- Email: [email protected]
The use of high stimulus rate auditory brainstem responses in the estimation of hearing threshold
Hearing Research 123 (1998) 201^205
The use of high stimulus rate auditory brainstem responses in the estimation of hearing threshold Sze-mun Leung a , Antoinette Slaven b; *, A. Roger D. Thornton b , Graham J. Brickley a
a
Hearing and Balance Centre, Institute of Sound and Vibration Research, University of Southampton, High¢eld, Southampton SO17 1BJ, UK MRC Institute of Hearing Research, Royal South Hants Hospital, Brinton's Terrace, O¡ St Mary's Road, Southampton SO14 0YG, UK
b
Received 10 January 1998; revised 6 June 1998; accepted 19 June 1998
Abstract This normative study investigates the efficiency of using the maximum length sequence (MLS) technique applied to auditory brainstem evoked response (ABR) testing to estimate hearing thresholds. Using a commercially available system, ABRs were recorded in sixteen subjects at two conventional rates ^ 9.1 and 33.3 clicks/s ^ and six MLS rates between 88.8 and 1000 clicks/s. Each subject was tested at five stimulus levels from 60 down to 10 dBnHL. The wave JV amplitude input-output (I/O) functions, relative signal to noise ratio (SNR) and speed of test were calculated for all conditions. The JV amplitude and detectability decrease as the stimulus rate increases and level decreases. The latency of JV increases as the stimulus rate increases and the intensity decreases. While the slope of the amplitude I/O function was maximal at 200 clicks/s, at 300 clicks/s it was comparable with that obtained at conventional rates. At higher rates, the slope of the I/O function decreases. When compared with the conventional recording rate of 33.3 clicks/s there is a small improvement in SNR for MLS rates between 200 and 600 clicks/s at levels above 30 dBnHL. The calculated speed improvement at 300 clicks/s is a factor between 1.4 to 1.6 at a screening level of 30^40 dBnHL. It is felt therefore that there may be a small advantage to using MLS in screening and that the optimal rate for this lies at around 200 to 300 clicks/s. However even at these rates, as a consequence of the adaptation of the response with both rate and level, the improvement in SNR or speed of test would be modest when estimating threshold. z 1998 Elsevier Science B.V. All rights reserved. Key words: Auditory brainstem response; Maximum length sequence; Threshold estimation
1. Introduction One important clinical use of ABR is in the estimation of hearing threshold in individuals from whom it is not possible to obtain, for whatever reason, accurate thresholds by other means. When click stimuli are employed, the hearing threshold in the 2^4 kHz region can be estimated to within 10 to 15 dB of behavioural measures. However, the conventional method of ABR testing can be relatively time consuming and one way of overcoming this is to increase the stimulus rate. Whilst the amplitude of the ABR shows adaptation with increasing stimulus rate (Picton et al., 1992; Thornton * Corresponding author. Tel.: +44 (1703) 637946; Fax: +44 (1703) 825611; E-mail: [email protected]
and Slaven, 1993 ; Thornton et al., 1994 ; Lina-Granade et al., 1994), it is likely that at rates beyond those attainable by conventional means, there still lies the possibility of reducing test time. The maximum length sequence (MLS) technique is a mathematical method which allows extraction of evoked responses at stimulation rates at which the stimulus and response overlap, and at which it would not be possible with conventional averaging to extract the original waveform. The principles, advantages and disadvantages of using the MLS technique to record the ABR have been discussed by various authors (Eysholdt and Schreiner, 1982 ; Li et al., 1988 ; Burkard et al., 1990 ; Chan et al., 1992; Marsh, 1992; Picton et al., 1992 ; Thornton and Slaven, 1993 ; Thornton et al., 1994 ; Lina-Granade et al., 1994). In their audiological
0378-5955 / 98 / $19.00 ß 1998 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 5 9 5 5 ( 9 8 ) 0 0 1 1 4 - 2
HEARES 3078 24-8-98
S. Leung et al. / Hearing Research 123 (1998) 201^205
202
manifestation, MLS stimuli are trains of clicks and silences. The length of an MLS is 2n 31 where n is the order of the MLS and the number of clicks within that MLS is 2n31 . Thus an order 4 MLS comprises 15 `click opportunities', order 5 comprises 31 `click opportunities', order 6 comprises 63 and so on. Approximately half of those `click opportunities' would be taken by clicks and the remainder would be silences. So, an order 4 MLS would comprise eight clicks and seven silences. The MLS click rates referred to throughout this paper represent the maximum rate occurring, i.e. the reciprocal of the minimum time between clicks. The average rate is approximately half of the maximum rate for the orders used. The aim of this study was to establish if indeed there are advantages to using the MLS method in threshold estimation and at what stimulation rate most bene¢t is to be gained. In order that threshold can be determined accurately, it is necessary that the transition between presence and absence of a response is clear cut. It is therefore preferable that the JV amplitude input/output (I/O) function is as steep as possible. In addition, the signal to noise ratio (SNR) and speed of test relative to the conventional conditions will be calculated. 