Electroencephalography
5th Edition

7
EEG Recording and Operation of the Apparatus
Edward L. Reilly
When the first edition of this book was published, this chapter stressed the importance of technical considerations in obtaining high-quality electroencephalogram (EEG) recordings. In the interval between the original and the current edition, there has been improvement in equipment, more integrated circuits, greater computerization, machine self-monitoring, machine automation, and less noisy electronics.
These changes can be great assets in attempting to produce a reliable diagnostic record, but as with many innovations, there can be a negative side. In the early EEG machines, a minor error in technique, whether a bad electrode or leaving some sort of electrical device such as a lamp in the room, created 60-cycle noise. Improved machinery is far more resistant to artifacts. In older machines, failure to plug the EEG cable to the jack box resulted in 60-cycle artifact of such severity that anyone in the room with the machine noticed it. An unplugged jack box in modern machines provides an extremely low voltage record that might be interpreted to represent extremely low voltage, but still plausible, brain activity if one was expecting severe damage to the central nervous system (CNS) of the patient or even electrocerebral silence.
In the office, such a record, of course, will be questioned and the error quickly picked up. But in an intensive care unit, where such an error is much more likely to happen, it may continue for minutes or longer before it is apparent and only after deliberate production of artifacts as a means of electrode verification.
Quality recordings continue to require the use of fail-safe techniques on the part of the technologist. In some instances, “smart” equipment is relied on but is run by less trained operators who have no troubleshooting skills. In reality, it may be necessary to have increasingly talented operators as we develop increasingly sophisticated equipment, if quality is to be maintained.
If one doubts the tendency to put less skilled operators with better equipment, one has only to go to the grocery store and compare the present-day cashier running a talking, adding, and subtracting register to the people who used to be hired to make change from their own calculations. In a few of the following subsections, specific points regarding the changing times will point out that the published guidelines for careful technology have become more critical, especially as related to evoked potentials or long-term ambulatory monitoring.
EEG Recording Electrode Placements
When EEG was first successfully carried out on humans by Hans Berger, electrodes were placed on the front and back of the head. Berger continued that method for a number of years and viewed what he saw as a measure of global cortical activity (Berger, 1929). It was discovered by others that, in fact, EEG activity varied in different locations on the head (Adrian and Matthews, 1934; Adrian and Yamagiwa, 1935).
By the mid-1930s, as the number of laboratories investigating EEG increased, there was a rapid proliferation of techniques and interpretations of the activity recorded. Multiple channels allowed investigators to record simultaneously from different scalp areas, and the presence of localized activity such as alpha rhythm and sleep spindles was discovered.
These observations were, in turn, followed by increased attempts to place electrodes at points where they might particularly enhance the observation of one or another type of activity (Gibbs and Gibbs, 1984). As awareness of pattern distribution improved and as new patterns were discovered, increased attempts were made to separate and discriminate between patterns and find electrode placements that would be most advantageous for demonstration of a particular pattern, such as sleep spindles, or demonstrate alpha activity not contaminated by the vertex activity. The observation that different types of activity occurred simultaneously encouraged the use of more electrodes for more channels of simultaneous recording. This was followed by attempts to place electrodes in a standardized manner so that a patient’s record could be compared over time and different patients could be compared to each other. Initially, there was wide diversity from place to place in established methods and standard placements.
A committee of the International Federation of Societies for Electroencephalography and Clinical Neurophysiology recommended a specific system of electrode placement for use in all laboratories under standard conditions (Fig. 7.1) (Jasper, 1958). Their recommendation was the system now known as the International 10–20 system. Specific measurements from bony landmarks are used to determine the placement of electrodes. Many of the systems had done this earlier, but they generally used a specific standard interelectrode distance on every patient. The breakdown of such a system is apparent if the application of electrodes to a microcephalic patient is compared to application to a hydrocephalic patient using the same number of centimeters from landmark locations or between electrodes.
The International 10–20 system attempted to avoid both “eyeball” placement and unvarying distances by using specific anatomic landmarks from which the measurements
P.140

would be made and then using 10% or 20% of that specified distance as the electrode interval. Such electrode placement can be replicated consistently over time and can be replicated between laboratories. Although the measured distances change as the young individual grows, the basic placement remains much more consistent than is sometimes achieved without the use of measurements (i.e., on a properly measured head, a deviation of 1 cm is not necessary and should be corrected if it occurs). The consistency and replication over time may prove to have significant value, particularly in growing children and in longitudinal studies of evoked potential measurements. It is expected that this is more likely to be seen in the longer latency-evoked potential, in which near-field characteristics make placement more critical than in short latency far-field records.
Figure 7.1. International 10–20 system placement and letter-number designation. Odd numbers on the left, even on the right, and Z, or zero, in the midline.
The standard placement recommended by the American EEG Society for use in the International 10–20 system is for 21 electrodes. The system is designed to allow the use of additional electrodes with predictable and easily repeatable placement. The standard numbering system in the 10–20 system places odd-numbered electrodes on the left and even-numbered electrodes on the right, with the letter designating the anatomic area. As can be seen in Fig. 7.1, numbers are deliberately skipped so that the midline or zero electrodes are flanked by electrodes numbered 3 on the left and 4 on the right, allowing insertion of an electrode in the same anatomic designation numbered 1 or 2 if narrower spacing is wanted. Similarly, an electrode can be placed between the left frontal electrode (F3) and the left anterior temporal electrode (F7), which would be designated F5. As the diagram illustrates, the F7 electrode is really over the frontal lobe in a posterior and inferior position, and it is just anterior to the temporal pole. It is clear that it is a sensitive electrode for temporal lobe activity. It has been suggested repeatedly that a still closer representation of temporal activity is desirable or at least might demonstrate maximum amplitude at points that are not currently positions for standard 10–20 system electrodes (Binnie et al., 1982). Greater specificity of temporal pole activity can be obtained with the use of specialized electrodes, but they are generally not necessary in standard recordings (see Chapter 36).
The American EEG Society, in 1991, added electrode placement nomenclature guidelines that designate specific locations and identification of 75 electrode positions along five anterior posterior planes lateral to the midline chain of 11 specific sites (Fig. 7.2). Similarly four coronal chains are anterior and four posterior to the chain of 13 electrode sites identifiable between the earlobe electrodes upthrough the midline C electrode (American EEG Society, 1994a,b,c,d,e).
The avoidance of a lockstep measurement system has been strongly advocated as one of the major advantages of the method now described as the anatomical placement of EEG electrodes (APEEGE) (Gibbs and Gibbs, 1984). Visual observation and application of electrodes without measurement, especially with distorted skulls, leads to individual decision making. These decisions may vary from test to test and observer to observer, so that one is less confident about whether a change in the EEG over time is due to a simple difference in electrode placement or to a change in the location of the activity being measured. Similarly, rules that require that an amplitude difference must be greater than 50% to be significant may be related more to common variation in electrode distances than to amplitude variation as an actual commonplace physiological feature.
The argument about the number of electrodes needed is based on many factors, including the montages used, the population routinely studied, the type of activity considered
P.141

important, the number of channels available, and the skills of the reader. Some systems are clearly more useful if one expects less skill and training for the technologist attempting to carry out the electrode application swiftly. Some of these combinations also allow more consistency throughout a recording, so the reader has a simpler task. Sometimes fewer electrodes are applied, as in infants, because it is believed that electrical activity is not recorded well from their heads because of small electrode distances; in fact, the voltage recorded on the heads of young children is not lower than in adults and in many cases is much higher. It is for this reason that the American EEG Society (1994b) guideline 2 noted that frequently a reduced sensitivity is needed to record from children. Similarly, the argument that fewer electrodes must be used in children because there is not room for the full 21 electrodes of the 10–20 system is spurious. As can be seen in Fig. 7.3, if collodion attachment of electrodes is used so that there is no smearing of the paste, the full 10–20 system can be applied. It provides very good reliability in heads with a circumference of as little as 23 cm. There is room for full electrode use on even smaller heads.
Figure 7.2. Modified combinatorial nomenclature. (Copyright 1990 American Electroencephalographic Society, with permission.)
Some deviation from the International 10–20 system is seen in the research setting (Barrett et al., 1976; Halliday, 1978). Although the advantages sometimes apply to the test in question, issues of economy of time and lack of training of the technician are more often a factor than one would wish in attempting to find a common consistent method of applying electrodes.
Studies with magnetic resonance imaging (MRI) and computed tomography (CT) have demonstrated that the anatomical correlates of the scalp electrodes are imperfect (Myslobodsky et al., 1990) and the distortion may be more evident and lateralized in some sites such as the occipital region (Myslobodsky and Bar-Ziv, 1989).
Montage and Reasonable Use of Channels
The issue of montage and channels in a standard record is important because it has great bearing on the style of recording and the number of electrodes that are used. The term montage refers to the particular combination of electrodes examined at a particular point in time (Chatrian et al., 1974; Harner, 1977).
The montage is a tool used for a specific purpose; in most instances, multiple montages are more useful than single montages for long periods. It is probably useful to change the montage every 2 to 3 minutes unless there is clear indication to stay with the ongoing montage, e.g., when a procedure is expected to produce a specific change. The variation from baseline is important, and montage change is self-defeating for technologist and reader in such a case. In hyperventilation, the montage should be maintained for at least 1 minute before the procedure and at least 2 minutes after its completion for comparison. Another instance when montage variation should be done cautiously is in the midst of a seizure. Generally, here again one wants a consistent montage in order to follow the evolution of the seizure. For someone in seizure status who has continued to show the same activity for 5 or 10 minutes, it may become useful to vary the montage to get information about what is going on in other areas or from a different perspective. Endarterectomy, temperature, and anesthesia changes lend themselves to the single montage recording (Barlogie et al., 1979; Kaufman et al., 1977; Presbitero et al., 1980; Reilly et al., 1974, 1980).
Technologists and electroencephalographers use montages as one of the tools by which they record EEG activity. The montage is meant to carry out two simultaneous functions. One is to record from all areas of the scalp. The standard
P.142

