Auditory Brainstem Response

INTRODUCTION TO AUDITORY EVOKED RESPONSES
 

            An auditory evoked response (AER) is activity within the auditory system that is produced or stimulated by sounds. In simplest of terms, AERs are brain waves (electrical potentials) generated when a person is stimulated with sounds.

        Since the stimulus is a sound, it is clear that the response arises from the auditory system. The specific source of the response within the auditory system is often difficult or impossible to pinpoint. Nonetheless, by analyzing the pattern of the response and by calculating the time period after the stimulus in which the response occurs, it is usually possible to determine the regions in the auditory system generating the response and, sometimes, specific anatomic structures. The time after the stimulus at which AERs occur is invariably less than 1 second. Therefore, the post-stimulus times (latencies) of peaks in the response pattern (waveform) are described in milliseconds. A millisecond is one-thousandth of one second.

            Responses with the shortest latencies are generated by the inner ear and the auditory nerve. A few milliseconds later, there are unique response patterns reflecting activity within the auditory brainstem. Recorded still are the response patterns due to activity in higher auditory  portions of the brain, such as the cerebral cortex.

AUDITORY BRAINSTEM RESPONSE

           

            From the first description of the human ABR, independently by Jewett and Williston (1971) and Lev and Sohmer (1972), the response has been described with a variety of terms and acronyms, and different schema has been used to denote the wave components. Auditory brainstem response (ABR) audiometry is a neurologic test of auditory brainstem function in response to auditory (click) stimuli. It is the most common application of auditory evoked responses.

            As noted by Goldstein (1984), all AERs can be described along various different dimensions such as:

·       Time (early versus middle versus late responses)

·       Speed (fast versus slow)

·       Anatomy (electrocochleography versus auditory brainstem response)

·       A general property of response generation (exogenous versus endogenous)

·       Some more specific generator property (stimulus-related).

            The auditory brainstem response (ABR) is one component of the auditory evoked potentials (AEPs), which also include electrocochleography, the auditory middle latency response, and the auditory late response.

            The ABR consists of a series of seven positive to negative waves, occurring within  ~ 10 msec following stimulus onset. There are several classes of auditory evoked potentials, that is, small changing voltages elicited using auditory stimuli. Within the cochlea, there are several types of electrical potentials elicited by auditory stimulation. The ABR consisting of 5 to 7 peaks occurring within the first 10 ms of stimulus onset depending on intensity has been labelled a far-field recording, since surface electrodes are attached to the scalp distant from the generator source. The response is termed as Brainstem Auditory Evoked Potentials (BAEP) to reflect the fact that various sites within the auditory brainstem pathway are generators of BAEP waves.

            ABR audiometry refers to an evoked potential generated by a brief click or tone pip transmitted from an acoustic transducer in the form of an insert earphone or headphone. The elicited waveform response is measured by surface electrodes typically placed at the vertex of the scalp and ear lobes. The amplitude (microvoltage) of the signal is averaged and charted against the time (millisecond), much like an EEG. The waveform peaks are labeled I-VII. These waveforms normally occur within a 10-millisecond time period after a click stimulus presented at high intensities (70-90 dB normal hearing level [dB nHL]).

            Although the ABR provides information regarding auditory function and hearing sensitivity, it is not a substitute for a formal hearing evaluation, and results should be used in conjunction with behavioral audiometry whenever possible

 

            In 1979, Hallowell Davis formally introduced the term “ABR”. There are inconsistencies in the polarity of the response for the vertex electrode (negative versus positive). With the Roman numeral labeling system as introduced by Jewett and Williston (1971), vertex positive waves are plotted upward. That is, the electrode at the vertex (or high forehead) is plugged into the positive voltage input of the amplifier while the earlobe (or mastoid) electrode is plugged into the negative voltage input. This produces a typical ABR waveform. However, some investigators, usually Japanese or European, reverse this electrode arrangement (negative voltage input is at the vertex or high forehead), and major peaks in the resulting waveform  are plotted downward.

PHYSIOLOGY    PPT

GENERATOR SITES

Several investigators have stated that each wave of the BAEP waveform is generated by specific nuclei along the auditory pathway (Lev &. Sohmer, 1972). These investigators indicated that wave I is generated from the auditory nerve, wave II is generated by the cochlear nuclei, wave III is generated from the superior olivary complex, and waves IV and V arc generated from the lateral: lemniscus and the inferior colliculus, respectively.

Moller and Jannetta (1985) contended that hypotheses or theories regarding the origin of the BAEP components IN humans based on experiments in animals such as cats, rats, and guinea pigs should be seriously questioned. That is, the BAEP components (except peak I) in animals are generated from sites different from those in humans and the auditory nerve in humans is much longer (2.5 cm long in humans) than in cats, rats, and guinea pigs (0.3 to 0.5 cm long). Since the conduction velocity of the nerve fiber is slow in both humans and animals, the conduction time is much longer in humans than in animals (1 ms as opposed to 0.1-0.3 ms). Bearing this in mind, several investigators (Moller & Jannetta, 1981) conducted studies using microelectrodes comparing the BAEPs recorded at far field (vertex and earlobe) and the BAEPs recorded directly along the auditory pathway.

            Auditory brainstem response (ABR) audiometry typically uses a click stimulus that generates a response from the basilar region of the cochlea. The signal travels along the auditory pathway from the cochlear nuclear complex proximally to the inferior colliculus. ABR waves I and II correspond to true action potentials. Later waves may reflect postsynaptic activity in major brainstem auditory centers that concomitantly contribute to waveform peaks and troughs. The positive peaks of the waveforms reflect combined afferent (and likely efferent) activity from axonal pathways in the auditory brain stem.

Information on the anatomic origins of the ABR is less precise and more conflicting for later components (III, IV, V, VI) than for the earlier components (I, II). Traditionally with experimental lesion studies in small animals, the ABR generators were as follows,

•  Wave I- eighth nerve (auditory)
• Wave II – cochlear nuclei
•  Wave III – SOC
•  Wave IV-  nucleus of Lateral leminiscus (LL)
•  Wave V- inferior colliculus (IC)
• Wave VI- medial geniculate body of the thalamus

            However, there are, in the human brainstem, numerous examples of auditory cell groups and nuclei that are smaller, or not even observed, when compared to brainstem anatomy of lower animals. Recent views believe in multiple generator sites for the ABR peaks.

Waveform components

Wave I: The ABR wave I response is the far-field representation of the compound auditory nerve action potential in the distal portion of cranial nerve (CN) VIII. The response is believed to originate from afferent activity of the CN VIII fibers (first-order neurons) as they leave the cochlea and enter the internal auditory canal.

Wave II: The ABR wave II is generated by the proximal VIII nerve as it enters the brain stem. A proximal eighth nerve generator site (the end of the nerve near the brainstem) for wave II is supported by the relationship between the latency of waves I and II and the relatively slow conduction time expected for the auditory nerve (10-20 msec), which is on the average 25 mm in length with a diameter of approximately 2 to 4 micrometers in adults. In small children, wave II is not consistently recorded. Moller attributes this observation to shorter eighth nerve length that results in the fusion of I and II waves. Based on the estimations of eighth nerve velocity and  synaptic delays, wave II must reflect activity of the first order neuron (i.e., the eighth nerve itself versus the auditory brainstem). Peaks I and II are generated by the auditory nerve. Peak I is generated at the distal part of the auditory nerve and peak II is generated primarily from the proximal part of the auditory nerve. Recent studies (Wada & Starr, 1983) supported the findings of Moller and his colleagues regarding the generation of waves I and II.

Wave III: The ABR wave III arises from second-order neuron activity (beyond CN VIII) in or near the cochlear nucleus. Literature suggests wave III is generated in the caudal portion of the auditory pons. The cochlear nucleus contains approximately 100,000 neurons, most of which are innervated by eighth nerve fibers. On the basis of  experimental lesion studies in small animals wave III was traditionally associated with neural activity in the superior olivary complex (SOC) within the Brainstem contralateral to the side of stimulation. Subsequent studies showed conflicting results. In humans wave III arises from second order neuron activity (beyond the eigth nerve) in or near the cochleur nucleus, whereas the negative trough following wave III appears to arise from the trapezoid body. Peak III is generated mainly by the ipsilateral cochlear nuclei and may receive a small contribution from the eighth-nerve fibers entering the cochlear nuclei.

 

Wave IV: The ABR wave IV, which often shares the same peak with wave V, is thought to arise from pontine third-order neurons mostly located in the superior olivary complex, but additional contributions may come from the cochlear nucleus and nucleus of lateral lemniscus. The precise anatomic location is complicated due to the multiple crossings of midline (decussations) for auditory fibers beyond the cochlear nucleus. In far field recordings, wave IV often appears as a leading shoulder on wave V and is refereed as IV/V complex. Peak IV is generated by the third-order neurons, that is, the superior olivary complex. Also contributing to wave IV are the cochlear nuclei and probably the lateral lemniscus nuclei.

Wave V: Generation of wave V likely reflects activity of multiple anatomic auditory structures. The ABR wave V is the component analyzed most often in clinical applications of the ABR. Although some debate exists regarding the precise generation of wave V, it is believed to originate from the vicinity of the inferior colliculus. The second-order neuron activity may additionally contribute in some way to wave V. The inferior colliculus is a complex structure, with more than 99% of the axons from lower auditory brainstem regions going through the lateral lemniscus to the inferior colliculus. The origin of the positive component of wave V is the lateral lemniscus. Other nuclei in the vicinity have a minor contribution to the positive component of wave V. The origin of the negative component of wave V is the inferior colliculus.

Wave VI and VII: Thalamic (medial geniculate body) origin is suggested for generation of waves VI and VII, but the actual site of generation is uncertain.

            In terms of clinical applications, since waves III, IV, and V have major and minor generators, if the latency and amplitude of these waves are substantially affected, one can assume that the major generators are impaired. When the latency and amplitude of waves III-V are only slightly affected, a conclusion cannot be drawn regarding the generator site damaged since either the major or minor generator may be responsible for the latency and amplitude effects.

ANATOMY AND PHYSIOLOGY: PRINCIPLES OF AUDITORY EVOKED RESPONSES

            The distribution of current flow in the extracellular (extraneural) space is a potential field. The Transmembrane ionic current flow of the cell, in the case of evoked responses, the neuron is the origin of voltage potentials that underlie AERs. Transmembrane current flow can be associated with an action potential that travels along an axon of a neuron or with synaptic activity between two or more neurons. Current flow in to one portion of the cell creates depolarization and is related to a negative charge potential in the nearby extra-cellular region, often termed the current “sink”. This current inflow is balanced by the current outflow at another portion of the cell, which is related to the positive voltage potential. The resulting electric field with a negative polarity at one end and a positive polarity at the other end is called a dipole. A positive potential, therefore, forms the leading edge of propagated neuronal activity, that is, neuronal transmission. Following activation of the sensory organ, the cochlea for AERs, such sequential Transmembrane current flow evoked by the stimulus occurs in neurons in a peripheral to central direction, i.e., from the ear to the cerebral cortex.

            The concept of volume conduction is important in understanding the neuroanatomy and neurophysiology underlying AERs and sensory evoked responses in general. Responses arising from the anatomic pathways deep within the brain are conducted to rather distant electrodes through a complex media consisting of diverse substances, including fluids, brain tissue, bone and skin. A number of factors contribute to an effect of surface-recorded, volume conducted AER. One factor detecting volume-conducted AERs if the location of recording electrodes relative to the electrical fields is dipole. Evoked responses recorded by electrodes close to or within the potential field, as in intracranial depth electrode studies are called Near-field responses. Evoked response detected at a relatively great distance from the source, such as the scalp-recorded ABR, and Far-field responses. Increasing the distance between the recording electrodes and the voltage source has two major effects on the volume conducted evoked responses.

• First, spatial resolution is reduced. The evoked response waveform from a single electrode on the scalp and distance from a dipole generator is a broad, less distinct waveform, whereas a wave recorded from the electrode closer to the electrical field is sharper. Also, a response recorded at some distance from two intracranial sources appears to consist of a single wave component. The individual contribution of each dipole cannot be distinguished. With the electrode located closer to the two electrical fields, the waveform is resolved into two distinct fields.
• The second effect of increased distance between the recording electrode and the voltage source is a dramatic reduction in response amplitude. Amplitude of near field responses is large, but decreases sharply when the electrode is moved even a small distance from the generator. In contrast, amplitude of far field responses is very small but does not change appreciably with large alterations in electrode location.

            The generators of AERs located within the skull, either within the temporal bone or the cranium. Clinically, however most AERs are recorded with far field techniques, i.e with electrodes located outside of the skull, and usually at the scalp except for intraoperative monitoring and during transtympanic recording of ECochG

            The geometric orientation of the activated neurons also exerts an important influence on volume conduction. Over 50 years ago Lorente de No (1947) noted that synchronous depolarization of a group of neurons that are oriented in the same direction produces an enhanced electrical field (dipole) that can be detected relatively great distance whereas with cells that are orients in different directions, there may be cancellation of the electrical fields. Lorente de No (1947) referred to these as open and closed potential fields. The orientation of fiber tracts, and of dendrites within nuclei plays an important role in the generation of far field responses. In fact with the closed field neuron orientation, it is conceivable that a sensory stimulus could evoke neuronal activity that is not measurable by volume conduction.

            The spatial and temporal characteristics of the current flow for individual neurons also play a role in generation of evoked responses. Evoked responses are directly dependent on the temporal synchronization of neuronal activity. AERs are optimally generated by action potentials or synaptic potentials arising almost simultaneously from many neurons within specific anatomic regions. When Transmembrane current flow occurs in a very restricted part of the neuron, special summation of extracellular potential field from many neuron is less likely. Similarly when Transmembrane current flow produces a rapid transient voltage potential, the likelihood of temporal summation of extracellular potential fields (synchronous activity among neurons) is reduced. Consequently volume conducted evoked responses are greatest for neurons with relatively widespread and extended duration changes in voltage. Spatial and temporal characteristics apply to Transmembrane neuronal activity in the region of the cell body i.e synaptic transmission but are not important for the propagation of the action potential along the long axon.

INSTRUMENTATION AND SIGNAL PROCESSING PPT 333888

            The magnitude of the BAEP is very small, approximately 0.01 – 1 µV. This small potential is masked by the larger background activity generated by several sources such as the random, ongoing electrical activity (EEG) within the brain, muscular (myogenic) activities in the skull region, electrical radiation from electronic devices in the environment such as the 60- Hz hum, and other artifacts produced while generating the stimuli or recording the potentials.

A typical BAEP device is shown in the figure 1 below. The system usually consists of three major components

·       The component labelled 1 is the signal generating component, which generates the type of signal desired, for example, click, tone pip, or tone burst.

·       The second component, labelled 2 is the amplifier and the filter component which is designed to amplify the BAEP from the scalp so the BAEP can be processed easily by the signal average and filter.

·       The third component labelled as 3 is the computer averager whose function is to store electrical responses that are time-locked to the stimulus and to cancel out the ongoing EEG activity.

AMPLIFIER AND FILTER

The physiological amplifier must be able to eliminate the 60-Hz hum. This could be done using a differential amplifier which has two stages.

·       PREAMPLIFIER STAGE- it receives three inputs.

#   The first input to the preamplifier is from the vertex of the scalp and called the non-inverting, or positive, waveform. It is called the non-inverted waveform because the signal polarity is unchanged when it reaches the preamplifier stage. This input has also been referred to as the positive waveform since the first wave has a positive-going direction. It has also been called the active input.

#   The second input, from the reference site of the scalp to the preamplifier, is the inverted or negative waveform. It is called the inverted input because the signal polarity is reversed when it reaches the preamplifier stage. This input has also been referred to as the negative input since the first wave has a negative- going direction.

The inverted input is then added to the noninverted input, each referenced to the common input. This resulting waveform, the output of the preamplifier stage, is the input to the second amplifier stage.. If the noninverted and inverted inputs to the first stage of amplification are identical, they will cancel. Since the 60-Hz hum is equal in amplitude but opposite in polarity at the noninverting and inverting electrodes, the 60-Hz hum will be canceled out, so the input to the second stage of amplification will consist of only the resultant BAEP waveform (inverted waveform added to the noninverted waveform. Since the BAEP waveform at the noninverting electrode is larger than that at the inverting electrode, there is only partial cancellation of the BAEP waveform when the inverted and noninverted inputs are added.

Fig 1

 

The second stage of the amplifier amplifies the input approximately 100,000 times (105). The output of the second stage passes through the filter. This filter is designed to increase the signal-to-noise ratio prior to signal averaging. The bandpass of the filter is adjusted to reject background activity unrelated to the potential. For example, the BAEP device can be set to reject frequencies-below 100 Hz and above 3000 Hz. Thus, only frequencies between 100 and 3000 Hz will be passed.

COMMON MODE REJECTION RATIO

The common mode rejection ratio (CMRR) is used to describe the extent to which the common inputs, such as the 60-Hz hum at the inverting and noninverting electrodes, are canceled out as shown in fig 2. A CMRR of 100 dB means that the common input is attenuated by a factor of 100,000 (every 20 dB is equivalent to a factor of 10) relative to the differential input. A 60-Hz hum of 1.0 V would be amplified by a differential amplifier gain of 10s to a value of 100,000 V which is then reduced by a factor of 100,000 (CMRR of 100 dB) to 1.0 V. A differential input of 10µV is amplified by a differential gain of 10⁵ to a value of 1.0 V. Thus, the common input has to be as large as 1V to have the same effect as a differential input as small as 10µV. An electrode impedance imbalance will decrease the effect of the CMRR.

Fig 2

SIGNAL AVERAGER

            The computer or signal averager's function is to store electrical responses that are time-locked to the stimulus and to cancel out the ongoing EEG activity. The principles of the computer averager are based on the concept that a response which is time-locked to the stimulus is consistent in polarity and that electrical activity of the background EEG is randomly changing in polarity. If acoustical stimulation is repeated over and over and a series of time-locked potentials are obtained, the addition of these time-locked potentials and ongoing random electrical activity results in the preservation of die time-locked responses (since they are similar in polarity) and elimination of the ongoing EEG activity (since the activity is random in polarity and addition of random activity leads to cancellation).

The figure 3 shows the major components of the signal averager. The first part of the averager is the analog-to-digital (A—D) converter which converts the responses (potentials) into numbers representing the amplitude of the response, which are then summed and placed in the digital memory of the data processor (computer). In order for the A-D converter to accurately convert the continuous responses into discrete number equivalents and the number equivalents back into analog form (the continuous waveform), the A-D converter must have a sufficient intersample interval (the dwell time or horizontal resolution) and adequate vertical (amplitude) resolution. The continuous response is accurately represented as discrete numbers and vice versa if the intersampling interval is fast. A short intersample interval leads to better analog representation of the waveform. In order for the intersample interval to be considered sufficiently short, it must be much smaller than the duration of a wave within the waveform divided by the number of points needed to represent the wave. For example, if the duration of peak I is 1 ms and 100 points are required to faithfully represent this wave, then the required intersample interval should be less than or equal to 1/100 ms or 0.01ms. Thus, the time between the points at which analysis occurs is 0.01ms. The second requirement for an A—D converter is adequate vertical resolution (analysis of the waveform amplitude). The number of bits in the computer averager determines how accurately the amplitude of the waveform of interest can be measured and the smallest amplitude of the waveform which can be detected. Let assume that the A—D converter of a computer average has 4 bits. This means that the voltage range of the waveform is divided into 16 parts. If the range of the electrode voltage is 1µV, the smallest potential which can be detected the computer averager if 1/16 or 0.06 µV. Thus, for a 4-bit computer averager, potentials with values less than 0.06µV cannot be detected.

Fig 3

FACTORS EFFECTING ABR

 STIMULUS PARAMETERS

INTENSITY

Intensity related duration and amplitude changes

Chiappa et al. (1979) reported that waves I, II, III, and V can be detected more than 90% of the time at 60 dB SL for clicks presented at a repetition rate of 10/s or 30/s. Also, at that sensation level, waves IV and VI can be detected 88 and 84% of the time, respectively, when clicks are presented at a repetition rate of 10/s, but can be detected only 67% of the time when clicks are presented at a repetition rate of 30/s. At the repetition rate of 70/s, only wave V continues to be highly recognizable (99% recognizability). At the repetition rate of 70/s wave III is the next most recognizable of BAEP components, having an 85% frequency of detectability. The frequency of recognizability of the other waves was below 80%, approaching 34% for wave IV.

Kavanagh and Beardsley (1979) reported that, in their group of 15 normal-hearing subjects, wave V was present at intensity levels as low as 10 dB HL. Other waves could not be elicited at low intensity levels. The presence or other waves was always accompanied by the presence of wave V. Wave II .was observed only at high intensity levels. Worthington and Peters (1980a) reported that wave III was present at intensity levels as low as 30 dB SL in approximately 50% of their subjects. Wave I was present at intensity levels as low as 50 dB SL in 75% of their subjects.

Therefore, In Order To Elicit A Waveform Which Contains Wave I As Well As The Other Waves, Intensity Levels Of At Least 70 dBnHL And Low Repetition Rates Should Be Employed.

