Clinical Evoked Potentials: An Illustrated Manual

© Springer Nature Switzerland AG 2020

O. N. MarkandClinical Evoked Potentialshttps://doi.org/10.1007/978-3-030-36955-2_1

1. Basic Techniques of Evoked Potential Recording

Omkar N. Markand¹ 

(1)

Professor Emeritus of Neurology, Department of Neurology, Indiana University School of Medicine, Indianapolis, IN, USA

Keywords

Stimulus-related evoked potentialsAD conversionSignal enhancementPrinciples of averagingEP nomenclatureNormative EP dataTolerance limitsNear- and far-field responses

The electrical activity arising from peripheral and central nervous systems may be subdivided broadly into spontaneous activity, e.g., electroencephalogram (EEG), and evoked potentials (EPs). The latter are time-locked and follow a fixed time period in response to a discrete sensory stimulus. Evoked potentials are mainly of two types:

1.

Stimulus-relatedEPs : They occur with short latency (greater than 1.50 msec) and elicited by stereotype stimuli of specific modality. Examples are brainstem auditory, visual, somatosensory, and motor evoked potentials.

2.

Event-relatedEPs : These are late potentials (e.g., P300), which are dependent on the information content of the stimulus. They appear when a subject attends to a meaningful stimulus. These EPs are not stimulus specific; equivalence in task relevance is the major determining factor in their elicitation.

The stimulus-related EPs are utilized most for diagnostic purposes. These will constitute the main subject for discussion in this text.

Basic Principals of EP Recording

The evoked potential recording equipment consists of three conceptually different components, although they may be all housed in the same cabinet:

1.

Stimulator to provide specific stimulation, e.g., visual, acoustic, or somatosensory stimuli

2.

Amplifier/filter system

3.

Analog to digital converter/averager

The stimulator is physically connected and thereby time-locked to the averager. A block diagram of EP recording equipment is shown in the Fig. 1.1.

Fig. 1.1

A simplified diagram of evoked potential equipment. AMP differential amplifier, LF and HF low- and high-frequency filters, respectively, ADC analog to digital conversion, DAC digital to analog conversion

Stimulus Characteristics

Stimulus of a specific modality is presented, e.g., auditory, visual, or somatosensory, to elicit brainstem auditory evoked potentials (BAEPs) , visual evoked potentials (VEPs) , and somatosensory evoked potentials (SSEPs), respectively. Certain characteristics of the stimulus are critical in eliciting satisfactory EPs. The stimulus should be:

Abrupt and short duration

Exactly controlled for rate and intensity

Constant and consistent

Short, rectangular electrical pulses delivered to the peripheral nerves are most commonly employed to elicit SSEPs rather than the more physiologic tactile stimulation of the skin because the former stimuli can be precisely controllable in its rate, duration, and intensity. Frequency of stimulus presentation during averaging needs to be properly chosen. It should neither be too fast nor too slow. Presentation of very fast rates usually deteriorates EP waveforms, whereas the slow rates prolong the EP acquisition time with undesirable change in the subject’s state of alertness or cooperation .

Amplification

EPs are very low in amplitude (0.1 μV–10 μV). On the other hand, a relatively large voltage (up to 1 V) is usually required by the AD converters and the averager, so that the input analog signal (containing EP) must first be amplified. Amplifications of 10³–10⁶ are available in most EP recording systems.

Filtering of the Analog Signal

Amplifiers have a wide range of LF and HF filters to properly condition the input signal prior to AD conversion. Careful filtering is essential so as to effectively shorten the number of averages required to extract EPs from the signal noise. Commonly available filters are:

LF: 0.01, 0.1, 0.3, 1.0, 3.0, 30, 100, 300 Hz

HF: 100 Hz to several kHz

60 Hz filter

Analog to Digital Conversion

Output of the differential amplifier is an analog signal, i.e., it continuously changes with time. In other words, there is an amplitude value at every point in time. On the other hand, the EP equipment uses digital processors capable of handling only binary or digital signals. Hence, analog output of the amplifier needs to be converted into digital format, which is accomplished in analog to digital converter (ADC) . AD conversion consists of two functions:

A.

Sampling: It involves sampling the analog waveform at regular time intervals (called intersample interval, sampling period, or sampling rate) and hold the sampled amplitude values for the next function of quantization to start. Sampling cuts the analog signal horizontally at regular intervals as shown in the Fig. 1.2b.

