|
| |||
|
Conventional and High Rate Otoacoustic Emissions in Normally Hearing and Hearing Impaired Subjects by Jemma Hine D.Phil. (2005), MRC Institute of Hearing Research, Royal South Hants Hospital, Southampton, UK. This white paper is a synopsis of the article: Hine J.E. &
Thornton A.R.D. (2005) Transient Evoked Otoacoustic Emissions recorded
using Maximum Length Sequences from Patients with Sensorineural
Hearing Loss. Hearing Research 203, 122-133.
INTRODUCTION Numerous investigations have established that transient evoked otoacoustic emissions (TEOAEs) can be used to separate normally hearing and hearing impaired populations (see Harris and Probst, 2002 for a review). The general consensus is that TEOAEs will be present when all hearing levels from 250 to 8000 Hz are better than 20 dB HL and absent when all hearing levels are poorer than 40 dB HL due to a purely cochlear loss. Whilst the majority of studies into the screening potential of TEOAEs has examined the response parameters of the nonlinear TEOAE response, there are those that have focused on the so-called 'linear' recordings. However, the nonlinear method is usually favoured because it should cancel the stimulus artefact by eliminating the linear portion of the response, thus leaving behind the nonlinear cochlear emission. So, whilst much consideration has been given to determining which of the response parameters of nonlinear TEOAEs, such as waveform reproducibility or response level, best identify the hearing impaired, little research has investigated the effects of altering the stimulus parameters, such as click rate, on the evaluation of TEOAEs from patients with hearing loss. That is, nearly all previous studies involving hearing loss patients utilize TEOAEs recorded using conventional averaging at low stimulation rates. The MLS technique enables TEOAEs to be recorded at high stimulation rates where the time between click stimuli is less than the duration of the response (e.g. Thornton, 1993; Thornton et al., 1994). Studies have shown that increasing the stimulus presentation rate from the conventional value of 40 clicks/s up to 5000 clicks/s leads to a reduction in emission amplitude that reaches an approximate asymptote at about 1500 clicks/s (Hine and Thornton, 1997). Despite this decrease in amplitude, the very large number of responses obtainable at high rates means that both the signal-to-noise ratio (SNR) and the detection sensitivity increase as the click rate increases. The question arises whether this improvement in detection sensitivity for MLS emissions in normally hearing subjects translates into a change in the hearing loss levels capable of giving repeatable emissions in patients. The aim of the current study was to compare conventional and MLS TEOAEs in normal and hearing-impaired subjects. METHODS Subjects One hundred and twenty eight ears from 76 hearing impaired volunteers were tested in the study. The patients were selected to cover a range of sensorineural hearing losses (SNHL) from mild to profound arising due to various etiologies. All SNHL ears had pure tone air-conduction and bone-conduction thresholds > 20 dB HL for at least three of the octave frequencies between 250 to 8000Hz, with an air-bone gap less than 10 dB. They did not have any indication of retrocochlear lesions from their case histories and all ears were required to give normal otoscopic and tympanometric results to exclude any middle or external ear disease. Data from 19 normally hearing ears of 12 volunteers were also included in the study. All normals were required to have audiogram < 20dB HL across all frequencies as well as normal otoscopic and typanometric results. Recording of TEOAEs TEOAEs using click stimuli were recorded both conventionally and using the MLS technique. Conventional averaging was carried out at 40 clicks/s and MLS averaging was performed at 5000 clicks/s. For both rates, at two stimulus levels (76 and 66 dB peSPL intrameatal level), responses were averaged for 50 s (2000 clicks at a rate of 40 /s, 60000 clicks at a rate of 5000 /s). Two runs were performed at each rate and level. No judgement was made about the presence or absence of an emission at the data collection stage. Off-line data analysis All the directly recorded 'linear' TEOAE waveforms were visually inspected off-line before further analysis to check for the presence of obvious artefacts. For the conventional and MLS TEOAEs two types of nonlinear TEOAEs were then derived from the linear waveforms. The level nonlinear (LNL) responses were calculated by a subtraction technique from two linear waveforms evoked by clicks 10 dB apart (i.e. the measurement made at the higher click level was rescaled according to the difference in stimulus level and then subtracted from the lower measurement). In contrast, the rate nonlinear (RNL) responses were obtained by subtracting a linear emission recorded at 5000 clicks/s from one recorded at 40 clicks/s, both at a fixed stimulus level. As pointed out by Rasmussen et al. (1998), when they first derived RNL components, it is the rapid suppression of emission amplitude as the stimulus rate increases that is the basis for this nonlinear response which also features stimulus cancellation. Presenting clicks at different rates, yet for the same test time, inevitably means that high rate recordings contain much less noise, simply because many more clicks are presented. Correcting the response level by taking account of the noise allows a fair comparison between the MLS and conventional responses. The average RMS of the repeat waveforms contains both signal and noise. In order to obtain a better estimate of the signal alone, an estimate of the noise is removed. The method used is the same as that of the ILO Otoacoustic Emissions Analyzer where the noise is estimated by taking the RMS of the difference between repeat waveforms. Subtracting this noise estimate from the average of the repeat waveforms then gives a corrected response level. All values presented are those of corrected response level, expressed as dB SPL. RESULTS Repeat waveforms at each level (66 & 76 dB peSPL) and both rates (40 & 5000 clicks/s) were successfully recorded from 111 (out of 128) SNHL ears. Data were also available from the 19 normally hearing ears. In these selected results, only data from the 9 - 13 ms section of the waveforms have been considered. Directly recorded or 'linear' waveforms Figure 1 shows some typical linear waveforms recorded at 40 and 5000 clicks/s from one normally hearing ear (top pair of traces), seven SNHL ears and a 2cc cavity (bottom pair of traces). Each ear's audiogram is also shown to display the typical range of hearing losses, from mild to profound, under test in the study. Replicates are shown for each condition and, clearly, for all 8 ears, the waveforms are highly reproducible both within and between the two stimulus rates.  