montages and use of all the International 10–20 system electrodes accomplishes that goal. The second function is to record activity in such a manner that it is easily perceived by the reader. To some extent the various montages are used in an attempt to display more clearly activity that may, in fact, be present in several montages. The activity may be there but presented in a way not easily seen, unless some visual tricks are used to make it more evident for the reader. EEG, stored in memory and manipulated by computer at a later date, can be used to vary montages in different ways for a specific segment of data, but it is unlikely that the need for technician skill in changing for “cause” will decline.
Figure 7.3. Collodion attachment of silver-disk electrodes (0.8 cm in diameter) on a premature infant with a head 23 cm in circumference.
There has long been debate within the EEG community as to the particular advantages or disadvantages of reference recording versus the scalp-to-scalp bipolar linkage. That debate, in turn, hinges on the quality of the technique involved in placing electrodes and on the type of abnormality being sought.
If one is mainly interested in a striking pattern, such as a spike-and-wave complex, then the display of such a pattern in a way that makes it stand out at a fairly large amplitude is desirable. For such a pattern, which is generally predominantly parasagittal, the use of the earlobe, whether ipsilateral or joined, is generally an excellent montage. If, on the other hand, the abnormality looked for is a fairly low-voltage limited pattern, perhaps involving only the midtemporal electrode and the reference electrode itself, there is great danger that the pattern can cancel out between the involved electrodes and not show up well. The use of reference electrodes also can lead to some rather complex situations in which there appears to be out-of-phase waveforms occurring both anteriorally and posteriorally (Garvin and Gibbs, 1971). In at least some instances, this phenomenon does not represent a physiological reversal of activity but the contaminated reference electrode effect on uninvolved electrodes (Reilly and Seward, 1980).
Reference Recording and Its Variants
The montages generally used in EEG are divided into two major categories. In one style of recording, the electrodes in their various placements over the scalp are all referred to one single electrode, one common electrode on each side of the head, or the electrically combined activity from two or more electrodes (Fig. 7.4). When all electrodes are referred to a common electrode that, it is hoped, had little brain activity in it, it was convenient to call it “monopolar,” but it is far better to recognize that such montages are reference and that it is virtually impossible to make either one of the electrodes in a pair so consistently neutral that the recording is indeed occurring in a monopolar fashion to a totally inactive reference electrode.
The value of a truly inactive electrode has always been evident, but the need is even greater with increased use of computed electrophysiological recording where subsequent display and analysis are quite vulnerable to reference influence.
A reference montage can be set up in two general ways. A common design for reference recording is to have alternating left and right electrodes so that comparable electrodes from each side are recorded next to each other. This sort of combination is particularly valuable when the major part of the abnormality is an asymmetry. Placing comparable electrodes next to each other allows for the observation and detection of subtle differences between the two sides. This is particularly useful if the activity observed might be acceptable if it were symmetrical in amplitude. Often waves are only abnormal because of an asymmetry. This situation is extremely common with activity observed in the temporal region, where one side demonstrates higher voltage and more persistent theta or delta activity than the other side. An asymmetry in the temporal region is frequently seen in drowsiness and light sleep at the time when there is a normally seen increase in amounts of theta and delta activity bilaterally. The asymmetry allows the reader to recognize existing abnormality, although some temporal differences have been reported in a number of normal or apparently normal individuals (Busse and Obrist, 1965; Drachman and Hughes, 1971; Kooi et al., 1964; Obrist and Busse, 1965; Obrist and Henry, 1958; Silverman et al., 1955; Yoshii, 1971). Interpretation of asymmetries will vary with the
P.143

reader, but the visualization of the asymmetry, if it is present, depends on appropriate montages recorded at appropriate times. An asymmetry enhanced by drowsiness and maximal in the temporal regions may not be evident in longitudinal bipolar or ear reference records because of cancellation. Such an asymmetry is seen easily with coronal montages or reference montages using the vertex electrode (Morgan, 1968). These asymmetries will not be seen if the record is turned off while waiting for sleep or if the vertex electrode is considered contraindicated in sleep because of the vertex sharp transients. The latter is a striking large wave but not really an interpretive problem.
Figure 7.4. A: A representation of bipolar (scalp-to-scalp) montage in a longitudinal or anteroposterior (AP) direction. This montage is designed for easy comparison of left-right differences in the parasagittal or temporal area, but a montage with strict left-to-right sequence of sets of four could provide the same information. B: This is a typical bipolar (scalp-to-scalp) montage in the transverse or coronal direction. C: This is a reference (monopolar) montage. This particular sequence is designed to allow a front-to-back sequencing of channels to provide anatomic continuity.
A sequential reference montage can be used to allow interhemisphere comparison if this is desirable. When it is useful, a montage might involve the left temporal electrodes with reference to the right ear, followed by the right temporal electrodes referenced to the left ear. Comparable electrodes from the two sides are further apart on the page, making comparison of very subtle differences more difficult, but the montage allows left-to-right comparison at any point in time at which the possible abnormalities appear. Such a “hemisphere sequence” is generally more valuable with bipolar montages than with reference, but it has its value with both types of recording.
One of the earliest and one of the staunchest advocates of monopolar recording was F. A. Gibbs (Gibbs and Gibbs, 1952, 1964). Some of the arguments for different electrode placement and the use of monopolar montages have been reviewed (Gibbs and Gibbs, 1984). In his various writings, Gibbs points out that many patterns are well delineated and were first seen on monopolar montages. These include sleep patterns, the 14/6 positive spike, and many of the specific seizure patterns. In contrast, Gibbs believes that bipolar montages number among their disadvantages the tendencies to (a) reduce voltages sometimes to the vanishing point and create complex waveforms, (b) make it difficult to recognize a “true” electrical sign, and (c) create out-of-phase relations that are hard to distinguish from true physiological out-of-phase voltages (Gibbs and Gibbs, 1964). False phase reversals are also created by involvement of the reference electrode and may be equally deceptive and require confirmation or elaboration by the use of bipolar techniques (Reilly and Seward, 1980).
One of the arguments for the theory that the earlobe electrode, or the two shorted together, makes a neutral reference is that it is, at times, relatively devoid of cortical activity. Unfortunately, one of the major sources of dysfunction and intermittent waves of high voltage is the temporal region, from which such activity is easily reflected in the earlobe electrode. In epileptics, temporal lobe abnormality may be present in as many as 70% of the patients; thus one or both earlobes can be “contaminated” in patients in whom the documentation and understanding of intermittent activity is quite important.
In an attempt to avoid the use of the earlobe electrodes in such patients, a balanced noncephalic electrode was developed by putting one electrode over the right sternoclavicular junction and another on the spine of the first thoracic vertebra. These electrodes are connected together through a 20,000-ohm variable resistor that can be changed in its resistance until the electrocardiogram is virtually canceled (Stephenson and Gibbs, 1949).
In a similar manner, many or all of the scalp electrodes were connected to a common reference lead of 1.5 megohm resistance (Goldman, 1949). Here again, the attempt is being made to average out the individual deviate voltages that may show up in only one or two electrodes. The average electrode concept works well unless there is some transient voltage that is either very high or involves a great many electrodes. In such a case, the transient voltage shows up with reverse polarity and much attenuated amplitude in a manner that may make it difficult to recognize. Of particular note is an eye blink artifact that very often gives a very high-voltage deflection in several electrodes. However, a transient, high-voltage abnormal waveform showing up in several electrodes simultaneously might be displaced in its apparent location and quite distorted in its actual amplitude. Various techniques to use electronics or computers to reconstitute signals in a more accurate way have been attempted (Hjorth, 1975, 1976, 1979; Ishiyama et al., 1982; Sorel and Ranwez, 1984).
Two of these methods are attempts to derive a more truly monopolar derivation (Ishiyama et al., 1982; Sorel and Ranwez, 1984). The latter article is also a good critique of some of the problems associated with bipolar recordings. Neoelectroencephalography (NEEG) is a recording with a true monopolar derivation that is derived by computer-reconstituted signals that use a reference electrode obtained from two electrodes on either side of the neck; the activity is subsequently subjected to Fourier analysis with continuous comparison of phase to determine whether or not the reference electrode is active. When it is determined that the reference electrode is active, it is automatically subtracted from the value of all other reference channels.
Source Derivation
From the early days of EEG, it has been recognized that the various combinations of electrodes represent an interplay of activity from surrounding areas and are particularly influenced by the two or more electrodes connected at any point in time. As noted above, a number of attempts have been made to reduce or eliminate the activity from one or two points so that the remaining observed discharge would become a truer reflection of the electrical activity at that particular point. A method of particular interest has been described by Hjorth (1975, 1976, 1979) with the intent of improving localization of focal activity from the scalp. This technique, which has been referred to as source derivation, has been examined not only for its potential use in EEG interpretation but for the examination of evoked potentials as well (Thickbroom et al., 1984). The technique replaces the visual estimation made by reviewing the various patterns from various combinations of electrodes with an on-line process designed to derive the source activity of each particular electrode (Hjorth, 1975). The method appears to enhance the activity that is unique to a particular electrode. It raises the possibility that one could make interpretations strictly using a single montage and review of the isolated activity from the various points involved in the EEG, rather than reviewing the various
P.144

