International Journal of Audiology 2014; 53: 629–632
Technical Report
The effect of notch filtering on the waveform of the newborn auditory brainstem response Guy Lightfoot∗,†, Inga Ferm‡, Amanda Hall§ & Kathryn Evans# *Department
of Medical Physics & Clinical Engineering, Royal Liverpool University Hospital, Liverpool, UK, †Clinical Advisory Group, English Newborn Hearing Screening Programme, London, UK, ‡Audiology Department, Croydon University Hospital, Croydon, UK, §Children’s Hearing Centre, St Michael’s Hospital, Bristol, UK, and #Paediatric Audiology Department, Leighton Hospital, Crewe, UK
Abstract Objective: To identify whether the use of a notch filter significantly affects the morphology or characteristics of the newborn auditory brainstem response (ABR) waveform and so inform future guidance for clinical practice. Design: Waveforms with and without the application of a notch filter were recorded from babies undergoing routine ABR tests at 4000, 1000 and 500 Hz. Any change in response morphology was judged subjectively. Response latency, amplitude, and measurements of response quality and residual noise were noted. An ABR simulator was also used to assess the effect of notch filtering in conditions of low and high mains interference. Results: The use of a notch filter changed waveform morphology for 500 Hz stimuli only in 15% of tests in newborns. Residual noise was lower when 4000 Hz stimuli were used. Response latency, amplitude, and quality were unaffected regardless of stimulus frequency. Tests with the ABR stimulator suggest that these findings can be extended to conditions of high level mains interference. Conclusions: A notch filter should be avoided when testing at 500 Hz, but at higher frequencies appears to carry no penalty.
Key Words: Auditory brainstem response; notch filter, distortion
Recording the auditory brainstem response (ABR) presents a technological challenge; the electrodes record not only the sub-microvolt response but also record much larger, unwanted, signals including those generated by the patient (neurogenic and myogenic) and electrical interference. A major source of electrical interference originates from the power supply (mains), either through the instrument itself or via capacitive or inductive coupling between the patient or the electrode leads and the power supply. This signal (60 Hz in the Americas and usually 50 Hz elsewhere) is present on all parts of the patient but the ABR signal that we seek to record varies from point to point on the body. Careful placement of the electrodes and amplification of the difference in the detected voltages provides a large degree of immunity to signals which are common to both recording electrodes. These so-called common mode signals are therefore effectively rejected when differential amplification is used. Amplifiers are imperfect and their performance in rejecting common mode signals is described by their common mode rejection ratio
(CMRR). The value of CMRR can be high, typically 100 dB, if the contact impedances of the electrodes are low and similar but the CMRR drops dramatically if the electrode impedances are unequal. Under these circumstances, common mode signals such as mains interference are not rejected and becomes problematic. It is for this reason that testers must endeavour to obtain low and equal electrode impedances. To ensure that all electrode leads pick up the same mains interference it is good practice to gather or twist the leads together; if they are physically separated then the differing amounts of interference induced in the leads will be amplified and contaminate the recording. Mains interference is not usually an issue with well-designed equipment and good clinical practice but occasionally, if the electrode contact is not ideal or if the recording environment is electrically hostile, mains interference can be problematic. When excessive, it can exceed the artefact rejection level and no recording is possible. Adopting a more lax artefact rejection level would
Correspondence: Guy Lightfoot, Department of Medical Physics & Clinical Engineering, Royal Liverpool University Hospital, Prescot Street, Liverpool, L7 8XP, UK. E-mail: [email protected] (Received 21 October 2013; accepted 10 February 2014 ) ISSN 1499-2027 print/ISSN 1708-8186 online © 2014 British Society of Audiology, International Society of Audiology, and Nordic Audiological Society DOI: 10.3109/14992027.2014.894644