B.

Quantization: It is the second process in which the amplitude values of the sampled signal are assigned discrete amplitude levels with digital or binary coding, as shown in Fig. 1.2c.

Fig. 1.2

Shows the two major processes involving analog to digital conversion of an analog signal (a). It is first sampled at regular time intervals, to produce a sampled signal (b). Amplitude values are then assigned and encoded digitally to generate a quantized signal (c)

The horizontal resolution of the ADC depends upon the sampling rate or sampling period. Sampling of the analog EP signal has to be performed at short enough intervals, so that the statistical property of continuous signal is preserved. This is illustrated in Fig. 1.3. The Nyquist sampling theorem states that the sampling rate should be at least twice the highest frequency component of the analog signal. The EP signal is usually a complex signal composed of multiple frequency components. The highest-frequency component determines the required sampling rate. Sampling rate must be at least twice the highest frequency present in the data. If the EP data has the highest-frequency component, say, 2 msec in duration (i.e., frequency of 500 Hz), the sampling rate must be at least 500 × 2 = 1000 Hz. Usually a rate more than 4–5 times is recommended so as to attain digital representation as close to the analog input as possible. If the data is sampled at a rate slower than 2 times the highest-frequency content, a wave distortion or aliasing artifact is introduced. This artifact consists of false frequencies appearing in the output signal that were not part of the original analog signal (Fig. 1.4).

Fig. 1.3

Shows the effect of different sampling rates on the analog signal (a). With increase of the sampling rate or shortening of the intersample interval (ISI) as in (b, c) and (d), the sampled signal represents more closely the original analog signal in (d)

Fig. 1.4

Shows the wave distortion or aliasing artifact when an analog signal is sampled below the Nyquist frequency. The signal of 500 Hz is satisfactorily resolved using a sampling rate of 1000 Hz (Nyquist frequency) or faster. If the signal is sampled using sampling rates below 1000 Hz (last two graphs), there is an artifactual introduction of slower frequencies, which have not even been a part of the original analog signal

The vertical resolution of the ADC relates to quantization. ADC is usually designated as 4-, 8-, 10-, or 12-bit converter. A 4-bit (2⁴) convertor has 16 amplitude levels whereas an 8-bit has 256 levels, etc. An 8-bit converter would provide 256 amplitude levels with resolution of 1/256. As for example, a 10 μV peak-to-peak input signal can be resolved to a level of 10/ 256, i.e., 0.04 μV. The size of the converter which would ensure a good EP recording depends upon the lowest amplitude EP component relative to the overall amplitude of the signal. Usually, the number of converter levels needed would be equal to background amplitude/EP amplitude of the smallest component. If the EP has the smallest component of 0.1 μV, and the overall signal of 100 μV in amplitude one would need an AD converter with at least 1000 levels, that is, a 10-bit convertor (2¹⁰ = 1024). Most EP instruments have at least 11- or 12-bit converters to assure that the quantized digital output is a good representation of the original analog signal (Fig. 1.5).

Fig. 1.5

Shows the effect of number of bits or levels in the ADC on the quantization of the analog signal (a). A 3-bit converter is unable to resolve the last three low amplitude components (b), a 4-bit converter resolves only one of these (c), and it is only the 5-bit converter with 32 vertical levels which resolves all of the three components (d)

Signal Enhancement

Signal in EP recordings is the EP, whereas everything else is the noise. The latter consists of EEG, electromyogram (EMG), electrooculogram (EOG), electrocardiogram, electrode, electronic noise, etc. Evoked potential, which is a tiny signal, is thus submerged within the high amplitude noise. Signal/noise ratio varies with different modalities of EPs from 1/5 to 1/100. One of the main challenges of recording an EP is to improve signal/noise ratio to the extent that the EP becomes identifiable without confusing it with the elements of the noise. Signal averaging is the technique used to improve signal/noise ratio, which it does by selectively decreasing the amplitude of the noise.

During signal averaging, the signal is evoked many times, and the electrical activity (EP plus noise) is summated by the averager and then normalized (∑/N). Signal averaging takes advantage of the fact that EP follows the stimulus at a constant time interval, i.e., EP is time- locked to the stimulus; whereas the noise is random. Hence as the averaging progresses, the noise being random progressively decreases approaching close to zero, whereas the EP remains of the same amplitude.