The salient features of the effect of intensity on wave V latency in this series of waveforms become more clearly apparent when latency values are plotted as a function of intensity. The latency-intensity function for wave V is the most common graphic display of clinical ABR data.

For low to moderate intensity levels, there is normally a systematic and rather abrupt shortening of latency values (up to 0.50 or 0.60 ms of latency per 10 dB of intensity, or 0.06 ms/dB) for wave V up to approximately 60 dB nHL. As wave V, even in normal hearers, is generally not detected visually at intensity levels below 10 dB nHL, at these apparent ABR threshold levels, normal wave V latency is 7.5 to 8.0 ms or more. For intensities from 60 to 95 dB nHL, the slope of the latency-intensity function is more gradual, producing a rate of change of only about 0.10 to 0.20 ms/10 dB. Indeed, as intensity is increased at the highest click signal intensity levels, there is little decrease in wave V latency. Overall, the latency-intensity slope is on the order of 0.38 ms/10 dB. Because the latency-intensity function has a bend at about 60 dB, it is not linear. According to Picton, Stapells, and Campbell (1981), the latency-intensity function can be calculated with the following power function:

log 10 (V latency in ms) = -0.0025 (click intensity in dB) + 0.924

At high intensity levels (75 to 95 dB nHL), wave V latency is normally in the region of 5.5 to 6.0 ms (Wolfe, Skinner, & Burns, 1978). As noted previously, wave V latency increases by about 2 ms over the range from high intensity levels (e.g., 80 dB) to threshold levels (e.g., 20 dB). That is, over this intensity range absolute latency increases from about 5.5 ms to 7.5 ms. At high signal intensity levels, it's possible to identify wave I and wave III (and even wave II, wave TV, and wave VI). In normal hearers, at high intensity levels (>80 dB nHL) for the click signal, wave I latency is about 1.5 ms, wave III is about 3.5 ms, and, as just noted, wave V is in the 5.5 ms region. Interwave latency values are approximately 2.0 ms for both the wave I to III interval and the wave III to V interval. Added together, this results in an average wave I to V interval on the order of 4.0 ms. Subject factors, such as gender or body temperature, may influence interwave latency values.

Although wave V can be detected for intensity levels down to 10 dB, and even lower, the lowest level at which wave I and wave III are usually visible in adult normal-hearing subjects with conventional recording techniques is about 25 to 35 dB nHL. Partly, this is because wave V normally has relatively larger amplitude. In addition, however, the rate of decrease in amplitude with decreasing intensity is more rapid for the earlier waves. The smallest amplitude that can be reliably detected visually is usually about 0.05 µV, but this is heavily dependent on the amount of background noise in the recording. On the average, at high intensity levels, the amplitude of wave V is 0.50 µV and for wave I it is 0.25 to 0.35 µV, producing a normal wave V:I amplitude ratio of 1.50. At low signal intensity levels, wave I latency is around 3.5 to 4 ms (versus the wave V latency at 7.5 to 8 ms). It would appear, then, that the wave I to V latency interval of about 4.0 ms generally holds constant across this range of stimulus intensity levels for normal subjects. While this is sometimes the case, the increase in latency with decrease intensity may not be precisely parallel for wave I versus wave V, or among waves in general (Pratt & Sohmer, 1976). A greater (increased) latency shift with declining intensity may occur for wave I than for wave V in normal adults, resulting in a net shortening of the interval between wave I and wave V latency by about 0.20 ms, or even more at lower intensity levels. The clinically important wave I to V latency intensity interval may also be influenced by subject characteristics (such as age, in children), by stimulus parameters other than intensity, by audiogram configuration, and, of course, by retrocochlear and brainstem auditory dysfunction.

ABR amplitude, even for wave V at very high intensity levels, rarely exceeds 1.0µV. As intensity level is decreased, amplitude for all wave components steadily diminishes.

As with latency, the trend is usually not linear, although some researchers have reported an essentially linear relationship (e.g., Starr & Achor, 1975: Wolfe et al., 1978). In addition, intensity-related amplitude changes are characteristically more variable than latency changes for subjects at all ages (Lasky, Rupert, & Walker, 1987). There are well-recognized interactions among stimulus intensity, rate, duration, and frequency. The preceding comments have pertained mostly to the influence of intensity on ABR latency and amplitude for click signals. Gorga, Ka-minski, and Beauchaine (1988) described latency-intensity functions for tone-burst stimuli (cosine2 gating functions) at frequencies from 250 Hz through 8000 Hz. Data were for 20 normal-hearing subjects. Latency was shorter for high versus low frequencies. Also, latency clearly decreased as intensity increased for all test frequencies. Intersubject variability was greater for lower than for higher frequencies. It appeared that the latency-intensity slopes were steeper for lower frequency stimuli. That is, the decrease in latency with intensity was greater for low than for high frequencies. Gorga and colleagues (1988) suggested that this convergence in functions among test frequencies at the highest intensity levels for low frequency stimuli was related to spread of activation to higher frequency regions.

 LATENCY-INTENSITY FUNCTION

Starr and Achor (1975) reported that the mean latency of wave V increases from 5.4 ms at 75 dB SL to 8.1 ms at 5 dB SL. The peak latency of wave III increases from 3.7 ms at 75 dB SL to 5.6 ms at 5 dB SL. From 75 to 25 dB SL, the peak latency for wave I increases from 1.4 to 2.9 ms.  Figure 10 shows the; change in peak latencies as a function of  intensity changes. Changes in intensity have a greater effect wave I than on the later components. Stockard et al (1979) reported that the peak latency of wave I is most affected by phase and the peak latency of wave V is least affected by phase.

Fig 10

The I-V IPL is also called the brainstem induction time or the central transmission time. Stockard et al.,  (1979) reported that the I-III and the I-V IPL decrease with intensity decrease from 70 to 30 dB SL, with latency shifts as large as 0.73 ms occurring for the I-V IPL. The III-V IPL shows a much smaller decrease with intensity. The decrease in the I-III and I-V IPLs with intensity decreases reflects the change in peak latency of wave 1. Wave I is particularly affected by intensity because it becomes bifid at higher intensities and because of the large transition between 40 and 60 dB SL. Stockard et al. (1979) concluded that the change in IPLs with intensity reflects the behavior of wave I, which is the surface reflection of the eighth-nerve action potential (AP). The two peaks of the action potential represent activity from neurons with high and low thresholds. At 40-60 dB SL (the transition zone), the thresholds for both neuron groups are activated and the AP amplitude dominance shifts from one peak to the other. This shift occurs to a lesser degree for wave III and hardly at all for the peak latency of wave V. Hence, the IPLs, particularly the I-III and the I-IV, increase with intensity.

Stockard et al. (1979) also showed that the inversion of phase from rarefaction to condensation clicks results in shifts in the I-III IPL by more than one standard deviation of the mean in the direction of smaller IPLs for the condensation clicks. This trend was reflected to a lesser extent in the I-V IPL and to a still lesser extent in the III-V IPL.

Stockard et al. (1979) also reported that the intersubject variability in the IPLs is large, ranging from 0.15 to 0.38 msec for one standard deviation. Eggermont, Don, and Brackmann (1980) reported that the distribution for the I—III and I-V IPLs was Gaussian, with a mean and standard deviation of 2.07 ms and 0.13 ms, respectively, for the former and 4.01 ms and 0.16 ms, respectively, for the latter IPL. Because of the larger intersubject variability and the effects from variation in the technical parameters, clinicians should develop their own IPL norms and not depend on published IPL norms.

The slope of the latency—intensity function is a measure of the steepness of the function. It is calculated by subtracting the latency at the higher intensity from that at the lower intensity and then dividing that remainder by the dB difference in the intensity levels. Slopes of less than or equal to 60µs/dB are considered to be normal. Values of less than 30µs/dB, seen in persons with high-frequency hearing loss, are also seen in normal-hearing persons when the slope calculation is made over the shallow portion of the latency-intensity function, that is at high intensities. Since the latency intensity function is nonlinear, even in normal-hearing subjects, it is an unreliable measure for differential diagnosis.

The latency-intensity change in newborns is in the range of 30 to 40 µsec/dB ( Lasky & Rupert, 1982). Age is also a determinant in the stimulus intensity level required for generating a just-detectable ABR. Among infants, there is some rather dated evidence that ABR threshold declines with age (Lasky, Rupert, & Walker, 1987). These authors reported a distinct reduction in the ABR threshold (on the order of 10 to 30 dB), even within the interval from 2 hours after birth to 50 hours after birth. ABR latency and amplitude values at these two times post-birth, when corrected for the threshold differences, were comparable. These studies did not, however, address several possible factors in the ABR threshold decrease with age, including imprecision in the placement of supra-aural earphones, collapse (closure) of the ear canal secondary to earphone cushion pressure on the cartilaginous portion, maturational changes in middle ear status, the likelihood of residual vernix caseosa in the external ear canal, and even mesenchyme in the middle ear space in the hours following birth. The latter two possible factors would essentially create a slight conductive hearing loss that could elevate thresholds and extend latencies for the ABR in newborn infants.

PHYSIOLOGIC EXPLANATIONS FOR THE LATENCY-INTENSITY FUNCTION

There are several explanations for the decrease in latency and increase in amplitude as intensity of a click stimulus increases. Even though a click contains energy across a broad frequency region, the site of ABR generation along the basilar membrane is, in part, related to intensity. Very high intensity levels activate the cochlea near the base. The site of cochlear activation then moves progressively toward the apex for lower intensity levels. At the lowest intensity levels (i.e., normal threshold), the portion of the cochlea representing frequencies 1000 to 2000 Hz generates the ABR. This apical shift in the primary place of stimulation along the basilar membrane produces roughly a 1 ms latency increase (Picton et al., 1981). Because wave V latency increases from 5.5 to 8 ms, a change of 2.5 ms, there must also be another mechanism in the intensity-latency function. One theory is that postsynaptic excitation potentials reach threshold faster for higher intensity levels, and therefore synaptic transmission time decreases.

The latency decrease with increasing transient (tone-burst or click) stimulus intensity is due to a progressively faster rising generator potential within the cochlea and a similarly faster development of excitatory postsynaptic potentials, or EPSPs (Moller, 1981). Compound AP latency is directly dependent on how quickly the generator potential and EPSPs increase and reach the threshold for firing. Higher intensity stimuli may also bring into play nonlinear activity within the cochlea and widening of tuning curves. These stimuli produce a shift in the place of maximal cochlear excitation along the basilar membrane toward the base. Travel time from the oval window to this more basal site is, of course, shorter and, therefore, so is compound AP latency.

One explanation of the physiologic mechanism underlying ABR, which takes into account stimulus intensity, frequency, and rate parameters, might be referred to as the "dual structure" or the "fast versus slow component" theory (Suzuki, Kobayashi, & Takagi, 1985). The ABR can be viewed as a broad, slow wave (i.e., slow component) upon which the characteristic wave components I through VI or VII (i.e., fast component) are superimposed. Spectral analysis of ABR suggests that the dominant energy around or below 100 Hz forms the slow component, and the energy peaks around 500 Hz and 900 Hz contribute mainly to the fast component (Hall, 1986). Presumably, slow-versus-fast components do not share the same neural generators (Davis, 1976). As stimulus (click) intensity is increased, amplitude of the slow component reaches a plateau in the 40 to 50 dB region, but the fast components (waves I through V at least) show the characteristic steady amplitude increase discussed previously.

One way of viewing this phenomenon is to think in terms of an amplitude ratio (slow-to-fast component amplitude). Intensity, therefore, produces an increase in the ratio of fast-to-slow components. The consequence of this dual component view for clinical ABR application is apparent. Namely, auditory threshold estimation, at low intensity levels, is dependent mainly on analysis of the slow component.

There are several theories on the physiologic basis for the two-segmented compound (whole nerve) AP with an increase of stimulus intensity. One theory is that there are two sets of primary fibers, a low- and a high-sensitivity population. The other viewpoint maintains that the slow-growing portion of the input-output function-(at lower stimulus intensity levels) reflects activity in the sharp tip portion of the tuning curves, and the rapidly growing portion of function for higher stimulus levels reflects recruitment of higher frequency neural units on the "tail" of the tuning curves (Oz-damar & Dallos, 1976).

AMPLITUDE

The amplitudes of the brainstem auditory-evoked potentials decrease as signal intensity decreases. There is considerable intersubjcct variability in amplitude. The amplitude of wave I increases gradually up to 55 dBSL and then increases rapidly at higher intensities. On the other hand, the amplitude of the IV—V complex increases linearly with intensity (Starr and Achor, 1975). In general, the amplitude of the IV-V complex is greater than that of wave 1. For example, Chiappa et al. (1979) reported that, when clicks were presented at a repetition rate of 10/s at 60 dB SL, the mean amplitude of the IV-V complex was 0.47 µV, the mean amplitude of wave 1 was 0.28 µV, and the mean amplitude of wave III was 0.23 µV. Chiappa et al. (1979) reported however, that the amplitude of wave I exceeded that of the IV—V complex in approximately 10% of his subjects. As the repetition rate increases, the amplitude of all the waves decreases.

Several researchers (Chiappa et al, 1979) have reported that the absolute amplitude of wave V is too unreliable for assessing the normalcy of wave V, Starr and Achor (1975) suggested using the IV-V:I amplitude ratio (amplitude of the IV—V complex divided by the amplitude of wave I). The IV-V:I amplitude ratio should be measured at moderate intensity levels, not exceeding 70 dBnHL. Starr and Achor (1975) reported that at intensities less than or equal to 65 dB SL, a IV-V:I amplitude ratio less than 0.5 was abnormal.

RELATION BETWEEN ABR AMPLITUDE AND LOUDNESS

One concept that is relevant in this discussion is the possible relationship between ABR amplitude-latency parameters and the behavioral perception of loudness. Intensity is a physical property of sound, and loudness is the perceptual correlate of intensity. Loudness is a subjective estimation of the strength of the stimulus. As an aside, there is a comparable correspondence between frequency (another physical property of sound) and the perception of pitch. Pratt and Sohmer (1977) attempted to correlate latency and amplitude for ABR and other AERs with psychophysical magnitude estimates of click stimuli. Amplitude, but not latency, was correlated with loudness estimation. In a follow-up study, Babkoff, Pratt, and Kempinski (1984) essentially replicated the earlier research and also reanalyzed the original results with a correction for the nonlinearity of the latency-intensity function. The second set of data and reanalysis showed closer agreement between electrophysiologic and psychophysical parameters, but did not alter the original conclusions. Similarly, Darling and Price (1990) failed to find a clear connection between ABR and the perception of loudness. On the other hand, Thornton and colleagues (Thornton, Farrell, & McSporran, 1989; Thornton, Yardley, & Farrell, 1987) describe a clinical protocol for estimating loudness discomfort level (LDL) with ABR wave V latency versus intensity data. Also, in a group of adult subjects (18 to 65 years), Serpanos, O'Malley, and Gravel (1997) "established a relationship between loudness and the ABR wave V latency for listeners with normal hearing, and flat cochlear hearing loss", whereas there was poor correlation between behavioral and electrophysiological intensity functions for patients with sloping sensorineural hearing loss.

REPLICABILITY

Hecox and Galambos (1974) evaluated the peak latency of wave V in 3 young normal adults over an 8 month period. The combined intrasession, intersession, and intersubject variability ranged from 0.09 to 0.31 ms; the larger end of the range of variability occurred near threshold. Hecox and Moushegian (1981) contend that peak latencies should be replicable within 0.2 ms for clicks presented at low to moderate intensity levels at click rates between 30 and 90 per second. Better replicability can be obtained at lower repetition rates and higher intensity levels. Chiappa et al. (1979) reported that, when 8 subjects were retested between 2 and 9 months later, no statistical differences in peak latencies or amplitudes resulted. Thus, the BAEPs are stable over time.

• TYPE OF STIMULUS

The ABR, as typically measured clinically with 0.1 ms click signals, is generated by higher frequencies in the click spectrum, at least in normal ears. An ABR evoked by moderately intense (e.g., 60 dB nHL) click signals that are delivered with conventional audiometric earphones, for example insert (ER-3A) earphones, reflects activation of the high-frequency region of the cochlea, roughly from 1000 though 8000 Hz. There is lack of agreement among investigators regarding the frequency region most important for generation of the ABR—that is, whether the ABR reflects activation of the 1000 to 4000 Hz region, 4000 to 8000 Hz region, or just frequency regions of the cochlea above 2000 Hz, above 3000 Hz, or above 4000 Hz. It is likely that differences among the studies in both methodology and subject characteristics account for the reported variability in the relationship between minimum response level for click stimuli and pure tone audiogram.

More apical (lower frequency) regions of the cochlea are also activated by the click, but these regions do not contribute to the ABR, at least in normal hearers. There are two reasons for this.

• First, the response to cochlear activation has already occurred in the higher frequency regions by the time the traveling wave has traversed the basilar membrane from the base to the apex of the cochlea to activate hair cells in this region.
• Second, the leading "front" of the traveling wave is more gradual (less abrupt) when it reaches the apical region and, consequently, the traveling wave is not as effective in generating synchronous firing of many eighth-nerve afferent fibers over a concentrated portion of the basilar membrane. Instead, smaller numbers of afferents sequentially fire over a more dispersed stretch of the basilar membrane.

            In persons with an impairment of auditory sensitivity for the higher frequency region, ABR generation may not necessarily follow this pattern. In addition, it is likely that the portion of the cochlea contributing to the ABR may vary as a function of response components (e.g., wave I vs. wave V) and stimulus intensity. For example, wave I appears to reflect more basal activation, whereas wave V may reflect activity from a more apical region. Also, at high stimulus intensity levels, there is spread of activation toward the apex, whereas at lower intensity levels, activation is limited more to the basal region. These points are important for meaningful clinical interpretation of the ABR.

In considering stimuli for clinical ABR measurement, it is important to keep in mind two general principles.

• First, the frequency specificity of a stimulus (i.e., the concentration of energy in a specific frequency region) is indirectly related to duration (Burkard, 1984). With very brief stimuli, energy tends to be distributed over more frequencies, whereas stimuli with longer duration (including rise/fall times and plateau time) are spectrally constrained.
• Second, there is generally a direct relationship between duration of the response and duration of the stimulus. That is, slower (longer latency) responses are activated best by slower (longer onset and duration) stimuli, whereas faster (shorter latency) responses require faster (shorter onset and duration stimuli).

Although the ABR can be most effectively elicited with click signals, their lack of frequency-specificity is a major drawback for clinical electrophysiological assessment of auditory function in infants and young children and, particularly, for estimation of auditory sensitivity a different 1 frequency regions. Currently, the use of tone-burst signals is the preferred technique for frequency-specific estimation of auditory function. The demand for an electrophysiologic technique for estimation of auditory sensitivity has increased markedly with the emergence of universal newborn hearing screening (UNHS). Newborn infants who do not pass hearing screening must be followed closely during the first few months after birth. If the hearing screening failure is confirmed, then diagnostic audiometry is indicated. A critical component of the diagnostic process is estimation of auditory sensitivity at different frequencies within the range of 500 to 4000 Hz. For the infant, hearing sensitivity within this frequency region is important for speech perception and for speech and language acquisition. Timely, accurate, and frequency-specific estimation of auditory sensitivity within the first two to four months after birth is an essential prerequisite for optimal audiologic management of infants with hearing impairment. The information on auditory sensitivity is critical for successful hearing aid fitting.

Duration

Synchronous firing of many neurons, which is a general physiologic underpinning of the ABR, is very dependent on an abrupt stimulus onset. The two practical consequences of this principle are that (1) the ABR is not heavily dependent on stimulus duration (Gorga et al., 1984) and (2) an almost instantaneous onset (almost always 0.1 ms or 100 µsec) click stimulus is routinely used in clinical ABR recordings, although click durations as short as 20 µsec (Coats, 1978) and as long as 400 µsec have been reported. For many clinical applications, the default stimulus duration is typically 0.1 ms.

Appreciation of the spectral characteristics of the click has traditionally been important for accurate interpretation of ABR in audiologic assessment, but not for neurodiagnostic or neurologic evaluation. Studies of the effects of stimulus duration on ABR have yielded mixed and rather unimpressive results. One of the earliest reports by Hecox, Squires, and Galambos (1976) described alterations in the ABR to changes in the duration (on time) and interburst interval (off time) of white-noise-burst stimuli in six normal-hearing female subjects. These investigators found an increase in ABR wave V latency (0.5 ms) and decreased amplitude as duration was increased from 0.5 to 30 ms, but these changes were not observed when the stimulus off time was lengthened (i.e., with a longer recovery period). On the basis of this observation, the authors concluded that the wave V component of the ABR was strictly an onset response, that is, the ABR changes were due to response recovery processes, not to duration. Gorga et al. (1984) estimated ABR and behavioral thresholds for 2000 Hz tone-burst stimuli (with 0.5 ms rise/fall times) ranging in duration from 1 to 512 ms. They demonstrated that stimulus duration does not affect ABR threshold for normal or hearing-impaired subjects, whereas behavioral thresholds decreased (improved) on the order of 10 to 12 dB per decade of time for normal subjects. Subjects with sensorineural hearing impairment showed less change in behavioral threshold with increased stimulation duration (5 dB per decade of time). The findings of this study are consistent with psychophysical data on temporal integration.