630
G. Lightfoot et al.
Method
Abbreviations ABR CMRR
Auditory brainstem response Common mode rejection ratio
allow the recording to proceed but at a high cost: the interference is allowed into the recording. Since the frequency of mains interference is known, one solution is to filter it out using a notch filter. This is a band-stop filter with a narrow and deep spectral notch at the mains frequency, rejecting only that frequency and passing everything else, as illustrated in the upper panel of Figure 1. Whilst this appears ideal, analogue notch filters have a non-linear phase response, shown in the lower panel of Figure 1. The consequences of this can be to change the shape, amplitude, and latencies of signals such as the ABR. A secondary issue is that mains-related interference can include harmonics of the mains frequency, which would not be attenuated by a mains notch filter. It is for these reasons that the normal expert advice is to avoid the use of notch filters, their primary application being limited to the identification of mains-frequency signals as the source of interference rather than a means of attenuating the interference (Hall, 1992; Thornton, 2007). The latency shifts associated with notch filters could pose a significant problem in the neurological ABR, where the measurement of absolute and inter-peak latencies is the primary means of analysis. However when the ABR is used for threshold estimation, waveform analysis is one of response detection, where signal to noise ratio is the issue. Although latency and amplitude are often recorded, minor changes would not invalidate the outcome. It is possible therefore that the use of a notch filter may be acceptable in threshold ABR testing, providing it creates no major change to response morphology, latency, or amplitude. The purpose of this study was to investigate this and so inform future clinical practice.
ABR test centres in Bristol, Liverpool, Crewe, Croydon, and Wirral, UK, participated in the collection of waveforms recorded from sleeping babies under the age of three months who were being tested as a routine component of their newborn hearing screening. ABR data were collected using 4000 Hz, 1000 Hz, and 500 Hz 5-cycle tone burst stimuli (most babies were tested at only one frequency) at a stimulus level in the range 35 to 75 dBnHL (Re: ISO 389-6). In all cases the stimulus level chosen was supra-threshold and evoked a clear response. Stimulus rates were 45.1/s for 4000 Hz (21 ms timebase) and 35.1/s for 1000 Hz & 500 Hz (26 ms timebase). Test parameters and methodology was in accordance with the English Newborn Hearing Screening Programme (NHSP) guidance for ABR testing in babies (2013). This included analogue filtering between 30 Hz and 1500 Hz (Butterworth response with ⫺ 12 dB/octave roll-off) with an artefact rejection level of ⫾ 5 μV, 2048 to 3072 sweeps, and no notch filter. Equipment was the Bio-logic Navigator Pro system (Natus Medical Incorporated, San Carlos, USA), software version 6.3.0 or later. Upon completion of testing, if time allowed and the baby was still sleeping, one of the stimulus levels previously used in the test to obtain a clear response was retested using the same number of sweeps but with the notch filter applied. In the Biologic system the notch filter is a programmable 50/60 Hz line filter with a Twin-T architecture and a quality factor (Q) of 10. The Q factor is the centre frequency (in this case 50 Hz) divided by the half-power bandwidth (in this case 5 Hz). Wave V latency, the wave V to SN10 amplitude, residual noise, and Fsp were recorded from the pair of waveforms with and without the notch filter. Fsp is a measurement of the quality of the recorded response, related to the response to noise variance ratio (Elberling & Don, 1984). Since clear ABRs had to be recorded without the notch filter, in none of the data was mains interference an issue. A total of 96 waveform pairs (32 at each frequency) were recorded from 65 babies.
Figure 1. The amplitude (upper panel) and phase (lower panel) of a typical notch filter.
Notch Filtering the threshold ABR
simulator to the ABR system to intentionally degrade the CMRR performance of the system, making it vulnerable to electrical interference. Firstly, recordings were made with and without the notch filter but with the connecting (electrode) leads grouped closely together, thus minimizing any induction of mains interference. Secondly, this was repeated but with a high level of mains interference produced as follows: the electrode leads were physically separated and the lead with the 3-kΩ resistor was placed near a mains power cable in order to induce a degree of 50-Hz interference. The incoming voltage of the ABR system was monitored whilst the position of the power cable was adjusted so that the interference was just below the ⫾ 20 μV rejection level. This ensured that no epochs were rejected and that the amplifier was not driven into saturation by the incoming signal. Recordings were repeated with and without the notch filter. The effect of the notch filter on response morphology could therefore be assessed in conditions of low and high 50-Hz mains interference.