combinations, in order to make determinations about the origin of activity. In some particular clinical settings, such focal activity can be enhanced and made apparent when not seen with either bipolar or reference techniques, as has been demonstrated (Hjorth, 1976). Reviews of source derivation in clinical EEG in comparison to standard visual techniques indicated that abnormality detected by the source method but not conventional derivations actually occurred in only a small number of instances. The focal abnormalities, however, were often more pronounced with this technique (Wallin and Stalberg, 1980). There are still questions about the possibility of distortion with this technique, as with others (Gibbs and Gibbs, 1984). Subsequently, use of these techniques to look at visually evoked potentials also suggested that visually evoked potentials showed less spread to adjacent activities when examined by source derivation as well. Thus, the technique may have some localizing value for evoked potentials (Clement et al., 1985).
Nonreference Recording
In a further attempt to eliminate the influence of reference electrodes on the analyses of electrophysiological activity Lehmann et al. (1986) used vector diagram best fits to replicate an analysis without reference electrode or channel.
Some suggest that the closest technical fit to this may be the use of tied electrodes at the muscle free end of the mandible on each side rather than at the earlobe or mastoids (R. N. Harner, personal communication).
Bipolar Variants
In attempts to examine closely the symmetrical areas of the brain and to compare one side of the head with the other, there has been a tendency to use scalp-to-scalp linkages that are particularly useful in demonstrating a marked change in polarity in the so-called phase reversal. The latter phenomenon is greatly overrated and sometimes has been assumed to be evidence of an abnormality in itself. In fact, it is nothing but a tool to show the point of maximum deflection of one polarity or the other. Using bipolar scalp-to-scalp electrodes, it may be demonstrated in two electrode pairs over a common electrode (Fig. 7.5) or it may be seen only over three electrodes, with a middle pair of electrodes in fact canceling out and the phase reversal only being seen if one looks at the electrode pair on either side of the two equally involved electrodes (Fig. 7.6). It is this latter phenomenon that causes some individuals so much difficulty with bipolar montages. If the reader is looking only for patterns (which are relatively rare) or an event in a single electrode, then this tendency for the focus to become low voltage and minimized in the most involved electrodes with bipolar techniques can be a source of confusion. If the phenomenon is recognized, looked for, and learned, it is a useful observation, but it requires the reader to use montages that put broader areas of the hemisphere close enough together to make the observation. Thus, when using bipolar linkage it is generally more desirable to place the channels for one hemisphere together on some part of a page where they can be compared to a similar sequence from the other hemisphere. Obviously, phase reversals and other phenomena can be seen if the left-right sequence of channels is being used, but it involves mastering a much larger part of the page while simultaneously separating the channels mentally into left and right side. For most readers, this is harder than looking at a hemisphere examined in toto and compared to the other hemisphere examined in toto.
Figure 7.5. Classic surface negative phase reversal over the C3 electrode which is 70 μV negative to both the F3 and the P3 electrode.
Figure 7.6. This is a classic phase reversal with equal surface negative activity in the C3 and P3 electrodes so that no deflection is seen in that channel, and the major deflections are seen between the uninvolved F3 electrode and the heavily involved central electrode and between the heavily involved parietal electrode and the uninvolved occipital electrode.
The subject of montages over the years has shown a great deal of variability and a great many different strategies. Some assume that, once one has observed the activity from particular electrodes, one can extrapolate and deduce what any combination of electrodes would show. It certainly is the experience of most electroencephalographers that different montages may in fact make particular kinds of abnormalities more evident or more visible. An abnormality is not seen, perhaps, in other montages until a particular combination makes the activity optically more unmistakable, and in retrospect the existence of the abnormality in the montages where it was subtle and initially missed can be confirmed.
It is sometimes suggested that the montages must be presented in an unchanging consistent sequence or the reader will be slowed down or confused. On the contrary, it is probably more likely that activity will not be missed if the technologist is encouraged to change the sequence of montages or devise montages to enhance and more clearly demonstrate what the technologist sees in the record up to that time.
P.145

Montage selection is improved ifthe technologist has a good idea of the extent of the abnormal activity in question. This is the distribution or the field of abnormal activity. One attempts to find montages that have some of the electrodes in the field and some outside of the field. The importance of doing this relates to the way EEG amplifiers work. The amplifier does not measure the absolute electrical voltage under any one electrode but, rather, measures the difference in the electrical voltage from two electrodes connected to a particular amplifier. If two electrodes are demonstrating the same 100-μV delta wave and there is no activity in either electrode that is different from the other, connection of the two electrodes results in a flat line. In this instance, the flat line does not represent inactivity but shows that the activity is of equal potential (equipotential) in both electrodes (Chatrian et at., 1974). The term isopotential has been applied to this circumstance, but it implies to some the idea of no voltage. This is generally an error.
There is a general rule of thumb that higher voltages are seen as interelectrode distances increase. This is best seen as one records a bipolar montage and switches to a reference montage. The voltages appear to be much higher in the reference montage, and often it is necessary to decrease the sensitivity of the machine. It may be necessary to go back to higher amplification when returning to bipolar montages. It is clear that the voltage of brain activity is not changing.
This is a function of the montage change. The display of different voltages between the two montages is related to the fact that the bipolar anteroposterior (AP) chains of electrodes are close to each other, and the activity is similar in these relatively closely spaced electrodes. In most instances, there is a far greater difference between the voltage in the parasagittal electrodes and that from the earlobes. The latter may be inactive if truly neutral, but there is at least activity of a different voltage and frequency, which results in a higher voltage difference with the reference than the bipolar montages.
One may be attempting to work out the distribution of activity that is clearly abnormal. In this instance, a reference montage with electrodes from one hemisphere in sequence may be useful if the reference electrode is selected as one clearly out of the distribution of the abnormality (Binnie and Lloyd, 1970; Yoshii et al., 1966). This type of sequence is particularly useful with individuals who have focal epileptiform activity or those with a recognizable area of focal delta activity. This is easier to interpret if superimposed on a distinctly different background with a different frequency and voltage, so that the abnormal wave stands out and is recognizable simply when it appears and not just when it is compared to the other hemisphere.
The American EEG Society (1994d) has proposed that standard montages be used, not exclusively, but at least as part of every record. The recommendations are also based on the premise that the other minimal technical standards of EEG are being carried out. The montage recommendations are designed around the idea that at least eight channels are being used, and 16 or more channels are encouraged. It is assumed that the full 21 electrodes of the 10–20 system are being used.
The recommendations suggest that both bipolar and referential montages be used in a record. It is recommended that the electrode connections for each channel be indicated at the beginning of each montage, rather than using a cryptic number or letter as the only montage mark or attaching a copy of the montages on the front of a record. It is suggested that the bipolar connections should run in straight unbroken line with interelectrode distances equal. It is also recommended that over the course of a record, bipolar recordings be run both in the longitudinal direction (Fig. 7.4A) and the transverse direction (Fig. 7.4B). Referential recording should be done as well (Fig. 7.4C).
In accomplishing this, the guidelines suggest that when 16-channel or larger machines are used, at least one montage from each of the three classes will be needed. For 8- or 10-channel recordings, at least seven montages will be needed from the standardized list to meet the same requirements. Although it is still the belief in many laboratories that referential recordings alone or bipolar recordings alone are the only ones necessary, perhaps increased experience in laboratories using both will demonstrate to readers that in many instances it takes the combination to make clear which of at least two possibilities is correct when there is a particular asymmetry or unusual activity recorded.
Electrode Impedance
In the initial years of checking electrode attachments, it was common to measure resistance rather than impedance. Even now, when checking individual electrodes that are not attached to the patient’s head, it is still sometimes simplest to take two electrodes, measure the resistance, and determine whether the resistance remains low when the two electrodes are slightly stretched. This has the advantage of picking up the “make-break” problem that is sometimes seen in an electrode that is conducting most of the time but is showing little intermittent breaks in conduction as the cable is moved. This is usually an electrode in which the inner wire is snapped but is not yet totally pulled apart. The stretching helps accentuate the break. Obviously, if this is done with excessive vigor, it may shorten the life of the electrode, which is costly and undesirable.
Once the electrode is attached, there can be discomfort and hazards in measuring electrode resistance (Seaba et al., 1973). As a result, it is useful to measure impedance rather than resistance once electrodes are attached to the patient (Seaba, 1985). To greatly minimize the hazards of measuring resistance, two major concerns can be cited: (a) the passage of current in the test may actually polarize or charge one of the electrodes, and the discharge of that electrode’s potential buildup may show up in the record at some later point, looking like a plausible event from a focal source; (b) the other problem in measuring resistance through electrodes on the patient’s head is that, if it is done with standard resistance meters rather than those specifically designed for patient use, it is likely that the current used is going to be felt by the patient with a sensation ranging from mild discomfort to clear pain.
Although resistance measurement depends predominantly on the material used for the electrode and on the ability of current to pass through the skin, the latter depends quite directly on the preparation of the skin before the electrode is attached.
P.146

In contrast, impedance involves characteristics of resistance capacitance and inductance. As with resistance, electrode impedance is lower if the skin has been prepared by cleansing or rubbing to remove surface oil and superficial layers of the epidermis. Over the years, a wide array of materials, including acetone, alcohol, and even soapy water, have been used for this purpose. Now it is common to use materials that are themselves conducting electrolytes. The rubbing of the cleansing material requires some firmness, and one must be familiar with the material being rubbed in. A few of the currently available cleansing electrolyte materials have some abrasive substances embedded in the material. If such a cleanser with an abrasive is rubbed with the same enthusiasm and vigor as a cream without an abrasive, it is possible that there may actually be a break in the skin with some blood or later scab formation. This is undesirable. Often a small piece of gauze rubbed against the skin both carries the electrolyte material and has ridges sufficient to help remove the surface materials of the skin. The effect of the gauze can be enhanced by wrapping the piece of gauze around an eraser on the end of a pencil. The gauze can be held in place with a rubber band. The give under pressure on the soft eraser makes it less likely to injure the skin. Holding the pencil gives the technologist good control and a greater amount of pressure against the skin. A similar combination of both firmness and softness can be obtained with a cotton-tipped applicator. It remains imperative to look for the possibility of lack of similarity between impedance in different channels or changes in impedance over the course of the record. Such changes can result in record artifacts (Gordon, 1980). If skin preparation and electrode selection is done well, the impedance can be reduced below 3,000 ohm, and quite often 1,000-ohm impedance is reached without difficulty. An impedance over 5,000 ohm in any electrode should not be accepted as adequate (American EEG Society, 1994a; American Medical EEG Association, 1988).
It is important that impedance actually be measured. It cannot be assumed that routine attempts to reduce impedance and to apply the electrodes with reasonably consistent application will ensure low impedance. Failure to check impedance routinely leads to a variety of artifacts that can be interpreted as brain activity.
Even though the electrode has proper impedance at the beginning of the record, it can deteriorate during the recording. This point is given less recognition than is desirable. Technologists need to be encouraged to recheck impedance at regular intervals. Many machines have excellent methods of checking impedance in one channel or in all channels simultaneously. The output varies depending on the machine. Some instruments measure the difference in impedance between electrodes involved (lead imbalance); other machines measure total impedance of the two electrodes involved. The latter is preferable. It is essential that technologists and readers understand the method and implications of the method used for a particular machine. Whether the displayed output is the sum of the impedance of the involved electrodes or the difference between the impedance of each electrode must be understood by all those who work with the machine.
One often-neglected indicator of rising impedance is the development of a 60-cycle artifact from a particular electrode. This artifact is considered an unwanted interference in the recording, but there is a tendency for electrodes with rising impedance to show this and other artifacts. Such a development tells the technologist to test and fix an electrode before great difficulty develops. It can be of value if this is understood and if routine use of the 60-cycle notch filter is avoided.
Electrode Application
Once skin impedance has been appropriately lowered, the electrode can be attached to the skin in many ways. Attempts have been made to develop self-adhering or self-attaching electrodes. The needle electrode lies under the dermis and generally stays in place without additional means of holding it. It may fall out from its own weight, and it certainly falls out if the head is moved abruptly. Spring-loaded electrodes with flat surfaces or teeth are painful and not tolerated over the scalp, but a spring-loaded, flat, clip-type electrode on the earlobe can be used as the reference electrode. Most commonly, the conducting metal electrode is attached to the skin surface of the scalp, rather than embedded in the skin. Some relatively thick pastes with conducting properties are commercially available. A common method of attaching electrodes is to take an electrode and scoop up a ball of electrolyte paste; the electrode and paste are then firmly pressed onto the skin. The pressing of the electrode pushes the paste and electrode against the scalp. The paste folds around the edge of the electrode, and the attachment of the paste to the electrode is enhanced by pressing a cotton ball or a piece of gauze on top of the electrode and paste combination. This type of electrode is relatively easy to prepare, and good impedance recording can be obtained with it. One hazard of the method is spreading of the paste that makes electrode contact. The recording site is the size of the ball of paste on the scalp. If an unduly large amount of paste is used, the recording is from a much larger area than with the simple metal electrode. Masses of paste can spread as the head is moved and may touch each other and short out between electrodes. Also, the paste tends to dry. Contact becomes variable as the electrode becomes less consistently adherent. Such drying can occasionally be seen in less than 20 minutes but may not be seen in recordings extending well beyond 45 minutes. The smaller the amount of paste, the faster the drying, but large paste quantities increase the other problems described above.
To provide a more secure method of electrode attachment, collodion is applied to actually glue electrodes to the head. This can be done with a small ring of collodion at the edge of the electrode at the skin-electrode junction. More commonly, the collodion is used to soak a piece of gauze, which is then spread over the electrode and skin (Fig. 7.7A,B). This provides a large area with glue. Compressed air is used to speed the drying (Fig. 7.7C). It is critical for laboratories using this method to be aware that collodion, not flexible collodion, must be used. The latter is used more commonly in hospitals for dressings. In that situation, it is
P.147

