The Auditory Brainstem Response (ABR) is a valuable measurement for assessing hearing thresholds in young patients and adults who are unable to cooperate with behavioural testing or subjective audiometry (Mason et al, 1995; Stevens et al., 2010; Don & Kwong, 2002). This important measurement is negatively affected by excessive noise and continuous research has been conducted to reduce the tainting effect that the noise has on the averaged recording (Sanchez & Gans, 2006; Don & Elberling, 1994).
As the measurements made using ABR are derived from the peripheral auditory pathway and lower brainstem nuclei (Jewett & Williston, 1971; Kevanishvili, 1980) they are of a very small potential, meaning that excessive noise has a great effect on their recording. Research has found that noise reduction techniques (NRT) such as filtering, averaging and artefact rejection have been employed to facilitate the extraction of the evoked potential (EP) from unwanted noise (Sanchez & Gans, 2006; Don & Elberling, 1994).
The purpose of this report is to give background information on what ABR testing is, what its parameters mean, the meaning of noise, methods of removing noise and the effect of removing noise on the overall waveform recorded. This will help the reader understand the purpose of my project, which is to, "Investigate the effect of different Artefact Rejection Levels (ARL) on ABR recordings". I aim to find an optimum level which can then be applied to the paediatric application of the test.
NHSP (Newborn Hearing Screening Program) protocol recommends a starting level of ±5µV which can be altered within the range of ±3µV and ±10µV (Stevens et al., 2010). However, departments within the region have been found to increase the level above this in order to gain better, quicker recordings and as an effect maximising time efficiency; however by doing so they are potentially causing a detrimental effect to the quality of the final waveform obtained. My research will determine whether their decision to use an alternative figure is a good idea or whether it is causing the quality of their recordings to suffer.
This research is clinically significant due to the constant daily use and application of ABR testing in the clinical Audiology environment nationwide.
Auditory Brainstem Response (ABR)
Auditory Brainstem Response audiometry is a neurological test of auditory brainstem function in response to auditory stimuli (Bhattacharyya, 2009; Zerlin 1988; Lightfoot, 1992). The two main applications of the test remain to be objective hearing threshold estimation and the detection of retro cochlear pathologies (Lightfoot, 1993). The waveform produced is an EP generated by a brief click or tone pip transmitted from an acoustic transducer in the form of an insert earphone or headphone (Stevens et al., 2010). The response is measured by surface electrodes, typically placed at the vertex of the scalp and mastoids. The amplitude of the signal is averaged and compared against the time, much like an Electroencephalograph (EEG). The waveform peaks are labelled I-VII; the peaks are derived from their respective nuclei as the signal advances up the auditory pathway and brainstem (Ponton et al., 2002). These waveforms normally occur within a 10-millisecond time period after a click stimulus presented at high intensities (70-90 dB normal hearing level (nHL)).
Patients being tested via ABR are not required to perform or actively do anything in order for the test to be performed. The recordings are taken electronically via electrodes and the patient is requested to fall asleep to ensure better recording conditions, due to lower levels of activity.
The neural generators involved in the production of the waveforms have been debated over by researchers for a while, mainly as the ABR waveform is a far-field recording. This means that the electrode site is at a great distance from the site of the EP (Jewett & Williston, 1971). This makes it difficult to locate the exact generator of a component of the waveform, meaning that a peak could be due to a summated response from two generators. The generators generally agreed to be the source of the waves are I - Distal portion of the VIII cranial nerve, II - Proximal portion of the VIII cranial nerve, III - Cochlear nuclei, IV - Superior Olivary Complex, V - Lateral Leminiscus, VI & VII - Inferior Colliculus (Hall, 2007; Don & Kwong, 2002).
Wave V is deemed to be the most robust, due to greater amounts of neurological bodies contributing to its response, (Lightfoot, 1993). Wave V's latency and amplitude is checked to determine whether a response is present or not. If it has reduced amplitude or a prolonged latency then this could be indicative of a possible retro cochlear pathology (Stevens et al., 2010; Aoyagi et al., 1990). Other comparisons made on the waveform of the ABR is the overall latency of the response between each ear, the inter-peak interval between waves I-V, the absolute latency of wave V and the absolute latencies of inter-peak intervals of waves I-III, III-V and I-V (Hall, 2007). The waveform is also checked to see if it is repeatable and its morphology checked.
There are two types of stimuli used when performing ABR measurements and they are chosen depending on what information you require. There is a click stimulus with a centre-frequency of 2-4kHz. This stimulus is normally used for general neurological screening of a patients hearing as it stimulates the whole cochlea and produces a much stronger amplitude response, meaning that diagnosis and interpretation of the waveform is made easier (Stevens et al., 2010). But due to the overall stimulation of the cochlea is does not produce frequency specific information. The other stimulus is a tone pip, this is used to obtain more frequency specific information, i.e. for fitting a hearing aid to a patients ABR data. It is beneficial as it allows an objective measurement of the patients hearing threshold as a set component of their hearing rather than a summated response. This stimulus focuses on a small frequency area and is gated using Blackman gating to reduce frequency splatter; this is the leakage of unwanted frequencies into the stimulus. Due to focusing only on a small area of the basilar membrane, as opposed to the whole basilar membrane using a click, the EP generated using tone pips is of a lesser amplitude and greater latency, depending on the frequency of the stimulus (Durrant & Boston, 2007).