Figure 2. An example of the effect of notch filtering on waveform morphology. Upper panel: no important effect; lower panel: considerable effect.
Results All newborn waveform pairs, with and without the notch filter enabled, were superimposed and visually inspected for any subjectively judged change in response morphology of an extent that might compromise response identification. None of the waveform pairs at 4000 Hz and 1000 Hz suffered from morphology change but in 5 (15%) of the 32 pairs at 500 Hz there was a noticeable change when the notch filter was applied. Figure 2 shows two examples of 500 Hz waveforms, with and without an obvious morphology change. These waveforms were recorded from the right and left ears of the same baby. In all five cases where such an effect was seen, the filtered waveform appeared to have low frequencies attenuated, that is, were made flatter.
This methodology has a potential drawback: the results may differ when a significant degree of mains interference is present. To investigate this issue, an experiment was conducted using an evoked potential simulator and the Bio-logic Navigator Pro system with and without the notch filter applied. For the experiment an ABR waveform was recorded from the simulator using the test parameters described above but with an amplifier gain of 100 000, an artefact rejection level of ⫾ 20 μV, and 2048 sweeps. The simulator was set to produce a waveform mimicking that of a suprathreshold click-evoked ABR together with 10 μV of noise. A 3-kΩ resistor was inserted into the non-inverting lead connecting the
Table 1. Results of the paired t-test statistical analysis comparing the ABR waveforms recorded from babies with and without the notch filter. Paired differences 95% confidence interval of the difference
4000 Hz Latency Amplitude Fsp Noise 1000 Hz Latency Amplitude Fsp Noise 500 Hz Latency Amplitude Fsp Noise
631
Mean
Std. dev
SEM
Lower
Upper
t
df
Sig. (2-tailed)
.00609 ⫺.00404 .25564 ⫺.10092
.02215 .19911 .61459 .18292
.00391 .03520 .10864 .03234
⫺.00190 ⫺.07583 .03406 ⫺.16687
.01407 .06774 .47722 ⫺.03497
1.555 ⫺.115 2.353 ⫺ 3.121
31 31 31 31
.130 .909 .025 .004
.00403 .03747 .07971 ⫺.01783
.02819 .20602 .48979 .14928
.00498 .03642 .08658 .02639
⫺.00614 ⫺.03681 ⫺.09688 ⫺.07165
.01419 .11175 .25630 .03599
.808 1.029 .921 ⫺.676
31 31 31 31
.425 .312 .364 .504
.00946 ⫺.00333 ⫺.13976 ⫺.04314
.03847 .43838 .72161 .11729
.00680 .07750 .12756 .02073
⫺.00440 ⫺.16138 ⫺.39992 ⫺.08543
.02333 .15473 .12041 ⫺.00086
1.392 ⫺.043 ⫺ 1.096 ⫺ 2.081
31 31 31 31
.174 .966 .282 .046
Std dev: standard deviation; SEM: standard error of the mean; df: degrees of freedom; Sig: significance (p-value). Units: Latency (ms); Amplitude (μV); Noise (nV). The Bonferroni correction was such that the p-value had to be less than 0.0125 to achieve significance (values shown in bold).