not necessary to have the quick-drying characteristic needed for EEG application. The oil in flexible collodion makes adequate drying impossible. The electrolyte under the electrode can be scooped into the electrode before the electrode is glued in place. A more reliable method is to use electrodes that contain a hole (Fig. 7.8) so that the electrolyte can be added after the electrode is glued tightly on the head (Fig. 7.7D). Collodion-attached electrodes with a hole in the surface have the added advantage that they can be left in place for periods of up to 5 days without affecting the skin. Electrolyte can be added for each recording, and this allows rerecording to be carried out with greater ease and greater speed than if the electrodes are reapplied at each test. The electrolyte must not contain calcium if long-term contact is expected without skin effects. The collodion-attached electrodes as described here are probably the best for use in any short- or long-term recording session, whether those sessions take place over minutes or days. They are generally the best for surgical situations. This type of attachment appears to be the most reliable over hours, and it is generally much more feasible to gain access to an electrode and add more electrolyte solution through the hole in the electrode than it is to totally remove and reglue an electrode during surgery.
Figure 7.7. A: The electrode is placed in the position on the head, which has been previously measured and has previously had the impedance lowered. B: A piece of gauze soaked with collodion is placed over the top of the electrode. C: Compressed air is sprayed on the collodion-soaked piece of gauze as the gauze is smoothed against the head. This causes rapid drying of the glue against the skin. A point in the compressed air applicator positions the air in the center of the electrode and helps hold the electrode against the skin. D: A blunt-tipped needle that fits the hole in the electrode relatively tightly is passed through the hole in the gauze into the electrode and an electrolyte solution is added to fill the cup. This blunt metal tip can be used to scratch the skin and improve impedance if required.
Figure 7.8. The electrode has a hole so electrode paste can be added and skin can be reached to reduce impedance. The size of this 0.8-cm electrode is compared to the United States penny.
P.148

Calibration
The calibration signal simultaneously presents several bits of information. The calibration signal provides an assessment of the sensitivity of individual amplifiers. To be sure the machine is accurate, the technologist must be aware of the signal size expected, must provoke a signal of adequate size to measure, and must be able to measure the signal obtained (Fig. 7.9). In the illustration, three deflections have been obtained. One is a deflection in excess of the 12 mm from the baseline considered accurate and reliable on the particular machine used. The top calibration signal shows a distinct overshoot, which would not be included in the measurement even if this deflection were obtained within the adequate range. The deflection excluding the pen overshoot should be measured; this should be the appropriate deflection for input. If not, the amplifier should be adjusted. The lowest of the three calibrations in the illustration is so small that accurate measurement is difficult, if not impossible; it is best to try to arrive at calibration signals with deflections somewhere between 5 and 12 mm on most machines. With machines with narrow pen spacing, actual deflections cannot exceed 10 mm without exceeding the accurate arch of the pen. In Fig. 7.10, a page of calibrations is shown with the high and low linear filter settings marked on the sides. Only one or two square wave pulsations provoking the classically seen upward and then downward deflection of the alternating current (AC) are required. The tendency to recalibrate across the page for 10 or 20 seconds is unnecessary. The major use of the calibration signal is to provide a sample of the deflection, the rise time, the decay as modified by the low linear frequency setting or time constant, and a display of the sharpness of the square wave for some suggestion of the effect of the high-frequency filters.
The most critical part of calibration usage is that the deflections be carefully observed and measured in all channels at least once a day and preferably before the start of each record. The next most critical use of the calibration signal is as an assessment of pen alignment and time axis. Both of these are best seen if the calibration signal can be lined up with one of the time axis lines down the paper, as is seen in the initial upward deflection in Fig. 7.10. If a single pen is leading and pens both higher and lower on the paper than that particular pen are on the time axis line, the pen is misaligned and needs to be adjusted. If pen alignment has been adequately carried out, it is often found that the pens at the top of the page may be on the line; however, there is progressive deviation of the pens from the time axis line as one moves down the paper. The pens may lead or follow the line by the time the last channel is reached. This suggests that the paper is going through the machine crooked compared to the actual pen alignment. Readjustment of paper is sometimes difficult, but both pen and paper alignment are very important if the reader is to make an accurate assessment of onset of closely spaced events.
Figure 7.9. Calibration deflections in response to three signals but with a constant sensitivity. Only the center calibration has a deflection in a range where the pen is not exceeding its limits but is large enough for easy and accurate measurement.
Two other types of calibration should be carried out. A low-voltage calibration such as a 5-μV input with a sensitivity of 7 or 7.5 μV/mm produces a very small deflection, less than 1 mm. The technologist and reader examine this to see if the onset of the wave is rounding or if there is absence of the deflection in one channel. The deflections are not measured, just examined. This is a very sensitive indicator when pen galvanometer difficulty begins. The ball bearings of the pen motor start to flatten, and the small signal is insufficient to make these flattened bearings move. The larger signal is sufficient to overcome that initial friction, and the large signal may show no apparent reduction in the deflection. The record will show loss of low-voltage signals, such as superimposed
P.149

beta activity. Regular use of this small-voltage calibration signal is desirable to detect this potential problem at an early stage of development.
Figure 7.10. A series of deflections using a variety of low linear frequency settings (20% attenuation at the designated number) and a variety of high linear frequency settings (20% attenuation at the high-frequency number). The time axis marked on the paper lines up with the initial upward deflection.
A “biological calibration” is essential. The “machine calibration” signal is internal to the machine; it gives no indication of the actual connection of the electrodes and does not actually involve the patient. The initial calibration signal tests only the machine, the amplifier, and the pen alignment. An additional calibration with the same two scalp-attached electrodes in all channels simultaneously demonstrates that all the amplifiers respond equally and correctly to a variety of frequencies and not just to a direct current (DC) signal. This tests the electrode cable and the patient’s cerebral activity as well.
At the end of a recording, it is desirable to calibrate again, using all of the sensitivities and all of the filter settings used throughout the recording (American EEG Society, 1994a). In instances in which records of electrocerebral inactivity have been carried out, it is necessary to calibrate using a 2-μV calibration signal with the most sensitive amplification used in the recording. This is a demonstration of the deflection that would occur if this minimal level of EEG activity were recorded. The suggestion of using a 2-μV calibration signal is included in the minimum technical standards for EEG recordings of suspected cerebral death (American EEG Society, 1994c).
Filters
The use or misuse of filters provides one of the major sources of contention among electroencephalographers and technologists. The range of neurophysiological activity extends as high as 2,000 cycles/sec, as in the cerebellum. Slower frequencies of cortical origin extend down to one third of a cycle per second and probably to one quarter of a cycle per second. Still slower frequencies arise from the brain and are recorded under special circumstances, such as studies of the contingent negative variation (CNV) or studies of negative DC shifts prior to seizures. Slow brain frequencies are difficult to differentiate from the very slow potential changes that occur on the basis of sweat artifact or due to change in the galvanic skin response (GSR). Fortunately, study of evoked potentials brought the attention of neurophysiologists back to portions of the frequency band that had been ignored for a period of time. Some will argue for measuring the full range of provoked activity.
In general, one can assume that the greater the recorded frequency band, the greater the fidelity of reproduction of the actual activity. In theory, this is true, but recording a larger frequency band increases the amount of outside interference and noise in the signal. Filters are used to make a compromise between reduction of extraneous signals and preservation of as much as possible of the fidelity of the brain waves one particularly desires to observe. If the frequency band recorded for EEG is compared to that recorded for evoked potentials, a striking discrepancy in bands of value is apparent. In EEG there is little of value in the signal over 50 cycles/sec, whereas in recording short latency far-field evoked potentials, it is common to filter out frequencies under 100 or 150 cycles/sec but to continue recording up to 3,000 or possibly 30,000 cycles/sec (Calloway et al., 1978; Cracco and Cracco, 1976; Stockard et al., 1979). One may insist on recording slow activity down to cycles/sec (direct current) but define it as acceptable to eliminate faster activity. In contrast, one may need to see all the fast activity but not care much about slower frequencies. In each case, one defines a frequency band. The component in equipment that eliminates unwanted frequencies and defines such limits is the filter. A filter range is the frequency that will appear without significant distortion at some determined level of accuracy. The question of significant distortion is answered variably for different kinds of studies.
For EEG activity, it is accepted that a frequency displayed at 70% or more of its actual voltage is acceptable. If the loss or attenuation to a particular frequency by the filter reduces the voltage more than 30% of its real value, the distortion is significant and puts the activity outside of the frequency band. This is an arbitrary number found acceptable by the EEG field and the various societies involved. Certainly, there is no reason why the societies cannot decide that 40% attenuation is acceptable or that 10% attenuation is too great. Electrocardiogram (ECG) equipment is allowed attenuation of only 10% or less within the defined frequency spectrum. Each industry or type of activity defines acceptable amounts of distortion at the limit of defined frequency bands.
The frequency band investigated in most detail by clinical neurophysiologists doing EEG has included frequencies under 50 cycles/sec. This limitation of the frequency leaves one with activity normally having the highest amplitude. Such frequencies were easier to record when equipment was less sensitive. The other reason for attention to slower frequencies (less than 100 cycles/sec) relates to limitation not of amplifiers, but of pens. Pens provide poor reproduction of activity over 100 cycles/sec. With the conventional pen and ink writing machine, there is little point in having an amplifier attempt to display activity faster than the pen will record.
One must consider what high-frequency and low-frequency activity is required in a particular record. A few of the most limiting filters overlap, but as a general rule filters used to influence low-frequency activity have no effect on the high-frequency waves. Those affecting the high-frequency activity do not do anything one way or the other to low-frequency waves. This is in contrast to manipulation of sensitivity or gain, which affects all frequencies recorded.
High-Frequency Filters
The filter that affects the high-frequency activity is described by several terms, including “high-frequency filter.” Generally, a particular number is specified, such as high-frequency filter 70 or high-frequency filter 35. The designated number indicates which particular frequency has been reduced or attenuated in amplitude by the maximum attenuation allowed. The percentage attenuation cannot be taken for granted and is only know if the reader is familiar with the particular electroencephalograph machine used. In the United States, two common machines sold had different amounts of attenuation at the designated number. The Grass Instrument Company told the reader that the designated frequency would be attenuated 20% from its true amplitude.
P.150