2. Materials and methods The right ears of 16 normally hearing female subjects, aged 21 to 29 years (mean = 24.4; S.D. = 2.4) were tested in this study. All subjects underwent otoscopic examination and tympanometry (Grason-Stadler GSI-33). Normal hearing ( 6 20 dB HL) was con¢rmed by pure-tone audiometry (Kamplex KC50) at octave intervals between 250 and 8000 Hz. Both conventional and MLS ABR were recorded using a Nicolet Spirit1 system. The stimuli were rarefaction clicks delivered through a TDH 49P headphone. Contralateral masking was provided by a Kamplex KC50 clinical audiometer at 10 dBSL through a TDH 39P headphone for all test conditions. A recording win-
dow of 12 ms was used and the responses were bandpass ¢ltered from 30 to 3000 Hz. Electrodes were placed on the vertex (active), the mastoid ipsilateral to the stimulus (reference) and the forehead (ground). Two responses were recorded for each condition. Averaging was carried out at conventional rates of 9.1 and 33.3 clicks/s and using MLS rates of 88.8, 200, 300, 400, 600 and 1000 clicks/s. These values represent the maximum stimulation rate occurring within an MLS which unlike the average rate is not a¡ected by the order of the MLS. Once the window and stimulus rate have been selected, the Nicolet Spirit uses an algorithm (described in the operators manual) to select the order of the MLS. The orders for the rates used here were : 88.8 clicks/s order 4; 200 clicks/s order 5; 300 and 400 clicks/s order 6; 600 and 1000 clicks/s order 7. The clicks were presented at levels of 60, 40, 30, 20 and 10 dBnHL. The order in which the rates were presented was balanced and randomised. Recordings were obtained at all stimulus levels, presented in descending order (60 down to 10 dBnHL), before moving on to the next rate. A test time of 100 s was used for each recording. This meant that 910 clicks were averaged at 9.1 clicks/s ; 3330 at 33.3 clicks/s ; 592 MLS sequences at 88.8 clicks/s; 645 sequences at 200 clicks/s; 476 sequences at 300 clicks/s; 635 sequences at 400 clicks/s; 472 sequences at 600 clicks/s and 787 at 1000 clicks/s. To ascertain whether a response was present for each recording pair, the data were examined by three independent observers. Wave V was accepted as present if two out of the three observers agreed. The two repeat waveforms were digitally ¢ltered o¡-line (band-pass ¢ltered from 100 to 1500 Hz) to reduce noise and then they were averaged. The amplitude and latency of JV were measured using the cursors and readout of the Nicolet Spirit. The amplitude was calculated as the difference between the JV peak and the most extreme data point occurring within 1.49 ms after the peak (Hall, 1992). The experiment was performed in accordance with the guidelines of the Declaration of Helsinki.
Table 1 The rate of wave JV detection at each level for conventional and MLS stimulation rates Rate (clicks/s)
60 dBnHL
40 dBnHL
30 dBnHL
20 dBnHL
10 dBnHL
9.1a 33.3a 88.8 199.9 299.9 400.1 599.4 999.5
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 15 13
15 16 14 16 15 14 13 11
9 11 12 9 9 9 6 5
a
Conventional rates.
HEARES 3078 24-8-98
S. Leung et al. / Hearing Research 123 (1998) 201^205
203
Fig. 1. Sample waveforms obtained for conventional rates of 9.1 and 33.3 clicks/s and MLS rates of 200 and 1000 clicks/s at a stimulus level of 60 dBnHL.
3. Results Fig. 1 shows a sample of waveforms obtained at a stimulus level of 60 dBnHL for the conventional rates of 9.1 and 33.3 clicks/s and the MLS rates of 200 and 1000 clicks/s. Wave JV is clearly identi¢able and can be seen to increase in latency and decrease in amplitude as the stimulus rate increases. Table 1 shows the number of subjects with an identi¢able wave JV at each stimulus rate and level. At the highest stimulation levels, JV was detected in all subjects at all rates. As the stimulus level decreases and the rate increases, so the JV detection rate falls. Fig. 2 shows the mean wave JV latency versus rate at each stimulus level. The latency of JV increases with increasing stimulus rate, this change being greater at the higher stimulus levels. Fig. 3A shows the mean JV amplitude versus rate at each stimulus level. The amplitude of JV decreases as the stimulus rate increases with the change being greater at higher stimulus levels. A 2-way ANOVA showed a signi¢cant e¡ect of both rate (F = 69.80, df = 4, P 6 0.001)
Fig. 2. Wave JV latency by rate for both conventional and MLS stimulation rates.