632
G. Lightfoot et al.
The four ABR parameters (latency, amplitude, Fsp, and residual noise) were subject to statistical analysis. The data were inspected and a positive skewness was apparent, particularly in the Fsp data. The data were therefore subject to a natural log transformation which removed the skew and allowed the application of parametric analysis. At each frequency, paired t-tests were conducted on the pairs of records with and without the notch filter for the four measurements: latency, amplitude, Fsp, and residual noise, with a Bonferroni correction applied for multiple comparisons. To achieve significance at the 0.05 level the observed p-value had to be 0.0125. Table 1 summarizes the results. It is apparent that the application of the notch filter had no significant effect on latency, amplitude, or Fsp at any frequency. For a stimulus of 4000 Hz only, the application of the notch filter significantly reduced residual noise. This parameter contributes to the calculation of Fsp but the effect of filtering on Fsp for a stimulus of 4000 Hz did not quite reach significance. The experiment using the ABR stimulator with low and high levels of mains interference confirmed that the characteristics of the notch filter did not appear to change with the level of mains interference. This suggests that the main findings of the study, using waveforms recorded in low noise conditions, can be extended with validity to conditions in which considerable levels of mains interference is present, providing that the amplifier is not driven into saturation.
frequencies. Why this effect should be seen at only 4000 Hz is unclear. This effect is not unwelcome and need not be interpreted as a reason for the avoidance of notch filtering when 4000 Hz stimuli are used. In conclusion, for threshold ABR in newborns, the use of a notch filter is not recommended for a 500 Hz stimulus but for higher frequency stimuli this study appears to suggest that a notch filter may be employed without serious consequences when faced with considerable mains-related interference. In such circumstances the tester should first try to identify and if possible eliminate the cause of the interference rather than routinely use the notch filter. The filter should not be relied on to mitigate the effects of poor clinical practice such as careless electrode application or routing of electrode or transducer cables. Note that these findings are applicable to the Bio-logic ABR system; they may be extended to other ABR systems only if their notch filter design is broadly similar. This technical report did not include the condition of a high level of mains interference in tests on real babies. The simulation component of this study was designed to ensure that our main findings could be extended to conditions of mains interference. However to be entirely confident that there is no unexpected non-linear interaction among mains interference, biological noise, and the notch filter a further study including this condition is recommended.
Acknowledgements Discussion The morphology changes seen in 15% of ABR waveforms evoked by 500-Hz stimuli make the use of a notch filter inappropriate for this stimulus since it carries the potential for misinterpretation of response presence. It was anticipated that any effect of notch filtering would be most pronounced for low frequency stimuli, where the longer response latencies associated with the cochlear travelling wave delay bring the lower edge of the response spectrum closer to the notch frequency (50 Hz) than for responses to higher frequency stimuli. In an attempt to understand the mechanism underpinning the observed morphology changes in 15% of 500 Hz cases the waveforms of those cases were studied. They did not have untypically high or low numbers of rejected sweeps or levels of residual noise, suggesting the mechanism was not associated with the degree of physiological noise in the recordings. Similarly their wave V latencies (and therefore the primary peak of their response spectrum) were not untypical within the 500 Hz group. The mechanism of the morphology change is therefore unclear. No significant effect of notch filtering was evident on the latency or amplitude of the ABR for any of the stimuli. The residual noise of 4000-Hz evoked responses was lower when notch filtering was applied but this effect was not apparent for the lower stimulus
The authors are indebted to Dr John Stevens for technical advice, to Prof. Azzam Taktak for statistical advice, and to the audiologists who contributed to data collection. Declaration of interest: The authors report no conflicts of interest.
References Elberling C. & Don M. 1984. Quality estimation of averaged auditory brainstem responses. Scand Audiol, 13, 187–197. Hall J.W. (ed.) 1992. Handbook of Auditory Evoked Responses. Boston, USA: Allyn and Bacon. ISO 389-6 (2007). Acoustics - Reference zero for the calibration of audiometric equipment - Part 6: Reference threshold of hearing for test signals of short duration. Geneva, Switzerland: International Organization for Standardization. Newborn Hearing Screening Programme (England) 2013. Guidance for auditory brainstem response testing in babies. Version 2.1. NHSP Clinical Group. (G. Sutton & G. Lightfoot, eds.) http://hearing.screening.nhs.uk/ audiologicalassessment Thornton A.R.D. 2007. Instrumentation and Recording Parameters. In: R.F Burkard, J.J. Eggermont, M. Don (eds). Auditory Evoked Potentials. Baltimore, USA: Lippincott Williams & Wilkins, pp. 73–101.