The Beckman Instrument Company told the reader that the attenuation at the designated number would be 30% or 3 dB. The latter degree of attenuation was the convention on machines some years ago, but flatter curves are now possible in modern amplifiers. With the many changes in companies and the variety of machines, the reader and technologist must find out the real effect on their particular machine.
Figure 7.11. A sleeping patient demonstrates a train of multiple spikes blending into multiple spike and slow wave complexes. The individual spike is considerably faster than its repetition rate would suggest.
A filter does not have a static or single level of attenuation at different frequencies. The high-frequency filter will affect the designated frequency by 20% or 30% and is expected to affect frequencies slower than the designated number at a progressively lesser degree (always less than the defined 20% or 30% attenuation) until there is no attenuation at all. In contrast, frequencies above the designated filter number will have progressively more than 20% or 30% attenuation, so that some higher frequencies may be attenuated to zero amplitude and be eliminated. The 1971 minimum standards of the American EEG Society originally recommended that 50 cycles/sec should not be attenuated by more than 50% of true amplitude. In the more recent versions of the standards (1994a-e), this recommendation has been changed. The 1994 standard suggested that 70 cycles/sec activity should not be attenuated by more than 30% of its true amplitude (attenuation of 3 dB).
Filter Issues in the Recording of Spikes
The upper frequency expected of spike discharges (those with a 20-msec base) is 50 cycles/sec, and the concern for this frequency is quite rational. For the individual interested in seizure disorders, the appropriate recording of high frequency may be of greatest concern. The filters on this end of the spectrum can be misused to provoke both underreading and overreading of the recording. Spikes by definition have a base of 20 to 70 msec, which ranges from 14.3 to 50 cycles/sec. Most recognize that spikes in the tonic phase of generalized seizure occur 13 to 20/sec, although Gibbs and Gibbs (1952) report that they can repeat as fast as 40/sec. Observation clearly shows that these are often spikes with a base, indicating a faster frequency than the number seen in a second. The important concept is to recognize that a spike is defined by its base (Fig. 7.11). Readers are describing not the base, but the repetition rate. Thus, a 14 to 20/sec train of spikes may, in fact, consist of 40 to 50 cycles/sec spikes if each is measured individually (Fig. 7.12).
This is not merely an academic point. It is of clinical importance when considering the use of high-frequency filters. Fortunately, the term muscle filter has become less used, but some advocate the advantage of a rapid 30 cycles/sec “roll-off filter” to eliminate muscle faster than 30 cycles/sec. A high-frequency filter eliminates fast activity regardless of its origin. A 40 cycles/sec muscle potential and 40 cycles/sec spike appear the same to the filter and are attenuated and modified to the same degree by the filter. Both technologists and electroencephalographers need to develop understanding and familiarity with the filter curves for their machines.
Figure 7.12. The frequency counter shows the repetition rate, but an uninvolved baseline exists between each spike.
Figure 7.13. The bottom two lines show the same spike and demonstrate the reduction in the amplitude and great modification of the spike pointedness of the wave as it occurs in channel 4 compared to its appearance in channel 5.
P.151

A 35 cycles/sec filter reduces 50 cycles/sec frequencies to only about 60% of their true amplitude. Some machines have 60 cycle “notch filters” that reduce 50 cycles/sec spikes 50% of their true amplitude. The 60-cycle filter and the high linear 35 cycles/sec filters have about equal effect on 40 cycles/sec activity, in that they both reduce it to about 60% of its original amplitude. The high linear filter 15, by definition, reduces 15 cycles/sec spikes by either 20% or 30% depending on the machine involved. It reduces spikes of 30 to 35 cycles/sec to less than two thirds of their true amplitude (Fig. 7.13).
It is critical that filters be used in such a way that they do not eliminate this wanted and desired spike activity. More specifically, the filters must not be used if a spike is to be clearly identified, because high-frequency filters also affect waveform. They round off the point crucial in the identification of a spike (Fig. 7.14). Once rounded off, it becomes difficult to separate a train of spikes from a high-voltage train of beta waves. The reader who consistently berates the technologist for running records with excessive muscle artifact should be aware that technologists sensitized to use filters or to turn off the machine to avoid muscle activity will quite likely eliminate more than muscle artifact from records. This is antagonistic to the idea that seizure onset must be recorded. Seizures are preceded at times by multiple spikes. Spikes are easier to differentiate from muscle artifact with an open high-frequency filter than in a highly filtered and thus uninterpretable recording.
Figure 7.14. This close-up of the train of spikes and slow waves demonstrates a well-defined spike in the bottom line, but after filtering of the high-frequency activity, the point has been rounded off. It is less clear in the top channel that these waveforms could be defined unequivocally as spike discharges and not merely as high-voltage beta waves.
High-frequency filters are, in some instances, referred to as “low-pass filters,” in that they pass low-frequency activity without affecting it. Thus their major influence is on the faster frequencies. This is certainly an accurate definition of their function, but there is more useful identification of their function when they are specified as high-frequency filters and the actual frequency they affect is specified.
Low-Frequency Filters
At the other end of the frequency band, there are low-frequency filters that are designated by numbers representing 20% or 30% attenuation at the particular frequency named. There are settings of low-frequency filter 1, low-frequency filter 0.1, or low-frequency filter 0.16, designating the frequency at which the expected attenuation occurs at the specified percent. The effect of a low-frequency filter is determined by the time constant involved. This feature in simpler, older amplifiers was determined by the combination of the resistance and capacitance in the amplifier. Resistance multiplied by capacitance equals the time constant in actual numbers. In more complex amplifiers, this ratio is not as linear as in the simpler circuits, but these components still are the features that determine the time constant. Time constant can be more operationally defined as the time it takes for a square wave deflection to drop 63% from its peak. That is the same as saying that time constant is the time it takes for a square wave signal deflection to drop within 37% of the original baseline. The technologist can make this measurement directly from the calibration signals that are, in fact, square wave deflections.
Figure 7.15. A series of shorter time constants measuring activity from the same two electrodes in each channel. UV, microvolts.
P.152

The terms time constant and low linear frequency are used interchangeably in considering the effect on slow activity, but the number is not the same where the terms are varied. A time constant of 1 second represents a low linear frequency filter of 0.1 or 0.16, depending on the machine involved. The time constant of 0.3 second may represent a low linear frequency filter of 0.5. It is critical to make the distinction that the same information can be conveyed using low linear frequency or time constant, but the number is not the same.
It has been stated that the time constant does not actually eliminate any of the slow frequencies, but only attenuates them. The effect, however, is that the reader finds the slow waves uninterpretable. Figure 7.15 demonstrates delta waves that are quite evident with longer time constants. The particular downward deflections of the delta components are evident even with the shortest time constant, but a reader is not apt to read or even see the slow activity at the shortest time constant, which is the equivalent of a low linear frequency 5 setting. In Fig. 7.16, it is observed that the delta waves on the left and the right side of the illustration are attenuated to different degrees. This is because they are different frequencies of delta activity. The right-hand wave is slower and is attenuated more sharply. This reemphasizes the point that all frequencies below the designated frequency are not attenuated equally.
Figure 7.16. The same two electrodes are in all four channels but the time constants are different. It is observed that the slower delta wave on the right is attenuated more sharply than the faster delta wave on the left.
Figure 7.17. The same two electrodes with four different time constants demonstrate the apparent peak (marked to the left of the time axis bar). The peak seemingly occurs earlier with the short time constant, but in reality the pen is returning to baseline sooner with a progressively less accurate demonstration of the actual peak.
An additional effect of low linear frequency or time constant is to change the apparent peak of the observed slow wave. Figure 7.17 shows that the peak is closer to the vertical bar (time axis) with the longest time constant and appears to occur earlier with the shorter time constants. This effect is called “phase shift.” Because the phase shift is dependent on the frequency of the wave attenuated, the degree of shift is different for each frequency and has the effect of modifying the relationship of waves to each other. This distorts relationships of different frequencies to each other. This is a major problem when the effect of the filter not only modifies the record but also modifies different components to a different degree, producing significant distortion.
The choice of the appropriate time constant can be critical in locating some asymmetrical discharges. It is general practice to start a record with the low linear frequency set at about 1, which suggests a time constant of 0.12 and 0.16 second. This is an appropriate time constant to use at the start of a record, but it is not the best; it is, in fact, a shorter time constant than is desirable in most records. At the start of a record, there is generally a good bit of patient movement; this relatively short time constant minimizes the artifact that is usually present in this part of the record. It should be routine to use a longer time constant; the 0.3- or 0.4-second time constant is appropriate in almost all records. When the patient is cooperative, it may be desirable to use even longer time constants. A time constant of 1 second, which approximates a low linear frequency of 0.1 cycle/sec, is useful. There is sometimes the feeling that, if an abnormality such as diffuse slowing is already evident in the record, nothing else needs to be done (Fig. 7.18). Readers and technologists both should be aware that records may have bilateral and relatively symmetrical 3 cycles/sec activity with
P.153