Fig. 3. A: Wave JV amplitude by rate for both conventional and MLS stimulation rates. B: Wave JV amplitude I/O function for conventional and MLS stimulation rates.
and level (F = 108.82, df = 7, P 6 0.001) with signi¢cant interaction (F = 1.619, df = 28, P 6 0.025). In Fig. 3B the same data are plotted as JV amplitude versus stimulus level at each rate. The decrease in amplitude between 60 and 40 dBnHL is less steep than that from 40 dBnHL downwards, at all rates. In this latter part, the I/O function is steepest for the rate of 200 clicks/s (9.3 nV/dB). The slopes become shallower as the rate increases thereafter, with a change of 2.6 nV/dB at 1000 clicks/s (10 dB value omitted from regression line ¢t due to small number of ears at this point). The conventional rates of 9.1 and 33.3 clicks/s show slopes of 6.6 and 7.6 nV/dB respectively. The standard deviation of the mean is not shown in these ¢gures for the sake of clarity. The JV amplitude data can be used to calculate the SNR at each stimulation rate relative to a conventional rate of either 9.1 clicks/s or 33.3 clicks/s and at each stimulus level. The SNR was equal to cWkWk(r/r0 ) where c equals the proportion of clicks in the MLS (2b31 /2b 31), k is the JV amplitude at the MLS rate divided by the amplitude of JV recorded at the conventional rate, at the same stimulus level, r is the MLS rate and r0 is the conventional rate against which the comparison is being made. This equation is derived in full in Thornton and
HEARES 3078 24-8-98
204
S. Leung et al. / Hearing Research 123 (1998) 201^205
Slaven (1993). Fig. 4 shows the SNR calculated relative to the conventional rates of 9.1 and 33.3 clicks/s. When compared to recordings made at 9.1 clicks/s (Fig. 4A), a better SNR could be obtained for all MLS rates, at all stimulus levels. At 40 and 60 dBnHL an optimal SNR was obtained at 200 clicks/s, while at lower levels the optimum SNR was obtained at the slightly higher rate of 300 clicks/s. The picture is slightly di¡erent when the MLS responses are compared with conventional recordings made at 33.3 clicks/s (Fig. 4B). At 30, 40 and 60 dBnHL there is a small improvement in SNR at rates between 200 and 600 clicks/s, optimal at 200 and 300 clicks/s. However, at the lowest stimulus levels there is no advantage to be had, in terms of SNR, at any MLS rate. Fig. 5 shows the speed of test relative to the conventional rate of 33.3 clicks/s, calculated using the formula c2 Wk2 W(r/r0 ) (Thornton and Slaven, 1993). This is the speed with which the test can be carried out when averaging to the same SNR and takes into account the reduction in JV amplitude seen with increasing rate and decreasing stimulus level. At 300 clicks/s the relative speed of test is greater than unity at all stimulus levels. As the rate increases and the level decreases, so
Fig. 4. A: SNR relative to the conventional rate of 9.1 clicks/s. B: SNR relative to the conventional rate of 33.3 clicks/s.
Fig. 5. Speed of test relative to the conventional rate of 33.3 clicks/s at levels from 10 to 60 dBnHL.
the relative speed of test decreases, until by 1000 clicks/s testing to the same SNR takes longer than the conventional case at all stimulus levels. 4. Discussion The JV detection rate falls as the stimulus rate increases and level decreases. However, the detection rate is comparable to that seen conventionally at all stimulus levels until around 300 clicks/s. Using a sample of 40 premature newborns, Weber and Roush (1993) found a similar decrease in JV detection rate with increasing stimulus rate. They were surprised however that the percentage of babies with a detectable wave JV was higher at an MLS rate of 227.3 clicks/s than at a conventional rate of 33.3 clicks/s. This was also true for waves JI and JIII. Since JV is usually considered the most robust of the ABR components, it would be expected that the earlier components would fare worse. However, in a later paper using a larger group of babies (50) they were unable to repeat this ¢nding and suggest that MLS does not provide a consistent improvement over traditional ABR screening and therefore that its use is not warranted (Weber and Roush, 1995). They do conclude however that MLS can o¡er advantages when the conventional response is poorly de¢ned as a consequence of high levels of either baby or environment related noise. When compared with a conventional rate of 9.1 clicks/s, the SNR was greater than unity for all stimulation rates and levels and maximal at 200 to 300 clicks/s. This con¢rms the ¢ndings of Thornton and Slaven (1993) who using a stimulus level of 80 dBnHL, compared their MLS data with that obtained at a conventional rate of 9.1 clicks/s. However, when comparison is made with a conventional rate more usual in threshold estimation (33.3 clicks/s), the MLS technique fares less well. The SNR is higher at