The rate at which the stimulus is delivered also determines the shape of the waveform. Typically for threshold ABR audiometry a faster rate is used as the only interest in the waveform is wave V (Stevens et al., 2010), so a faster rate is used as wave V is very robust. For neurological ABR testing a slower rate is used as it is important to get a strong response from all neural generators. Both rates must take into account the refractory period of the neurones involved in the response because if the rate is faster than this then adaptation to the signal occurs, meaning that the amplitude of wave V will decrease and therefore making interpretation of the recorded waveform difficult (Lightfoot, 1993).
The electrodes used for the detection of the response are Silver/Silver Chloride and should have a skin contact impedance of 5Ω or less. Each electrode should also be within 1Ω of another. The better the contact the better the accuracy and reliability of the recordings obtained. Therefore the skin is cleaned using an abrasive gel to remove dead skin, oils and debris prior to testing. (Stevens et al., 2010). They are positioned strategically to improve the quality of the measurement obtained; their position can enhance the amplitude of the EP or its components (Don & Elberling, 1994).
ABR is used extensively in the NHSP in England. This is an initiative set up in 2001 which ensured all children born in England are screened for their hearing ability with the incentive to diagnose and intervene as early as possible, should there be a problem.
There are many subjective factors which can affect the recordings made using ABR. These involve patients' age, gender and head diameter (Aoyagi et al., 1990). ABR response data is compared to normative values to determine whether it is normal or not, meaning that a sample of a population has been selected and tested to interpret what is a normal response and averaged.
Age is a significant factor as people have different levels of neural maturation meaning that the developments of their neurological bodies, with regards to myelination, vary. The more matured their auditory pathway and brainstem is the better the conduction of the signal and the faster the response occurs. Naturally as babies are less matured they should have a longer latency response, meaning that wave V should have a greater latency than that of a 21 year old. Age is a significant factor as babies tend to have poorer conduction which improves as they age; maturation peaks once a patient is middle aged but then as they get older the pathway tends to diminish again.
The gender of a patient has a subtle difference also as development rates between the sexes tend to vary, meaning that normative data for a similar aged man and woman can differ. Therefore this is another factor to take into account when considering normative values.
Head diameter is considered a factor due to the assumption that the volume of the brain is larger for larger heads meaning that the auditory pathway is a greater size, which could cause a latency delay (Aoyagi et al., 1990). This is a factor that correlates with gender and age.
Noise
ABR recordings are taken via far-field electrodes, meaning that the electrodes are placed at a great distance from the actual evoked site (Jewett & Williston, 1971). Also the EP generated is of very small amplitude (0.1-1µV), the frequency range of the ABR response is 30-3,000Hz (Sokolov et al., 2005). With these two points in mind it is understandable that the recordings made are very susceptible to external interference and noise.
Noise sources can be both extraneous and physiological. Physiological noise tends to be of an electrical nature and is derived from the following sources; the brain (EEG), the eyes (Electrooculogram (EOG) and Electronystagmogram (ENG)), the heart (Electrocardiogram (ECG)) and the skeletal muscles (Electromyogram (EMG)). Each of the sources have different effects on the trace at different amplitudes and frequencies, which will be explained to emphasise the effect of noise on the recordings.
EEG activity has an amplitude when awake of 70-100µV within a frequency range of 3-40Hz. However in light sleep the amplitude can rise to 400µV and have a centre frequency range of 10Hz (alpha-waves) (Sokolov et al., 2005). This sleeping state can saturate the amplifier meaning that it is impossible to make a recording and is therefore filtered out of the response, which will be explained further into this report.
The eye houses an electric potential field which can be described as a fixed dipole with a positive pole at the cornea and a negative pole at the retina (Ennever et al., 1971; Sokolov et al., 2005). This potential can be between 400-1,000µV in the frequency range of 0.5-10Hz but is only generated once the eye rotates (Halgren et al., 1998). This is another reason for why ABR testing is best performed with the patient asleep with minimal eye movement.
ECG activity occurs within 1-50Hz with amplitude up to 500µV (Sokolov et al., 2005). In infants this can reach higher frequencies due to their twofold heart-beat rates and have larger amplitude due to closer proximity of the electrodes to the heart.
Muscular activity (EMG) generates very strong artefacts up to 200µV, with the strongest artefact coming from the face and neck muscles. EMG artefacts are within the frequency range of 30-500Hz. Their effect can reduce with sleep however they will always be present in a relaxed state so must re moved by ARL's do to their sporadic nature.