Copyright of International Journal of Audiology is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
Technical Report
The effect of notch filtering on the waveform of the newborn auditory brainstem response Guy Lightfoot∗,†, Inga Ferm‡, Amanda Hall§ & Kathryn Evans# *Department
of Medical Physics & Clinical Engineering, Royal Liverpool University Hospital, Liverpool, UK, †Clinical Advisory Group, English Newborn Hearing Screening Programme, London, UK, ‡Audiology Department, Croydon University Hospital, Croydon, UK, §Children’s Hearing Centre, St Michael’s Hospital, Bristol, UK, and #Paediatric Audiology Department, Leighton Hospital, Crewe, UK
Abstract Objective: To identify whether the use of a notch filter significantly affects the morphology or characteristics of the newborn auditory brainstem response (ABR) waveform and so inform future guidance for clinical practice. Design: Waveforms with and without the application of a notch filter were recorded from babies undergoing routine ABR tests at 4000, 1000 and 500 Hz. Any change in response morphology was judged subjectively. Response latency, amplitude, and measurements of response quality and residual noise were noted. An ABR simulator was also used to assess the effect of notch filtering in conditions of low and high mains interference. Results: The use of a notch filter changed waveform morphology for 500 Hz stimuli only in 15% of tests in newborns. Residual noise was lower when 4000 Hz stimuli were used. Response latency, amplitude, and quality were unaffected regardless of stimulus frequency. Tests with the ABR stimulator suggest that these findings can be extended to conditions of high level mains interference. Conclusions: A notch filter should be avoided when testing at 500 Hz, but at higher frequencies appears to carry no penalty.
Key Words: Auditory brainstem response; notch filter, distortion
Recording the auditory brainstem response (ABR) presents a technological challenge; the electrodes record not only the sub-microvolt response but also record much larger, unwanted, signals including those generated by the patient (neurogenic and myogenic) and electrical interference. A major source of electrical interference originates from the power supply (mains), either through the instrument itself or via capacitive or inductive coupling between the patient or the electrode leads and the power supply. This signal (60 Hz in the Americas and usually 50 Hz elsewhere) is present on all parts of the patient but the ABR signal that we seek to record varies from point to point on the body. Careful placement of the electrodes and amplification of the difference in the detected voltages provides a large degree of immunity to signals which are common to both recording electrodes. These so-called common mode signals are therefore effectively rejected when differential amplification is used. Amplifiers are imperfect and their performance in rejecting common mode signals is described by their common mode rejection ratio
(CMRR). The value of CMRR can be high, typically 100 dB, if the contact impedances of the electrodes are low and similar but the CMRR drops dramatically if the electrode impedances are unequal. Under these circumstances, common mode signals such as mains interference are not rejected and becomes problematic. It is for this reason that testers must endeavour to obtain low and equal electrode impedances. To ensure that all electrode leads pick up the same mains interference it is good practice to gather or twist the leads together; if they are physically separated then the differing amounts of interference induced in the leads will be amplified and contaminate the recording. Mains interference is not usually an issue with well-designed equipment and good clinical practice but occasionally, if the electrode contact is not ideal or if the recording environment is electrically hostile, mains interference can be problematic. When excessive, it can exceed the artefact rejection level and no recording is possible. Adopting a more lax artefact rejection level would
Correspondence: Guy Lightfoot, Department of Medical Physics & Clinical Engineering, Royal Liverpool University Hospital, Prescot Street, Liverpool, L7 8XP, UK. E-mail: [email protected] (Received 21 October 2013; accepted 10 February 2014 ) ISSN 1499-2027 print/ISSN 1708-8186 online © 2014 British Society of Audiology, International Society of Audiology, and Nordic Audiological Society DOI: 10.3109/14992027.2014.894644