Figure 2 illustrates just how correlated the linear waveforms are at 40 (A) and 5000 (B) clicks/s for all the SNHL patients and the normally hearing subjects. Waveform reproducibility has been plotted against the best audiogram threshold. At both rates many SNHL ears produce highly reproducible waveforms and, in fact, at 5000 clicks/s the vast majority produce waveforms as highly correlated as those recorded from normals. It is obvious, therefore, that reproducibility alone cannot be used to separate the normal from hearing loss ears. Figure 2 also shows the amplitude of the response plotted against best threshold for the 40 clicks/s (C) and 5000 clicks/s (D) data. Conventional recordings at 40 clicks/s yield only 3 SNHL ears whose linear response amplitude was in the range produced by 95% of the normals. These 3 ears all had their best threshold at 1000 Hz and two of them belonged to patient 39 who had a matched loss in both ears and whose left audiogram is shown in figure 1. In contrast, MLS recordings at 5000 clicks/s show a big overlap in response amplitudes between the SNHL patients and normally hearing subjects that includes results from patients with profound hearing losses, such as the two shown in figure 1 (ear 5R and 6R).  
Characteristically, the amplitude of TEOAEs recorded from normally hearing subjects decreases significantly as the stimulus rate is increased from 40 to 5000 clicks/s. An example of this can be seen clearly on the top two waveforms of figure 1. In patients with a SNHL however, this expected decrease in amplitude of TEOAE with rate is much less evident. In fact, in many cases there appeared to be little or no rate related reduction in amplitude (see for example ear 5R & 6R on figure 1). Consequently, the 'rate effect', defined here as the decibel reduction in response amplitude between 40 and 5000 clicks/s, is much reduced in the hearing loss patients. Figure 3 shows that the rate effect from normal ears (triangles) was greater than 7 dB for 95% of the time. In contrast, only 5 ears from the SNHL group exhibited a rate effect of normal magnitude.  
'Rate derived nonlinear (RNL) waveforms The linear data above (Figure 3) showed a reduced rate effect in the majority of SNHL compared to normal ears. In addition to describing the rate effect in terms of a single value calculated from a response at 40 and one at 5000 clicks/s, it is also possible to derive RNL waveforms that can be compared to both the linear and the LNL waveforms. RNL waveforms obtained from the same ears shown in figure 1 and 4 are shown in Figure 6.  Figure 6: Rate nonlinear (RNL) waveforms from the same ears shown in figure 1 & 5, i.e., one normally hearing ear and three SNHL ears. Replicate waveforms recorded using a 76 dB peSPL stimulus are shown for each condition. As seen for the linear and LNL data, Figure 7A shows that RNL waveform reproducibility alone could not be used to successfully separate normally hearing and hearing impaired subjects since many ears from the SNHL group gave waveforms as reproducible as those of normals. Once again a much better tool is response level. Figure 7B shows that only 2 SNHL ears (both belonging to patient 39) gave a response level the same as the normals.   Percentage of ears correctly classified For each method of analysis, the focus so far has been on the actual numbers of ears from the SNHL group (defined at study outset as ears who have an audiogram with at least three pure tone frequencies > 20dB HL) that gave evoked OAE responses within the range produced by the normal group i.e. those that, according to their audiogram are hearing impaired but who produced some normal evoked OAE responses. Alternatively, the results can be presented in terms of the percentage of ears from the SNHL group that were correctly classified as hearing impaired using their evoked OAE responses. This is shown in table 1 where the criterion used to classify the responses as abnormal was any data (either repeat waveform reproducibility or corrected response level) that fell below the 5th percentile of the normal group. The response levels of the RNL waveforms, the rate effect and the conventional linear and LNL responses all successfully classify more than 95% of the SNHL group as having abnormal evoked OAE responses.  DISCUSSION In agreement with previous research, the present study found that conventionally recorded TEOAEs can be used to separate normally hearing and hearing impaired subjects using either the late components of linear waveforms or derived nonlinear waveforms. For both these types of conventional TEOAEs, no more than 4 ears from the SNHL group (3.5%) gave waveforms of amplitude that fell within that produced by 95% of the normal group. All 4 ears had a best threshold between 500 to 4000 Hz of < 40 dB HL. This finding is similar to that of Collet et al. (1993) in their study of 931 SNHL ears. Conventionally recorded nonlinear TEOAEs were absent when best hearing frequency between 250 to 8000 Hz was above 40 dB HL. In contrast, the current results suggest that, when recorded for the same time, MLS TEOAEs are not as successful as conventional TEOAEs at separating normal and impaired ears largely because of the greater