Funasaka and Ito (1986) investigated the effect of 3000 Hz tone bursts with durations of 5, 10, 20, and 30 ms on ABR. Subjects were 20 young adults. Rise/fall times were constant at 1 ms. Interstimulus intervals ranged from 80 to 140 ms. As stimulus duration was lengthened, there were increases in latency and amplitude of waves V and VI. Wave III latency remained unchanged, but amplitude decreased. The authors argue that these effects are not a function of the recovery process limitation. Instead, the duration differentially affects the slow wave (frequency) component of the ABR and not the fast component.

ABR latency increases directly with stimulus rise times, beginning with instantaneous (0 ms) onset stimuli, at least for normal-hearing subjects (Hecox, Squires, & Galambos, 1976). When rise time exceeds 5 ms, identification of earlier ABR wave components, such as wave I, becomes difficult. The physiologic basis for this general effect is a reduction in the amount of neural units that fire synchronously (Spoendlin, 1972). Also, because the traveling wave is slower, there is an increased contribution of the more apical regions of the cochlea to the ABR (Kiang, 1975). An additional possible factor is activation of more basal cochlear regions by the increased proportion of spectral energy in higher frequency regions for briefer versus longer stimuli.

STIMULUS OFFSET ABR 

Basic studies of the auditory CNS have provided evidence of a variety of functional neuron types (Tsuchitani, 1983). Two of these types are onset neurons, which fire only at the onset of a stimulus, and offset neurons, which fire only at the offset of a stimulus (when the stimulus is turned off). As typically recorded, ABR is thought to reflect synchronous firing of onset neurons. For a click stimulus with the conventional duration of 0.1 milliseconds, stimulus onset and offset occur almost simultaneously, and identification of any offset contribution to the response is impossible. Over the years, papers have sporadically appeared describing AERs generated by the offset portion of stimuli. Early work in this area was conducted with the ALR (Rose & Malone, 1965). An offset ALR resembling the onset ALR was recorded in all subjects showing an onset response and it was not systematically affected by stimulus frequency or rise/fall time. Prolonged stimulus duration of 850 to 1500 ms was required to elicit the offset response (Rose & Malone, 1965).

Studies of ABR measurement with offset stimuli are not conclusive and, in fact, the existence of a true offset ABR is somewhat controversial (Antonelli & Grandori, 1984). The offset ABR is generally less distinct than the onset response. A long-duration stimulus (tone burst or noise burst) is necessary to separate in time the offset response from the onset response, yet with a tone burst of 15 to 20 ms, the offset ABR may be obscured by the AMLR generated by stimulus onset.

Clinical studies, conducted with modest numbers of normal-hearing subjects, suggest that in comparison to onset responses, the offset responses (a) are not as robust (70 to 80% smaller amplitude) or as reliably recorded, (b) have a higher (poorer) threshold (by 10 to 20 dB), and (c) may be reversed in polarity (showing downward peaks) with white-noise-burst stimuli . The offset response is recorded with a stimulus of extended duration (e.g., 10 ms duration or longer) to prevent overlapping with the invariable onset response. The problem with this method, at least when rise/fall times are very brief (less than 5 ms) is interference of offset identification by AMLR activity. There is also some concern in human investigations that what is thought to be an offset response is, in fact, produced by acoustic transducer ringing that follows stimulus onset (Brinkmann & Scherg, 1979). In short, offset ABRs are poorly understood, at best. More normal descriptive research is needed on the relationship of stimulus parameters, such as intensity, duration (rise/fall time, plateau), presentation rate, the type of stimulus (noise versus tone burst), and response acquisition parameters (e.g., filtering) to offset ABRs. Nonetheless, with careful stimulus selection, including stimuli characterized by no ringing artifact, it is possible to record a reliable ABR for the offset of tonal stimulation.

SUMMARY

Click duration does not have a marked influence on ABR latency or amplitude. There is no latency change for stimulus durations ranging from 0.25 to 100 µsec, and an increase in latency of 0.2 ms, at most, can be expected for durations ranging from 100 through 400 µsec. Nonetheless, click duration should be routinely specified and used in a consistent manner in clinical ABR measurement. A complete discussion of stimulus duration, although seemingly straightforward, actually leads to concerns about the possible effects of related stimulus characteristics. For example, duration directly influences frequency content of the stimulus and to the audibility of the stimulus. Duration effects also interact with the envelope of the rising portion of the stimulus and whether the onset slope is constant or variable. Finally, current understanding of stimulus duration effects is limited to data obtained from young, normal-hearing subjects. There is no published study of click duration in older subjects and/ or in those with hearing impairment, even though these and perhaps other subject characteristics might be expected to interact with duration

OTHER STIMULUS TYPES

1. FILTERED CLICKS

In addition to clicks and tone bursts, miscellaneous additional types of acoustic stimuli have been reported in ABR measurement, often in animal models rather than in patients. For example, filtered clicks are produced when a wide-spectrum click (e.g., the usual unfiltered or raw click resulting from delivering a rectangular electric pulse to a transducer) is passed through a set or series of filters to produce transient stimuli, with energy centered at desired frequencies (Arlinger, 1981). ABRs (Lehnhardt, 1982) and later latency AERs have been elicited with chirps or linear frequency ramps. These stimuli consist of a sweep through a defined frequency range (e.g., 1200 to 1700 or 4200 to 4700 Hz) over a defined period of time (e.g., 10 ms). The sweep frequencies can be rising or falling. Intensity level is determined at the center frequency of the ramp.

2. PAIRED CLICKS

Ernest Moore and colleagues   describe another component of the ABR, a potential evoked by presentation of two closely spaced click stimuli (0.1 ms duration). With the paired-click stimulus paradigm, the presentation of a "standard" click is followed soon by a second click. The time difference (delta t) between the two clicks in the pair is manipulated, with interstimulus intervals ranging from 4.0 ms down to only 0.1 ms. As stated by Davis-Gunter et al. (2001), "These time intervals were chosen to be shorter than, encompass, as well as exceed the duration of the absolute (1.0 ms) and relative (4 to 5 ms) refractory periods of the VIIIth nerve. The first click, of course, generates combined (ensemble) action potentials (AP) in the distal portion of the auditory nerve, the ABR wave I. If the second click is presented before the auditory nerve fibers have fully recovered from firing (during their refractory period), it presumably will not generate an AR The second click will, however, produce excitatory postsynaptic potentials (EPSPs) that are reflected within the activity measured as an ABR. To isolate and identify the EPSP activity, the authors utilize a derived response technique, i.e., the waveform for the first (standard) click is subtracted from the waveform for pair of clicks. In theory, the derived response (the difference wave) consists of only EPSP activity. When Davis-Gunter et al. (2001) performed the paired stimulus and derived response analysis technique with three normal-hearing adult subjects, the authors identified two waves—1⁰ and I¹—appearing before the conventional wave I. As reported previously (Moore et al., 1992), the average latency for wave I¹ was 0.97 ms, whereas conventional wave I latency was 1.83 ms. The authors speculate "that peaks 1⁰ and I¹ represent the summating potential and the generator potential, generated by the cochlea and VIII th nerve dendrites, respectively" (Davis-Gunter et al., 2001).

3. PLOPS

In a study of novel stimulation and analysis techniques, Scherg and Speulda (1982) recorded ABRs with conventional clicks (100 µsec square-wave pulses) and with Gaussian-shaped impulses centered around 1000 Hz, referred to by the authors as "plops." Stimuli were presented separately for three polarity modes (alternating, rarefaction, and condensation), via a TDH-39 earphone. The envelope of the acoustic waveform for the plop, resembled that of a click, but it lacked the ringing (and added frequency components) of the click waveform. It appeared that absolute latency values for wave I, wave III, and wave V were greater for the "plop" versus the click, while other ABR parameters (absolute amplitude of wave V and interwave latencies) were similar for the two stimuli types. A latency delay for a stimulus with a center frequency of 1000 Hz versus a click is expected because the click activates a more basal region of the cochlea.

4. CHIRPS

There is consensus that the ABR evoked by conventional click stimulates on is dominated by activation of the basal region of the cochlea, mostly frequencies above 2000 Hz. Attempts to enhance the contribution of other regions of the cochlea to ABR generation include the generation of rather unique types of stimuli, such as "chirps" and sophisticated recording techniques, such as "stacked ABR." The chirp stimulus is designed mathematically "to produce simultaneous displacement maxima along the cochlear partition by compensating for frequency-dependent traveling-time differences" (Fobel & Dau, 2004). In theory, the chirp will optimize synchronization across a broad frequency region at high and low intensity levels, yielding a more robust ABR than the conventional click stimulus.

Fobel and Dau (2004) designed two chirp stimuli for elicitation of the ABR. One—referred to as the O-chirp—was derived from previously published group-delay data (Shera & Guinan, 2000) from stimulus frequency OAEs (the term "O" chirp refers to the derivation from an OAE stimulus). The other stimulus— referred to as the A-chirp—was designed with reference to data demonstrating the relationship between tone burst frequency and ABR latency (the term "A" chirp refers to the derivation from ABR data). ABRs generated by these two chirp stimuli were compared also to a previously developed type of chirp (Dau et al., 2000)—the M-chirp—based on a model (the "M" refers to "Model") for producing a flat-spectrum stimulus. Fobel and Dau (2004) recorded ABRs from 9 normal-hearing adult subjects with the different chirp stimuli, and with conventional click stimuli. Since the frequency composition of the chirp stimuli covered a range from 0.1 to 10,000 Hz, durations for the chirp stimuli were remarkably long in comparison to clicks. Duration for the O-chirp was 13.52 ms, whereas duration for the M-chirp was 10.32. For the A-chirp, however, duration varied as a function of stimulus intensity, from 12.72 ms at 10 dB SL (sensation level) to 5.72 ms at 60 dB SL. Each of the chirp stimuli, however, evoked ABRs with larger amplitude values than those elicited by conventional click stimuli. Among the three types of chirps, the A-chirp produced the most robust ABR waveforms, and it "is particularly effective at very low [intensity] levels where wave-V amplitude is about three times as large as for the click" (Fobel & Dau, 2004). The authors speculate on the potential benefits of the A-chirp for clinical application of ABR, especially for estimation of auditory thresholds.

TONE BURSTS

Tone bursts are now regularly used in clinical ABR measurement for estimation of auditory sensitivity for discrete frequency regions, especially in infants and young children. The minimum response level for the wave V component of the ABR evoked by tone-burst stimuli is recorded within 10 dB of the behavioral threshold for a comparable pure-tone frequency for the majority of patients with sensory hearing loss, with over 90 percent of patients yielding a difference within ±20 dB.  A number of investigators have examined with ABR  forward masking phenomenon. Walton, Orlando, and Burkard (1999) utilized tone-burst maskers and probe signals in an investigation of the recovery from forward masking as a function of age in adults. Using the forward masking paradigm, the authors found an age difference in the latency shift of wave V with short intervals (e.g., 16 ms) between the masker and the probe for higher tone-burst frequencies (e.g., 4000 and 8000 Hz), but not for a lower frequency (1000 Hz). Recovery from masking, that is, baseline latency values for the ABR, was always complete under all stimulus conditions and age groups with an interval of at least 64 ms between masker and probe signal.

MODULATED TONES

There are also descriptions of AER generation with stimuli that are frequency modulated (Eggermont & Odenthal, 1974) and with stimuli that are amplitude modulated ( Milford & Birchall, 1989). Legendre sequences and maximum length sequences (MLS) of pulse trains have also been described in stimulation of ABR. Although these techniques may potentially increase efficiency of ABR data collection and reduce test time, clinical confirmation is lacking.

STIMULUS TRAINS

Tietze (1980) reported two clever techniques for simultaneous stimulation and recording of the ABR and ALR. The methodological problem in simultaneously recording these two AERs is that the ABR requires stimuli with rather abrupt onset (e.g., 2 to 4 cycles), a similarly brief duration, and short interstimulus intervals (e.g., 25 ms), whereas the ALR is best elicited with stimuli having relatively leisurely rise/fall times (e.g., 8 to 30 ms), plateau durations (30 t o 500 ms) and interstimulus intervals of approximately 2.5 seconds. With one technique, trains of tone pips are presented with an interval of 2.5                  seconds between each train. Each train has the effect of a single stimulus unit in eliciting the ALR. However, within each train, individual tone pips at intervals of 25 ms serve as the stimuli for the ABR. The second technique is similar. The individual tone pips continue to serve as the ABR stimuli, and between each group of tone pips, a tone burst of a slightly lower intensity level is inserted, to evoke the ALR. Because the ALR is a larger amplitude response  (i.e., a more robust signal), fewer stimuli must be averaged to obtain a stable waveform. As a consequence, although relatively few trains of tone pips versus numerous individual tone pips are presented per given unit of time, the number of stimuli averaged is equally adequate for the ABR and ALR.

Trains with multiple stimuli (20, 56, and more), often presented at very rapid rates, are also used to evoke the ABR in an attempt to minimize test time (Henry et al., 2000). One undesirable outcome often associated with the rapid presentation of a train of stimuli (clicks or tone bursts) is a reduction in the response amplitude and prolongation of latency, presumably secondary to adaptation within the auditory system.

NOISE STIMULI

Noise signals, presented alone or in the presence of click or tone-burst stimuli, are often described in the clinical ABR literature. Reasons for incorporating noise as the stimulus, or with the stimulus, vary among studies of the ABR and include investigation of fundamental auditory phenomenon (e.g., temporal response properties of the cochlea and auditory nerve) or more clinically applicable objectives (e.g., achieving greater frequency specificity for clinical assessment of infants and young children). Noise stimuli used to measure gap detection, a psychophysical procedure for assessing temporal auditory processing, or temporal resolution, are also effective as stimuli for generation of the ABR (Poth, Boettcher, Mills, & Dubno, 2001). An ABR is evoked by an initial noise burst with a duration of >15 ms, and then within milliseconds a second noise burst is presented as a stimulus for a second ABR. The silent interval—the gap—is usually varied over the range of 0 ms to over 100 ms. A basic assumption underlying the electrophysiological measurement of gap detection is that the ABR for the second of the two noise bursts will be unchanged if the silent gap is fully processed by auditory system—that is, the gap is equal to or exceeds the interval required for temporal resolution. Normative data are collected for the ABR evoked by a noise burst presented following another noise burst, i.e., the normal gap detection threshold is defined. Changes in the latency of ABR wave V, or absence of an ABR, for a stimulus presented after a silent gap that is detected by normal subjects, i.e., a gap duration that is long enough to not interfere with the ABR, are associated with deficits in temporal auditory processing. Normal changes in the ABR with shorter gap durations include latency prolongation and amplitude reduction. A detectable ABR is present in young normal subjects with gap durations as short as 8 ms, whereas the ABR may not be present when the silent gap is a short as 4 ms (Poth et al., 2001).

Werner et al. (2001) investigated application of the ABR as an electrophysiological measure of temporal processing with a gap detection method. Subjects were 35 young normal adult subjects and 30 infants, ten who were 3 months old and 20 who were 6 months old. Stimuli were a pair of 15 ms bursts of broadband noise separated by silent gaps ranging in duration from 0 to 125 ms. In one experiment, Werner et al. (2001) found that gap detection thresholds determined with ABR (2.4 ms) were on the average similar to those obtained with conventional psychophysical methods (2.9 ms). In another experiment, the authors recorded for subjects with sloping high-frequency sensorineural hearing loss higher gap detection thresholds (longer thresholds for silent gaps) with both the ABR (12.7 ms) and the psychophysical techniques (10.7 ms). In contrast, data recorded from infants revealed a difference in the gap detection thresholds as measured with the electrophysiological versus psychophysical methods. Temporal resolution was immature for infants (longer silent gaps were required for detection) as measured with the psychophysical method, whereas developing age did not influence ABR gap threshold. According to Werner et al. (2001), these findings "suggest that it is not immaturity at the level of the brainstem that is responsible for infant's poor gap detection performance" .

Poth et al. (2001) described the use of broadband noise with silent gaps as a stimulus for investigating auditory temporal processing with the ABR. The electrophysiological study of gap detection using noise bursts was a clinical follow-up to experiments conducted by the authors with various animal models. Stimuli consisted of 50 ms broadband noise bursts interrupted by a silent period varying in duration from 4 to 64 ms. This ABR stimulus paradigm is comparable to the stimuli used in psychophysical measurement of temporal resolution. Poth et al. (2001) reported that ABR amplitudes were reduced, and in a group of older subjects (> 60 years), the proportion of subjects yielding measurable responses was diminished. In other words, longer gaps of silence were required for generation of a normal ABR for older subjects.

STIMULATION OF "STACKED ABR."

            Dr. Manny Don and colleagues at the House Ear Institute developed the "stacked ABR" technique with the goal of detecting with greater accuracy and sensitivity than convention click stimuli retro-cochlear auditory dysfunction, particularly small acoustic tumors (Don, Masuda, Nelson, & Brackmann, 1997). The technique is an outgrowth of previous investigation by the authors and others of the effects of ipsilateral high-pass masking on "cochlear response times," i.e., traveling wave distance and velocity along the basilar membrane.

The click stimuli used to evoke the ABR include energy from a wide frequency region, yet the response is mostly generated by stimulus-related activity for the high-frequency region of the cochlea and, correspondingly afferent fibers in the eighth cranial (auditory) nerve serving this region. Development of the stacked ABR technique was motivated by appreciation that small acoustic tumors often have negligible impact on functional integrity of high-frequency neural fibers and, therefore, little or no influence on the click-evoked ABR. Even when an acoustic tumor does impinge on high-frequency neural fibers with resulting disruption in their capability to fire synchronously, an ABR may still be generated by frequencies in another portion of the broadband click spectrum. In either case, the result is a normal ABR in a patient with an acoustic tumor—i.e., a false-negative diagnostic outcome. To circumvent this limitation of conventional click stimuli, the stacked ABR technique calls for the use of bands of ipsilateral masking to derive a series of ABRs, each generated by cochlear activity within a designated frequency region and, therefore, activity of a tonotopically defined set of nerve fibers.

Measurement of the stacked ABR begins with conventional click stimuli of 0.1 ms duration and rarefaction polarity, presented at interstimulus intervals of 22 ms and an intensity level of 93 dB peak-to-peak SPL. Then, bands of pink noise are presented ipsilaterally at levels sufficient to mask the ABR evoked by the click stimulus. High-pass masking noise with five different low-frequency cutoffs (8000, 4000, 2000, 1000, and 500 Hz) is sequentially presented with the click stimulus. A total of 6 ABR waveforms are thus obtained— one for the conventional click and then one waveform generated by a different frequency region, i.e., only frequencies above 8000 Hz, then only for frequencies above 4000 Hz, and so forth. In other words, with each change in the cutoff for the ipsilateral masking, more of the lower frequency region is masked and its contribution to the ABR removed. Once the ABRs are recorded,     ABRs for different frequency regions are derived by digitally subtracting the ABR waveform for one ipsilateral masking condition from the ABR waveform generated by the previously recorded condition. The ABRs for each of the derived bands reflect activity of the cochlea region and auditory nerve fibers that serve the frequencies within the band, with characteristic frequencies of 11,300 Hz, 5700 Hz, 2800 Hz, 1400 Hz, and 700 Hz. Wave V latencies of the ABRs for each of the derived bands vary predictably. The derived band with the highest characteristic frequency (11,300 Hz) produces the ABR with the shortest wave V latency, and wave V latency is progressively longer for subsequent lower frequency bands.

After the ABRs for each frequency region are derived, they are aligned according to wave V latency, that is, lower frequency ABR waveforms are shifted to the left until wave V latencies for all bands are the same as the wave V latency for the 11,300 Hz band. The term "stacked ABR" is therefore quite appropriate as the result of the manipulation of ABR waveforms according to wave V latency produces a stacking of waveforms. Each waveform represents ABR energy generated by stimulation of a band of frequencies. If all of the frequency bands are normally represented in the overall ABR—that is, the cochlea and the auditory nerves are functionally intact within each frequency region—then adding them together should result in an ABR equivalent in amplitude to the click-evoked ABR. If, however, an acoustic tumor is affecting neural integrity within one or more of the frequency regions, the amplitude sum of the five individual stacked ABRs will be less than the amplitude of the click ABR. The influence of an acoustic tumor on the auditory nerve, therefore, is inferred by a reduction in summed amplitude of the stacked ABRs versus normal expectations (the "standard" ABR for the conventional unmasked click).

SPEECH STIMULI

            Ample evidence now exists confirming the feasibility of generating an ABR and a frequency-following response (FFR) with speech stimuli. Nina Kraus and colleagues at Northwestern University have published a series of articles describing in detail characteristics of the ABR and FFR evoked by speech sounds (Wible, Nicol, & Kraus, 2005). These researchers further introduce the application of speech-evoked ABR and FFR as a tool for investigating the neural representation of speech processing at the brainstem and for documenting neural plasticity with auditory training. The 40 ms speech stimulus (/da/) in each of the studies just cited was a synthetically generated stop consonant (/d) and shortened vowel (/a/) consisting of a fundamental frequency and five carefully defined formants. The speech stimuli were typically presented via insert earphones with alternating polarity at an intensity level of 80 dB SPL and in trains of four stimuli with an interval of 12 ms between the offset of one train and the onset of a consecutive train. Subjects were distracted during the ABR recording by watching a videotape of a movie or cartoon with a low-intensity sound track.