630
G. Lightfoot et al.
Method
Abbreviations ABR CMRR
Auditory brainstem response Common mode rejection ratio
allow the recording to proceed but at a high cost: the interference is allowed into the recording. Since the frequency of mains interference is known, one solution is to filter it out using a notch filter. This is a band-stop filter with a narrow and deep spectral notch at the mains frequency, rejecting only that frequency and passing everything else, as illustrated in the upper panel of Figure 1. Whilst this appears ideal, analogue notch filters have a non-linear phase response, shown in the lower panel of Figure 1. The consequences of this can be to change the shape, amplitude, and latencies of signals such as the ABR. A secondary issue is that mains-related interference can include harmonics of the mains frequency, which would not be attenuated by a mains notch filter. It is for these reasons that the normal expert advice is to avoid the use of notch filters, their primary application being limited to the identification of mains-frequency signals as the source of interference rather than a means of attenuating the interference (Hall, 1992; Thornton, 2007). The latency shifts associated with notch filters could pose a significant problem in the neurological ABR, where the measurement of absolute and inter-peak latencies is the primary means of analysis. However when the ABR is used for threshold estimation, waveform analysis is one of response detection, where signal to noise ratio is the issue. Although latency and amplitude are often recorded, minor changes would not invalidate the outcome. It is possible therefore that the use of a notch filter may be acceptable in threshold ABR testing, providing it creates no major change to response morphology, latency, or amplitude. The purpose of this study was to investigate this and so inform future clinical practice.
ABR test centres in Bristol, Liverpool, Crewe, Croydon, and Wirral, UK, participated in the collection of waveforms recorded from sleeping babies under the age of three months who were being tested as a routine component of their newborn hearing screening. ABR data were collected using 4000 Hz, 1000 Hz, and 500 Hz 5-cycle tone burst stimuli (most babies were tested at only one frequency) at a stimulus level in the range 35 to 75 dBnHL (Re: ISO 389-6). In all cases the stimulus level chosen was supra-threshold and evoked a clear response. Stimulus rates were 45.1/s for 4000 Hz (21 ms timebase) and 35.1/s for 1000 Hz & 500 Hz (26 ms timebase). Test parameters and methodology was in accordance with the English Newborn Hearing Screening Programme (NHSP) guidance for ABR testing in babies (2013). This included analogue filtering between 30 Hz and 1500 Hz (Butterworth response with ⫺ 12 dB/octave roll-off) with an artefact rejection level of ⫾ 5 μV, 2048 to 3072 sweeps, and no notch filter. Equipment was the Bio-logic Navigator Pro system (Natus Medical Incorporated, San Carlos, USA), software version 6.3.0 or later. Upon completion of testing, if time allowed and the baby was still sleeping, one of the stimulus levels previously used in the test to obtain a clear response was retested using the same number of sweeps but with the notch filter applied. In the Biologic system the notch filter is a programmable 50/60 Hz line filter with a Twin-T architecture and a quality factor (Q) of 10. The Q factor is the centre frequency (in this case 50 Hz) divided by the half-power bandwidth (in this case 5 Hz). Wave V latency, the wave V to SN10 amplitude, residual noise, and Fsp were recorded from the pair of waveforms with and without the notch filter. Fsp is a measurement of the quality of the recorded response, related to the response to noise variance ratio (Elberling & Don, 1984). Since clear ABRs had to be recorded without the notch filter, in none of the data was mains interference an issue. A total of 96 waveform pairs (32 at each frequency) were recorded from 65 babies.
Figure 1. The amplitude (upper panel) and phase (lower panel) of a typical notch filter.
Notch Filtering the threshold ABR
simulator to the ABR system to intentionally degrade the CMRR performance of the system, making it vulnerable to electrical interference. Firstly, recordings were made with and without the notch filter but with the connecting (electrode) leads grouped closely together, thus minimizing any induction of mains interference. Secondly, this was repeated but with a high level of mains interference produced as follows: the electrode leads were physically separated and the lead with the 3-kΩ resistor was placed near a mains power cable in order to induce a degree of 50-Hz interference. The incoming voltage of the ABR system was monitored whilst the position of the power cable was adjusted so that the interference was just below the ⫾ 20 μV rejection level. This ensured that no epochs were rejected and that the amplifier was not driven into saturation by the incoming signal. Recordings were repeated with and without the notch filter. The effect of the notch filter on response morphology could therefore be assessed in conditions of low and high 50-Hz mains interference.