underlying lateralized 0.5 or 1 cycle/sec delta activity obscured unless a long time constant is used to allow the asymmetry to appear (Fig. 7.19). The search for subtle underlying asymmetries is of particular value in head trauma patients, who often have both diffuse slowing and focal changes. This is not the only etiology that shows this phenomenon; tumors may provoke similar dysfunction.
Figure 7.18. A record of a patient after a head injury. With a low linear frequency setting of 1, the record shows what appears to be reasonably symmetrical delta activity.
Figure 7.19. With the use of a time constant representing a low linear frequency of 0.1, there is very slow delta activity on the left anterior temporal midtemporal region, much more persistent than that observed in a comparable area on the right.
On occasion it is quite useful to deliberately eliminate much of the slow activity by use of the shortest time constant such as 0.03 or 0.35 second (Fig. 7.20). This attenuates most of the delta range activity. Some individuals believe that reducing the delta activity alone allows one to see faster frequencies more clearly. This may be more of a visual impression than a reality, because the faster frequencies, if observable at the particular sensitivity, are usually seen superimposed on the delta waves. The major value of reducing the delta activity is that it allows an increase in the sensitivity to a level that would be impossible without simultaneous attenuation of the delta waves. The mechanism of this effect is relatively simple, in that slower frequencies tend to be of higher voltage than faster frequencies. As a general rule, an
P.154

individual will demonstrate the lowest voltage for beta activity, somewhat higher voltage for alpha waves if they are present, higher voltages still for theta activity, and the highest voltage for delta waves. In some illnesses this sequence may be modified, but usually modification occurs only with relatively serious involvement; in the typical situation, the typical relationships are preserved. Because delta waves are generally of highest voltage, these peaks are likely to push the amplifier to its limit and require the use of a lower sensitivity. In Fig. 7.20, the sensitivity is 15 in the lower figure; the sleep spindles superimposed on the delta activity are visible but of quite low voltage, and their morphology is difficult to discern. The time constant was reduced to the low linear 5 setting, which means a time constant of 0.03 second. It was possible to increase the sensitivity to 7 μV/mm, which was not practical with the longer time constant. Now the hemispheric asymmetry and loss of spindles on the right compared to the left are more appreciable. An attempt to use sensitivity 7 without having reduced the time constant most likely would have resulted in amplifier and pen blocking and a barely readable record.
Figure 7.20. The bottom portion of the illustration is at S-15. Bilateral slow activity can be observed with some asymmetry, but the difference in the fast frequencies is only suggested. In the upper portion of the illustration, amplification was doubled after the removal of the delta components. The sleep spindles are quite visible and well defined, and the asymmetry is evident.
Sensitivity
The technologist has the ability to modify amplification in all channels of the EEG machine. Such changes can be different in each channel if this proves advisable, but this must be done with great caution.
Most EEG machines are linear, in that a specified number of microvolts deflects the pen 1 mm up to the point where the pen and/or the amplifier has reached its limit. Some older machines have compression amplifiers in which the sensitivity, in essence, decreases as the pen moves. The first millimeter from baseline might require 7 μV of electrical potential, but to move the pen from the 10th to the 11th mm from baseline might require 50 μV/mm. The virtue of this latter type of amplification is that there is less need for intervention by the technologist, and the pens are less likely to block even in the event of abrupt changes in voltage of EEG activity. The serious drawback of compression amplification is that clear and accurate measurement of each of the different types of activity is difficult.
There is no set guideline as to the amount of amplification desirable. Some electroencephalographers prefer to have sensitivity varied as little as possible. This lets one judge amplitude against a relatively constant baseline. Reading ability is limited by not using the full range of machine capability to display activity as clearly as possible. Another philosophy of amplification encourages the technologist to use the maximum amplification that can be employed at any time while avoiding consistent blocking of the pens or amplifiers and record distortion. Each change in sensitivity must be clearly and prominently marked on the EEG record so that the electroencephalographer can be aware of the voltage as the record continues. This latter approach allows for technically better recording of ongoing activity. The low-voltage record is further amplified; it becomes apparent whether or not the attenuation or low voltage involves all frequency bands or only the dominant frequency. Often the maneuver of increasing amplification can be critical in an asymmetrical record in determining whether the side with low voltage dominant activity is the better side in contrast to higher voltage slower frequencies or, in fact, is the more damaged side, because the loss of activity occurs in all bands and represents attenuation of voltage, as can commonly occur with vascular compromise and trauma.
The visual reader has the ability to see changes that involve a 0.5-mm deflection. Undoubtedly, there is individual variability, but the lower limit probably cannot be a greater visual sensitivity than 0.25 mm, because that is the width of the line drawn by an ink pen. If one assumes these estimates are reasonably accurate, a recording with a sensitivity of 7 μV/mm allows the reader to see changes of 3.5 μV when they occur, but not below that. Certainly, the reader of such a record would not see changes at the level of 1.75 μV, which is represented by only a 0.25-mm deflection. If this is kept in mind, the crucial importance of adequate sensitivity in particular circumstances, such as the recording of electrocerebral inactivity, becomes evident. The guidelines for minimal technical standards of EEG recording in suspected cerebral death (American EEG Society, 1994c) note that electrocerebral inactivity is defined as no activity over 2 μV. To have any chance to see 2-μV activity, the reader must have a sensitivity of 4 μV/mm. To display this activity, not just at the visual limit, but well above, would require a sensitivity of at least 2 μV/mm to make such change a readable 1 mm deflection. Present-day machines allow recordings of 1 μV/mm and 0.5 μV/mm sensitivity. These machines have a noise level of 2 μV/mm, but these are peak-to-peak extreme deflections. It is possible with experience to separate electrical noise from brain activity at levels of 1 and 1.5 μV. The argument is made that maximum amplification amplifies noise; this is quite correct. Discrimination between 1 μV noise and 1 μV cortical electrical activity is difficult but, at less than 2 μV/mm sensitivity, a distinction simply cannot be made. Increased sensitivity permits visualization of the activity but does not simplify the decision (Bennett et al., 1976). Some investigators contend that cortical activity at such low levels does not occur (Weiss et al., 1975), but there is reason to suspect that these lower levels of EEG activity have validity in determining ongoing brain function at the cortical level (Reilly et al., 1974, 1978, 1980).
Sensitivity, amplitude, voltage, deflection, and gain are all terms used as parameters with relationships that allow interchange if two of the three major features are known. The most useful formula is that voltage equals the pen deflection times the sensitivity (V = D × S). One can calculate sensitivity (S), deflection (D), or voltage (V) if the other two are known. In every record, during calibration and patient recording, these relationships are used at least indirectly.
Sensitivity is defined as a ratio of input voltage to output pen deflection in an EEG channel (Chatrian et al., 1974). It is described as microvolts per millimeter (μV/mm). Gain isan older term and is the ratio of output signal voltage compared to input signal voltage (Chatrian et al., 1974). It is a multiplication factor. This is usable information in clinical EEG when instrument deflection can be converted to voltage. If input and output voltage are known, sensitivity can be determined. Voltage is the electric potential or potential difference between two points or, in EEG, between two electrodes.
P.155

In EEG it is expressed in microvolts. Deflection is the vertical distance between two points, and the amplitude of a wave is the voltage of a wave measured peak to peak (Chatrian et al., 1974). Amplitude and voltage are essentially synonymous when indicated in measurements such as microvolts, but both are derived from deflection. The amplitude of a wave is sometimes described in millimeters, but this is incorrect even if commonly done. Amplitude is not the millimeter height of the wave but is the potential voltage that caused the pen to deflect that much. The relation established is that the amplitude is calculated by measuring the deflection, but it is not, in itself, deflection.
Use of Paper Drive Controls
An electroencephalograph displays patterns in a manner that becomes familiar to readers. The display can provide essentially the same information at a range of paper speeds, but the ease with which one makes determinations about the activity visually varies greatly depending on the paper speed used. Slow paper speed, such as 15 mm/sec, allows for economy of paper and economy of time in reviewing the record. This can be valuable if the activity recorded is carried out over a great length of time, as in surgery or sleep studies. Concentration of activity in a relatively small space may be quite adequate if the only changes of importance are activities that stand out from the ongoing activity because of some characteristic of the waveform. Very slow paper speed can be used to monitor repetitive episodes of high-voltage spike-and-wave discharge. Similarly, sudden drops in voltage, such as periods of attenuation or marked change in the frequency, may be evident at slow paper speeds.
Clinical laboratories in the United States and many other countries traditionally use a speed of 30 mm/sec. Clinical EEG machines are generally capable of running at half and twice this paper speed. A much wider range of frequencies is available for research machines. For some research applications, 25 mm/sec is as much of a standard as 30 mm/sec is for clinical use. This is particularly true with polygraph studies.
Slower Paper Speed
The paper speed that in some countries is considered slow (15 mm/sec) has been the conventional paper speed in other countries, particularly France. Comfort with one conventional speed over the other is, to a great extent, a matter of training and experience. Concern with the morphology and definition of very fast components such as spikes or with minimal asymmetries in the onset of spike discharges leads one to view 15 mm/sec paper speed less favorably. Particular interest in some of the slower components and in periods of attenuation or rhythmicity over longer periods of time makes slower speeds attractive. They have been particularly popular for studies of premature infants and records of neonatal sleep activity.
Possible subtle asymmetries of slow activity have been enhanced by changing from 30 to 15 mm/sec (Fig. 7.21). This maneuver does not change the amplitude of the wave in question, but it does change the ratio of amplitude to width and allows the reader to visualize asymmetries of slow activity that otherwise might not have been defined as clearly.
Figure 7.21. The top four lines show repetitive 1 cycle/sec delta activity. The bottom four lines show the same asymmetry; particularly in channel 2, the delta activity stands out and becomes more visible because of the change in its height/width ratio.
Faster Paper Speed
Increase in the paper speed will spread out various components. It is not infrequent for the technologist or reader to be concerned as to whether or not fast activity under observation is merely 60-cycle artifact or whether it represents muscle activity. The use of faster paper speed spreads fast frequencies out so that the metronome-like rhythmicity of 60-cycle artifact, versus the clear irregularity of the muscle activity defines the true origin.
Another important use of fast paper speed is the study of asymmetrical onset of seemingly synchronous activity. It is clinically important to distinguish secondary bilateral synchrony following onset from a lateralized focus from primary bilateral synchrony. When bilaterally synchronous epileptiform activity occurs, the onset may have hemispherical variation of 15 to 25 msec, but large (70–100 msec) differences suggest secondary bilateral synchrony (Tukel and Jasper, 1952).
Artifact
Artifacts fall into several major categories. The major class is machine and impedance artifact. The most common ones relate to problems with the electrode, such as the electrode itself being broken or improperly attached to the head. The next most common artifact in the machine category is the presence of 60-cycle artifact, either from nearby equipment or the very common ground loop. The latter often occurs when the patient has been grounded more than once and there is a difference between the grounds, causing 60-cycle artifact. Another source of 60-cycle artifact results when the ground electrode is shorted to one of the active electrodes. This is a particularly disconcerting artifact, in that the 60-cycle activity appears in many different channels. The channels
P.156

