were collected for each cavity. A t-test statistic suggested that the mean of the processed responses was significantly different than the mean of un-processed linear responses at 1.0 , 2.0, 3.0 and 4.0 kHz ( p < 0.003 and p < 0.001 for the 2cc and 5cc cavities respectively) .
As it was expected, the linear recordings from the tested SNHL ears showed lack of emissions. The first ms of the TEOAE response were corrupted by a low frequency artifact. After the post-processing (i.e. filtering and windowing), the artifact was totally suppressed and the average amplitude of the TEOAE response was within levels of random noise. A typical example of a SNHL recording evoked by a click stimulus of 72 dB SPL is showing in Figure 2. Panel A shows the response prior to any processing and Panel B shows the processed result. The stimulus artifact manifests as a low frequency waveform (Panel A) spanning at least 4.2 ms (this was the largest value observed in all tested SNHL ears). The TEOAE response in the second panel shows that the artifact has been suppressed / eliminated . The amplitude of the TEOAE response in the second panel remains below a value 60 mPa, data in accordance with the hearing status of the tested ear (i.e. absence of emissions).
Figure2: Results from simulating TEOAE recordings in two coupler cavities, in order to evaluate the efficiency of the post-processing applied to linear TEOAE responses. The upper panel (A) shows the processed TEOAE response recorded in a cavity of 2 cc and elicited by a stimulus of 74 dB SPL. The lower panel B shows the processed TEOAE response recorded in a cavity of 5 cc, and elicited by the same amplitude stimulus as in (A). Both recordings were high-pass filtered at 830 Hz (second order filter, provided on-line by the ILO software) and windowed by a 3.80-13.8 ms window with a 0.6 ms rise-time. The x-axis in both panels shows time (0- 13) in ms
Section 3: Comparison between linear and nonlinear responses
The results from the repeated measures model indicated that for all tested parameters, except the TEOAE response and noise, the mean linear recording values were significantly larger than the values from the two nonlinear datasets. As it was expected, the mean nonlinear recording values for the TEOAE response were larger due to the difference in the stimulus intensity (84 vs.( 84 vs. 72 dB SPL for the nonlinear and linear protocols respectively). The significant differences for the TEOAE noise variable verify the assumptions regarding the noising-effects of the fourth click in the nonlinear stimulus train. The results from the linear-nonlinear data comparisons are summarized in Table 1.
| Variable | Interpretation |
| Response | D > N > L |
| A-B (Noise) | D > N > L |
| Corrected Signal | L> N > D |
| Correlation % | L> N > D |
| SN_1 kHz | L> N > D |
| SN_2 kHz | L> N > D |
| SN_3 kHz | L> N , D |
| SN_4 kHz | L> N , D |
| SN_5 kHz | L> N , D |
| L = linear post-windowed data from 3,8 to 13, 8 ms N = nonlinear post windowed data from 3,8 to 13,8 ms D = nonlinear post-windowed data from 2,5 to 19,5 ms |
Table 1: Results from the comparison between post-processed responses from linear and nonlinear TEOAE protocols. The first column indicates the TEOAE variable and the second shows the relationship between the means. For the majority of variables, the means for the linear recordings are significantly larger than those for nonlinear recordings. The symbolÒ>Ó is used to indicate the order in terms of magnitude between the means.
Section 4: Scoring Criteria.
For the scoring criteria we have considered mainly the S/N ratios and the correlation estimate which is used traditionally to indicate the quality of the TEOAE response ( or the absence of noise in the TEOAE recording). We have used two levels of confidence at 95% and 90%. The various option-scenarios are presented in Table 2 .The most advantageous choice of variables (shaded section of Table 2) can be summarized as follows: (i) we are 90% confident that at least 91% of the tested populationÕs TEOAE linear recording values will present a S/N at 2 kHz ³ 13 dB; (b) we are 90% confident that at least 91% of the tested populationÕs TEOAE linear recording values will present a S/N at 3 kHz ³ 11 dB; and (iii) we are also 90% confident that at least 91% of the population TEOAE linear recording values will present a correlation value ³ 91%. For the scoring criteria the values of S/N ratio at 1.0 kHz were very low (close to zero or negative) and for this reason they are not reported in Table 2.