Figure 2. An example of the effect of notch filtering on waveform morphology. Upper panel: no important effect; lower panel: considerable effect.
Results All newborn waveform pairs, with and without the notch filter enabled, were superimposed and visually inspected for any subjectively judged change in response morphology of an extent that might compromise response identification. None of the waveform pairs at 4000 Hz and 1000 Hz suffered from morphology change but in 5 (15%) of the 32 pairs at 500 Hz there was a noticeable change when the notch filter was applied. Figure 2 shows two examples of 500 Hz waveforms, with and without an obvious morphology change. These waveforms were recorded from the right and left ears of the same baby. In all five cases where such an effect was seen, the filtered waveform appeared to have low frequencies attenuated, that is, were made flatter.
This methodology has a potential drawback: the results may differ when a significant degree of mains interference is present. To investigate this issue, an experiment was conducted using an evoked potential simulator and the Bio-logic Navigator Pro system with and without the notch filter applied. For the experiment an ABR waveform was recorded from the simulator using the test parameters described above but with an amplifier gain of 100 000, an artefact rejection level of ⫾ 20 μV, and 2048 sweeps. The simulator was set to produce a waveform mimicking that of a suprathreshold click-evoked ABR together with 10 μV of noise. A 3-kΩ resistor was inserted into the non-inverting lead connecting the
Table 1. Results of the paired t-test statistical analysis comparing the ABR waveforms recorded from babies with and without the notch filter. Paired differences 95% confidence interval of the difference
4000 Hz Latency Amplitude Fsp Noise 1000 Hz Latency Amplitude Fsp Noise 500 Hz Latency Amplitude Fsp Noise
631
Mean
Std. dev
SEM
Lower
Upper
t
df
Sig. (2-tailed)
.00609 ⫺.00404 .25564 ⫺.10092
.02215 .19911 .61459 .18292
.00391 .03520 .10864 .03234
⫺.00190 ⫺.07583 .03406 ⫺.16687
.01407 .06774 .47722 ⫺.03497
1.555 ⫺.115 2.353 ⫺ 3.121
31 31 31 31
.130 .909 .025 .004
.00403 .03747 .07971 ⫺.01783
.02819 .20602 .48979 .14928
.00498 .03642 .08658 .02639
⫺.00614 ⫺.03681 ⫺.09688 ⫺.07165
.01419 .11175 .25630 .03599
.808 1.029 .921 ⫺.676
31 31 31 31
.425 .312 .364 .504
.00946 ⫺.00333 ⫺.13976 ⫺.04314
.03847 .43838 .72161 .11729
.00680 .07750 .12756 .02073
⫺.00440 ⫺.16138 ⫺.39992 ⫺.08543
.02333 .15473 .12041 ⫺.00086
1.392 ⫺.043 ⫺ 1.096 ⫺ 2.081
31 31 31 31
.174 .966 .282 .046
Std dev: standard deviation; SEM: standard error of the mean; df: degrees of freedom; Sig: significance (p-value). Units: Latency (ms); Amplitude (μV); Noise (nV). The Bonferroni correction was such that the p-value had to be less than 0.0125 to achieve significance (values shown in bold).
632
G. Lightfoot et al.
The four ABR parameters (latency, amplitude, Fsp, and residual noise) were subject to statistical analysis. The data were inspected and a positive skewness was apparent, particularly in the Fsp data. The data were therefore subject to a natural log transformation which removed the skew and allowed the application of parametric analysis. At each frequency, paired t-tests were conducted on the pairs of records with and without the notch filter for the four measurements: latency, amplitude, Fsp, and residual noise, with a Bonferroni correction applied for multiple comparisons. To achieve significance at the 0.05 level the observed p-value had to be 0.0125. Table 1 summarizes the results. It is apparent that the application of the notch filter had no significant effect on latency, amplitude, or Fsp at any frequency. For a stimulus of 4000 Hz only, the application of the notch filter significantly reduced residual noise. This parameter contributes to the calculation of Fsp but the effect of filtering on Fsp for a stimulus of 4000 Hz did not quite reach significance. The experiment using the ABR stimulator with low and high levels of mains interference confirmed that the characteristics of the notch filter did not appear to change with the level of mains interference. This suggests that the main findings of the study, using waveforms recorded in low noise conditions, can be extended with validity to conditions in which considerable levels of mains interference is present, providing that the amplifier is not driven into saturation.