HEARES 3078 24-8-98
S. Leung et al. / Hearing Research 123 (1998) 201^205
low MLS rates since the rate related reduction in amplitude is less here. The best SNRs were obtained across the range of levels at 300 clicks/s. The speed of test at the MLS rates relative to the conventional condition is only greater than unity for all stimulus levels at 300 clicks/s. The advantage is greatest at the highest stimulus levels and minimal at 10^20 dBnHL. In the clinical situation it is common practice to average fewer responses at the higher stimulus levels where any response would be clear cut and more at the lower levels where the response would be less well de¢ned. Any potential bene¢t in terms of test time saved would therefore be modest but since normal hearing as de¢ned by ABR is indicated by the presence of a response at 30 dBnHL, there would be some advantage. Wave JV amplitude decreases with increasing stimulus rate and decreasing stimulus level. The slope of the amplitude I/O function, measured between 40 and 10 dBnHL because of non-linearity at higher levels, was steepest at a rate of 200 clicks/s. At 300 clicks/s the slope was 7.0 nV/dB intermediate between the conventional rates of 9.1 (6.6 nV/dB) and 33.3 clicks/s (7.6 nV/dB). When estimating threshold it is important that the cut-o¡ between presence and absence of the wave is clear and hence it is desirable that the I/O function is steep. On those terms, the MLS rate of 200 clicks/s would appear optimal although it would still be acceptable at 300 clicks/s. There is some argument in the literature as to the e¡ect of higher stimulation rates on this function. Lasky et al. (1993) compared conventional ABR with MLS ABR in ten normally hearing adult subjects. They reported that the slopes of the I/O functions for both JV latency and amplitude were less steep for MLS ABR than for the conventional response. However, Lina-Granade et al. (1994) tested 20 normal subjects and found no signi¢cant di¡erence in either I/O function. 5. Conclusions Although when compared with a conventional response recorded at 9.1 clicks/s the MLS rates appear to o¡er modest improvement in SNR at all stimulus levels, the picture is less favourable when compared to
205
the more usual rate of 33.3 clicks/s. The detection rates, taken in combination with the results of the calculated improvements in SNR and speed, suggest that MLS ABR would not be particularly suitable for threshold estimation but that there may be some advantages when using the technique for screening purposes. The optimum rate would be about 300 clicks/s when a screening level of 30^40 dBnHL was to be used. This would allow the test to be carried out 1.4 to 1.6 times faster. References Burkard, R., Shi, Y., Hecox, K.E., 1990. A comparison of Maximum Length and Legendre Sequences for the derivation of brain-stem auditory-evoked responses at rapid rates of stimulation. J. Acoust. Soc. Am. 87, 1656^1664. Chan, F.H.Y., Lam, F.K., Poon, P.W.F., Du, M.H., 1992. Measurement of human BAERs by the maximum length sequence technique. Med. Biol. Eng. Comp. 30, 32^40. Eysholdt, U., Schreiner, C., 1982. Maximum Length Sequences - a fast method for measuring brain-stem-evoked responses. Audiology 21, 242^250. Hall, J.W. (1992) Handbook of Auditory Evoked Responses. Allyn and Bacon, Newton, MA, p. 302. Lasky, R.E., Perlman, J., Hecox, K., 1993. Maximum length sequence auditory evoked brainstem responses in human newborns and adults. J. Am. Acad. Audiol. 3, 383^389. Li, H.F., Chan, F.H.Y., Poon, P.W.F., Hwang, J.C., Chan, W.S., 1988. Maximum length sequences applied to the measurement of brainstem auditory evoked responses. J. Biomed. Eng. 10, 14^24. Lina-Granade, G., Collet, L., Morgon, A., 1994. Auditory-evoked brainstem responses elicited by maximum length sequences in normal and sensorineural ears. Audiology 33, 218^236. Marsh, R.R., 1992. Signal to noise constraints on maximum length sequence auditory brain stem responses. Ear Hear. 13, 396^400. Picton, T.W., Champagne, S.C., Kellett, A.J.C., 1992. Human auditory evoked potentials using maximum length sequences. Electroencephalogr. Clin. Neurophysiol. 84, 90^100. 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. Thornton, A.R.D., Slaven, A., 1993. Auditory brainstem responses recorded at fast stimulation rates using Maximum Length Sequences. Br. J. Audiol. 27, 205^210. Weber, B.A., Roush, P.A., 1993. Application of maximum length sequence analysis to auditory brainstem response testing of premature newborns. J. Am. Acad. Audiol. 4, 157^162. Weber, B.A., Roush, P.A., 1995. Use of maximum length sequence analysis in newborn hearing testing. J. Am. Acad. Audiol. 6, 187^ 190.