The rule of averaging is that the noise decreases proportional to the square root of the number of trials, i.e., noise α 1/√N. Therefore, signal/noise roughly equals signal amplitude divided by noise amplitude multiplied by square root of N, i.e., signal/noise = signal/noise x √N.

The principal of averaging is illustrated in Fig. 1.6. Because of the square root relationship of the noise and the number of averages, it is obvious that greatest improvement of signal/noise is achieved at the beginning of a set of averages. Signal/noise improvement by a factor of 5 can be obtained by 25 averages, whereas a factor of 10 will require additional 75 averages (a total of 100 averages). Remember, that signal averaging improves the signal/noise ratio essentially by decreasing the amplitude of the noise and not in any way enhancing the signal.

Fig. 1.6

Demonstrates basic principle of signal averaging . (a) The figure on the top diagrammatically elucidates that the recorded signal consists of the background noise and the EP. As the averaging progresses, the noise being random progressively decreases making the EP stand out above the noise level. (b) The figure below shows that the noise decreases by square root of number of averages. As the averaging progresses, the noise decreases enough making the EP identifiable by its emergence above the noise level

Limitation of Signal Averaging

A.

The noise may not be random; hence it may not get eliminated according to the square root of the number of averages.

B.

If the rate of stimulus presentation is harmonic or subharmonic of an artifact (e.g., 60 Hz), the artifact may get enhanced and deteriorate the quality of the recorded EP.

C.

Irrelevant signals may occur time-locked with the stimulus, e.g., EMG activity may be triggered by the stimulus. Such time-locked signals may contaminate the EP.

D.

There may be excessive artifacts, e.g., eye movements, blink, EMG, and these may not get fully eliminated even by increasing the number of averages.

E.

Averaging may degrade the signal if variability occurs in the individual’s EPs elicited from one stimulus to the other giving rise to (a) latency jitter, (b) amplitude variation, and (c) both amplitude and latency variations (Fig. 1.7).

Fig. 1.7

Shows the effects of variability in the latency (a) and amplitude (b) of the individual EP from one stimulus to the other. These degrade the final average EP

Strategies to Improve EP Recording

A.

It cannot be overemphasized that noise must be reduced to start with by proper electrode application, ensuring patient’s comfort (to reduce EMG, EOG, and movements), and, if necessary, by using sedation for BAEP and SSEPs.

B.

Train of stimuli must be presented at a rate unrelated to 60 Hz, e.g., for VEP at a rate of 2.1/sec, for BAEP 11/sec, and for SSEP 4.7/sec. In the case of BAEP, a stimulus artifact commonly occurs with high-intensity stimulation, which can be effectively eliminated by using alternate polarity clicks.

C.

Individual trials having amplitude above predetermined levels must be removed by using automatic artifact rejection. This will reject a trial contaminated with high-amplitude artifact not to become a part of the averaging process. It is achieved by a window discriminator. If a sweep generates an input signal outside the set window, it is eliminated from subsequent processing by the AD converter. Artifact rejection is an important method to prevent artifactual responses becoming a part of the ultimate averaged response.

D.

Adequate filtering must be used to reduce noise frequencies which do not attenuate the evoked potential per se.

E.

Two or more replications must be obtained and plotted upon top of each other. Only those waveforms, which are consistent between the recorded replications, can be unequivocally considered to be stimulus related (Fig. 1.8).

Fig. 1.8

Demonstrates the need for two or more replications. X and Y are not stimulus evoked components because they are not consistent on the two replications. They are probably part of the noise. If there was only one replication, these waveforms could have been easily mistaken for EP components

Common Problems Encountered During EP Recordings

Remember that EPs are very tiny responses in the microvolt range. To record optimal responses for interpretation is very often challenging especially in difficult patients. This is particularly true of the SSEPs especially the posterior tibial elicited responses. A close working relationship between the technologist and the clinical neurophysiologist ensures higher percentage of high-quality studies. At my institution, I will always like to look at the EPs before the patient leaves the laboratory to make sure that optimal recording has been achieved. Some of the common difficulties encountered are as follows:

EPs may not be recordable. The causes may be several such as inadequate stimulation, high impedance electrodes, high-amplitude artifacts (from the subject, instrument, or environment), incorrect use of recording electrodes, or incorrect electrode placement.