The ABR waveform evoked by the speech stimulus (/da/) consisted of an initial segment within 10 ms that contained the typical positive ABR peaks (e.g., wave III and wave V, and sometimes wave I) and a negative wave (labeled wave A) that resembled the SN10 wave. After the transient portion of the ABR, there was within the next 30 to 40 ms a complex FFR waveform—the "sustained response"—that contained two rather stable waves, referred to as peak C (latency of about 18 ms) and peak F (latency of about 40 ms). As noted by Russo et al. (2004), "The defining feature of the sustained portion of the response is its periodicity, which follows the frequency information contained in the stimulus". Analysis of the waveform included calculations of interpeak latency intervals, and the amplitude, slope, and area of individual peaks. The speech-evoked ABR was recorded in quiet (no background noise) and also in the presence of noise to evaluate speech processing in an adverse listening condition. The authors of the above-noted papers describe five analysis techniques for extracting speech stimulus related information from the complex FFR waveform:

(1) calculation of root mean square (RMS) amplitude

(2) calculation of the amplitude for a spectral component evoked by the fundamental frequency of the stimulus

(3) calculation of the amplitude for a spectral component evoked by the first formant frequencies of the stimulus

(4) correlations of the stimulus-to-noise response

 (5) correlations between the ABR and FFR recorded in quiet versus noise.

 An additional waveform analysis technique ("wavelet-denoising") was required to identify speech-evoked peaks in the noise condition. The speech-evoked brainstem response faithfully reflects many acoustic properties of the speech signal. In the normally perceiving auditory system, stimulus timing, on the order of fractions of milliseconds, is accurately and precisely represented at the level of the brainstem.

BONE CONDUCTION STIMULATION IN ABR MEASUREMENT

            Bone conduction ABR is an essential component of the test battery for auditory assessment of infants and young children. Comparison of threshold estimation cased air versus bone conduction stimulation permits objective documentation of the degree of air-bone gap, and differentiation among conductive, sensorineural, and mixed hearing losses, even in patients who cannot be properly evaluated with behavioural audiologic techniques. Evidence in support of bone conduction ABR as a clinically viable technique in infants dates back to the 1980s. (Hooks and Weber, 1984). Hooks and Weber (1984) assessed 40 premature infants with both air- (TDH-49 earphone) and bone-conduction (Radioear B-70A vibrator) click stimuli. In 36 out of 40 infants, a mastoid bone-vibrator placement was used, and forehead placement was used in the remainder. A significantly larger proportion of infants showed an ABR for a stimulus-intensity level of 30 dB nHL for bone-coonduction (93%) than for air-conduction (73%).

            Similar bone- versus air-conduction statistics were found at a 45 dB intensity level. Because of technical problems (mostly excessive stimulus artifact), two subjects had an interpretable ABR for air but not for bone conduction. Contrary to the expectations for adult subjects, latencies for ABR wave I, wave III, and wave V were shorter (by about 0.30 to 0.45 ms) for bone-conduction stimuli than for air-conduction stimuli. These authors (and later Yang et al., 1987) speculate that the earlier bone-conduction latencies are due to the pattern of cochlear development in the newborn. In the immature cochlea, responsiveness to low-frequency stimuli develops initially in the basal regions, which are the place for high- frequency responsiveness in the adult cochlea (Rubel & Ryals, 1983).

            In the study by Yang et al., (1987), ABRs with bone conduction signals were recorded from three sets of patients: adults, 1 –year old children, and healthy neonates tested between  24 and 72 hours after birth. The intensity levels were  15, 25, and 35 dB nHL. ABR results for three vibrator surface placements were analyzed:

1.  On the frontal bone (midline forehead)
2. On the occipital bone (I cm lateral the ipsilateral occipital protuberance), and
3.  On the temporal bone (superior post auricular area).

            Spectra for the bone vibrator versus the TDH-39 earphone were described.

Some of the important findings of the Yang et al., (1987) study were as follows. Latencies varied as a function of

1.  Air – versus bone-conduction stimulation,
2.  Vibrator placements for bone conduction, and as expected,
3.  Age.

In adults, wave V latency was the shortest for air-conduction stimulation. The temporal-bone placement yielded the next shortest latency values, while frontal and occipital bone placement latency values were longer and comparable. As noted, effective stimulus intensity for a brief duration stimulus decreased by about 7 dB (Boezman et al., 1983) when the vibrator was moved from the mastoid to the frontal bone. The latency patterns in the Yang etal., (1987) study was varied somewhat for 1 year old infants, in that frontal versus occipital bone placements were associated with different latencies.  A remarkable finding of this study was the very unique latency versus placement patterns observed for the neonates. For temporal bone placement, wave V latency was markedly shorter than for the other two bone vibrator locations and was slightly shorter than even the air-conduction latency values.

            Tucci DL, Ruth RA, Lambert PR (1990) evaluated the Wave I component of the BC-ABR and determined  the utility of this response in assessment of cochlear reserve. The source of Wave I has been shown to be the distal eighth nerve. It was postulated that the presence or absence of this component would provide ear specific information useful for determination of cochlear integrity. In order to test this hypothesis, patients with a documented unilateral hearing loss were studied. Stimulus presentation was via the Radioear B-70 bone vibrator used in conventional audiometric assessment. Evoked potential responses were recorded at four presentation levels. Subjects had either normal hearing bilaterally or normal hearing in one ear and a mild-to-profound sensorineural hearing loss in the opposite ear. Their  data indicated  that the Wave I response, when measured in this fashion, is ear specific. Ear specificity was shown to be aided by good waveform morphology, as typically observed in younger subjects, and by a relatively large discrepancy in hearing thresholds between the normal and hearing-impaired ears. This technique may be of value in determination of cochlear reserve in patients with problematic masking dilemmas. With bone conduction ABR wave V latency at 60 dBnHL is on an average 0.59 ms greater than air conduction values for adults and 0.67 ms greater for infants (Cornacchia, Martini and Morra, 1983).There is clinical evidence that the bone conduction ABR measurement can be useful in circumventing the Masking dilemma even in patients with maximum conductive impairment as in case of aural atresia because  of the wave I component obtained from an electrode on or near the ear ipsilateral to the stimulus conforms the contribution of the stimulated ear to the response, regardless of  whether masking is presented to the non- test ear.

PITFALLS

• The maximum effective intensity level for bone conduction stimulation is about 55 dBnHL
• The effective range of intensity to obtain a relatively large V wave in bone conduction ABR is in the order of 30 to 40  dB [55 dB maximum minus the 20 dB ABR threshold (difference between the behavioural and the ABR)]
• The electromagnetic energy radiating from the bone vibrator can cause serious stimulus artifacts especially when the mastoid is the location for the electrode as well as the bone vibrator
• Using the click stimuli in determining the air – bone difference in ABR may underestimate the low frequency thresholds as most of the conductive impairment are in the regions below 1000 Hz.
• Masking dilemma in serious bilateral conductive impairment.

 STIMULUS RATE

As the click rate (number of clicks per second, also referred to as the repetition rate, RR) increases, the absolute latencies of all the BAEP components and the interpeak latencies (IPLs) increase. Nevertheless, the absolute latency of wave V is not substantially prolonged until RRs exceeding 30 Hz are obtained.

At high RRs, waves IV and V often merge. Also, as the RR increases, the amplitude of the BAEP components decreases, particularly for the earlier components (Pratt &: Sohmer, 1976). At high RRs, the IV-V:I amplitude ratio increases since the amplitude of wave I is more adversely affected than that of the IV-V complex by the rate increase.

The BAEP waveform may be unidentifiable at high RRS. Repetition rates below 33 Hz are recommended for routine clinical use, particularly for the identification of wave 1 in neurotologic diagnosis. To prevent the appe arance of the 60-Hz hum in the BAEP waveform, the stimulus rate should not be a multiple or harmonic of 60Hz. Thus, RRs of 11.4 or 33.1 are acceptable. RRs of 30 or 10 Hz are unacceptable.

When click rate is increased, the absolute latencies of waves I and V a re increased. Wave V is prolonged to a greater extent than wave I so the I-V IPL increases. The increase in IPL with stimulus rate increases is smaller and moderate (50 dB SL) than at high (70 dB SL) intensities, since wave I is more affected than wave V by the rate increase at moderate than at higher intensities.

Fig 6

Chiappa (1979) reported that the frequency of recognizability of the earlier waves decreases at high RRs. The stimulus rare has little effect on the BAEP threshold. Figure 6 shows the effect of RR on the BAEP waveform.

            Some investigators have suggested that the use of high repetition rates of stimulation to stress the auditory system may reveal subtle abnormalities. Since an increased repetition rate may result in increased peak latencies and decreased peak amplitudes in normal-hearing persons, a shift in latency or decrease in amplitude as a result of an increase in repetition rate can be considered significant only if the change is greater than what would be seen in normal-hearing persons.

            Musiek and Gollegly (1985) use the following formula to determine whether a shift in the peak latency of wave V as a result of the increased repetition rate is significant

v For even 10 Hz increase in the repetition rate, 0.1 ms is added to the tolerance limit and a variability of 0.2 ms around the final result is acceptable.

          

Hecox (1980) uses the formula

v 0.006 x high rate – low click rate + 0.4ms = maximum acceptable latency shift

There is disagreement in the literature concerning the clinical utility of varying repetition rate to elicit an abnormality.

• Hecox (1980) reported that only 4% of the neurologically impaired have BAEP abnormalities at high but not at low repetition rates.
• Elidan et al. (1982) reported that, in their group of patients with multiple sclerosis, there was no case in which a normal BAEP was obtained for standard click stimuli (75 dBnHL presentation level, 10-20 clicks/s and an abnormal response was obtained when the repetition rate was increased to 50 clicks/s or 80 clicks/s and/or the the presentation level was lowered. Nevertheless, the intensity .and repetition rate manipulations did make any abnormality seen at standard click intensity and repetition rate more pronounced.
• Discrepant findings were obtained by Stockard, Stockard, and Sharbrough (1977) who evaluated the BAEPs in 100 multiple sclerosis patients. They reported that, as the repetition rate increased, the number of patients showing BAEP abnormalities was increased.

RATE-RELATED ABR FINDINGS IN AUDITORY PATHOLOGY

Some investigators have suggested increased stimulation rate as an effective technique for detecting subtle auditory neuropathology (Don et al., 1977), presumably because the nervous system is stressed beyond its functional capacity. Pratt et al. (1981) speculated that the sensitivity of high-rate ABRs to neuropathology was specific to white versus gray matter. In support of this contention, abnormal latency shifts or disappearance of later waves at very rapid stimulus rates have been reported in various types of peripheral and CNS pathology, including eighth-nerve tumors, head injury, hypoxia, mixed CNS diseases, and multiple sclerosis (MS). Other authors confirmed a relatively greater degree of abnormality for the faster stimulus rates in neuropathology. However, they generally found that abnormal ABRs also were recorded at the conventional rate.

Mechanisms for the excessive latency shifts with increased rate in auditory pathology, when they occur, are proposed but not yet confirmed. For example, Yagi and Kaga (1979) note that the abnormal rate-related latency changes may have an essentially different basis than normal latency shifts, citing data from experimental studies of induced axon demyelination or neuron synapse disorders. Other authors, however, describe comparable degrees of latency shift at high stimulus levels for ABR waves in auditory pathologies, such as cochlear impairment of various etiologies (Fowler & Noffsinger, 1983) and different neuropathologies.

PHYSIOLOGIC BASES OF RATE EFFECTS

Several investigators have speculated on possible neurophysiologic mechanisms underlying the divergent effect of increased rate on ABR latency versus amplitude. One physiologic explanation offered for the overall rate effect is a cumulative neural fatigue and adaptation, and incomplete recovery, involving hair cell-cochlear nerve junctions and also subsequent synaptic transmission. The effect of rate would, by this theory, be additive as the number of synapses increased (from wave I through wave V). Pratt and Sohmer (1976) have attempted to reconcile this discrepancy, theorizing that adaptation may not be precisely uniform for all neurons. This would result in desynchronization of the response and prolonged latency. Temporal summation would remain adequate for amplitude preservation. The role of divergence and convergence of auditory neurons from lower to higher order auditory neurons has also been implicated.

According to a number of investigators ( Suzuki, Kobayashi, & Takagi, 1985), the ABR consists of two major spectral components—a slow component (energy at frequencies of 100 Hz and below) and a fast component (energy mostly at frequencies in the regions of 500 and 900 Hz). This dual nature of the ABR is easily appreciated by inspecting the typical ABR (recorded with wide filter settings). The ABR is a slow wave on which the fast components (waves I through VII) are superimposed. There is a physiologically based distinction in the effects of stimulus rate, intensity, and frequency on these fast-versus-slow ABR components. Suzuki, Kobayashi, and Takagi (1985) recorded ABRs for signal rates of 8/second up to 90.9/second, then performed power spectral analysis, and then digitally separated the ABR waveforms into a slow component (0 to 400 Hz) and fast component (400 to 1500 Hz). Slow-component amplitude was relatively constant across this range of stimulus rates, whereas amplitude of ABR waves I through V (the ABR fast component) decreased. Latency of each component increased with rate. Interestingly, slow component amplitude, which did decrease very slightly with increasing rate, paradoxically showed an amplitude increase at a rate of 40 Hz. These authors point out that the differential effect of rapid stimulus rate on ABR latency (an increase) versus amplitude (essentially no change) reported by others may be explained by this dual nature of the ABR. ABR amplitude is especially resistant to rate effects when stimulus intensity is maintained below about 50 dB.

 STIMULUS POLARITY

            Rarefaction (R), condensation (C), or alternating (A) polarity (phase) signals are employed in BAEP assessment. Rarefaction clicks tend to be associated with shorter absolute peak latencies than condensation (C) clicks presented at 70 dB SL (Stockard et al, 1979). The difference between the peak latencies obtained with C and R clicks is greatest for wave I and smallest for wave V.

R Clicks and C Clicks:

• Shorter latencies for wave I for R clicks
• Some investigators have reported increased amplitude and resolution of wave I with R clicks (Kevanishvili 1981).
• Because of the shorter latencies obtained with the R clicks for wave I and the lesser effect of polarity on the peak latency of wave V; the interpeak latencies involving wave I are longer for R clicks than for C clicks.
• At 50 dB SL, a moderate intensity level, waves I and III often appeared as broad and double peaked with R clicks.
• No morphologic differences with R as opposed to C clicks were apparent at 30 dB SL (Stockard et al, 1979).
• Stockard et al (1979) also reported that click phase is an important factor contributing to intrasubject variability in amplitude, morphology and interpeak latencies of the BAEPs.

• 70% of their subjects, wave IV was more prominent than wave V with R clicks whereas wave V was more prominent than wave IV with C clicks presented at 70 dB SL.
•  The wave V amplitude was increased with C clicks compared with R clicks presented at 70 dB SL in 80% of their subjects.
• The increase in interpeak latency with increases in stimulus rate is more pronounced for R than for C clicks
• At high RRs, wave I peak latency remains essentially unchanged or decreased in latency with R clicks whereas it is increased with C clicks.
• Because of the differential effects of stimulus polarity at high RRs, the use of alternating polarity clicks may result in a spurious wave I, reflecting the summation of the prolonged wave I for C clicks and the early wave II response for R clicks

            Some clinicians use alternating polarity clicks when the electrical artifact and the cochlear microphonic impinge on wave I; this effect is more apparent at high stimulus presentation levels. The use of separate recordings of both R and C clicks to facilitate resolution of the BAEP components has been suggested (Stockard et al, 1979). According to Stockard et al, (1979), phase effects are enhanced in the hearing impaired.

Although Stockard et al (1979) and others have observed polarity effects on the latencies and amplitudes of the BAEP components, other investigators (Rosenhamer 1978) have failed to observe any polarity effects. Stockard et al (1979) suggested that many of the previous studies on the effects of stimulus polarity were done using small sample sizes and that discrepancies may reflect differences in stimulus intensities and ages of the subjects. However, recently Gorga (1991) in a convincing article demonstrated that stimulus phase effect on wave V latency is frequency dependent. The effect was shown to be strongest at low frequency (250 Hz) stimulus and diminished as the frequency of the stimulus decreased (2000 Hz). This suggests that the effect of click polarity on ABR latency will depend on the configuration of the hearing.

AUDITORY SYSTEM PATHOLOGY

            Borg and Lofqvist (1982) conducted a comprehensive study of stimulus polarity. The stimulus in this study was a 2000 Hz sine pulse presented at an intensity level of either 75 or 35 dB nHL and at a rate of 20 millisecond. Data were collected for 65 normal ears, 20 ears with conductive hearing impairment, 29 with steep sloping high frequency loss, and 17 with retro cochlear auditory dysfunction. For normal ears, latency of wave V was on the average 0.1ms shorter for rarefaction versus condensation clicks, but 30 percent of the ears showed the opposite effect. Recall that, with the ½ wavelength rule, a 0.25 ms difference would be expected for a 2000 Hz stimulus. Comparable polarity differences for wave V latency were found in the ears with conductive hearing impairment. Rarefaction clicks produced progressively shorter latencies as wave V latency increased and also as the frequency decreased, beginning at the frequency at which the hearing loss began. These findings for ears with high frequency (presumably cochlear) impairment were not observed with stimulation of ears with retro cochlear pathology.

  IPSILATERAL MASKING

Burkard and Hecox (1983) investigated the effect of ipsilateral broad-band noise (BBN) masking on the latency and amplitude of the click-evoked BAEP. They found that above approximately 10 dB EML effective masking or noise level required to perceptually mask a click of the same nominal intensity level in dBnHL.

• The ipsilateral noise resulted in increased wave V latencies for click intensities between 20 and 60 nHL (0 dB EML was equivalent to 26 dB SPL).
• The wave V amplitude decreased as a function of ipsilateral noise level above approximately 20 dB EML for click intensities between 20 and 60 dBnHL.
• As click RR was increased, the absolute ipsilateral noise-induced wave V latency shift decreased.

Burkard and Hecox (1983a) concluded that these effects of ipsilateral noise had implications for BAEP assessment in nonsound attenuated test environments. That is, the wave V latency and amplitude are affected whenever BBN levels (ambient noise levels) measured under the earphones exceed approximately 10dB EML (36 dB SPL) and 20 dB EML (46 dB SPL), respectively. Since the earphones provide approximately 30 dB attenuation of ambient noise, noise levels in the test suite would have to reach 66 and 76 dB SPL to affect the BAEP latencies and amplitudes, respectively. Another implication of their study concerns the effect of crossover of masking noise from the nontest ear to the test ear. That is, when crossover yields, approximately 40 dB SPL of BBN in the test ear, the amplitude and latency of wave V will be prolonged and so cause a. higher false-positive rate to occur.

Burkard and Hecox (1983b) investigated the effects of ipsilateral BBN on the 1000- and 4000-Hz toneburst on the BAEPs. They found that as the ipsilateral noise level increased, the identifiability and response reliability of wave V decreased. Wave V latency was markedly prolonged for ipsilateral noise levels of 20 dB EML and 30 dB EML for the 1000- and 4000-Hz tonebursts, respectively. The amplitude of wave V decreased as a function of noise level above 20 dB EML for both toneburst stimuli. Burkard and Hecox, (1983b) also reported that the noise-induced shifts in latency and amplitude were greater for the higher derived; bands than for the lower derived bands when clicks were; employed as stimuli (i.e., the derived band subtractive masking technique was employed to yield responses for the specific location of the cochlear partition—derived bands).

 CONTRALATERAL MASKING

Finitzo-Hieber, Hecox, and Cone (1979) recorded the BAEPs from two adults with unilateral profound hearing impairment. BAEPs were absent even when the hearing impaired ear was stimulated up to 117 dBSPL. They concluded that contralateral masking was unnecessary in -BAEP assessment.

Chiappa et al (1979) obtained findings that contradicted those of Finitzo-Hieber et al (1979). They recorded the' BAEPs from a few patients with unilateral profound hearing impairment. The BAEPs were present when the poors ear was stimulated but were abolished with BBN masking in the nontest ear at 60 dB SL (the reference level for the SL was unspecified). The contralateral masker did not affect the latencies or amplitudes of the BAEP components in normal listeners, although in several subjects the contralateral masking resulted in a change in waveform morphology. Therefore, they concluded that contralateral masking is  needed to be employed during BAEP assessment.

Humes and Ochs (1982) recorded the BAEPs for elicit using contralateral masking with BBN in four unilateral, deaf subjects. They reported that the BAEPs recorded the deaf ear was abolished when sufficient masking V applied to the good ear. The interaural attenuation values (the difference between the behavioral thresholds of the two ears) ranged from 70—75 dB. They concluded that contralateral masking with BBN does not affect the latency or amplitude of the BAEP. High masking levels are unlikely to ever be employed clinically since—assuming an average-IA of 70-75 dB, an average behavioral threshold for clicks of 42 dBpeSPL in normal listeners, and a maximum diet intensity of 135 dBpeSPL—more than sufficient masking can be obtained with just 40 dB EML of BBN (one could mask a 40 dB SL click in normal-hearing listener).