HEARES 3078 24-8-98
The use of high stimulus rate auditory brainstem responses in the estimation of hearing threshold Sze-mun Leung a , Antoinette Slaven b; *, A. Roger D. Thornton b , Graham J. Brickley a
a
Hearing and Balance Centre, Institute of Sound and Vibration Research, University of Southampton, High¢eld, Southampton SO17 1BJ, UK MRC Institute of Hearing Research, Royal South Hants Hospital, Brinton's Terrace, O¡ St Mary's Road, Southampton SO14 0YG, UK
b
Received 10 January 1998; revised 6 June 1998; accepted 19 June 1998
Abstract This normative study investigates the efficiency of using the maximum length sequence (MLS) technique applied to auditory brainstem evoked response (ABR) testing to estimate hearing thresholds. Using a commercially available system, ABRs were recorded in sixteen subjects at two conventional rates ^ 9.1 and 33.3 clicks/s ^ and six MLS rates between 88.8 and 1000 clicks/s. Each subject was tested at five stimulus levels from 60 down to 10 dBnHL. The wave JV amplitude input-output (I/O) functions, relative signal to noise ratio (SNR) and speed of test were calculated for all conditions. The JV amplitude and detectability decrease as the stimulus rate increases and level decreases. The latency of JV increases as the stimulus rate increases and the intensity decreases. While the slope of the amplitude I/O function was maximal at 200 clicks/s, at 300 clicks/s it was comparable with that obtained at conventional rates. At higher rates, the slope of the I/O function decreases. When compared with the conventional recording rate of 33.3 clicks/s there is a small improvement in SNR for MLS rates between 200 and 600 clicks/s at levels above 30 dBnHL. The calculated speed improvement at 300 clicks/s is a factor between 1.4 to 1.6 at a screening level of 30^40 dBnHL. It is felt therefore that there may be a small advantage to using MLS in screening and that the optimal rate for this lies at around 200 to 300 clicks/s. However even at these rates, as a consequence of the adaptation of the response with both rate and level, the improvement in SNR or speed of test would be modest when estimating threshold. z 1998 Elsevier Science B.V. All rights reserved. Key words: Auditory brainstem response; Maximum length sequence; Threshold estimation
1. Introduction One important clinical use of ABR is in the estimation of hearing threshold in individuals from whom it is not possible to obtain, for whatever reason, accurate thresholds by other means. When click stimuli are employed, the hearing threshold in the 2^4 kHz region can be estimated to within 10 to 15 dB of behavioural measures. However, the conventional method of ABR testing can be relatively time consuming and one way of overcoming this is to increase the stimulus rate. Whilst the amplitude of the ABR shows adaptation with increasing stimulus rate (Picton et al., 1992; Thornton * Corresponding author. Tel.: +44 (1703) 637946; Fax: +44 (1703) 825611; E-mail: [email protected]
and Slaven, 1993 ; Thornton et al., 1994 ; Lina-Granade et al., 1994), it is likely that at rates beyond those attainable by conventional means, there still lies the possibility of reducing test time. The maximum length sequence (MLS) technique is a mathematical method which allows extraction of evoked responses at stimulation rates at which the stimulus and response overlap, and at which it would not be possible with conventional averaging to extract the original waveform. The principles, advantages and disadvantages of using the MLS technique to record the ABR have been discussed by various authors (Eysholdt and Schreiner, 1982 ; Li et al., 1988 ; Burkard et al., 1990 ; Chan et al., 1992; Marsh, 1992; Picton et al., 1992 ; Thornton and Slaven, 1993 ; Thornton et al., 1994 ; Lina-Granade et al., 1994). In their audiological
0378-5955 / 98 / $19.00 ß 1998 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 5 9 5 5 ( 9 8 ) 0 0 1 1 4 - 2
HEARES 3078 24-8-98
S. Leung et al. / Hearing Research 123 (1998) 201^205
202
manifestation, MLS stimuli are trains of clicks and silences. The length of an MLS is 2n 31 where n is the order of the MLS and the number of clicks within that MLS is 2n31 . Thus an order 4 MLS comprises 15 `click opportunities', order 5 comprises 31 `click opportunities', order 6 comprises 63 and so on. Approximately half of those `click opportunities' would be taken by clicks and the remainder would be silences. So, an order 4 MLS would comprise eight clicks and seven silences. The MLS click rates referred to throughout this paper represent the maximum rate occurring, i.e. the reciprocal of the minimum time between clicks. The average rate is approximately half of the maximum rate for the orders used. The aim of this study was to establish if indeed there are advantages to using the MLS method in threshold estimation and at what stimulation rate most bene¢t is to be gained. In order that threshold can be determined accurately, it is necessary that the transition between presence and absence of a response is clear cut. It is therefore preferable that the JV amplitude input/output (I/O) function is as steep as possible. In addition, the signal to noise ratio (SNR) and speed of test relative to the conventional conditions will be calculated. 