change from one montage to another. No matter how frequently one tries to correct individual scalp electrodes, the artifact continues. Testing the electrodes shows impedances that are adequate. It is only when the bridge between ground and scalp electrodes is removed and the area cleansed that the problem is eliminated.
Figure 7.22. The top line represents the actual electrocardiogram (ECG), and the bottom two lines show relatively regular discharges that might be mistaken for repetitive EEG if the patient’s own heart rate was not readily observable.
The next major class of artifacts is physiological. There are a number of separate categories. The major and most common of the physiological artifacts are cardiac or oculographic in origin.
With cardiac artifacts, the most frequent and most troublesome is ECG artifact resulting from the QRS complex. This part of the ECG wave can have very rapid upward deflection resembling a sharp wave or spike in the EEG. The constant rhythm expected is not a sufficient discriminator (Fig. 7.22). The ECG complex is frequently intermittent even in a perfectly regular ECG; a particular ORS complex will be observed, then several complexes will be missed, and then another complex will be visible. This intermittent presentation complicates the distinction between ECG artifact and spikes. It is crucial to have an ECG monitor applied to every patient. A wrist lead is not necessary, and an EEG electrode on the shoulder referenced to the ear is adequate. If the monitor is attached, it can be initiated without changing the level of consciousness of the patient (Fig. 7.23). If it has to be added at the time the question arises, the addition of the monitor may change the level of consciousness of the patient, and the questionable activity may no longer be present (Fig. 7.24). It remains uncertain whether the activity was ECG artifact or the arousal resulting in loss of the spike discharges.
Figure 7.23. On the left, two channels of EEG are seen with sharp waves that may be cardiac in origin. Addition of an ECG on the right demonstrates that this is not the case.
Figure 7.24. Cortical spike discharges have a great tendency to be sensitive to level of consciousness. The sharp activity on the left could have been cortical or occasional ECG complexes. Once the patient’s level of consciousness is changed, as on the right, it is easy to make the discrimination, but the activity is no longer evident.
Another cardiac-related artifact is pulsation artifact. This results from the blood pulsing through a vessel under an electrode. Recording ECG will demonstrate that this is time-locked to the pulsation artifact, which presents as a rhythmic slow wave. Touching the appropriate electrode demonstrates a pulsation. Slight movement of the electrode in whatever direction necessary to move it off the pulsating vessel is sufficient to eliminate the artifact. It is desirable to actually move this electrode; the artifact should not be allowed to continue through the record just because it is identified as pulsation artifact.
A less common cardiac-related artifact is the ballistocardiographic artifact, provoked by the rocking movement of the patient’s entire body each time the heart beats. It is not due either to the ECG itself or to the blood vessel pulsation, but to actual movement of the head and body. This artifact is identified as time-locked to the heart by use of the ECG monitor. It can be eliminated by moving the patient’s head or by putting a pillow under the patient’s neck so that electrode movement against the bed is minimized.
Eye movement artifact is of most concern but easiest to document when it develops from vertical eye movement, in which it appears anteriorally and suggests the possibility of bilaterally synchronous delta waves. Proof of this type of eye movement artifact requires minimum expenditure of the technologist’s time. Electrodes are placed under the center of the eye. (If the patient is looking straight ahead, the electrodes are under the pupil.) Referencing the frontal polar electrode to the ear and, in the next channel, referencing the ipsilateral under-eye lead to the ear will demonstrate in-phase waves between the two channels if the activity is coming from the frontal lobe (Fig. 7.25). Cortical activity has the same phase both above and below the eye. This montage will demonstrate out-of-phase waves if there is eye movement in the vertical plane (Matsuo et al., 1975). Outward phase reversal occurs, because the eye is a relative dipole with the anterior part (cornea) more positive than the posterior
P.157

(retinal) portion, which is negative. With vertical movement, the cornea must move between the frontal polar electrode above the eye and the electrode below the eye. The electrode getting closer to the cornea becomes relatively positive; the electrode the cornea is moving away from becomes relatively negative. These electrodes are both input electrodes in their respective channels, and a phase reversal occurs.
Figure 7.25. Simultaneous reference recording from above and below the eye gives out-of-phase responses to eye blink, but the synchronous delta activity of cortical origin would be in phase.
In other instances, it is useful to record all eye movements. This requires electrode combinations that record lateral as well as vertical eye movements. One of the simplest methods of recording all eye movements is to connect an electrode at the lateral canthus of an eye to the frontal polar electrode ipsilateral to that eye. In the next channel, record from the other frontal polar electrode to the lateral canthus of the second eye. This montage does not distinguish between cortical activity and eye movement, but picks up eye movement deflections in any and all directions.
Troubleshooting
Identification and elimination of difficulties in a record relates to elements of recording discussed earlier in this section. Artifacts frequently are identifiable because of high voltage that is sufficient to cause the amplifiers to reach their limits and because they are frequently limited to a single electrode. Sudden voltage change with an almost instantaneous change from baseline to full voltage is uncharacteristic of true brain activity and alerts the reader to artifactual origin, but not the specific type of artifact occurring.
Troubleshooting requires consideration of the most common possibilities. It then involves a sequence of steps to eliminate the probable or possible causes. Because electrode application and electrode movement are the most common problems in EEG recording, the first step in eliminating a possible artifact is to check electrode impedance with the electrode test or lead imbalance features of the EEG machine. If there is high impedance, it is apparent that an electrode needs to be fixed. If there is not high impedance, it is possible that an intact electrode is making contact with the patient’s pillow or is being touched or rubbed. Readjustment of the patient’s position either by turning the head or placing a pillow or towel roll under the patient’s neck to lift all the electrodes into the air may become necessary. Sometimes it is necessary to move the jack box into which the electrodes are plugged. Simply moving the electrode cables as they go from the jack box to the patient’s head can help. Touching the electrodes and verifying that they are attached tightly to the scalp is a valuable measure.
Computed Topographic Mapping and Analysis
The proliferation of recording systems that allow computation and display of electrophysiological activity in a display that uses the recorded activity to calculate and display the activity or its analyzed components continuously between the few points of actual data recorded has greatly increased the need for attention to almost all of the technical points related to records, including the number and consistent replacement of electrodes (Duffy et al., 1994).
Some of the procedures that at times are considered part of quantitative EEG are being used in more detailed analysis of the standard EEG activity. In some of these approaches to the standard EEG detection may include automated spike detection, using a dipole localization for artifact rejection (Flanagan et al., 2003), or multistage approach for automatic detection of epileptiform EEG (Liu et al., 2002).
Concluding Remarks
In recent years, there has been a tendency in some circles to devote more attention to the apparatus than to the actual operation of the machine and the recording of an EEG record or the evoked potential. As long as unpredictable episodic events continue to play a major role in the patient population generally seen in a clinical neurophysiology laboratory, it will remain necessary for attention to be given to the recording climate and to the skills of the people operating the machinery. This chapter presented a number of variations that cannot be incorporated into a standard record, but can be incorporated into the repertoire of the technologist recording the EEG, evoked potential, or operative record. If technologists are to become familiar with these various
P.158