Moreover, 40 dB EML can never result in interference with the BAEP from the test ear. Humes and Ochs do not recommend assuming that the IA value for all clicks is 70-75 dB since the clicks vary in spectra and the spectra of the maskers differ from instrument to instrument. They attributed the difference in the latency, amplitude, and morphology between the BAEPs obtained at maximum click intensity from the poor ear and the BAEPs obtained from the good ear with click intensities at an SL equivalent to the maximum click intensity minus the 1A to the frequency dependency of the 1A, suggesting that the high-frequency energy of the click stimulus would be attenuated by a greater amount than the low-frequency energy. Thus, the crossed stimulus has much more of its energy in the low frequencies and may be expected to produced ABRs of differing latency, amplitude, and morphology. If clicks are used as stimuli, broad-band noise can be used to mask the nontest ear. If frequency-specific stimuli are used, narrow-band noise can be used to mask the nontest ear provided the spectral envelope is wide enough to mask the sidebands of the frequency-specific stimuli (Gorga & Thornton, 1989).

Fig 8

Insert earphones, such as the Etymotic ER-3A, used with BAE testing reduce the electrical stimulus artifact and eliminate the problems with collapsed canals. They are also more comfortable to wear than the supra-aural headphones. With insert earphones, however, the peak latencies are delayed by approximately 0.9-1.0 ms because of the tubing length. Van Campen, Sammeth, and Peek (1990) observed that insert earphones do not provide appreciably increased interaural attenuation for BAEP testing as they do for pure-tone behavioral assessment. Van Campen et al. (1990) hypothesized that the lack of increased interaural attenuation for insert earphones as compared with supra-aural earphones may be related to the dependency of the click-elicited BAEPs to the 2000-4000 Hz region of the - cochlea. At this frequency region, insert earphones used for pure-tone behavioral assessment .provide little additional interaural attenuation beyond that offered by supra-aural earphones.

 INTENSITY REFERENCE

Three scales of measurement have been employed to describe the amplitude of the click or other short-duration stimulus: sound-pressure level (SPL) or peak-equivalent SPL (pe SPL), sensation level (SL), or normal-hearing level (dB nHL). Many sound level meters cannot measure impulse sounds such as clicks since they do not have a peak hold capacity. Consequently, it is usually not possible to get true dB SPL measures. Therefore, the stimulus is displayed on an oscilloscope. A pure-tone signal is also fed to the oscilloscope and its amplitude (cither peak or peak-to-peak) is adjusted until it matches that of the stimulus. Then a sound level meter is used to measure the SPL of the pure-tone signal that has an amplitude set to match that of the click or other short-duration stimulus. This SPL is therefore referred to as dBpeSPL (dB peak-equivalent SPL).

The dBnHL scale is similar to the dB HL reference intensity employed in conventional audiometry. That is, 0 dBnHL is equivalent to the average of the behavioral perceptual thresholds for the click or other short-duration stimulus in a group of young normal-hearing adults. Thus, 60 dBnHL is 60 dB above the average behavioral perceptual threshold for the BAEP stimulus. The dBnHL value ' must be determined in each clinic by measuring the behavioral perceptual thresholds for the BAEP stimulus in a group of 10-20 young normal-hearing adults. This value is -approximately 35 dB.SPL for clicks having a duration of 100 µsec.

With the dB SL scale, the stimulus intensity is referenced to the individual's behavioral perceptual threshold for the stimulus. Although sensation level equates stimulus intensity in persons with conductive hearing impairment, it does not do so in subjects with cochlear hearing impairment.

The dB peSPL and dB SPL scales are absolute whereas the dB SL and dBnHL scales are relative.

BINAURAL STIMULATION

In binaural stimulation, clicks are presented simultaneously to both ears and the responses recorded monaurally. Binaural stimulation results in increased amplitudes of the later waves at all intensities. Since binaural stimulation increases the amplitude of wave V but not wave I, the IV-V:I amplitude ratio is increased with binaural as compared with monaural stimulation (Stockard et al, 1978b). The difference between the summed monaural responses from each ear (the predicted binaural waveform) and the binaural evoked responses should be obtained in order to observe the binaural interaction (Dobie & Berlin, 1979). If each ear and its neural connections functioned independently of the other ear and its neural connections, the summed monaural waveform would be equal to the waveform obtained by binaural stimulation. When the predicted binaural waveform is subtracted from the binaurally evoked waveform, the difference waveform is polyphasic, consisting of two positive peaks (P1 and P2), each followed by a negative peak (N1 and N2). This polyphasic waveform representing binaural interaction may reflect, in a complex way, neural activity underlying binaural processes such as localization and lateralization of binaural stimuli.

SUBJECT PARAMETERS

1. GENDER

Stockard et al. (1979) reported that the interpeak latencies of male adults exceed those of female adults. This effect was attributed to differences in head/brainstem size, length of the external auditory meatus, and auditory nerve diameter. Jerger and Hall (1980) reported that the absolute peak latencies are shorter and the peak amplitudes are larger-in females than in males. On average, the wave V peak latency is approximately 0.2 ms shorter and the wave V amplitude is approximately 25% larger in females than in males. Thus, separate norms should be generated for males and females. If gender is not taken into consideration, then the false-negative rate for females and the false-positive rare for males may be increased.

2. DRUGS

Drugs that do not affect ABR

It has been reported that sedation does not affect the BAEPs. In fact, sedatives such as chloral hydrate, secobarbitol, and DPT (Demerol, Phenergan, and Thorazine) reduce the muscle artifact, thereby enhancing the BAEP waveform. Starr and Achor (1975) reported that the BAEPs are unaffected even in cases of drug-induced coma. Anticonvulsants such as Dilantin, administered at therapeutic levels, essentially have no effect on the BAEPs (Stockard 1977). Anesthetics such as halothane, erhrane, isoflurane, fentanyl, nitrous oxide, meperidine, thiopentene, and diazepam have little effect on the BAEPs. Since anesthesia has a minimal effect on BAEPs, intraoperative monitoring of BAEPs is being increasingly employed. It has also been reported that neuromuscular blocking agents such as pancuronium do not affect the BAEPs (Harker et al., 1983).

Drugs that affect ABR

            Acute alcohol intoxication is associated with an increased I-V IPL. Lidocaine, a local anesthetic, has been found to affect the absolute and interpeak latencies and amplitudes of the BAEPs (Shea & Harrell, 1978). Similarly, phenytoin, an anticonvulsant, has been found to cause increased interpeak latencies (Green 1982). Cholinergic drugs affect the peak amplitudes of the BAEPs.

            In conclusion, BAEP measurement is unaffected by most drugs including anesthetics. Thus, BAEPs can be accurately obtained in sedated children and adults: Acute alcohol intoxication should be ruled out before administering the BAEP test.

3. TEMPERATURE

Hypothermia is associated with increased interpeak latencies. Lutschig, Pfenninger, Ludin, and Fassela (1983) suggested subtracting 0.15 msec from the obtained I-VIPL for every °C that the body temperature falls below 36°C. Decreased body temperatures are commonly observed during surgical procedures involving cardiopulmonary bypass or total circulatory arrest, and coma and drug intoxication cases. The BAEPs may be affected when increased body temperature is induced in patients suspected of having multiple sclerosis. Esophageal temperature measured at the level of the left atrium corresponds better with BAEP latency changes than: rectal temperature. In summary, body temperature does not affect the BAEP unless it drops below 36°C. The effect of temperature needs to be considered only during special surgical procedures and coma and drug intoxication cases.

4. AGE

The effect of age on amplitude and latencies of the BAEPs is inconclusive. Rowe (1978) reported that the peak latencies and the I-III interpeak latency increased with age as shown by their comparison of 25 young (17-33 years) and 25 older (51-74 years) adults. Thomsen, Terkildsen, and Osterhammel (1978) reported that the peak latency of wave V increases by approximately 0.1 ms per decade. Jerger and Hall (1980) examined the effect of age on the BAEP amplitudes and latencies in 98 normal-hearing adults and 221 sensorineural hearing-impaired adults ranging in age from 20 to 79 years. They reported that the peak latency-of wave V increased by 0.20 ms, on average, over the age range from 25 to 55 years. The peak amplitude of wave V decreased, on average, by about 0.05 µV over the same age range. The age effect was present in both males and females. Maurizi et.al., (1982) reported that the average peak latencies were increased by approximately 0.20 ms in their old subjects when compared with their young normal-hearing subjects. Similar age effects on the absolute and interwave BAEP latencies were also reported by other investigators (Allison, Hume, Wood, & Goff, 1984)

Rosenhamer, Lindstrom, and Lundborg (1980) reported that the gender effect present in young adults was absent in their older (50-65 years) adults. They also found that the peak latencies but not the interpeak latencies were increased in the older female adults compared with the younger female adults; no differences between the peak or interpeak latencies were found between the younger and older male adults. Jerger and Johnson (1988) found an interaction among age, gender, and high-frequency hearing sensitivity. In young adults, females had shorter absolute wave V latencies than males. As the hearing sensitivity at 4000 Hz increased, however, the absolute latency of wave V in females remained constant whereas it increased in males. Similar findings were obtained in the elderly. In contrast with the young adults, however, the gender effect, although present, was reduced in the normal-hearing older adults.

Since some of the studies reported a slight age effect (on the order of 0.2 ms) the age effect must be considered when collecting normative data. As shown by Jerger and Hall (1980) and Maurizi et al. (1982), the age effect becomes apparent after age 50. Therefore, separate norms should be obtained for subjects under 50 years of age and subjects over 50 years of age.

5. SIZE OF NORMATIVE DATA BASE

Most clinical laboratories have based their normative data base on a small sample size of 10-30 subjects. Chiappa (1983) suggested a sample size of 35 whereas Stockard et al. (1978b) recommended an N of 100 to ensure a normal distribution.

NORMATIVE DATA

A. BAEP COMPONENTS

Hecox and Galambos (1974) reported that wave V is the most stable of the BAEP components. Starr and Achor (1975) obtained similar findings and found that the IV-V complex was present even at intensity levels as low as 5 dB SL in normal-hearing subjects. Stockard et al (1978b) observed that a bifid wave III and a wave VI with a sharp downward slope are common normal variant morphologic patterns.

Chiappa et al (1979) evaluated the BAEPs in 50 normal adults. They identified 6 morphologic patterns (Types A through F) in which the shape of waves III and the IV-V complex varied. Figure 9  shows these 6 morphologic patterns.

Ø  Pattern A was characterized by a single peak for the IV—V complex.

Ø  Pattern B was characterized by separate peaks for waves IV and V with the wave IV peak lower than the wave V peak.

Ø   Pattern C was similar to pattern B except that the wave IV peak is higher than the wave V peak.

Ø  In pattern D, the peaks for IV and V are superimposed, with wave V appearing as a bump on the downward slope following wave IV.

Ø  In pattern E, the peaks of waves IV and V are also superimposed except that wave IV appears as a bump on the upward slope preceding the peak of wave V.

Ø  In pattern F, the peaks for waves IV and V are superimposed with bumps of equal heights.

The most frequently occurring patterns were B and C (occurring 38 and 33% of the time, respectively) with patterns D and F occurring least frequently (1 and 4% of the time, respectively). Subjects did not always have the same pattern in both ears—only 42% had the same pattern, bilaterally. A bidfid wave III occurred in 5.8% of the subjects and the first peak or the midpoint of the two peaks corresponded to the mean peak latency for wave III seen in other subjects. Chiappa et al (1979) were unable to posit a hypothesis to account for the various common normal variant morphologic patterns.

Fig 9

                        On the other hand, Stockard et al. (1979) recorded the BAEPs in 64 normal-hearing adults using rarefaction clicks presented at 70 dB SL. They found that the morphology was dependent upon the click phase. That is, wave IV was more prominent than wave V in 70% of the BAEPs to rarefaction clicks.

Hecox and Moushegian (1981) observed that waves IV and V are fused at high intensities with ipsilateral recordings and that wave III has an amplitude exceeding that of wave V at intensities above 70 dBnHL. They also observed that wave I is often notched or bidfid, reflecting the N₁ and N₂ from the eighth-nerve action potential; N₁ is more apparent at intensities above 60 dBnHL whereas N₂ is more apparent at intensities below 60 dBnHL. This factor must be considered when determining the peak latency of wave 1.

RECORDING PARAMETERS

FILTER SETTING

            The choice of filter setting is important. The purpose of the filter is to eliminate the contaminating effects of electromyographic noise and to reduce the amplitude of the ongoing electroencephalographic activity without significantly affecting the brainstem auditory evoked potentials.

As the low-frequency cutoff of the bandpass filter is increased to 300 Hz, the latencies of all the waves decrease and the amplitude of wave V decreases relative to that of wave IV (Laukli & Mair 1981). The effect on the amplitude of wave V is most marked as the low-frequency cutoff is increased from 100 to 300 Hz. Thus a low-frequency cutoff of 100-150 Hz is preferred (Stockard et al, 1978).

As the high-frequency cutoff of the bandpass filter is increased from 300 to 3000 Hz, the latencies of all the waves decrease and there is improvement in the resolution of waves IV and V. As the high-frequency cutoff increases beyond 3000 Hz, the resolution of waves IV and V is not improved further and high-frequency noise is added to the waveform (Stockard et al, 1978b). The effect of various filter settings on the BAEP waveform is shown in Figure 4.

Most clinicians and researchers use a bandpass filter of 100 to 3000 Hz or 150 to 3000 Hz. A wide bandpass filter is recommended to avoid distortion of the BAEP amplitudes and latencies (Laukli & C Mair, 1981).

                        Fig 4

 SAMPLE SIZE

            The electrical activity recorded following auditory stimulation consists of both the time-locked response to the signal and the ongoing electroencephalographic activity (the "noise"). In the averaging process, auditory stimuli are repeatedly presented to an car and the waveforms elicited by each stimulus are averaged. If one assumes the noise to be random and the time.-locked stimulus to be constant, then the averaging process should reduce the noise amplitude and thereby improve the signal-to-noise ratio. The decrease in the noise amplitude related to the number of sweeps is given by the formula I/(N^1/2). Thus, as the number of sweeps increases from 1 to 2, the noise is reduced in amplitude by a factor of 0.707. As the number of sweeps increases from 2000 to 4000, the noise amplitude is reduced by a factor of 0.0066. The greatest enhancement in the signal-to-noise ratio occurs in the first 500-1000 samples. Stockard et al. (1978b) and others recommend the use of at least 2000 samples per average. Figure 5 shows the effect of sample size on the BAEP waveform.

                                    Fig 5

.  ANALYSIS TIME

The analysis time or sweep time is the number of milliseconds after the stimulus onset that the signal averager continues to sample the responses. In adults, the recommended analysis time is 10ms. In neonates, because of the prolonged peak latencies compared with those of adults, the preferred analysis time is 14-15ms. Some researchers suggest an analysis time of 20 ms in neonates.

ELECTRONIC FILTERING

Electronic filtering causes a shift in the peaks over time. The extent of the shift is related to the spectrum of the peaks and is not the same for all peaks (Doyle & Hyde, 1981). Moreover, the peak spectrum has considerable inter-subject variability. The Bessel filter, an electronic filter with linear phase shift, and the digital filter have been developed to overcome the phase shift' limitations of electronic filtering (Doyle Hyde, 1981). Digital filtering is more flexible and economic than the Bessel filter (Moller, M., & Moller. 1985).

 ELECTRODE IMPEDANCE

Electrode impedance values should not exceed 5000 ohms in adults; below 3000 ohms is preferable. Differences between electrode impedances ideally should not exceed 1000 ohms. In infants, electrode impedance values may reach 10,000 to 15,000 ohms.

 PLOTTING CONVENTION

When the positive input to the signal averager has a greater magnitude than the negative input, the deflection is directed upward on the oscilloscope. In BAEP assessment, the vertex and ipsilateral earlobe (or mastoid) are common recording sites for the noninverting and inverting electrodes, respectively; thus, if the vertex electrode lead is plugged into the "+" input and the earlobe lead is plugged into the "-" input, the potential which is positive at the vertex relative to the earlobe will be represented as an upward deflection on the oscilloscope; this situation is referred to as "vertex positive and up." If the vertex electrode is plugged into the negative input and the earlobe lead is plugged into the positive input, a potential which is positive at the vertex relative to the earlobe will be represented as a downward deflection; this situation is referred to as "vertex positive and down."

 RECORDING MONTAGE

Various sites for placement of the noninverting, inverting, and common electrodes have been reported. Although electrode location can affect waveform morphology (Martin & Moore, 1977), investigators disagree on the preferred electrode location.

The vertex site is commonly employed for the noninverting electrode. Several investigators have reported that movement away from the vertex by 6-10 cm in any direction has essentially no effect on the brainstem auditory-evoked potential (Martin & Moore, 1977). Some investigators have recommended the upper forehead site for the non inverting electrode since that site is free of hair and convenient for cleaning and electrode application (Bcattie & Boyd, 1984). The upper forehead site is also more comfortable than the vertex site since it does not cause the electrode to be compressed between the skull and earphones.

The inverting electrode is commonly placed on the mastoid or earlobe ipsilateral to the car receiving the stimuli (Berlin & Dobie, 1979), the ipsilateral neck (Berlin & Dobie, 1979) and on the spinous process of the seventh cervical vertebra (Howe & Decker, 1980).

• Wave I is enhanced with the mastoid placement whereas wave V is enhanced with the neck placement.
• Wave I is enhanced (i.e., the trough following the peak is increased) with placement on the medial surface or the earlobe compared with other periaural recording sites such as the mastoid.
• Hecox (1980) reported that, with a horizontal recording in which the noninverting electrode is placed on the contralateral mastoid or earlobe and the inverting electrode is placed on the ipsilateral mastoid or earlobe with the common electrode on the forehead, the amplitude of wave 1 is enhanced.
• Ruth et al. (1982) found that this montage did not enhance wave I. (The vertex-to-earlobe arrangement is referred to as the vertical montage).

Electrically linked electrodes on the lateral part of the neck rather than electrodes on the mastoid or earlobe were preferred by Terkildsen and Osterhammel (1981) as the inverting electrode site because earphones placed over the ears do not interfere with electrode placement when the electrodes are on the neck.

• Nevertheless, Glasscock et al (1981) noted that the neck site was susceptible to neuromuscular potentials.
•  Keyanishvili (1981) reported that when the inverting electrode was placed on the spinous process of the seventh cervical vertebra, larger responses resulted than when the inverting electrode was placed on the mastoid.

             Typical electrode locations for the common electrode include the mastoid, earlobe, or neck on the side contralateral to the ear receiving the stimuli or on the forehead (Berlin & Dobie, 1979).

            Beattie, Beguwala, Mills, & Boyd (1986) evaluated the effect of electrode placement on the amplitude and latent of the brainstem auditory-evoked potentials when clicks alternating polarity were presented at 70 dBnHL in  adults. Ten electrode combinations were evaluated. The noninverting electrode was placed on the vertex for half the subjects and on the upper forehead for the other half the subjects. The location for placement of the invert electrode included the ipsilateral mastoid, the ipsilateral neck, or the seventh cervical vertebra. The various location for the placement of the common electrode included the contralateral mastoid, the lower forehead, and the contralateral side of the neck. Beattic et al reported that these electrode placements did not significantly affect the latencies of waves I, III. or V. Electrode placement did, however, affect the amplitudes of the brainstem auditory-evoked potentials. The vertex placement for the noninverting electrode resulted in larger amplitudes for wave V (X = 0.527 µV) than the lower forehead placement (X = 0.385 µV).

Also, for wave V, greater amplitudes were obtained with the following electrode placement locations:

(a) inverting on the seventh cervical vertebra and common on the flower forehead (X = 0.525µV);

 (b) inverting on the ipsilateral heck and the common on the lower forehead (X = 0.480 µV); and

(c) inverting electrode on the ipsilateral neck and the common on the contralateral neck, (X - 0.483 µV).

Lower amplitudes for wave V were obtained under the following conditions:

(a) inverting on the ipsilateral heck and common on the contralateral neck R(X - 0.393µV), and

 (b) inverting on the mastoid and common on the lower forehead (X = 0.401µV).

Electrode placement did not affect the amplitudes of waves I or III.

Since electrode placement did not affect the BAEP latencies for the 10 recording montages evaluated, Beattie et al(1986) concluded that "electrical activity is distributed simultaneously through the head and neck region". In contradiction to Beattie et al (1986), other investigators reported latency differences for various electrode placement sites (Barratt, 1980). These studies which found latency effects used electrode placement sites different from those employed by Beattie et al. (1986)

Beattie et al (1986) stated that the vertex (noninverting) 7^th cervical vertebra (noninverting) forehead (common) placement is preferred for maximizing the amplitude of wave V, since electrode leads do not have to he switched when switching test ears, this montage is employed for investigating binaural interaction; this placement is also preferred since only three electrodes are needed. Previous investigators have also reported increased amplitude for Wave V with non cephalic inverting electrode placement sites than with cephalic inverting electrode placement sires (Berlin & Dobie, 1979). Some investigators have related the non cephalic advantage to the fact that the non cephalic site may be more electrically silent than the cephalic site for the inverting electrode leading to a larger difference between the non inverted waveform and the inverted waveform shifted by 180°. The wave V amplitude difference may be accounted for by the varying impedances of the neural, fluid, bone, fat, and skin tissues of the head and neck regions (Abraham & Ajmone Marsan, 1958). Because electrode montage varies from clinic to clinic, each clinic must generate its own normative data for the selected electrode montage.