2. Materials and methods The right ears of 16 normally hearing female subjects, aged 21 to 29 years (mean = 24.4; S.D. = 2.4) were tested in this study. All subjects underwent otoscopic examination and tympanometry (Grason-Stadler GSI-33). Normal hearing ( 6 20 dB HL) was con¢rmed by pure-tone audiometry (Kamplex KC50) at octave intervals between 250 and 8000 Hz. Both conventional and MLS ABR were recorded using a Nicolet Spirit1 system. The stimuli were rarefaction clicks delivered through a TDH 49P headphone. Contralateral masking was provided by a Kamplex KC50 clinical audiometer at 10 dBSL through a TDH 39P headphone for all test conditions. A recording win-
dow of 12 ms was used and the responses were bandpass ¢ltered from 30 to 3000 Hz. Electrodes were placed on the vertex (active), the mastoid ipsilateral to the stimulus (reference) and the forehead (ground). Two responses were recorded for each condition. Averaging was carried out at conventional rates of 9.1 and 33.3 clicks/s and using MLS rates of 88.8, 200, 300, 400, 600 and 1000 clicks/s. These values represent the maximum stimulation rate occurring within an MLS which unlike the average rate is not a¡ected by the order of the MLS. Once the window and stimulus rate have been selected, the Nicolet Spirit uses an algorithm (described in the operators manual) to select the order of the MLS. The orders for the rates used here were : 88.8 clicks/s order 4; 200 clicks/s order 5; 300 and 400 clicks/s order 6; 600 and 1000 clicks/s order 7. The clicks were presented at levels of 60, 40, 30, 20 and 10 dBnHL. The order in which the rates were presented was balanced and randomised. Recordings were obtained at all stimulus levels, presented in descending order (60 down to 10 dBnHL), before moving on to the next rate. A test time of 100 s was used for each recording. This meant that 910 clicks were averaged at 9.1 clicks/s ; 3330 at 33.3 clicks/s ; 592 MLS sequences at 88.8 clicks/s; 645 sequences at 200 clicks/s; 476 sequences at 300 clicks/s; 635 sequences at 400 clicks/s; 472 sequences at 600 clicks/s and 787 at 1000 clicks/s. To ascertain whether a response was present for each recording pair, the data were examined by three independent observers. Wave V was accepted as present if two out of the three observers agreed. The two repeat waveforms were digitally ¢ltered o¡-line (band-pass ¢ltered from 100 to 1500 Hz) to reduce noise and then they were averaged. The amplitude and latency of JV were measured using the cursors and readout of the Nicolet Spirit. The amplitude was calculated as the difference between the JV peak and the most extreme data point occurring within 1.49 ms after the peak (Hall, 1992). The experiment was performed in accordance with the guidelines of the Declaration of Helsinki.
Table 1 The rate of wave JV detection at each level for conventional and MLS stimulation rates Rate (clicks/s)
60 dBnHL
40 dBnHL
30 dBnHL
20 dBnHL
10 dBnHL
9.1a 33.3a 88.8 199.9 299.9 400.1 599.4 999.5
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 15 13
15 16 14 16 15 14 13 11
9 11 12 9 9 9 6 5
a
Conventional rates.
HEARES 3078 24-8-98
S. Leung et al. / Hearing Research 123 (1998) 201^205
203
Fig. 1. Sample waveforms obtained for conventional rates of 9.1 and 33.3 clicks/s and MLS rates of 200 and 1000 clicks/s at a stimulus level of 60 dBnHL.
3. Results Fig. 1 shows a sample of waveforms obtained at a stimulus level of 60 dBnHL for the conventional rates of 9.1 and 33.3 clicks/s and the MLS rates of 200 and 1000 clicks/s. Wave JV is clearly identi¢able and can be seen to increase in latency and decrease in amplitude as the stimulus rate increases. Table 1 shows the number of subjects with an identi¢able wave JV at each stimulus rate and level. At the highest stimulation levels, JV was detected in all subjects at all rates. As the stimulus level decreases and the rate increases, so the JV detection rate falls. Fig. 2 shows the mean wave JV latency versus rate at each stimulus level. The latency of JV increases with increasing stimulus rate, this change being greater at the higher stimulus levels. Fig. 3A shows the mean JV amplitude versus rate at each stimulus level. The amplitude of JV decreases as the stimulus rate increases with the change being greater at higher stimulus levels. A 2-way ANOVA showed a signi¢cant e¡ect of both rate (F = 69.80, df = 4, P 6 0.001)
Fig. 2. Wave JV latency by rate for both conventional and MLS stimulation rates.