techniques and be able to use them in urgent situations, they must be given a degree of flexibility in recording so that they can try out these variations. The technologist, therefore, needs to be given latitude in routine situations and encouraged to try out different variations so that these tools can be used when a critical need arises.
If electrode integrity seems good, one is faced with the possibility that the artifact may in fact be related to the machine. It may be necessary to dial the same electrode combination into two amplifiers to see if they appear the same. Both amplifiers will show the artifact if it is from the electrode and/or jack box. If the waveform in the two channels differs, something is wrong with the amplifier. When an amplifier is seemingly disturbed, the settings of the sensitivity and filters in the individual amplifier should be reviewed. This is a common problem when, for some reason, two individual settings of an amplifier have been modified, making the response of the amplifier to the master control inoperative. Failure to return control to the master selector after a specific use is a common error.
References
Adrian, E.D., and Matthews, B.H.C. 1934. The interpretation of potential waves in cortex. J. Physiol. 81:440–471.
Adrian, E.D., and Yamagiwa, K. 1935. The origin of the Berger rhythm. Brain 58:323–351.
American EEG Society. 1994a. Guidelines (#1) in EEG. J. Clin. Neurophysiol. 11:2–5.
American EEG Society. 1994b. Guidelines (#2) in EEG. J. Clin. Neurophysiol. 11:6–9.
American EEG Society. 1994c. Guidelines (#3) in EEG. J. Clin. Neurophysiol. 11:10–13.
American EEG Society. 1994d. Guidelines (#7) in EEG. J. Clin. Neurophysiol. 11:30–36.
American EEG Society. 1994e. Guidelines (#13) in EEG. J. Clin. Neurophysiol. 11:111–113.
American Medical EEG Association. 1990. Electroencephalography Standards.
Barlogie, B., Corry, P.M., Yip, E., et al. 1979. Total-body hyperthermia with and without chemotherapy for advanced neoplasms. Cancer Res. 39:1481–1489.
Barrett, G., Blumhardt, L., Halliday, A.M., et al. 1976. A paradox in the lateralization of the visual evoked response. Nature 261:253–255.
Bennett, D.R., Hughes, J.R., Korein, J., et al. 1976. An Atlas of EEG in Coma and Cerebral Death. New York: Raven Press.
Berger, H., 1929. On the electroencephalogram of man. Arch. Psychiatr. Nervenkrankheiten. 87:527–570.
Binnie, C.D., and Lloyd, D.S.L. 1970. Letters to the editors. Common reference methods of derivation (response), Ed. J.W. Osseltron. Am. J. EEG Technol. 10:69–77.
Binnie, C.D., Dekker, E., Smit, A., et al. 1982. Practical consideration in the positioning of EEG electrodes. Electroencephalogr. Clin. Neurophysiol. 50:282–292.
Busse, E.W., and Obrist, W.D. 1965. Presenescent electroencephalographic changes in normal subjects. Gerontology 10:315–320.
Calloway, E., Tueting, P., and Coslow, S.H. 1978. Event-Related Brain Potentials in Man. New York: Academic Press.
Chatrian, G.E., Bergamini, L., Dondey, M., et al. 1974. A glossary of terms most commonly used by clinical electroencephalographers. Int. Fed. Soc. Electroencephalogr. Clin. Neurophysiol. 37:538–547.
Clement, R.A., Flanagan, J.G., and Harding, G.F. 1985. Source derivation of the visual evoked response to patten reversal stimulation. Electroencephalogr. Clin. Neurophysiol. 1:74–76.
Cracco, R.Q., and Cracco, J.B. 1976. Somatosensory evoked potential in man: far field potentials. Electroencephalogr. Clin. Neurophysiol. 41:460–466.
Drachman, D.A., and Hughes, J.R. 1971. Memory and hippocampal complexes. III. Aging and temporal EEG abnormalities. Neurology 2:1–14.
Duffy, F.H., Hughes, J.R., Miranda, F., et al. 1994. Status of quantitative EEG (QEEG) in clinical practice. Position Paper of the American Medical EEG Association on QEEG. Clin. Electroencephalogr. 25:VI-XXII.
Flanagan, D., Agarwal, R., Wang, Y.H., et al. 2003. Improvement in the performance of automated spike detection using dipole source features for artefact rejection. Electroencephalogr. Clin. Neurophysiol. 114(1): 38–49.
Garvin, J.S., and Gibbs, E.L. 1971. Focal frontal slow activity with physiological reversal of electrical sign in the occipital areas. Clin. Electroencephalogr. 2:218–223.
Gibbs, E.L., and Gibbs, T.J. 1984. Universal APEEGE (anatomical placement of EEG electrodes) system. Clin. EEG 15:1–21.
Gibbs, F.A., and Gibbs, E.L. 1952. Atlas of Electroencephalography. Vol. 2, Methodology and Controls. Reading, MA: Addison-Wesley.
Gibbs, F.A., and Gibbs, E.L. 1964. Exclusive monopolar recordings. Am. J. EEG Technol. 4:8–9.
Goldman, D. 1949. The use of a new type of “indifferent” electrode. Electroencephalogr. Clin. Neurophysiol. 1:523.
Gordon, M.R. 1980. Artifacts created by imbalanced electrode impedance. Am. J. EEG Technol. 20:149–160.
Halliday, A.M. 1978. Evoked potentials in neurological disorders. In Event-related Brain Potentials in Man, Eds. E. Calloway, P. Tueting, and S.H. Coslow, pp. 197–210. New York: Academic Press.
Harner, R.N. 1977. A recommendation for standard EEG montages. Am. J. EEG Technol. 17:105–114.
Hjorth, B. 1975. Technical contribution. An on-line transformation of EEG scalp potentials into orthogonal source derivations. Electroencephalogr. Clin. Neurophysiol. 39:526–530.
Hjorth, B. 1976. Localization of foci in the scalp field. In Quantitiative Analytic Studies in Epilepsy, Eds. P. Kellaway and I. Petersen. New York: Raven Press.
Hjorth, B. 1979. Multichannel EEG preprocessing. Analogue matrix operations in the study of local effects. Pharmakopsychiatr. Neuro-Psychopharmakol. 12:111–118.
Ishiyama, Y., Ebe, M., Homma, I., et al. 1982. Elimination of EGK artifacts from EEGs recorded with balanced noncephalic reference electrode method. Electroencephalogr. Clin. Neurophysiol. 53:662–665.
Jasper, H. 1958. Report of committee on methods of clinical exam in EEG. Electroencephalogr. Clin. Neurophysiol. 10:370–375.
Kaufman, H.H., Reilly, E.L., Porecha, H.P., et al. 1977. Cerebral ischemia during carotid endarterectomy with severe but reversible changes. Surg. Neurol. 7:195–198.
Kooi, K.A., Guevener, A.M., Tupper, J.C., et al. 1964. Electroencephalographic patterns of the temporal region in normal adults. Neurology 14:1029–1035.
Lehmann, D., Ozaki, H., and Pal, I. 1986. Averaging of spectral power and phase via vector diagram best fits without reference electrode or reference channel. Electroencephalogr. Clin. Neurophysiol. 64:350–363.
Liu, H.S., Zhang, T., and Yang, F.S. 2003. A multistage, multimethod approach for automatic detection and classification of epileptiform EEG. Electroencephalogr. IEEE Trans. Biomed. Eng. 49(12 patient 2):1557–1566.
Matsuo, F., Peters, J.F., and Reilly, E.L. 1975. Electrical phenomena associated with movements of the eyelid. Electroencephalogr. Clin. Neurophysiol. 38:507–511.
Morgan, P. 1968. Recording the temporal region spike. Am. J. EEG Technol. 8:7–22.
Myslobodsky, M.S., and Bar-Ziv, J. 1989. Location of occipital EEG electrodes verified by computer tomography. Electroencephalogr. Clin. Neurophysiol. 72:363–366.
Myslobodsky, M.S., Coppota, R., Bar-Ziv, J., et al. 1990. Adequacy of the International 10–20 electrode system for computed neurophysiologic topography. J. Clin. Neurophysiol. 7:507–518.
Obrist, W.P., and Busse, E.W. 1965. Application of electroencephalography in psychiatry. In The Electroencephalogram in Old Age, Ed. W. P. Wilson, pp. 185–205. Durham, NC: Duke University Press.
Obrist, W.P., and Henry, C.E. 1958. Electroencephalographic frequency analysis of aged psychiatric patients. Electroencephalogr. Clin. Neurophysiol. 10:621–632.
Presbitero, J.V., Ruiz, R.S., Rigor, B.M., Sr., et al. 1980. Intraocular pressure during influence and neuroplet anesthesia in adult patients undergoing ophthalmic surgery. Anesth. Analg. 59:50–54.
P.159

Reilly, E.L., and Seward, M.A. 1980. Reversal of electrical sign in the occipital area: physiological or montage artifact. Clin. Electroencephalogr. 11:57–66.
Reilly, E.L., Brumberg, J.A., and Doty, D.B. 1974. The effect of deep hypothermia and total circulatory arrest on the electroencephalogram in children. Electroencephalogr. Clin. Neurophysiol. 26:661–667.
Reilly, E.L., Kondo, C., Brunberg, J.A., et al. 1978. Visual evoked potentials during hypothermia and prolonged circulatory arrest. Electroencephalogr. Clin. Neurophysiol. 45:100–106.
Reilly, E.L., Barlogie, B., Seward, M.A., et al. 1980. Persistence of EEG activity with prolonged induced hyperthermic fever. Clin. Electroencephalogr. 11:22–27.
Seaba, P. 1985. The Importance of Electrode Impedance. The Oxford Observer. Clearwater, FL: Oxford Medilog. Inc.
Seaba, P., Reilly, E.L., and Peters, J.F. 1973. Patient discomfort related to measurement of electrode resistance. Am. J. EEG Technol. 13:7–12.
Silverman, A.J., Busse, E.W., and Barnes, R.H. 1955. Studies in the process of aging. Electroencephalographic findings in 400 elderly subjects. Electroencephalogr. Clin. Neurophysiol. 7:67–74.
Sorel, L., and Ranwez, R. 1984. Basis and use of a true monopolar derivation: the neo-electroencephalography (NEEG) Clin. Electroencephalogr. 2:71–82.
Stephenson, W., and Gibbs, F.A. 1949. Electroencephalograms recorded with non-cephalic electrodes as a reference. Electroencephalogr. Clin. Neurophysiol. 1:523.
Stockard, J.E., Stockard, J.J., Westmoreland, B.F., et al. 1979. Brainstem auditory-evoked responses. Arch. Neurol. 36:823–831.
Thickbroom, G.W., Mastaglia, F.L., Carroll, W.M., et al. 1984. Source derivation: application to topographic mapping of visual evoked potentials. Electroencephalogr. Clin. Neurophysiol. 59:279–285.
Tukel, K., and Jasper, H.H. 1952. The electroencephalogram in parasagittal lesions. Electroencephalogr. Clin. Neurophysiol. 4:481–494.
Wallin, G., and Stalberg, E. 1980. Source derivation in clinical routine EEG. Electroencephalogr. Clin. Neurophysiol. 50:282–292.
Weiss, M., Weiss, J., Cotton, J., et al. 1975. A study of the electroencephalogram during surgery with deep hypothermia and circulatory arrest in infants. J. Thorac. Cardiovasc. Surg. 70:316–329.
Yoshii, N. 1971. Electroencephalographic observation of apparently normal aged subjects. Tohoku J. Exp. Med. 103:202–215.
Yoshii, N., To, T.W., Murase, I., et al. 1966. A comparative study of average reference and routine monopolar ear lobe reference electrodes in EEG recording. Keio J. Med. 15:219–226.