Systematic checking of the electrode impedances, electrode placement, and electrode connection to the amplifiers needs to be undertaken. For SSEPs, one needs to make sure that the stimulating electrodes are appropriately placed on the skin overlying the nerve to be stimulated and that the stimulus intensity is sufficient to produce a visible muscle twitch. In the case of the BAEP recordings, the earphone may not be secured tightly across the ears or may have even got slipped down over the mandibles.

EPs are of suboptimal quality due to excessive artifacts. When the subject is tense or moving a lot, the artifacts may not be thoroughly removed by increasing the number of averages or by automatic artifact rejection. The signal-to-noise ratio in such situations cannot be improved enough to extract a robust and reproducible EP. In such conditions assuring the patient and identifying and eliminating the source of extraneous interference, use of sedation (for BAEP and SSEP) may be some of the strategies which may improve the quality of the recording.

At times, the only remedy may be just to reschedule the patient for testing at a later date, preferably at no extra charge. I have done that in some 2–5% of the patients, and this usually works in almost all by providing an optimal study at the time of retesting.

The EPs look odd. Not uncommonly the EPs are well resolved but look different and one of the reason may be that the electrodes in one or more recording channels are reversed. As for example, instead of using A2–Cz derivation to record BAEP from right ear, the recording is actually made from Cz to A2. This problem can be easily identified by holding the paper copy of the EP against bright light and reversing the paper which will have an effect of reverting the derivation to A2–Cz.

Final Recording, Display, and Storage of EPs

Averager converts the digitized waveforms to analog waveform through the process of digital to analog conversion, an opposite of AD conversion. The EP can then be displayed on visual monitor. Measurements are made using cursor. Also, hard copy of the EP can be made by printer and the responses stored on a digital storage device.

Nomenclature of EP Components

There have been three commonly used systems to name EP components . These are:

Ordinal, used primarily for BAEP components which are designated as Roman numeral, e.g., wave I, wave II, and so on

Polarity-ordinal, e.g., N1, P1, N2, P2

Polarity-latency, e.g., N10, P37, P100

The letters P and N in the second and third systems above indicate positive and negative polarity, respectively. The major problem with the ordinal and polarity-ordinal system is to identify components correctly if initial ones are absent due to pathology or technical issues; e.g., if P1 is absent, the P2 could be mistaken as a delayed P1. Furthermore, the numerals 1, 2, etc. provide no information regarding the latency of that component except denoting their order or sequence of occurrence. Often the EP components are named according to the generated sources, e.g., brachial plexus potentials or Erb’s potential for N9 component of the median SSEP and popliteal potential for the N7 of the posterior tibial SSEP.

The polarity-latency convention is preferred for naming the EP components because it provides information of the component’s polarity as well as its mean latency. Even with this system, there are several problems, e.g., latency of scalp-recorded P37 depends upon age, height, and site (ankle vs knee) of posterior tibial nerve stimulation. A recommendation by the American Clinical Neurophysiology Society to use a horizontal bar above the component to indicate a theoretical component has not gained much acceptance. Also, the polarity of an EP component may depend on the recording derivation. One example is that of the negative polarity component, N13, recorded from an electrode placed over the C5 spinous process on median nerve stimulation at the wrist. The same component, on the other hand, is of positive polarity, P13, when recorded from an electrode placed over the anterior neck. Hence, knowledge of the recording derivation is essential for the polarity to be relevant.

Measurements

Once the EP has been acquired, the latency of the obligate components, interpeak latencies as well as the amplitude of the main components are determined by curser technique. Latency is measured in milliseconds between the time of stimulus presentation and the EP component, usually its peak. This is often called absolute latency. Interpeak latency refers to the time interval in milliseconds that separates the two peaks. Since the absolute latency may be influenced by multitude of factors, interpeak latencies (which are less often affected) are recommended to assess clinically relevant abnormalities. Most EP equipments provide a time cursor to move manually to the point where the latency needs to be determined. Latency measurements pose no issues once the peaks are correctly identified. The interpreter must always make sure that the technologist has made the measurement at a correct point on the EP waveform before accepting that latency value. Amplitude of the EP component is measured in microvolts. The appropriate measurement of the amplitude of an EP component is not adequately standardized. Should it be measured relative to the baseline (often undefinable) or to the peak before or after the component? Another method to denote EP amplitude is by determining the area under the component. This is rarely used because it requires computer-driven measurements and poses difficulty also in correctly identifying the total area to be measured. The amplitude of an EP component is commonly expressed in reference to the points before or after. In the case of BAEP and SSEP, the amplitude of an EP component is measured to the following peak, whereas for the VEP components, the amplitude is measured to the earlier peak (N75) (Fig. 1.9). In general, the amplitude of EP components are relatively less useful compared to the latency values because the former have much greater intersubject variability in normal population. Significant differences in the amplitude between the two sides may constitute at best a soft abnormality.