Fig. 3. A: Wave JV amplitude by rate for both conventional and MLS stimulation rates. B: Wave JV amplitude I/O function for conventional and MLS stimulation rates.
and level (F = 108.82, df = 7, P 6 0.001) with signi¢cant interaction (F = 1.619, df = 28, P 6 0.025). In Fig. 3B the same data are plotted as JV amplitude versus stimulus level at each rate. The decrease in amplitude between 60 and 40 dBnHL is less steep than that from 40 dBnHL downwards, at all rates. In this latter part, the I/O function is steepest for the rate of 200 clicks/s (9.3 nV/dB). The slopes become shallower as the rate increases thereafter, with a change of 2.6 nV/dB at 1000 clicks/s (10 dB value omitted from regression line ¢t due to small number of ears at this point). The conventional rates of 9.1 and 33.3 clicks/s show slopes of 6.6 and 7.6 nV/dB respectively. The standard deviation of the mean is not shown in these ¢gures for the sake of clarity. The JV amplitude data can be used to calculate the SNR at each stimulation rate relative to a conventional rate of either 9.1 clicks/s or 33.3 clicks/s and at each stimulus level. The SNR was equal to cWkWk(r/r0 ) where c equals the proportion of clicks in the MLS (2b31 /2b 31), k is the JV amplitude at the MLS rate divided by the amplitude of JV recorded at the conventional rate, at the same stimulus level, r is the MLS rate and r0 is the conventional rate against which the comparison is being made. This equation is derived in full in Thornton and
HEARES 3078 24-8-98
204
S. Leung et al. / Hearing Research 123 (1998) 201^205
Slaven (1993). Fig. 4 shows the SNR calculated relative to the conventional rates of 9.1 and 33.3 clicks/s. When compared to recordings made at 9.1 clicks/s (Fig. 4A), a better SNR could be obtained for all MLS rates, at all stimulus levels. At 40 and 60 dBnHL an optimal SNR was obtained at 200 clicks/s, while at lower levels the optimum SNR was obtained at the slightly higher rate of 300 clicks/s. The picture is slightly di¡erent when the MLS responses are compared with conventional recordings made at 33.3 clicks/s (Fig. 4B). At 30, 40 and 60 dBnHL there is a small improvement in SNR at rates between 200 and 600 clicks/s, optimal at 200 and 300 clicks/s. However, at the lowest stimulus levels there is no advantage to be had, in terms of SNR, at any MLS rate. Fig. 5 shows the speed of test relative to the conventional rate of 33.3 clicks/s, calculated using the formula c2 Wk2 W(r/r0 ) (Thornton and Slaven, 1993). This is the speed with which the test can be carried out when averaging to the same SNR and takes into account the reduction in JV amplitude seen with increasing rate and decreasing stimulus level. At 300 clicks/s the relative speed of test is greater than unity at all stimulus levels. As the rate increases and the level decreases, so
Fig. 4. A: SNR relative to the conventional rate of 9.1 clicks/s. B: SNR relative to the conventional rate of 33.3 clicks/s.
Fig. 5. Speed of test relative to the conventional rate of 33.3 clicks/s at levels from 10 to 60 dBnHL.
the relative speed of test decreases, until by 1000 clicks/s testing to the same SNR takes longer than the conventional case at all stimulus levels. 4. Discussion The JV detection rate falls as the stimulus rate increases and level decreases. However, the detection rate is comparable to that seen conventionally at all stimulus levels until around 300 clicks/s. Using a sample of 40 premature newborns, Weber and Roush (1993) found a similar decrease in JV detection rate with increasing stimulus rate. They were surprised however that the percentage of babies with a detectable wave JV was higher at an MLS rate of 227.3 clicks/s than at a conventional rate of 33.3 clicks/s. This was also true for waves JI and JIII. Since JV is usually considered the most robust of the ABR components, it would be expected that the earlier components would fare worse. However, in a later paper using a larger group of babies (50) they were unable to repeat this ¢nding and suggest that MLS does not provide a consistent improvement over traditional ABR screening and therefore that its use is not warranted (Weber and Roush, 1995). They do conclude however that MLS can o¡er advantages when the conventional response is poorly de¢ned as a consequence of high levels of either baby or environment related noise. When compared with a conventional rate of 9.1 clicks/s, the SNR was greater than unity for all stimulation rates and levels and maximal at 200 to 300 clicks/s. This con¢rms the ¢ndings of Thornton and Slaven (1993) who using a stimulus level of 80 dBnHL, compared their MLS data with that obtained at a conventional rate of 9.1 clicks/s. However, when comparison is made with a conventional rate more usual in threshold estimation (33.3 clicks/s), the MLS technique fares less well. The SNR is higher at