Extraneous noise is artefact which is derived from an external source within the testing environment. It can include mains interference (50Hz), electromagnetic field interference and radio-frequency noise. The test is best performed in an area away from mains outputs and the wiring be kept free from any electromagnetic sources. Appliances such as mobile phones should also be kept away from the equipment during testing.
To try and improve the quality of the recordings made there are a few noise reduction techniques (NRT) to ensure the best reading possible can be made.
Noise Reduction Techniques (NRT)
NRT's such as filtering, averaging and artefact rejection all play a combined part in achieving the best signal-to-nose ratio (SNR) possible; this is the ratio between how much signal there is in relation to noise. The greater the SNR value the better the quality and reliability of the waveform recorded. The aim is to remove as much noise as possible whilst maintaining the signal, however by removing the noise we are also removing some important signal information (Stevens et al., 2010).
Filtering involves the complete removal of frequencies which can be detrimental to the recording. It works by only allowing a set range of frequencies pass which are relevant to the information we seek. NHSP protocol suggests the use of a high-pass filter of 30Hz and a low-pass filter of 1500Hz (Stevens et al, 2010); meaning only signals in the frequency range 30-1500Hz are recorded. Research shows that the main spectral bands involved in the frequency composition of the human ABR waveform are 0-350Hz, 350-700Hz and 700-1200Hz (Suzuki et al., 1983; Kevanishvili & Aphonchencko, 1979). The main frequency region that contributes to wave V's presence is below 350Hz (Sanchez & Gans, 2006). Therefore the application of filters is a compromise between passing too much low frequency noise, which is of high amplitude and can mask the recording, and passing low frequency signal enabling a better, more defined response (Laukli & Mair, 1981). Filtering alone has shown to provide minimal improvement of the SNR due to the frequency spectrum of noise overlapping the frequency composition of the EP (Doyle & Hyde, 1981).
Signal averaging involves the recording and summation of many recordings of the EP from ABR testing in order to achieve a better SNR. Noise is random relative to the stimulus, but long term averaging will allow summation of neural components whilst cancelling out the noise (Don & Elberling, 1994). Although this means that numerous recordings need to be made to generate a high SNR. Therefore this technique is used in conjunction with filtering and specifically, artefact rejection.
Artefact rejection evaluates the amplitude of the incoming noise from the electrodes for individual sweeps. If the recording exceeds a predetermined level then the recording is excluded from the computers memory and will not be involved in the averaging process (Sanchez & Gans, 2006). The principal behind this method is that the EP will never exceed 1µV in amplitude therefore recordings which do are assumed to be artefact. The current NHSP recommendation is a starting level of ±5µV which can be altered within the range of ±3µV and ±10µV depending on the recording conditions (Stevens et al., 2010). There isn't much research on artefact rejection levels for ABR however the ±5µV value is derived from a compromise between the averaged waveforms quality and time. This is because a high rate allows too much noise, meaning that more recordings are accepted but the quality of the averaged waveform would be poorer. A low rate rejects too many recordings, meaning that the averaged waveform is of a better quality but takes longer to achieve. Sanchez et al. (2006) found that strict artefact rejection levels can lead to a difficulty in ABR interpretation, which could result in diagnostic errors. This was because the artefact rejection level had an aggressive effect on the low frequency information in the EP, causing wave V's amplitude to reduce. Stevens et al. (2010) also suggests that the use of higher rejection levels are not recommended in difficult recording conditions as it leads to a poorer SNR.
Study
All NHS departments who perform ABR for paediatric testing in England should follow the protocol set out by the NHSP (Stevens et al., 2010). However it is known that not all departments follow the parameter suggestions for artefact rejection strictly, it is known for people to increase the ARL to levels up to and above ±20µV. This is done to improve the recording made and to decrease recording time. This could have a detrimental effect on the waveform quality derived from doing this, which has an adverse effect on the diagnosis made.
The study intends to investigate the effect of the different ARL's on the final recording made by the ABR testing. I shall do this by exposing normal hearing participants to three different levels of ARL (±5µV, ±10µV and ±20µV). Each condition will be exposed to a no-stimulus trial and a stimulus trial of 70dBnHL, which is standard NHSP procedure (Stevens et al., 2010). The residual noise from each condition will then be compared. It's expected that an increase in the rejection level will cause the amount of average residual noise to also increase.
ARL is the only parameter to be changed; every other parameter shall remain as per standard NHSP protocol recommended by Stevens et al., 2010.
There isn't much previous research on ARL's as it is generally known to be a compromise and thought that the lower level that you're able to use the better, which is manipulated depending on testing conditions. Sanchez & Gans (2006) compared two ARL's on their effect on the amplitude of wave V for active and quiet patient conditions. The paper indicates that strict ARL's can lead to difficulty in ABR interpretation due to the reduction in amplitude of the waveform, wave V and poor waveform morphology.
This study is clinically relevant as NHSP ABR testing is often performed in day-to-day operation of an Audiology department in England. It is also worthwhile providing evidence as to a figure to use as it is generally a compromise.