frequencies. Why this effect should be seen at only 4000 Hz is unclear. This effect is not unwelcome and need not be interpreted as a reason for the avoidance of notch filtering when 4000 Hz stimuli are used. In conclusion, for threshold ABR in newborns, the use of a notch filter is not recommended for a 500 Hz stimulus but for higher frequency stimuli this study appears to suggest that a notch filter may be employed without serious consequences when faced with considerable mains-related interference. In such circumstances the tester should first try to identify and if possible eliminate the cause of the interference rather than routinely use the notch filter. The filter should not be relied on to mitigate the effects of poor clinical practice such as careless electrode application or routing of electrode or transducer cables. Note that these findings are applicable to the Bio-logic ABR system; they may be extended to other ABR systems only if their notch filter design is broadly similar. This technical report did not include the condition of a high level of mains interference in tests on real babies. The simulation component of this study was designed to ensure that our main findings could be extended to conditions of mains interference. However to be entirely confident that there is no unexpected non-linear interaction among mains interference, biological noise, and the notch filter a further study including this condition is recommended.
Acknowledgements Discussion The morphology changes seen in 15% of ABR waveforms evoked by 500-Hz stimuli make the use of a notch filter inappropriate for this stimulus since it carries the potential for misinterpretation of response presence. It was anticipated that any effect of notch filtering would be most pronounced for low frequency stimuli, where the longer response latencies associated with the cochlear travelling wave delay bring the lower edge of the response spectrum closer to the notch frequency (50 Hz) than for responses to higher frequency stimuli. In an attempt to understand the mechanism underpinning the observed morphology changes in 15% of 500 Hz cases the waveforms of those cases were studied. They did not have untypically high or low numbers of rejected sweeps or levels of residual noise, suggesting the mechanism was not associated with the degree of physiological noise in the recordings. Similarly their wave V latencies (and therefore the primary peak of their response spectrum) were not untypical within the 500 Hz group. The mechanism of the morphology change is therefore unclear. No significant effect of notch filtering was evident on the latency or amplitude of the ABR for any of the stimuli. The residual noise of 4000-Hz evoked responses was lower when notch filtering was applied but this effect was not apparent for the lower stimulus
The authors are indebted to Dr John Stevens for technical advice, to Prof. Azzam Taktak for statistical advice, and to the audiologists who contributed to data collection. Declaration of interest: The authors report no conflicts of interest.
References Elberling C. & Don M. 1984. Quality estimation of averaged auditory brainstem responses. Scand Audiol, 13, 187–197. Hall J.W. (ed.) 1992. Handbook of Auditory Evoked Responses. Boston, USA: Allyn and Bacon. ISO 389-6 (2007). Acoustics - Reference zero for the calibration of audiometric equipment - Part 6: Reference threshold of hearing for test signals of short duration. Geneva, Switzerland: International Organization for Standardization. Newborn Hearing Screening Programme (England) 2013. Guidance for auditory brainstem response testing in babies. Version 2.1. NHSP Clinical Group. (G. Sutton & G. Lightfoot, eds.) http://hearing.screening.nhs.uk/ audiologicalassessment Thornton A.R.D. 2007. Instrumentation and Recording Parameters. In: R.F Burkard, J.J. Eggermont, M. Don (eds). Auditory Evoked Potentials. Baltimore, USA: Lippincott Williams & Wilkins, pp. 73–101.
Copyright of International Journal of Audiology is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