HEARES 3078 24-8-98
S. Leung et al. / Hearing Research 123 (1998) 201^205
low MLS rates since the rate related reduction in amplitude is less here. The best SNRs were obtained across the range of levels at 300 clicks/s. The speed of test at the MLS rates relative to the conventional condition is only greater than unity for all stimulus levels at 300 clicks/s. The advantage is greatest at the highest stimulus levels and minimal at 10^20 dBnHL. In the clinical situation it is common practice to average fewer responses at the higher stimulus levels where any response would be clear cut and more at the lower levels where the response would be less well de¢ned. Any potential bene¢t in terms of test time saved would therefore be modest but since normal hearing as de¢ned by ABR is indicated by the presence of a response at 30 dBnHL, there would be some advantage. Wave JV amplitude decreases with increasing stimulus rate and decreasing stimulus level. The slope of the amplitude I/O function, measured between 40 and 10 dBnHL because of non-linearity at higher levels, was steepest at a rate of 200 clicks/s. At 300 clicks/s the slope was 7.0 nV/dB intermediate between the conventional rates of 9.1 (6.6 nV/dB) and 33.3 clicks/s (7.6 nV/dB). When estimating threshold it is important that the cut-o¡ between presence and absence of the wave is clear and hence it is desirable that the I/O function is steep. On those terms, the MLS rate of 200 clicks/s would appear optimal although it would still be acceptable at 300 clicks/s. There is some argument in the literature as to the e¡ect of higher stimulation rates on this function. Lasky et al. (1993) compared conventional ABR with MLS ABR in ten normally hearing adult subjects. They reported that the slopes of the I/O functions for both JV latency and amplitude were less steep for MLS ABR than for the conventional response. However, Lina-Granade et al. (1994) tested 20 normal subjects and found no signi¢cant di¡erence in either I/O function. 5. Conclusions Although when compared with a conventional response recorded at 9.1 clicks/s the MLS rates appear to o¡er modest improvement in SNR at all stimulus levels, the picture is less favourable when compared to
205
the more usual rate of 33.3 clicks/s. The detection rates, taken in combination with the results of the calculated improvements in SNR and speed, suggest that MLS ABR would not be particularly suitable for threshold estimation but that there may be some advantages when using the technique for screening purposes. The optimum rate would be about 300 clicks/s when a screening level of 30^40 dBnHL was to be used. This would allow the test to be carried out 1.4 to 1.6 times faster. References Burkard, R., Shi, Y., Hecox, K.E., 1990. A comparison of Maximum Length and Legendre Sequences for the derivation of brain-stem auditory-evoked responses at rapid rates of stimulation. J. Acoust. Soc. Am. 87, 1656^1664. Chan, F.H.Y., Lam, F.K., Poon, P.W.F., Du, M.H., 1992. Measurement of human BAERs by the maximum length sequence technique. Med. Biol. Eng. Comp. 30, 32^40. Eysholdt, U., Schreiner, C., 1982. Maximum Length Sequences - a fast method for measuring brain-stem-evoked responses. Audiology 21, 242^250. Hall, J.W. (1992) Handbook of Auditory Evoked Responses. Allyn and Bacon, Newton, MA, p. 302. Lasky, R.E., Perlman, J., Hecox, K., 1993. Maximum length sequence auditory evoked brainstem responses in human newborns and adults. J. Am. Acad. Audiol. 3, 383^389. Li, H.F., Chan, F.H.Y., Poon, P.W.F., Hwang, J.C., Chan, W.S., 1988. Maximum length sequences applied to the measurement of brainstem auditory evoked responses. J. Biomed. Eng. 10, 14^24. Lina-Granade, G., Collet, L., Morgon, A., 1994. Auditory-evoked brainstem responses elicited by maximum length sequences in normal and sensorineural ears. Audiology 33, 218^236. Marsh, R.R., 1992. Signal to noise constraints on maximum length sequence auditory brain stem responses. Ear Hear. 13, 396^400. Picton, T.W., Champagne, S.C., Kellett, A.J.C., 1992. Human auditory evoked potentials using maximum length sequences. Electroencephalogr. Clin. Neurophysiol. 84, 90^100. 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. Thornton, A.R.D., Slaven, A., 1993. Auditory brainstem responses recorded at fast stimulation rates using Maximum Length Sequences. Br. J. Audiol. 27, 205^210. Weber, B.A., Roush, P.A., 1993. Application of maximum length sequence analysis to auditory brainstem response testing of premature newborns. J. Am. Acad. Audiol. 4, 157^162. Weber, B.A., Roush, P.A., 1995. Use of maximum length sequence analysis in newborn hearing testing. J. Am. Acad. Audiol. 6, 187^ 190.
HEARES 3078 24-8-98