Fig 7

Figure7 shows the International 10-20 system for electrode placement. Under this system, the vertex electrode site is labeled Cz, the forehead site is labeled Fz, the earlobe site is labeled A (A₁ for the left ear and A₂ for the right ear), and the mastoid site is labeled M.

10. Two-Channel Recordings

Simultaneous ipsilateral and contralateral (two-channel) recordings using the vertex as the noninverting site, the ipsilateral and contralateral earlobes (or mastoids) as the inverting site, and the forehead as the common electrode site can facilitate identification of the peaks. With the contralateral recording, peaks I and III are reduced in amplitude (although I, the negative trough following the positive peak of wave I, amplitude remains essentially the same), the II—III IPL is shortened, and the I-V IPL, the IV-V IPL, and wave V peak latency are increased with the contralateral compared with the ipsilateral recordings (Stockard, J. E. 1983). The increased IV-V IPL in the contralateral recording helps to resolve wave V in the ipsilateral recording since waves IV and V may be difficult to identify in the ipsilateral recording derivation. The preservation of wave II with the reduction in the amplitude of waves I and III in the contralateral recording similarly assists in resolving peaks I, II, and III in the ipsilateral recording derivation.

ANALYSIS AND INTERPRETATION

            Most ABR waveforms are plotted in the time domain—that is, the amplitude of ABR components (almost always (µvolt) is displayed over time (almost always in milliseconds, or ms). This is such a con-vention in ABR measurement that one might reasonably ask whether there is any other way a waveform can be displayed. There is another approach for describing ABR data. The ABR can be plotted also in the frequency domain, with amplitude (expressed in µvolts, µvolts², or dB), or phase (expressed in radians or degrees) displayed as a function of frequency (in Hz). Spectral composition of the ABR, the frequency response of the waveform, is revealed with this plotting approach.

• In the time domain, ABR waveforms are, simply put, a sequence of peaks (amplitude of greater voltage) and valleys (amplitude of less voltage) occurring-within a specific time period (the analysis period or epoch).
• Morphology of a waveform is the pattern or overall shape of these waves. Usually, morphology is described with reference to an expected normal appearance for the AER. For example, if an ABR waveform does not fit the clinician's expectation of normal appearance or two ABR waveforms recorded in sequence are not highly reliable, even though wave component latency and amplitude values may be within normal limits, morphology is often judged "poor." In routine clinical AER measurement, morphology currently remains a rather subjective analysis parameter.

Latency is the time interval between stimulus presentation, really the onset of the stimulus, and the appearance of a peak or valley in the ABR waveform. Latency of ABR waves is expressed in milliseconds (ms). Latencies calculated for any waveform depend on the analysis criterion used to precisely define each component, that is, where the peak is marked. Peak voltage (the very highest amplitude on the wave) is used for some waves and by some "clinicians. However, other clinicians do not always define ABR waves by the peak. Instead, some other portion of the wave, such as the shoulder, is used for selected components. Among audiologists, this is a common clinical practice for definition of ABR wave V. The ABR wave V often appears combined with ABR wave IV—the wave IV/V complex—and a clear peak is not apparent. In such cases, latency is calculated at toward the end of the combined wave IV and V complex. Latency is an absolute measure calculated from the stimulus onset to some point on or near the peak of an ABR component, as illustrated by the horizontal arrows in Figure11. Interwave latencies are relative measures calculated as the time between two different ABR waves. As illustrated in Figure 11, latency intervals commonly calculated between waves include the wave I to wave III, the wave III to wave V, and the wave I to wave V latencies. Latency varies indirectly with stimulus intensity. That is, as stimulus intensity is decreased the absolute latency of ABR waves increases, and waves with relatively smaller amplitudes (e.g., wave I and wave III) gradually disappear. Interwave latencies represent general indices of transmission times along the auditory pathways, from the eighth (auditory) nerve to the midbrain and beyond. Specifically, interwave latencies reflect delays associated with either axonal conduction time along neuron pathways and/or synaptic delay between neurons (e.g., Ponton, Moore, & Eggermont, 1996).      Fig 11

            Amplitude is the second major response parameter typically analyzed. Although amplitude is usually described in µvolts, different techniques are used for calculating amplitude for ABR waves, sometimes within a single waveform. One common technique is measurement of the voltage difference between the peak and the preceding trough the peak of a wave. This is typically the approach used for determining amplitude for ABR wave I and for wave III. Another clinically popular approach is calculation of the peak and following trough, typically used in determining wave V amplitude. A third approach, calculation of the difference in amplitude between peak voltage of the wave and some measure of a baseline voltage, is rarely applied in analysis of ABR waveforms. This approach is, however, commonly used in calculating the ratio of two wave components, such as the ratio of amplitude for the ECochG summating potential and action potential, or in calculating the absolute amplitudes of later latency auditory evoked response's; e.g., AMLR, ALR, and P300 response, as described" in subsequent chapters. The most commonly calculated relative amplitude measure in ABR analysis is the wave V to wave I ratio. Normally the value is at least 1.0, or considerably larger.

            A final fundamental concept in ABR waveform analysis is the direction of response polarity—that is, which way is up. Polarity of an ABR is dependent on the electrode location relative to the generator of the response and, of course, which electrode is plugged into the positive and negative voltage inputs of the differential amplifier. The clinician can record major waves of any ABR as negative or positive in voltage depending on which electrodes are plugged into which amplifier inputs. For most ABR recordings, the electrode located at a high forehead (or vertex) site are plugged into the positive (noninverting) voltage input of the amplifier, and the electrode located near the ear (e.g., earlobe) is plugged into the negative (inverting) amplifier input. With this approach, the resulting, and familiar, ABR waveform is characterized by positive voltage peaks (e.g., wave I, wave III, and wave V) plotted upward and the negative voltage troughs plotted downward. This convention for plotting waveform polarity is not followed consistently, however, as some investigators in Canada, Scandinavia, Israel, and the United States show these characteristic waves plotted "upside down," that is, with positive voltage downward.

            The relation between ABR waves and anatomic generators is complex and not entirely understood. Electrodes located on the scalp and ear are insufficient to resolve all auditory regions (nerve pathways and nuclei) that are activated with acoustic stimulation, due to the inherent imprecision of far field evoked response measurements. Different ABR components, reflecting activity of different anatomic structures, may occur within a single ABR wave. In addition, there is variability in the temporal relation between ABR waves and the caudal-to-rostral arrangement of potential anatomic generators due to differences in the number of synapses and differences in direct versus indirect ascending pathways within the auditory brainstem.

CONVENTIONAL ABR WAVEFORM ANALYSIS

Nomenclature

Beginning with the first descriptions of the human ABR, independently by Jewett and Williston (1971) and Lev and Sohmer (1972), different schema have been used to denote wave components. Initially, wave components were labeled by Roman numerals, positive (P) and negative (N) voltage indicators plus Arabic numerals, and simply with Arabic numbers. There are inconsistencies, as noted above, in vertex polarity (negative or positive), and even in the sequence of wave components. For example, with the Roman numeral labeling system as introduced by Jewett and Williston (1971), vertex positive waves are plotted upward. However some investigators, as noted previously, display waves with Jewett Roman numeral labels and negativity plotted upward. As typically recorded, there is often no clear distinction between wave IV and wave V in the ABR waveform. Probably for this reason, some investigators (Lev & Sohmer, 1972; Thornton, 1975) have labeled the wave IV-V complex with the number 4 and have labeled what is conventionally referred to as wave VI with number 5 (or P5 or N5)

Normal Variations

There are myriad normal ABR variations. In fact, ABR waveforms among individuals are quite distinctive, much like fingerprints. That is, rarely are identical ABR waveforms recorded from any two persons. There may, in addition, be differences in waveforms between ears for a single person. ABR waveforms in multiple birth newborn infants (twins, triplets, quadruplets) may even have distinctly unique waveforms under the same recording conditions. Subtle differences in waveforms among patients are not important clinically, so long as there are clear and consistent criteria for distinguishing a normal response from an abnormal response.

The various ABR components are largely time-dependent, each occurring within a limited time period following the stimulus. Latency values are remarkably consistent among audiologically and neurologically normal persons. However, even a latency analysis approach becomes difficult when identification of components is obscured, either because of normal variability, poor reliability, or the effects of auditory pathology. ABR analysis based on response amplitude is often problematic because amplitude normally tends to be highly variable. Finally, due to normal variability in ABR waveforms, analysis of morphology (versus latency and amplitude) rarely permits confident differentiation of normal versus abnormal findings.

Response Reliability and Analysis Criteria

The "textbook" normal ABR has clear and repeatable wave components. Waves I through V are each unequivocally present in two or more repeated waveforms recorded with the same stimulus and acquisition parameters. Repeatability of at least two ABR waveforms recorded in succession with the same measurement conditions (e.g., stimulus with no change intensity, rate, and polarity in one ear) is a typical prerequisite for waveform analysis. An exception to this definition of and requirement for reliability occurs in pediatric applications of the ABR when waveforms are successively recorded at different intensity levels, and repeatability is determined for waveforms for two stimulus intensities.

            The basic concept is simply that two or more averaged waveforms, when superimposed, are very similar. Ideally, the two waveforms are almost indistinguishable, except for slight differences in background activity or noise. In this case, even the inexperienced clinician can assess repeatability of the waveforms at a glance. As a rule, only waveforms (minimally two) that meet criteria for repeatability can be considered responses. While there are occasional exceptions to this policy, the clinician is well advised to routinely attempt replication of ABR waveforms.

There is in AER measurement the assumption that the response is perfectly time-locked to the stimulus and background noise is minimal and "stationary," that is, randomly distributed (a Gaussian distribution). According to this assumption, sequential averaged waveforms will be essentially identical. Clinically, it is extremely difficult to analyze AER waveforms that are contaminated by excessive artifact. The presence of large amplitude, relatively low-frequency artifact, usually related to patient movement, may seriously interfere with, or preclude, accurate identification of AER components. During the course of a test session, at least in a relatively relaxed normal subject, muscle (electromyographic, or EMG) activity decreases significantly as signal averaging progresses. Since muscle activity is considered noise in an AER recording, it will decrease with increased averaging. The noise level in an evoked response recording is inversely proportional to the square root of the number of samples (stimulus repetitions or sweeps). Because the ABR signal-to-noise ratio is the magnitude of amplitude for ABR waves divided by the amount of noise during measurement, a reduction in noise with averaging will result in a larger signal-to-noise ratio (SNR). Small amplitude, high-frequency artifact, whether electrical or myogenic in origin, can interfere with precise estimation of the wave component peak and, therefore, influence accuracy of latency calculations.

Other sources of variability in AER measurement, in order of their importance, are differences in the responses between subjects, between ears, from one test session to another (separated by days or longer intervals), and from one run to the next. Response latency and amplitude may change also during extended test sessions, lasting hours. Poor reliability of response amplitude is a well-appreciated problem in AER measurement)

Amplitude is highly influenced by EEG activity level and muscle artifact, as well as by measurement parameters such as stimulus intensity and filter settings. Amplitude ratios of, for example, wave V/I, vary with subject characteristics and stimulus factors. Greater amplitude for wave I than for wave V (a small V/I ratio) may be a normal finding in young children. Immature neurological development, reflected by reduced synchronization of neural firing and incomplete myelinization, is suggested as a basis for this phenomenon (Gafhi et al., 1980). A reduced wave V/I amplitude ratio may also be due to specific test conditions.

A second source of uncertainty in AER analysis involves, in Hoth's (1986) words, the "measuring device." Clinically, this usually means the tester, including analysis criteria used by the tester. There are reports of high agreement between and within interpreters in ABR analysis. Two interpreters using common criteria will be consistent in judging an ABR as normal or abnormal in over 95 percent of cases} (Ross-man & Cashman, 1985)(and a single interpreter will render the same judgment on repeated analysis of waveforms approximately 80 percent of the time (Kjaer, 1979). Interpreter agreement is decreased for waveforms that are less repeat-able. Repeatability, in turn, tends to decline as hearing loss increases. Nonetheless, the experienced clinician with a good understanding of the impact on the ABR waveform of manipulation of measurement parameters will for the majority of patients succeed in recording repeatable waveforms.

 The Art of "Peak Picking"

"Picking the peak," that is, consistent and accurate selection of the single representative data point on a waveform that will be used in labeling the wave and calculating latency and amplitude values, is an important clinical skill. There are two fundamental approaches to this type of wave analysis.

One is to select as the peak the point on wave component that produces the greatest amplitude. In waveforms with sharply peaked components, this selection is simple and unequivocal. Although intuitively appealing, this approach can present analysis problems. One problem occurs when the point of greatest amplitude clearly does not best represent the wave. Perhaps the most frequent example of this limitation, even in normal subjects, is found with patterns of the wave IV-V complex that do not have two actual peaks (i.e., one for wave IV and another for wave V).

Fig 12

With a prominent wave IV and relatively minor wave V pattern, selecting the maximum amplitude as the peak essentially substitutes wave IV latency for wave V latency. The clinical consequences of this type of waveform misinterpretation would include calculation of an unusually short latency on the suspect ear, a significant inter-aural latency difference for wave V and the wave I-V latency interval, and possibly the presumption that the nonsuspect ear is abnormal.

Another problem with defining peaks on the basis of maximum amplitude arises when the top portion of the wave is rounded or even a plateau, rather than sharply peaked. This morphology may occur spontaneously, or it may be the result of a restricted low-pass filter setting. An apparent solution to this problem is to take as the peak the point at which lines extended from the two slopes of the wave intersect. Several disadvantages of the technique are readily evident. First, the point of intersection of the two lines does not correspond to an actual peak. Also, slight variations in either the leading or following slope may produce important variations in the arbitrarily defined "peak."

The second fundamental "peak picking" approach is to select the final data point on the waveform before the negative slope that follows the wave. This point may be the final peak, or a plateau or shoulder in the downward slope. This technique virtually eliminates the incorrect selection of wave IV versus V, but introduces its own problems. Some waves have multiple shoulders on the downward slope, caused by background activity. Other waves have shoulders that are extremely subtle and ill defined.

Extra Peaks

A normal variant in AER morphology is when selected peaks are reliably recorded, but smaller in amplitude, than the major components in an ABR waveform. Edwards et al. (1982) carefully tallied up the number of peaks occurring between major ABR waves in a group of 10 normal subjects. They consistently showed repeat-able peaks between successive waves in the wave II through V region (between waves I and II, between waves II and III, and so on). Approximately 25 percent of the waveforms for the subjects in their study had one or more extra peaks. Extra peaks, especially in the early portions of ABR waveforms, may be partially related to the conventional high forehead-to-earlobe (or vertex-to-ipsilateral mastoid) electrode array used for ABR recording. McPherson et al. (1985) associated an extra component between waves I and II (referred to by them as "x") and another component between waves II and III (referred to as "y") when the ABR was recorded with an electrode on the mastoid ipsilateral to the stimulus. These extra components were most prominent in ABRs recorded from new born infants (McPherson et al., 1985).

A bifid wave I component, that is, a wave I with two closely spaced peaks, is sometimes observed in ABR wave forms. The latency separation between the two peaks is generally less than 0.5 ms, thus ruling out the possibility that the second peak is actually wave II. Factors that may increase the likelihood of recording a bifid wave I component are high stimulus intensity level, mastoid or earlobe electrode site, and, possibly, stimulus polarity. Slightly different wave I latencies are sometimes recorded with rarefaction versus condensation clicks. With an alternating click stimulus in which theoretically each polarity contributes to the averaged response, a shorter latency peak in a bifid wave I may be generated by one polarity (most often rarefaction) while the second peak is generated by the opposite polarity (typically condensation).

Ambiguities in accurate peak identification of wave V are, without doubt, the most common and most troublesome. Multiple-peaked wave V components present another major problem. Typically, with this waveform variation there are two or more distinct peaks superimposed on a broad wave within the expected latency region for wave V. The problem is that selection of one of the peaks results in a normal interpretation while selection one or more of the other peaks yields an abnormal interpretation. The first step in solving this dilemma is application of consistent analysis criteria. For example, if the analysis convention for wave V is to calculate latency from the point of the shoulder preceding the large negative slope, then this criterion should be applied in analysis of the multiple-peaked component. This is a reiteration of the axiom stated above: The first principle of waveform analysis is to achieve repeatability of waveforms. Manipulations of stimulus and acquisition parameters are often very useful for resolving confusion in identification of the true wave V.

            A related normal variation in ABR morphology is an unusually prominent wave VI that closely resembles the characteristic wave V. In some cases, wave V appears as a relatively minor hump on the initial slope leading to the wave VI, rather than a distinct wave component. The clinical problem presented by the prominent VI is similar to the bifid wave configuration just noted for waves I and III. If the earlier component is taken as wave V, the response is interpreted as normal, but if the apparent wave VI is reported as wave V then there is a markedly abnormal wave I to V latency interval. Minor extra peaks falling between major waves are curious, but not a concern or a factor in AER interpretation.

Fused Peaks

Fused peaks are, technically, two peaks combined into a single wave complex. Both peaks are distinct yet one of the peaks usually dominates. The peaks most often fused in nor-mal ABR waveforms are IV and V (Chiappa, Gladstone, & Young, 1979). In these cases, wave IV usually appears as a hump or short plateau before wave V, or, conversely, wave IV is a distinct peak and is followed by a shoulder or plateau. If there is single peak in the expected latency region for waves IV-V, it is typically labeled as wave V, and the wave IV is presumed missing. Wave IV is often not observed in normal ABR waveforms. Edwards et al. (1982), for example, found that about 50 percent of their subjects showed no wave IV. This was a consistent finding within subjects. That is, waveforms for selected subjects showed a clear wave IV and V on some runs and did not on others. Fusion of waves IV and V is more likely in cochlear pathology than in normal hearers.

Stimulus and acquisition parameters also influence fusion of ABR waves. Stimulus polarity can cause changes in wave component latency. Because wave V latency may be different for rarefaction versus condensation stimuli, configuration of the wave IV/V complex may vary with polarity (Borg & Lovquist, 1982). Electrode array, which exerts a very prominent effect on waveform morphology, is discussed separately in a following section. There is some evidence, at least in normal adult subjects, that differentiation of wave IV versus wave V is poorer (a fused complex is more likely) when high-pass filter settings are extended below 150 Hz. Put another way, raising the high-pass setting from 5 or 30 Hz to about 150 Hz appears to resolve separate waves IV and V (McPherson, Hirasugi, & Starr, 1985). This strategy for resolution of the wave IV and wave V complex is, however, not advisable with infants because the ABR is dominated by low-frequency energy. Increasing the high:pass filter setting for infants will reduce ABR amplitude and, in some cases, may remove the response itself.

Missing Peaks

Peaks often missing in normal ABR waveforms include wave IV, as just noted, and waves II and VI. In figure 13, waveform "A," waves I, III, and V are repeatable and well formed. Absolute and interwave latency values are within normal limits. Amplitude of each of these components is also within normal expectations (note the amplitude marking of 0.25  in the figure 13) and in the proper relationship (wave V/I amplitude ratio is greater than 1.00). By conventional latency and amplitude criteria, this can be classified as a normal ABR. There is no evidence, however, of waves II, IV, or VI. Waveform "B" shows a distinct wave I and a wave IV-V complex and also evidence of waves II and VI. There is a small deflection in the ABR waveform in the wave III latency region, but it is no larger in amplitude than the background activity observed elsewhere in the waveform.

Fig 13

The wave IV-V complex, in contrast to the pattern in waveform "A," is not sharply peaked but, rather, consists of multiple minor humps. At least three different sites on the wave could be selected for latency calculation, including the peak, a slight shoulder following the peak, and a repeatable small plateau on the slope after the peak. Waveform "C," also recorded from a normal-hearing subject, has only a clear wave V. As noted above, well-defined criteria for "peak picking" are required for consistency in ABR interpretation of these types of waveform. Such criteria include both consistency between clinicians and consistency for a single clinician for ABR interpretation among patients, or even from .one ear to the other. In waveform "C," there appears to be a wave I, wave II, and wave III, but they are not reliably recorded.