Confidence (M) | Probability (p) | Scoring Criteria |
| 95% | 93% | S/N 2.0 kHz >= 13 dB S/N 3.0 kHz >= 10 dB S/N 4.0 kHz >= 0 dB Correlation >= 90% |
| 89% | S/N 2.0 kHz >= 13 dB S/N 3.0 kHz >= 11 dB S/N 4.0 kHz >=; 5 dB Correlation >= 91% |
| 90% | 94% | S/N 2.0 kHz >= 13 dB S/N 3.0 kHz >= 10 dB S/N 4.0 kHz >= 0 dB Correlation >= 90% |
| 91% | S/N 2.0 kHz >= 13 dB S/N 3.0 kHz >= 11 dB S/N 4.0 kHz >= 5 dB Correlation >= 91% |
Table 2: The normative adult scoring criteria for a confidence range (M) of 90% to 95% and a probability (p) of 89% to 94%. The best option (shaded cell ) corresponds to a combination of M=90% and p=91%, showing that the minimum S/N ratios from normal hearing adults should be 13, 11 and 5 dB at 2.0, 3.0 and 4.0 kHz. To extend the scoring criteria to more frequencies a larger normally hearing population is necessary.
4. DISCUSSION
The comparison of linear and nonlinear responses in the interval from 3.2 to 5.2 ms (prior to any post-processing of the data) suggested that the linear and nonlinear recordings were similar, due to lack of any significant differences between their means.means, Based on this result we have defined a window function from 3.8 to 13.8 ms which we have applied to all linear TEOAE recordings. The fact that the linear and nonlinear responses are so is surprising, considering that the nonlinear protocol was invented in order to suppress stimulus artifacts. Prior to the analysis of the data, we postulated that the outcome of the analyses could be interpreted as follows: or (a) the linear responses are not contaminated by an artifact; or (b) the nonlinear responses contain linear components in the early segments of the recording. After the completion of the analyses we have concluded that both assumptions are probably correct. An earlier study by Grandori et al, 1994, using the growth patterns of TEOAE input-output curves, has presented evidence showing that the nonlinear responses contain linear components.
The lower limit of the window function was set at 3.8 ms despite the fact that the repeated measures model indicated no significant differences at time = 3.2, 3,4 and 3.6 ms. In our opinion In our opinion thisthis choice represents a compromise for the suppression of the artifact and the preservation of the highest possible signal bandwidth. The definition of the 3.8 to 13.8 window is a procedural improvement over older data suggesting response windows starting from 5.0 ms (Osterhammel et al, 1993;Grandori et al, 1994) or 6.0 ms (Lutman, 1993). By shifting the lower bound of the window function to a lower value we are more confident that we can capture a wider bandwidth of the TEOAE response. It is worth mentioning, that for adult subjects, the later TEOAE recording segments do not contribute significantly to the overall TEOAE response in terms of frequency components and signal energy (Grandori et al, 1994; Hatzopoulos et al, 2000b). A response averaged over a shorter window (3.8 to 13.8 ms) benefits from the absence of the noisier segments of the TEOAE recording ( latter segments > 10 ms), thus it presentingpresents higher S/N ratios than theS/N ratiosones obtained from recordings processed with previously proposed processing schemes. It should be noted, that with the ILO-292 system it is possible to perform the processing of the TEOAE linear responses (i.e. filtering and windowing) during the data-acquisition stage, by changing a number of default values of the ILO software.
One of the necessary requirements for the successful application of the linear protocol is the condition that the stimulus intensity should not exceed the mean value of 74 + 2 dB SPL (data derived from the 2 cc cavity simulations). To control the efficiency of the window function it is mandatory to control the intensity of the stimulus energy reaching the tympanic membrane. GivenGiven a good probe fit with no leakage effects (attenuation of the LF or the HF TEOAE components), this can be achieved by manipulating the ILO stimulus level to a level of a relative 72.0 to 74.0 dB SPL, an indication that is available to the ILO user prior the collection of data. The reader should be aware that the stimulus artifact associated with the linear protocol actually depends on the positional relationship between the TEOAE probe and the tympanic membrane. The longer the distance between the two the larger the latency of the stimulus artifact (the reflection energy takes more time to reach the transducer microphone of the probe). In this context, a larger distance between the probe microphone and the tympanic membrane ( caused in many instances by an erroneous placement of the probe) will result in a temporal prolongation of the stimulus artifact, i.e. the ringing will last longer in terms of ms. Such an effect was not observed in any of the normal subjects who participated in this study. In our clinical practice, when the probe fitting results in excessive stimulus ringing , we have found very useful to use the smaller adult ILO probe (like the neonatal version but in brown color) which offers superior fitting and less ringing in the spectrum of the TEOAE stimulus. An improper position of the ILO probe might result in ringing stimulus waveforms with spectral peaks around 4 kHz in adult subjects) , which will probably generate artifacts longer than 2.5 ms. The data of this study are suggesting that the combination of high-pass filtering and windowing attenuate significantly the artifacts (for practical purposes the artifact vanishes), but in such cases it is recommended to use lower intensity click stimuli (i.e. 70 dB SPL) to ensure the absence of any linear artifacts in the sampling window.