Fig. 1.9

Demonstrates the different methods of amplitude measurement of EP components

Normative Data

It is preferred that normative data is obtained in each laboratory engaged in performing significant number of EP studies. If one elects to use normal values from other institutions, it is mandatory that similar instrumentation and recording techniques are utilized as closely as possible for the borrowed values to be correctively used in interpreting the EPs. It is emphasized that of the three EPs, normative data for VEP must always be obtained in each laboratory because several variables associated with VEP recordings result in a large variability between different laboratories.

The normal value of EP latencies are commonly determined by doing the appropriate EP test in a sample of 30 or more normal subjects who do not have symptoms or signs pertaining to the afferent system responsible for that EP. The mean (x̄) and standard deviation (SD) are determined, and the upper limit of the latency is usually set at 2.5 SD or 3.0 SD above the mean. In clinical testing, the latencies at the lower end of the bell curve, that is, very short latencies, are not considered abnormal; only longer latencies are clinically relevant. Hence one tail statistics are adequate. For values determined in a very large sample (approaching population characteristics), 2 SD, 2.5 SD, and 3.0 SD above the mean will include roughly 97.7%, 99.4%, and 99.9% of the values in a normal population, using one-tailed statistics. Most laboratories use mean + 2.5 SD or mean + 3.0 SD to set their upper limits of normal latency values. Since the sample size to obtain normal data is usually smaller (around 30), the corresponding percentages for 2.0, 2.5, and 3.0 SD above the sample mean may be slightly lower. Another problem is that different normal samples would have slightly different sample means and standard deviations. Therefore, using a sample mean and adding 2.5–3.0 SD may not provide the most accurate reflection of the upper limit of the latency values in a normal population. The American Clinical Neurophysiology Society (2006) has therefore recommended using tolerance limits . Introduced be Shewhart in 1931, tolerance limits define an interval that includes a proportion of the overall population with a given confidence level. For EP latencies, only the upper limit is clinically relevant. Hence, establishing a one-sided tolerance limit is appropriate (Table 1.1). The concept of tolerance limits is somewhat complex, but it takes into account the variability of the x̄ and SD of different samples taken from a population. Usually 99% tolerance limit is determined with at least 95% confidence level. It is illustrated below:

Table 1.1

One-tail tolerance factor (K) for normal distribution

Tolerance factor K such that probability is γ that at least a proportion (1–∝) of distribution will be less than x̄ + Ks (or greater than x̄ – Ks) where x̄ and s are estimate of the mean and the standard deviation computed from a sample size n (Adapted from Lieberman 1957)

For example, suppose that the sample mean for IPL I–V is 3.97 msec with SD of 0.19 msec as determined in a sample of 30 normal subjects in a laboratory. A value of K (also called tolerance factor) is determined corresponding to n = 30, γ = 0.95 (confidence level 95%), and α = 0.01 (tolerance limit of 99%). For one-sided tolerance interval, K = 3.064 (Lieberman 1957). The required tolerance limit is given by x̄ + (K × SD) = 3.97 + (3.064 × 0.19) = 4.54 msec. What this value tells us is that we are certain that 99% of values for IPL I–V in the population fall below 4.54 msec and that we have 95% confidence in this statement. Obviously if one has a larger sample size, e.g., 100 or more, the value of K is lesser and the upper tolerance limit may be slightly shorter. Note that the value of K for one-sided tolerance limit is close to 3 for n = 30. Hence, if a laboratory uses a sample size of 30 subjects to determine the normative data, the x̄ plus 3 SD will give a value of 99% tolerance limit at 95% confidence level.

Another