Spectral Analysis

The earliest studies of evoked response frequency (spectral) composition were carried out with steady-state, as opposed to transient, visual evoked responses and were based on Fourier analysis techniques (Regan, 1966). The ABR waveform elicited by a series of repetitive stimuli is periodic, or repetitive, and composed of one and usually more frequencies, corresponding in some way to the frequency of the stimulus.  Most AERs are typically generated with a transient or very brief stimulus. The response is aperiodic. That is, it is not repetitive but, rather, one response waveform is produced for each stimulus. Electrical activity within the response is time-locked to multiple discrete stimuli and is averaged. The averaging process occurs over time (in the time domain). Waveform analysis consists of calculation of the latency between stimulus and wave components and the amplitude of the components. Even casual visual inspection of an AER waveform, such as the ABR, shows that it typically consists of prominent sharp, closely spaced peaks (occurring frequently) riding atop (superimposed on) more gradual widely spaced peaks (occurring less frequently). The presence of the rounded (versus peaked) waves depends mostly on the low-pass filter setting used to record the response. Also, depending on the low-pass filter setting, there may also be numerous (very frequent) tiny spikes or background noise in the waveform. By means of fast Fourier transform (FFT) techniques, it is possible to deconvolute, decompose, or separate out the relative contribution of major frequencies within the response waveform after it has been digitized. In this way, AERs are displayed in the frequency domain

SPECTRUM OF ABR

The greatest amount of energy is in a low-frequency region (below 150 Hz), with prominent energy regions also from about 500 to 600 Hz and 900 to about 1100 Hz. There is little spectral ABR energy above 2000 Hz. The frequency content of specific ABR waves has been inferred on the basis of spectral analysis and offline studies, i.e., ABR waveforms filtered with a digital technique. Some investigators suggest that later waves (IV through VI) consist of energy from the lower two energy regions (about 100 Hz and 500 Hz), wave III is dependent on 100 to 900 Hz energy, and the earliest waves (I and II) consist of relatively higher frequency energy in the range of 400 to 1000 Hz. Boston (1981) offered a slightly different explanation, finding a correspondence between energy in the 900 to 1100 Hz region and wave I, wave II and wave III; energy around 500 Hz to wave V; and lower frequency energy to the slow wave activity upon which these components are superimposed. The relationship between ABR spectrum and specific components, even in normal subjects, is not yet clear, the topic of some controversy (Elberling, 1979), and clearly requires further investigation.

Fig 14

FSP

A statistical variance ratio measure—Fsp—is a commonly applied algorithm for automatic and statistically confirmed quantification of the ABR signal-to-noise ratio, particularly in newborn hearing screening. The Fsp is based on the magnitude of the response when the stimulus is present (the "signal") divided by the magnitude of the response when the stimulus is not present (the "noise"). In an individual patient and for a given stimulus intensity level, the magnitude of an ABR is remarkably stable assuming the patient is quiet and not moving. The magnitude of the noise, however, varies widely depending on patient-related factors (e.g., muscle activity, EEG unrelated to the response), environmental factors (e.g., electrical artifact), and selected ABR stimulus and acquisition parameters (e.g., number of sweeps, filtering). Ongoing signal averaging reduces the noise and enhances detection of the ABR. As a general rule, noise is lowered as signal averaging is increased and, as a reflection of the larger SNR, the Fsp value increases (e.g., from 1.0 to 3.0). In patients who do not produce an ABR under adequate measurement conditions (e.g., a stimulus is presented and there are no technical problems), the measured response is approximately the same as the noise, with a resulting signal-to-noise ratio of approximately 1.0. On the other hand, the measured response for patients with who do generate an ABR is larger than the noise, as indicated by a signal-to-noise ratio value greater than 1.0. Generally, an Fsp value of > 2.1 is consistent with the presence of an ABR. For newborn hearing screening with ABR, the Fsp value of > 2.1 would indicate a "Pass" outcome. As first described (Don et al., 1984), the signal-to-noise ratio was evaluated with the F statistic using a value at a single point in the ABR waveform, hence the abbreviation Fsp. Statistical calculations of signal-to-noise ratio values in ABR with current evoked response systems are based not on a single data point but, rather, on multiple points. However, the term Fsp continues to be used in reference to the automatic analysis technique. The Fsp analysis technique is incorporated in most newborn hearing screening devices, and now is also applied for objective identification   of cortical auditory evoked responses, e.g., the MMN.

RECOMMENDED TECHNICAL PARAMETERS FOR BAEP TESTING

1.               The bandpass filter should be 100-3000 Hz or \50-3000 Hz.

2.               The sample size should be 2000 sweeps. When testing infants in the nursery and using low click presentation levels, the sample size may need to be increased to 4000 sweeps in order to compensate for the effect of low intensity and ambient noise on the BAEP.

3.               The RR should be below 15 Hz for neurologic purposes (11.4 is commonly employed) in order to maximize the clarity of the waves, particularly wave I. For hearing-loss prediction, the RR can be as high as approximately 40 Hz since this rate is more time saving and does not substantially affect wave V amplitude. To avoid the 60-Hz hum contamination of the BAEP, the RR should never be a multiple or harmonic of 60 Hz.

4.               The rarefaction polarity is preferred for clicks since it is associated with shorter peak latencies and increased amplitude and resolution of wave I than condensation clicks. When normative and clinical data are obtained, the polarity should be kept constant. Clinicians should also obtain normative data for alternating polarity for clicks at high presentation levels (e.g., 90 and 100 dBnHL) so that AP clicks can be employed during clinical testing at high click presentation levels when a stimulus artifact is observed. For short-duration tonal stimuli, alternating polarity should be employed in order to reduce the effect of stimulus artifact.

5.               The analysis time (time window) should be 10 ms in adults and 15—20 ms in neonates.

6.               Electrode impedance values below 3000 ohms are preferred and should not exceed 5000 ohms in adults. In children, high electrode impedance* values are common but should not exceed 15,000 ohms. In any case, differences between electrode impedances should not exceed 1000 ohms.

7.               The plotting convention for BAEP testing should be vertex positive and up.

8.               Based on Beattie et al (1986) and Kavanagh and Clark's (1989) results, electrode montage does not have a significant effect on BAEP latencies. Therefore, for neurologic purposes and single-channel recordings, we recommend the upper forehead (noninverting)—medial earlobe (inverting)-lower forehead (common) electrode array. For the purpose of prediction of hearing impairment, the electrode montage which maximizes the amplitude of wave V should be employed. Based on Beattie et al (1986) results, wave V amplitude is maximal with the vertex—7th cervical vertebra-lower forehead, vertex-ipsilateral neck-lower forehead, or the vertex—ipsilateral neck-contralateral neck electrode arrays.

9.               Two-channel recordings (ipsilateral and contralateral) are preferred over one-channel recordings (whenever possible) to aid in the identification of the BAEP peaks.

10.            Contralateral masking for BAEP testing should be employed using the initial masking formula for suprathreshold speech-recognition testing, conservatively estimating interaural attenuation for clicks to be 50 dB. (Each clinic should obtain its own norms for interaural attenuation of clicks but the 50 dB value can be used in the interim.)

CLINICAL APPLICATION

PREDICTION OF HEARING THRESHOLD LEVEL FROM THE BAEP THRESHOLD

Several investigators (Coats & Martin, 1977) reported that the BAEP threshold in sensorineural hearing-impaired subjects is correlated with the behavioral hearing threshold levels at frequencies between 2000 and 4000 Hz if the hearing sensitivity is best in that region. Thus, if there are normal-hearing thresholds between 2000 and 4000 Hz with hearing impairment above or below this frequency range, the BAEP threshold will generally reflect the normal-hearing region and hearing impairment will be missed at the other frequencies (Stapells, 1989).

The correlation between the click-elicited BAEP threshold and the hearing thresholds between 2000 and 4000 Hz is low, approximately 0.49 Hz (Jerger & Mauldin, 1978), because the 2000—4000 Hz region may not be the frequency region with the lowest hearing-threshold levels. When the 2000-4000 Hz frequency region is not the region of best hearing sensitivity, the BAEP threshold may be consistent with hearing thresholds in the frequency region of best hearing sensitivity, that is, frequencies not between 2000 and 4000 Hz (Stapells, 1989). For example, Glattke (1983) reported that the BAEP threshold in a patient with a rising configuration that was profound at the low frequencies and became normal-hearing at 8000 Hz was present at 25 dBnHL, reflecting the normal hearing threshold at 8000 Hz. Thus, the BAEP threshold often misses or under estimates hearing sensitivity in the low, mid, or high frequencies. If the BAEP threshold for the click is elevated, it can be concluded that there is hearing impairment at least of the degree indicated by the BAEP threshold in the frequency region with the best hearing sensitivity. Despite the limitations of the click-elicited BAEP threshold, it is a valuable tool for gross prediction of hearing impairment in the difficult-to-test. According to Stapells (1989), the click-elicited BAEP threshold is approximately consistent with the pure tone average within one category of normal, mild, moderate, severe, or profound hearing impairment in 90—95% of the cases. Stapells (1989) maintained that most of the time, the BAEP threshold for the click corresponds to the average of the hearing thresholds between 1000 and 4000 Hz or to the best hearing threshold in that range.

ABSENT BAEPS

Schulman-Galambos and Galambos (1979) found that, in one case, no BAEPs could be elicited despite the presence of normal-hearing sensitivity. They attributed their finding to undetected technical error. Worthington and Peters (1980b), however, presented four cases with no neurological evidence of brainstem dysfunction in which hearing-threshold levels were normal or no-worse-than-severe and BAEPs were absent even at maximum intensity levels on two separate occasions in two of the four cases. They suggested that an absent BAEP might indicate that "the ABR is generated only by a portion of the auditory system which is not essential for hearing. Because absent BAEPs can result despite normal-hearing sensitivity, they advise against employing BAEPs as the sole measure of hearing sensitivity. Davis and Hirsh (1979) reported that BAEPs could not be elicited in some children who responded to sounds at low or moderate intensities.

Davis and Hirsh (1979) proposed that cochlear impairment could result in desynchronizarion and consequently absent BAEPs. Some other proposed causes for absent BAEPs in cases with normal or near-normal hearing include absence of neural activity, a nerve-conduction block, or desynchronizarion or discharges in the eighth nerve. The hypothesis concerning desynchronizarion evolved from findings of abnormal BAEPs in multiple sclerosis (MS) patients with normal hearing sensitivity since MS is associated with demyelination or lack of myelination of the eighth nerve (Naunton & Fernandez, 1978).

Kraus, Ozdamar, Stein, and Reed (1984) reported on seven cases with absent BAEPs or absent wave III or V on repeat testing, absence of clinical signs of brainstem neuropathy, and normal to no-worse-than-moderate hearing-threshold levels. The results of testing with the ITPA (Illinois Test of Psycholinguistic Abilities) and the Wepman Perceptual test battery revealed below age-level scores and poorer performance on the subtests dealing with auditory skills than on those dealing with visual skills. All experienced speech and language delays and formal intelligence testing in five cases revealed four normal scores and one borderline normal score. The investigators suggested that an absent BAEP in the presence of no-worse-than-moderate hearing-threshold levels in cases with absence of clinical signs of brainstem dysfunction may be indicative of communicative and/or learning disorders which may result from subtle brainstem dysfunction. This finding has not yet been verified by other large-scale studies.

INFANTS AND YOUNG CHILDREN

There is a direct relationship between maturity of the CNS and the effect of rate on ABR. Stimulus rate has a more pronounced influence on ABR latency for premature than term neonates, for younger children (under age 18 months) than older children, and for older children (up to age 13 years) than adults (Cox, 1985). In these studies, changes in ABR latency as a function of signal rate are often expressed in units of 10 Micro sec per decade (of rate). Despland and Galambos (1980) stated that the slope of the latency-versus-rate function declined from about 270 µsec/ decade of rate in the 30-week gestational age preterm infant to about 110 µsec/decade in the term infant. These slopes are both considerably steeper than the linear latency-versus-rate slope in adults (approximately 35 to 40 µsec/decade in rate). The latency-versus-rate slopes are steeper for the 60 and 80 dB intensity levels than at 40 dB. The slope at 80 dB is generally about 130µsec/decade for neonates versus 70 µsec/decade for adults. The much greater increase in ABR latency values, and diminished amplitude values, with increased stimulus rate for infants versus adults are reflected in progressively steeper slopes for the latency-rate functions, particularly for later ABR waves (Jiang, Brosi, & Wilkinson, 1998). Still, a reliable ABR can be recorded from term, and also preterm, neonates for stimulus rates up to the 455/sec rate and even the 909.1/sec rate used in the maximum length sequence (MLS) recording technique. As a rule, the rate effect is greatest for wave V. This results in a combined effect of young age and rate on the wave I to V interval. Prolonged neural transmission in younger subjects, due to incomplete myelinization and reduced synaptic efficiency, is suggested as a general neurophysiologic basis for these age-rate-latency interactions (Hecox, 1975). Slow rates may be necessary to obtain age-independent AERs. Lasky and Rupert (1982) found no ABR latency difference for 40-week term infants between signal rates of 3/second versus 10/second. Preliminary data for 32-week infants, however, suggested that wave V latencies were less for a 5/second than for a 10/second stimulus rate. As noted previously, stimulus rate in the adult can be increased to at least 20/second with no resulting effect on ABR latency or amplitude. These age-rate-ABR interactions, along with the possible influence of even more factors (e.g., stimulus intensity and polarity), must be considered both in developing a normative database, in establishing clinical ABR protocols, and in interpreting ABRs clinically.

EFFECT OF AUDITORY PATHOLOGY ON ABR

 CONDUCTIVE HEARING IMPAIRMENT

Kavanagh and Beardsley (1979) evaluated the BAEPs elicited with clicks in conductive hearing-impaired subjects. The latency-intensity functions, for the seven subjects with conductive hearing impairments shown in Figure 15.   This figure15  shows that the wave V latency—intensity function for conductive hearing-impaired persons is shifted to the right of that for normal-hearing subjects. The slope of the function is the same as for normal-hearing subjects. The amount of shift in the latency - intensity function (and hence, the BAEP threshold) generally corresponds to the degree of hearing impairment in the high frequencies.

Because of the shift of the latency-intensity function to the right in conductive-impaired persons, the peak latencies of the BAEP components are prolonged at a given intensity level relative to those in normal-hearing subjects. This prolonging of peak latencies and elevation of the BAEP threshold occurs because a conductive hearing impairment reduces the effective intensity of a sound. Several investigators reported that the prolonged latency values in conductive hearing-impaired listeners are similar to those seen in retrocochlear-impaired listeners.

Although several researchers have Suggested that the amount of peak latency delay can be predicted by the magnitude of the conductive component, other researchers (Clemis Mitchell, 1977), on the other hand, have reported that the latency delay cannot be predicted from the size of the conductive component. The results of these studies, however, were confounded by technical and procedural factors.

Fig 15

McGee and Clemis (1982) obtained BAEP’s elicited with 1000, 2000, and 4000 Hz tonebursts (1 ms rise-decay rime, no plateau) in 23 subjects with various kinds of conductive pathologies. For each subject, the air—bone gap was calculated from the audiogram. The degree of conductive hearing loss was estimated from the latency—intensity function. The latency intensity function for tone-burst a: a given nominal frequency from a particular subject and the mean latency-intensity function for normal-hearing subjects were plotted on the same graph. The dB difference between a given latency on the conductive function and the same latency on the normal function was then measured. This was done for all the points on the conductive latency-intensity function. Then the dB difference values were averaged to obtain a mean latency—intensity function (LIF) shift value in dB. The results only grossly support the conclusion that the prolonged latencies reflect the size of the air-bone gap in conductive disorders resulting from middle-ear effusion and earplugs. There is a substantial lack of agreement between the LIF shift and the size of the air-bone gap in cases of conductive impairment resulting from ossicular chain disorders; the air-bone gap seen in the audiogram does not accurately reflect the attenuation in sound caused by the conductive hearing impairment because of the elevated bone-conduction thresholds resulting from purely mechanical factors. Hence, the air-bone gap will underestimate the true sound attenuation resulting from the conductive component and the LIF shift will reflect the actual sound attenuation.

McGee and Clemis (1982) claim that the accuracy in determining the LIF shift resulting from the presence of air—bone gaps is sufficient to enable the clinician to deter mine which part of the prolonged wave V peak latency, in cases of combined conductive pathology not due to ossicular chain discontinuity disorder and retrocochlear pathology, results from just the conductive component. Eggermont (1982) contends that the error in prediction of conductive hearing loss from the LIF can be as large as 20 dB since agiven latency value may be associated with a range of intensity values as large as 20 dB. Hence, one may not be able to detect the presence of a mild conductive impairment on the basis of the LIF.

McGee and Germ's (1982) suggest that, in cases in which the latency prolongation resulting from the retrocochlear pathology is slight, it is difficult to sort out the portion of the latency delay attributable to just the conductive pathology in cases of combined conductive and retrocochlear disorders. If the clinician attempts to predict the air-bone gap from the LIF in such cases, the retrocochlear disorder may not be detected. Thus, wave I needs to be recorded so the I—V IPL can be determined for differential diagnosis. McGee arid Clemis (1982) reported that wave I could be recorded in only 15 of their 32 ears with conductive pathology. In all of these 15 ears, the I-V IPL was essentially normal, as is also the case in cochlear-impaired ears, in contrast with retrocochlear-impaired ears, which are characterized by a prolonged I-V IPL.

In summary, the following conclusions can be drawn With respect to conductive-impaired ears:

1.               The latency-intensity function is shifted to the right of the LIF for normal-hearing subjects; the slope of the function is essentially similar to that for normal-hearing subjects.

2.               The LIF shift in dB (and hence, the BAEP threshold elevation) only roughly corresponds to the air-bone gap (within 20 dB), except in cases of ossicular chain disorders, in which case the LIF shift overestimates the apparent air—bone gap as observed on the audiogram.

3.               The I-V IPL is essentially normal, similar to that in cochlear-impaired ears. It is not always possible to obtain the I-V IPL since wave I often cannot be recorded in conductive-impaired ears.

4.               The latency-intensity function for wave V resembles that LIF in retrocochlear-impaired ears.

COCHLEAR HEARING IMPAIRMENT

Galambos and Hecox (1978) reported that the BAEP threshold in patients with cochlear hearing impairment is elevated, as is the case for patients with conductive hearing impairment. In general, the latency-intensity function for wave V has a steep slope, that is, the latency values are prolonged at low intensities and become or approach normal values at high intensities. This latency-intensity function is most characteristic of flat or mild-to-moderate sloping cochlear hearing impairment. Subjects with significant sloping hearing impairment often demonstrate a latency—intensity function which is two-legged (Galambos &: Hecox, 1978). The average value of the L-I slope in persons with significant sloping cochlear impairment may be similar to that for normal-hearing persons, since the function is two-legged—one leg with a steep slope followed by a leg with a shallow slope—and the slope is determined on the basis of latencies at a high intensity (where the slope is steep) and at a low intensity (where the slope is shallow). A two-legged latency-intensity function for a person with a high-frequency sensorineural hearing impairment is shown in Figure 16. Subjects with hearing impairment which is precipitously sloping above 1000 Hz may demonstrate a latency-intensity function that is shifted upward of that in normal-hearing persons (Stapells et al, 1985). Figure 16 illustrates a latency-intensity function obtained from a person with a cochlear hearing loss precipitously sloping above 1000 Hz. In such cases, the BAEP threshold may be obtained at normal threshold levels (although the latency will be prolonged), reflecting the contribution of intact nerve fibers from the apical end of the basilar membrane. The peak latency of wave V never approaches, values seen in normal-hearing persons since the response is always dominated by apical fibers. As the intensity increases, there is a basalward shift in the fibers dominating in the response, but the intensity never reaches a level sufficient to stimulate the basal fibers.

The figure showing latency intensity function of V peak in cochlear impairment with steep high frequency hearing loss                               fig 16

If the slope of the latency-intensity function exceeds 60µs/dB (the slope is obtained by subtracting the latency value obtained at the higher intensity from that obtained at a lower intensity and then dividing the difference by the decibel difference in the intensity levels, cochlear hearing impairment is likely to be present. If the slope is less than 30µs/dB, one can conclude that the patient either has normal hearing-threshold levels or a significant sloping high-frequency hearing impairment. If the slope is normal, that is, between 30 and 60µs/dB, then one can conclude that either the patient has a significant high-frequency cochlear hearing impairment and the slope was determined over the steep and shallow legs of the function or that the patient has normal hearing and the slope was determined based on the latency values are the higher intensities. The slope values for a person with a precipitously sloping cochlear hearing impairment are similar to those for normal-hearing persons. Thus, cochlear hearing impairment cannot he ruled out on the basis of a normal slope value. Also, slope values which are normal or less than 30µs/dB may also he encountered in cases or retrocochlear pathology. Thus, only slope values exceeding 60µs/dB can be considered diagnostically significant. The clinicians should carefully delineate the latency-intensity function over several intensity level, and a wide range in intensity.

Coats (1978) evaluated the latency—intensity function in subjects with cochlear pathology. He subdivided his 37 cochlear-impaired ears according to the average of the hearing threshold levels in the 4000 to 8000 Hz range. The latency-intensity functions for wave V, the I-V IPL, and wave I for these cochlear-impaired ears were then superimposed on the plot showing the limits of the latency-intensity function in normal-hearing ears. Figure 17 shows these latency-intensity functions for the cochlear-impaired ears subdivided into four hearing-loss groups on the basis o; the average hearing threshold levels in the 4000 to 8000 Hz range. This figure 17  illustrates the shift upward in the latency-intensity function as the magnitude of the hearing loss increases. The latency-intensity function for the I-V IPL, on the other hand, approaches the lower limit of the normal-hearing limits as the magnitude of the hearing loss increases. It can be seen that the effect of hearing impairment is most striking on the latency-intensity function for wave 1, which clearly shifts substantially above the upper' limit of normalcy as the hearing loss increases. The latency prolongation seen in. the latency-intensity functions for waves 1 and V is most noticeable at the lower stimulus intensities and approaches more normal values at the higher intensities. The steepness of the latency-intensity function increases as the magnitude of the hearing loss increases and is greater for wave 1 than for wave V. As a result of the high-frequency cochlear hearing loss effects, on the peak latencies of waves I and V, the tendency seen in normal-hearing persons for the I-V IPL to increase slightly as stimulus intensity increases is exaggerated in the cochlear-impaired group. Thus, the slight upward slope of the 1-V latency-intensity function in normal-hearing ears increase (steepens) as the hearing loss in the 4000 to 8000 Hz range increases.