The data from the ears with SNHL hearing losses have indicated that a low-frequency linear artifact is present in the unprocessed recording, but it is completely suppressed after the application of the proposed post-processing scheme. None on the tested SNHL ears generated responses which could be placed to or above the estimated 10th normative percentile of TEOAE correlation and S/N ratio at 2.0 , 3.0 and 4.0 kHz.
REFERENCES
AHEAD. Newsletter of the European Concerted Action on Otoacoustic Emissions, 1995, No 7: 1-4.
Berlin C.I., Hood L.J., Wen H., Szabo P., Cecola R.P., Rigby P., Jackson D.F. Contralateral suppression of non-linear click-evoked otoacoustic emissions. Hear. Res. 1993; 71: 1-11.
Berlin CI, Hood LJ, Hurley AE, Wen H, Kemp DT. Binaural noise suppresses linear click-evoked otoacoustic emissions more than ipsilateral or contralateral noise. Hear Res. 1995 Jul;87:96-103.
Cheng J . Time frequency signal representation of transient evoked otoacoustic emissions via smoothed pseudo-Wigner distribution. Karolinska Institute of Technical Audiology, Report TA129, 1993.
Cheng J . Time frequency analysis of transient evoked otoacoustic emissions via smoothed pseudo-Wigner distribution. Scand Audiol 1995; 24:91-96
Dircxx J J J, Daemers K, Somers Th, Offeciers F E and Govaerts P J. Numerical Assessment of TEOAE Screening Results: Currently Used Criteria and their Effect on TEOAE Prevalence Figures. Acta Otolaryngol (Stockh) 1996;116:672-679.
Fitzgerald TS and Prieve B. COAE Thresholds: 1. Effects of Equal amplitude versus subtraction methods. J Speech Lang Hear Res 1997, 40:1164-1176.
Gorga MP, Neely ST, Bergman BM, Beauchaine KL, Kaminski JR, Peters J, Schulte L, Jesteadt W. A comparison of transient-evoked and distortion product otoacoustic emissions in normal-hearing and hearing-impaired subjects. J Acoust Soc Am. 1993 Nov;94(5):2639-48.
Grandori F and Ravazzani P. Non-linearities of click-evoked otoacoustic emissions and the derived non-linear technique. Br J Audiol. 1993 Apr;27(2):97-102.
Grandori F, Ravazzani P, Tognola G, Hatzopoulos S. "Derived Nonlinear versus Linear Click-Evoked Otoacoustic Emissions", in F. Grandori, L. Collet, D. Kemp, G. Salomon, K. Shorn, R . Thornton (Eds) Advances in Otoacoustic Emission Vol I, Casa editrice Stefanoni, 1994, pp 48-64.
Hatzopoulos S., Mazzoli M., Martini A. "Identification of Hearing Loss using TEOAE-based descriptors: Preliminary results", Audiology, 34, 248-259, 1995.
Hatzopoulos S. , Martini M. , Cheng J., Grzanka A., Morlet T. and Martini M. ÒOptimization ofA. and Morlet T. ÒOptimization of the TEOAE recording protocols. A linear protocol derived from parameters of a time-frequency analysis: a pilot study onanalysis . a pilot study from neonatal subjects. Scand Aud , 2000a Mar ;29:21-27.
Hatzopoulos S. , Cheng J. ,Grzanka A. and Martini A. ÒTime-frequency analyses of TEOAE recordings from normal and SNHL casesÓ Audiology. 2000b;39:1-12.
Hatzopoulos S. Tsakanikos M, Grzanka A, Ratynska J, Martini A. ÒA comparison of neonatal TEOAE responses recorded with linear and QuickScreen protocolsÓ. Audiology, 2000c;39:70-79.