overlap between the amplitude of the MLS emissions from SNHL and normal ears. However, this is not the case if the results of the MLS and conventional recordings are combined, either through using the rate effect or by deriving RNL waveforms. When compared to the results for both conventional and MLS LNL waveforms, as well as the linear data or the rate effect, the present study found that it is the RNL results that best reflect the patients' hearing loss. For the 9 - 13 ms section of the RNL waveforms, only 2 ears from the SNHL group (2%) produced responses of comparable amplitude to the normal group (figure 7). Both ears belonged to patient 39 whose best threshold was 30 dB HL at 1000Hz and who therefore could be expected to produce genuine TEOAEs. Indeed, this patient was the only one that gave normal results on every analysis method used. Whilst the RNL responses were the most successful at correctly identifying the ears from the SNHL group as impaired, the conventional linear and LNL responses as well as the rate effect performed nearly as well. However, the RNL method can be viewed as having a slight advantage over each of the other methods. Compared to linear recordings, which only a few researchers favour for screening because of the possibility of stimulus artefacts affecting the response, RNL waveforms are derived in a way that results in stimulus cancellation. Also, unlike LNL waveforms, RNL waveforms are derived from two linear recordings both taken at the same, high, stimulus level. This inevitably means that the RNL derivation is done on recordings with a higher SNR. Given that the measurement of any TEOAEs is always constrained by the amount of noise in the recording session, a method that requires only one stimulus level should be an advantage. Lastly, compared to relying on a single value for the rate effect, the RNL responses are waveforms that can undergo a range of analyses. The current study has shown that, in linear mode, reproducible, TEOAE-like waveforms lasting the full 17 ms recording window can be obtained both conventionally and using the MLS technique from the majority (> 75%) of SNHL ears, even those with a severe or profound best threshold. Many of these responses are seen to behave linearly in that they produce neither rate nor level nonlinear waveforms (the derivation of both types of nonlinear waveforms should result in cancellation of linear components). The cause of these highly reproducible TEOAE-like waveforms recorded in linear mode from so many hearing loss patients (Figure 2A & B) is unclear but there are several possible contributing factors. Click artefact due to ringing in the ear canal and middle ear could play a part but, when probe fit is good, it is unlikely that this would affect the whole response window. In addition, ringing should be readily visualised as a decaying oscillation rather than the TEOAE-like waveforms seen here (figure 1). It is also possible that some of the responses were caused by residual outer hair cell activity, even in the presence of inner hair cell dysfunction or nerve damage. It could also be that the cochlear damage itself somehow manages to act as a reflection site capable of producing waveforms that resemble TEOAEs. Whatever the cause of the TEOAE-like waveforms recorded in linear mode, in excess of 29% of SNHL ears then gave reproducible (repeat waveform correlation > 0.5) level or rate derived nonlinear waveforms. It is only when the amplitudes of the nonlinear responses are compared to those from normally hearing subjects that a clear difference can be seen for ears in the SNHL group. This highlights the finding that reproducibility alone cannot be used to reliably indicate a hearing loss. REFERENCES Collet, L., Levy, V., Veuillet, E., Truy, E., Morgon, A., 1993. Click-evoked otoacoustic emissions and hearing threshold in sensorineural hearing loss. Ear Hear. 14, 141-143. Harris, F.P., Probst, R., 2002. Otoacoustic emissions and audiometric outcomes. In: Robinette, M., Glattke, T.J. (Eds), Otoacoustic Emissions - Clinical Applications (2nd Ed). Thieme Medical Press, New York 213 - 241. Hine, J.E., Thornton, A.R.D., 1997. Transient evoked otoacoustic emissions recorded using maximum length sequences as a function of stimulus rate and level. Ear Hear. 18, 121-128. Rasmussen, A.N., Osterhammel, P.A., Johannesen, P.T., Borgkvist, B., 1998. Neonatal hearing screening using otoacoustic emissions elicited by maximum length sequences. Br. J. Audiol. 32, 355-366. Thornton, A.R.D., 1993. High rate otoacoustic emissions. J. Acoust. Soc. Am. 94, 132-136. Thornton, A.R.D., Folkard, T.J., Chambers, J.D., 1994. Technical aspects of recording evoked otoacoustic emissions using maximum length sequences. Scand. Audiol. 23, 225-231. About the Author Since completing her doctorate at Oxford University in auditory physiology, Dr. Jemma Hine has worked as a clinical scientist at the Medical Research Council's Institute of Hearing Research (IHR) in Southampton where she is also an honorary research fellow in Clinical Neurosciences at the University of Southampton. Her recent research has focused on the potential of otoacoustic emissions recorded using the maximum length sequence technique. |
 |
 |