Fig 17

The above figure shows the latency intensity curves from 37 cochlear loss ears (fine lines) superimposed on estimated normal ranges (heavy lines). The four groups represent the average 4-8 KHz hearing level. First column- 17-30 dB HL, second- 32-42 db, third- 43-62 dB and the fifth- 63-117 dB HL.

The peak latency of wave V is related to the region of the cochlea which predominates in the response. In normal hearing persons and persons with mild-to-moderate flat or high-frequency hearing impairment, the latency of wave V at high intensities is determined by the basal region of the cochlea. At lower stimulus intensities in normal-hearing persons and persons with flat hearing impairment, the latency of wave V is dominated by slightly more apical regions of the cochlea. At lower stimulus intensities in persons with significant high-frequency hearing loss, the cochlear region which predominates in the response is shifted significantly in the apical direction so substantially prolonged latencies are obtained, reflecting longer travel times to the apical site where maximal disturbance of the basilar membrane occurs.

In summary, the following conclusions can be drawn about subjects with cochlear hearing impairment:

1.         The BAEP threshold is elevated as it is for conductive hearing-impaired persons.

2.         The latency-intensity function for wave V has a steep slope. That is, the latencies are prolonged at lower intensities and become or approach normal values at high intensities. This type of latency-intensity function is most commonly observed in persons with flat or mild-to-moderate sloping cochlear hearing impairment. Persons with more severe high-frequency hearing loss may demonstrate a two-legged latency-intensity function in which the first leg, over the lower intensities, is steep and the second leg, over the higher intensities, is shallow. The latency-intensity function in persons with hearing loss precipitously sloping above 1000 Hz (but sloping below 4000 Hz) often demonstrates a latency—intensity function which is shifted upward with respect to that seen in normal-hearing persons. The clinician is reminded that the likelihood of obtaining a latency—intensity function decreases as the hearing loss increases beyond approximately 80 dB HL and the BAEP may be absent even at the maximum output levels.

3.         If the slope of the latency—intensity function for wave V   exceeds 60µs/dB, it is diagnostic of cochlear pathology. Smaller slope values are nondiagnostic.

4.         If the latency-intensity function is obtained, it should be based on several intensity levels over a wide intensity range so that the shape can be accurately defined and appropriate intensity levels can be selected for determination of the slope value.

5.         The hearing impairment has a greater effect on the latency of wave I than on that of wave V, that is, the latency of wave 1 is prolonged more than that of wave Vas a result of the hearing loss. In cases of flat cochlear hearing impairment, the latencies of I and V are normal once threshold is exceeded. In cases of notched high-frequency cochlear hearing impairment, wave V is prolonged but wave 1 latency tends to be earlier than normal. 6. The 1-V IPL is reduced (as it is in conductive hearing-impaired persons) or normal compared with that in normal-hearing persons. Nonetheless, the 1-V interpeak interval may be prolonged in cases of notched high-frequency hearing impairment or normal in cases of flat cochlear hearing impairment.

RETROCOCHLEAR IMPAIRMENT

The use of BAEP testing for the identification of retro-cochlear site of pathology is becoming an increasingly popular challenge for the audiologist. Various BAEP measures, either singly or in combination, have been employed for the detection of the presence of retrocochlear pathology: (a) prolonged wave V, (b) long interaural wave V difference, (c) prolonged I-V interpeak latency (IPL), (d) absence of the later waves, (e) absence of the BAEP waveform, (f) absence of waveform reproducibility, (g) abnormal 1V-V:I amplitude ratio, (h) prolonged interaural I-V IPL, (i) abnormalities at high repetition rates, (j) contralateral ear effects, and (k) abnormal waveform morphology.

PROLONGED WAVE V LATENCY

Coats (1978) investigated the latency-intensity functions for wave V in 14 retrocochlear-impaired ears with varying degrees of hearing loss. Figure 18 shows the latency-intensity functions for the retrocochlear-impaired ear, superimposed against the two standard deviation limits for the cochlear hearing-impaired and normal-hearing ears. These results confirm the findings of previous investigators (Coats He Martin, 1977) that the effect of retrocochlear pathology is to prolong the peak latency for wave V. Also, the peak latency of wave V is prolonged in a retrocochlear ear with hearing impairment beyond that seen in a cochlear hearing-impaired ear with an equivalent degree of hearing impairment.

Fig 18

 

 The above figures show the latency vs intensity curve with the latency of the V peak in the y axis and intensity of the V peak along the x axis. Dashed lines indicate the normal curves, dashed lines- cochlear impairment and thin lines- retrocochlear pathology.

            Bauch, Rose, and Harner (1982) found that the presence of a bilaterally delayed wave V was the least clinically useful measure compared with the absence of the BAEP or the prolonged interaural wave V measure. Their findings showed that bilateral prolongation of wave V occurred in only 3% of their retrocochlear ears; an absent BAEP was present in 50% and the prolonged interaural wave V was present in 43% of their retro-cochlear-impaired ears. The primary factor accounting for their low rate was that the peak latency of wave V was not considered prolonged unless it was prolonged bilaterally. It is probable that they employed, this measure to detect the presence of bilateral tumors which would otherwise be missed with the interaural wave V latency difference.

            The results of the study by Clemis and McGee (1979) indicated that the unilaterally or bilaterally prolonged wave V latency measure was clinically useful; this measure detected 70% of retrocochlear-impaired ears, compared with the absence of the BAEP waveform measure, which detected only 15% of the retrocoehlear-impaired ears. Coats (1978) found that wave V latency was prolonged in 79 to 86% of his retrocochlear-impaired ears (depending upon whether the degree of hearing loss in the high frequencies was considered).

PROLONGED INTERAURAL LATENCY DIFFERENCE FOR WAVE V

            The presence of an interaural difference in the peak latency of wave V has been used as an indicator of retrocochlear pathology. With this measure, the ear not at risk for retrocochlear pathology serves as the control ear. Thus, the subject serves as his or her own control. Setters and Brackmann (1977) described the distribution of interaural latency differences (ILDs) for ears with retrocochlear pathology (tumors) and ears with cochlear hearing impairment. The most common ILD in the nontumor group was 0 ms, that is, many of the cochlear-impaired ears had equal latencies in both ears despite the presence of a unilateral cochlear hearing impairment. In contrast, in the retrocochlear-impaired group, there was a wide distribution of ILDs; 21 of the 46 retrocochlear-impaired ears had ILDs of at least 0.4 msec.

            Selters and Brackmann (1977) reported that the ILD measure was as good as the absence of response measure in detecting the presence of retrocochlear pathology; they reported that 46% of their retrocochlear-impaired ears had significantly prolonged ILDs. Bauch, Rose, and Harner (1982) obtained similar findings; the ILD measure was prolonged in 43% of their retrocochlear impaired ears. Musiek, Josey, and Glasscock (1986a) reported that more of their retrocochlear-impaired ears had a prolonged ILD than had a prolonged I-V IPL: the former measure identified 25% more ears than the latter measure. Clemis and McGee (1979) reported that a prolonged ILD was present in essentially the same proportion of retrocochlear-impaired ears as the prolonged absolute wave V peak latency measure.

            Once wave V can be measured bilaterally, the percentage of retrocochlear-impaired ears with significantly prolonged ILDs has been reported to be very high—as high as 100% as reported by Musiek et al. (1986b) and as high as 90% as reported by Eggermont et al. (1980).

            The ILD measure has a drawback in cases of bilateral retrocochlear impairment, which would have an effect on the peak latency of wave V in both ears, so the interaural difference would not be significant. Another disadvantage of the ILD measure is that it may be influenced by asymmetrical hearing impairment.

            Moller and Moller (1983) obtained the ILD for wave 111 as well as for wave V in their group of 23 patients with cerebellopontine angle tumors, in 11, the ILD for wave 111 was the same as for wave V. In 4 ears, the ILD for wave 111 was Jargcr than for wave V and in another 4 ears, the reverse finding was obtained. Thus, Moller and Moller suggested that the ILD for wave 111 may be a useful indicator of retrocochlear pathology.

            Musiek, Johnson, Gollegly, Josey, and Glasscock (1989) observed that in persons with bilaterally symmetrical hearing thresholds, an ILD for wave V exceeding 0.3 ms occurred in none of their normal-hearing cars, 21% of their cochlear-impaired ears, 1007u of their 15 cases with eighth-nerve or cerebellopontine-angle tumors, and 47% of their 15 cases with brainstem lesions. If the ILD of 0.4 ms was considered significantly prolonged, a significantly prolonged ILD occurred in none of their normal group. 6% of their cochlear-impaired group, 100% of their eighth-nerve cerebellopontine-angle group, and 40% of their brainstem group. bus, the ILD for wave V appears to be less sensitive to brainstem than to eighth-nerve or cerebellopontine-angle lesions.

I-V INTERPEAK LATENCY

            The presence of a prolonged I-V interpeak latency (IPL) or interwave interval (1\V1) has been commonly employed as an indicator of retrocochlear pathology. Coats (1978) investigated the  I-V latency-intensity functions in 14 retrocochlear impaired ears with varying degrees of hearing impairment. Figure  shows the individual functions superimposed against the upper normal and cochlear limits. This figure shows that the I-V latency functions are markedly shifted upward in comparison with the normal-hearing and cochlear hearing-impaired limits. Thus, at a given click intensity level, the I-V IPL is markedly prolonged in a retrocochlear-impaired ear in comparison with a normal-hearing ear. Moreover, the figure 19 shows the opposing effects of retrocochlear and cochlear hearing impairment on the I-V IPL. In retrocochlear-impaired ears, the I-V IPL is prolonged whereas in cochlear-impaired ears, the I-V IPL is reduced. The differing effects of retrococh-lear and cochlear pathology become more apparent as the degree of cochlear hearing impairment is increased. Coats (197S) employed electrocochleography in order to record wave I so the I-V IPL could be determined.

            The drawback of the I-V IPL is that it is often difficult to record wave 1 during BAEP testing, particularly when hearing impairment is present. Eggermont et al. (1980) reported that 30% of their 45 patients with cerebellopontine angle tumors did not show a wave I. Cashman and Rossman (1983) reported that interwave measurements were unobtainable in 86% of their-group of 35 patients with acoustic tumors. Hyde and Blair (1981) reported that wave I was present in only 42% of their patients with sensorineural hearing impairment. Bauch et al. (1982) did not employ the 1-V IPL measure because many of the early waves could not be detected, even in normal-hearing listeners (Bauch, Rose, & C Harner, 1980), and, in cochlear hearing-impaired ears, these components were even more difficult to identify.

Figure 19 showing ABR in left sided acoustic nerve tumour

Moller and Moller (1983) demonstrated that the peak, latency of wave I is normal although the peak latencies of waves II and III are prolonged in cases of acoustic neuroma. The presence of a prolonged peak latency for wave II or III is then reflected as a prolonged peak latency for wave V. Therefore, the I-V IPL is prolonged. The peak latency wave I is normal since it comes from the part of the auditor nerve which is distal to the brainstem and acoustic neuromas are usually proximal to the brainstem. Since wave II is generated from the proximal portion of the eighth nerve, the peak latency of wave 11 is prolonged in cases of tumors affecting the auditory nerve.

            Most investigators who evaluated the clinical utility of several BAEP measures within the same study found that the I—V IPL measure is one of the more powerful for differential diagnosis since more retrocochlear ears and fewer cochlear hearing-impaired and normal hearing ears are labeled positive by this measure than by the other BAEP measures. Although the ILD measure identified more retrocochlear-impaired ears than the 1-V IPL measure, the ILD test is positive in substantially more cochlear impaired and normal-hearing ears than the I-V IPL measure. In patients with high brainstem lesions, the III—V rather than the I-III IPL is commonly prolonged (Lvnn Verma, 1985).

Weber (1983) and Stockard and Stockard (1983), maintain that interaural threshold differences contaminate the interaural I-V interpeak latency difference. They reported that the 95th percentiles for interaural 1-V IPL do not exceed 0.2 ms in normal-hearing persons or persons with unilateral or symmetrical hearing loss and do not exceed 0.3 msec in patients with asymmetrical or noise-induced hearing loss. Eggermont et al. (1980) considered interaural 1-V IPLs exceeding 0.4 ms to be positive. Further study on the interaural 1-V IPL is needed before judgment can be made regarding the clinical utility of this measure.

 ABSENT BAEP

Absence of the BAEP waveform even at the maximum-output levels of the instrument has been interpreted as consistent with the presence of retrocochlear pathology. Clemis and McGee (1979) and Egger-mont et al. (1980) reported that the BAEP waveform could not be obtained, that is, all waves were absent, in approximately 15% of their retrocochlear-impaired ears. Of the retrocochlear-impaired ears investigated by Bauch et al. (1982), approximately half had absent BAEP whereas 43% had a prolonged wave V interaural latency difference. The BAEPs were also absent in 10% of Bauch et al.'s (1983) cochlear hearing-impaired ears.

IV/V-.I AMPLITUDE RATIO

The presence of abnormal amplitude ratio (AR) as an indicator of retrocochlear pathology has been employed by some investigators (Stockard &c Rossiter, 1977). Hecox (1980) reported that a IV—V:I AR was suggestive of retrocochlear pathology if it was between 0.5 and 1.0 and was diagnostic of retrocochlear pathology if it was less than 0.5. He recommended that the AR be calculated at intensity levels not exceeding 80 dBnHL, for example, 60-70 dBnHL, since the AR decreases at the higher intensity levels. Musiek et al. (1984) found that the IV-V:I AR was less than 1.00 in 44% of their retrocochlear-impaired ears. In only 4% of their cochlear hearing-impaired and normal-hearing ears was the AR less than 1.00. Hecox (1983) suggested that, if there is doubt concerning which of the early peaks is wave I, the AR can be considered normal as long as none of the early peaks has an amplitude which exceeds that of wave V. The AR is not a commonly employed criterion for differential diagnosis since it is very sensitive to recording and stimulation variables.

LACK OF TEST-RETEST REPLICABILITY

Some investigators have suggested that lack of test-retest reproducibility of the BAEPs is predictive of retrocochlear pathology. Musiek, Josey and Glasscock (1986a) reported that waves I, III, and V were not replicable at the same repetition rate, intensity, and polarity in 74% of their 61 retrocochlear-impaired ears. In many patients with multiple sclerosis, there is lack of replicability. Elidan, Sohmcr. Gafni, and Kahana (1982) found that in several of thru patients with multiple sclerosis, waveform morpholep changed from minute to minute, possibly reflecting the presence of intermittent conduction in a demyelinated fiber (1980) contended that lack of replicability should not be considered a powerful predictor of neurologic disease since application of this criterion makes it hard to maintain quality control. Lack of replicability should be considered an indicator of retrocochlear pathology only it all technical and subject factors can be ruled out.

ABSENCE OF THE LATER WAVES

The presence of wave I only or the presence of the early waves with the absence of the later waves has been employed as an indicator of retrocochlear pathology (House & Brackmann. 1979). Rosenhall er al. (1981) reported that wave V was absent in 57% of their retrocochlear-impaired subjects. Selters & Brackmann (1977) reported that in 46% of their retrocochlear-impaired ears and 0% of their cochlear-impaired ears, wave V was absent.

ABNORMAL WAVEFORM MORPHOLOGY

Starr and Achor (1975) reported that distorted BAEP waveforms may be consistent with the presence of retro-cochlear pathology. Eggermont et al. (1980) observed that: normal BAEP waveform morphology was rarely present in ears with acoustic tumors. Peak identification is a major 'problem in ears with retrocochlear pathology. Significant cochlear hearing impairment may also adversely affect waveform morphology.

FIGURE 20 SHOWING THE ABNORMAL WAVEFORMS IN AUDITORY NEUROMA CASES

CONTRALATERAL EAR EFFECTS

            Selters and Braekmann (1977) observed that, when large tumors were present, the peak latency of wave V was prolonged and the wave V peak amplitude was reduced for the ear contralateral to the side of the lesion. The delay in wave V was reflected by the prolonged III—V IPL in the ear contralateral to the side of the tumor. The mean 111—V IPL in the nontumor ear in cases in which the tumor size was < 3.0 cm was 1.87 ms as opposed to 2.14 ms in the nontumor ear in cases in which the tumor size was at least 3.0 cm. This difference was statistically significant. Selters and Braekmann (1977) hypothesized that a prolonged III—V IPL in the ear contralateral to the side of the tumor can result when the tumor is large enough to displace the brainstem and compress the contralateral auditory neurons. This hypothesis was confirmed by the results of computerized cranial tomograms.

Musiek, Sachs, Geurkink, and Weider (1980) reported on the BAEP findings in a patient with retrocochlear pathology in which the III—V IPL was prolonged in the ear contralateral to the side with the tumor. The hypothesis of a large tumor pressing against the brainstem was confirmed by radiology and at surgery. The presence of a profound hearing loss in the retrocochlear-impaired ear precluded BAEP testing in that ear; the tumor was detected by doing BAEP testing in the non-retrocochlear-impaired ear. Schwartz (1985) found that the I-V IPL was prolonged in the ear contralateral to the side with a large cerebellopontine angle tumor.

 COMBINED BAEP MEASURES

Several investigators judge the BAEP waveform as normal or abnormal on the basis of several rather than just one BAEP pleasure. The ear is considered abnormal if abnormal findings are obtained on at least one of the BAEP measures. Most investigators have reported hit rates exceeding 90% when the evaluation of the BAEP waveform is based on several BAEP measures. These high hit rates indicate that the BAEP test is highly sensitive to the presence of retrocochlear pathology. The false-positive rate ranges from 4 to 30%. Such differences may reflect differences in the hearing sensitivity of the control group, the use of different criteria BAEP measures, and the use of different recording parameters.

            Another factor which may contribute to the high false-positive rates reported by some investigators is the possibility of undetected retrocochlear pathology, such as head trauma, which may affect various levels of the auditory brainstem pathway in the cochlear-impaired and normal-hearing ears. Moller and Moller (1985) reported that vascular compression at the brainstem in older adults may result from looping of the anterior inferior cerebellar artery in the cerebellopontine angle and the audiovestibular and facial nerve bundles. If these loops become elongated In-cause of arteriosclerosis or deterioration of vascular collagen, BAEP abnormalities may result. Symptoms of the disorder include vertigo, unilateral sensorineural hearing impairment, and reduced speech-recognition ability. This disorder cannot be detected by any existing radiographic technique—only by vascular compression studies.

Because of the effect of hearing impairment on the BAEP waveform, the risk of a false-positive result increases as the magnitude of high-frequency hearing impairment increases beyond 50 dB HL. Another factor which may contribute to the false-positive rate is the inability to record wave 1 in many cases. Without wave I, the I-V IPL, the most powerful measure, cannot be obtained, leading the investigator to rely on less powerful BAEP measures that are more susceptible to the effects of peripheral hearing impairment.

 THE EFFECT OF TUMOR SIZE AND LOCATION ON THE BAEP

Selters and Brackmann (1977) found that tumor size was highly correlated with the ILD. The BAEPs were absent in 30% of the ears with acoustic tumors less than 2.5 cm in Size and were absent in 80% of ears with acoustic tumors greater than 2.5 cm in size. Clemis and McGee (1979) also reported the presence of a significant correlation between tumor size and the ILD for I wave V. Nevertheless, one patient with a small rumor had a substantially prolonged ILD. In cases when no response |was present, there was a wide range in rumor size. Thus, it is not only the size of the tumor but also the site of the tumor that determines the strength of the effect on the BAEP. Even a small tumor can have a marked effect on the wave V peak latency if it is strategically placed. In general, however, the larger the tumor, the greater its effect on the BAEP latency. 

Musiek, Kibbe-Michal, Geurkink, Josey, & Glasscock ; (1986b) found no correlation between any of the BAEP indices and tumor size in their group of 16 young adult normal-hearing patients with posterior fossa tumors. They suggested that other factors such as tumor consistency, rate of tumor growth, exact site of the rumor, and neural plasticity influence the BAEP in ears with posterior fossa tumors. In cases of low brainstem lesions, BAEP abnormalities are usually observed in the ear ipsilateral to the side of the pathology unless the lesion is large enough to compress the brainstem, in which case BAEP abnormalities may be observed bilaterally. In cases of high brainstem lesions, BAEP abormalities may be observed ipsilateral or contralateral to the site of the lesion or bilaterally.

REFERENCES

• Auditory Diagnosis by silman and silverman
• Handbook of Auditory Evoked Responses by James.W.Hall