Johnsen NJ, Bagi P, Parbo J, Elberling C.Evoked acoustic emissions from the human ear. IV. Final results in 100 neonates. Scand Audiol. 1988;17(1):27-34.
Kemp D T, Bray PB, Alexander L, Brown AM. Acoustic Emission Cochleography-Practical Aspects. Scand. Audiol 1986;Suppl. 25: 71-94.
Kemp D T, Ryan S, Bray P. Otoacoustic Emission Analysis and Interpretation for Clinical Purposes, in :F. Grandori, G. Ciafrone and D.T. Kemp (Eds): Cochlear Mechanisms and Otoacoustic Emissions, Karger, 1990; 77-98.
Kim DO. Active and nonlinear cochlear biomechanics and the role of outer-hair-cell subsystem in the mammalian auditory system. Hear Res. 1986;22:105-14.
Leeuw AR, Dreschler WA. The relation between otoacoustic emissions and the broadening of the auditory filter for higher levels. Hear Res. 1998 Dec;126(1-2):1-10.
Lutman ME. Reliable identification of click-evoked otoacoustic emissions using signal-processing techniques. Br J Audiol. 1993 Apr;27(2):103-8.
Lutman ME, Jennings K, Davis AC, Houston HG, Meredith R. Coloration of Click-Evoked Otoacoustic Emissions by characteristics of the Recording Apparatus in F. Grandori, L. Collet, D. Kemp, G. Salomon, K. Shorn, R . Thornton (Eds) Advances in Otoacoustic Emission Vol I, Casa editrice Stefanoni, 1994, pp 36-47.
Osterhammel PA, Rasmussen AN, Olsen S, Nielsen LH. The influence of spontaneous otoacoustic emissions on the amplitude of transient-evoked emissions. Scand Audiol.1996;25:187-92.
Patuzzi RB. A model of the generation of the cochlear microphonic with nonlinear hair cell transduction and nonlinear basilar membrane mechanics. Hear Res. 1987;30:73-82.
Prieve BA, Gorga MP, Schmidt A, Neely S, Peters J, Schultes L, Jesteadt W. Analysis of transient-evoked otoacoustic emissions in normal-hearing and hearing-impaired ears. J Acoust Soc Am. 1993;93:3308-19.
Shera CA, Guinan JJ Jr. Evoked otoacoustic emissions arise by two fundamentally different mechanisms: a taxonomy for mammalian OAEs. J Acoust Soc Am. 1999 Feb;105(2 Pt 1):782-98.
Thornton ARD. High Rate Otoacoustic Emissions. J. Acoust. Am. 1993; 94:132-136
Thornton AR, Shin K, Gottesman E, Hine J. Temporal non-linearities of the cochlear amplifier revealed by maximum length sequence stimulation. Clin Neurophysiol. 2001;112:768-77.
Tognola G, Ravazzani P, Molini E, Ricci G, Alunni N, Parazzini M, Grandori F. "Linear" and "derived" otoacoustic emissions in newborns: a comparative study. Ear Hear. 2001;22:182-90.
Welzl-Muller K, Stephan K. Confirmation of Transiently Evoked Otoacoustic Emissions Based on User-Independent Criteria. Audiology 1994;33: 28-36.
White K R, Vohr B, Behrens T R. Universal Newborn Hearing Screening Using Transient Evoked Otoacoustic Emissions: Results of the Rhode Island Hearing Assessment Project. Seminars in Hearing 1993; 14:18-29.
Whitehead M L, Jimenez A M, Stagner B B, McCoy M J, Lonsbury-Martin B L, and Martin G K. Time-Windowing of Click-Evoked-Otoacoustic Emissions to Increase Signal-to-Noise-Ratio. Ear and Hearing 1995;16:599-611.
Von Specht H, Ganz M, Pethe J, Leuschner S, Pytel J. Linear versus non-linear recordings of transiently-evoked otoacoustic emissions--methodological considerations. Scand Audiol Suppl. 2001;52:116-8.
Zinn C, Maier H, Zenner H, Gummer AW. Evidence for active, nonlinear, negative feedback in the vibration response of the apical region of the in-vivo guinea-pig cochlea. Hear Res. 2000;142:159-83.
Zwicker E. "Otoacoustic" emissions in a nonlinear cochlear hardware model with feedback. J Acoust Soc Am. 1986 Jul;80:154-62.
