Fig. 1
Example of simultaneous aEEG and cEEG recorded from a single patient. The aEEG with select source channels (top) is displayed at the bedside for the neonatal team to review in real time, while the full cEEG (bottom) is available for remote review by a neurophysiologist
Similar to the use of qEEG in older patients, aEEG display may also be used by neurophysiologists to screen large quantities of cEEG recording to quickly identify presence or absence of seizures before a more detailed review is undertaken. This allows for more targeted review of long periods of cEEG recording and expedited intervention in many cases. Typical aEEG settings allow display of 3–6 h of EEG on a single screen; an overview of a day of recording can be viewed in minutes.
For all of these reasons, even if neurophysiologists typically rely on cEEG for neonatal seizure detection, aEEG is often indicated either to supplement or in conjunction with cEEG for seizure detection.
Recording and Display
As mentioned previously, cEEG is widely considered to be the gold standard method for seizure detection. Full array cEEG uses 9 to 16 electrodes placed according to the international 10–20 system, modified for the smaller neonatal head (Fig. 2). Up to 16 channels result. This is generally thought to capture all but exceptionally rare spatially restricted seizures [3]. cEEG provides detailed information about background activity as well as the location, form, evolution, and migration of ictal patterns. The addition of video allows rapid identification of artifacts and correlation of electrographic seizures with clinically apparent phenomena. In contrast, aEEG records less information, typically from just two to four electrodes, which is then simplified and compressed, allowing for more rapid but more limited assessment of cerebral function, including by those without formal neurophysiology training.


Fig. 2
Illustration of electrode placement in cEEG vs aEEG. Open circles represent typical electrode placement for neonatal EEG. Shaded circles (C3, C4, P3, P4) represent typical electrode placement for dual-channel aEEG
A key feature is that aEEG uses a reduced array, with a minimum of three electrodes – two placed in the biparietal location (P3-P4) and one to serve as a ground to record “single-channel” EEG. Various types of electrodes are available. Subdermal needle electrodes have the advantage of being easily secured for long-term monitoring and having reduced impedance. However, concerns for needlesticks may make disk or sticker electrodes preferred by some users [7]. Hydrogel electrodes can be used in extremely preterm infants, though scalp preparation with abrasive cream is still typically required to achieve acceptable impedances [3, 7].
The reduced array of electrodes records limited channels of EEG, just as in conventional EEG. Originally, aEEG systems recorded only at P3–P4 to generate “single-channel” EEG. Increasingly, aEEG systems use four electrodes (placed at C3, P3, C4, and P4) with a ground [15] (Fig. 2). This “dual-channel” configuration provides information about laterality (using C3–P3 and C4–P4 as hemispheric channels) in addition to a cross cerebral channel (P3–P4). The use of dual-channel aEEG increases sensitivity for seizure detection as compared to single-channel aEEG, as discussed further below [13]. More electrodes can be used, though not all commercial systems have this ability. Some aEEG systems allow a machine to start recording limited array EEG for aEEG from a reduced number of channels (such as when a recording is started by a nurse in the middle of the night) and then “flex up” to a full array of EEG channels later in the recording (such as when a technologist becomes available).
Electrodes should be placed in the centroparietal region to maximize sensitivity for seizure diagnosis. This vascular “watershed” area is particularly susceptible to injury; most neonatal seizures arise from this region [7]. This area is also less affected by eye movements or scalp muscle artifact, which can be problematic with frontal or temporal electrode placement [7]. Furthermore, while electrode placement in the frontal area may be more convenient (given no hair interferes with the application), frontal electrodes show more artifact and have reduced sensitivity for seizures. Over half of seizures are missed with single-channel EEG using frontal electrodes [20].
This reduced array of electrodes records channels of EEG, just as in conventional recordings. Voltage differences between electrical potentials at two different scalp areas are recorded. The aEEG machine or software then processes the raw EEG tracing to facilitate its interpretation. The raw recording is filtered, removing signal with frequencies below 2 Hz and above 60 Hz. The filter parameters are designed to eliminate artifact, but the high pass 2 Hz filter can also eliminate normal and pathologic features, including low-frequency seizures. Recorded electrical potentials are rectified, with negative voltages converted to positive values [3, 7]. Amplitudes are then plotted on a time-compressed display, with the x-axis representing time and the y-axis a semilogarithmic representation of amplitude (linear from 0 to 10 μV and logarithmic above 10 μV) (Fig. 3). Amplitude data is plotted in consecutive, thin, vertical lines, with each line representing 15 s of recording. For each line, the top point on the y-axis represents the maximum amplitude recorded during the interval, while the bottom point on the y-axis represents the minimum amplitude during that interval. In neonates with impaired brain function, the majority of amplitudes will be between 0 and 20 μV, and the display emphasizes a finer degree of detail within this range. As the display progresses, the side-by-side vertical lines form the activity band [15], which serves as a graphical representation of amplitude of brain activity on a compressed time scale. While many systems allow adjustment of the time scale, typical aEEG displays 1 h of recording over 6 cm or 1 min of recording per millimeter. Of note, aEEG displays only the amplitude of the EEG signal over time. There is no information regarding frequency, power, or other features that may be included in other qEEG techniques.


Fig. 3
Example of dual-channel aEEG tracing, with time displayed on the x-axis and amplitude in uV semi-logarithmically displayed on the y-axis
Newer machines can display the raw EEG tracing corresponding to a particular area of the processed data, which is critical for identification of artifacts or subtle seizures (Fig. 4). Some machines can also display other metrics, such as impedance, with alarms for excessively high impedance during recording. There is commercially available software for automated seizure detection based on aEEG, though this has not been approved for use in all countries. As above, some cEEG acquisition systems allow display of aEEG at the bedside during cEEG recording, with cEEG displayed for a neurophysiologist at a review station [3].


Fig. 4
aEEG (bottom two panels) displayed along with source EEG (upper two panels). Line at left of aEEG panels indicates area of interest selected for review, with arrow pointing up to corresponding segment of source EEG, in this case, confirming presence of seizure
Interpretation
Background
Accurate detection of neonatal seizures on aEEG recordings requires a basic understanding of overall aEEG interpretation. As with conventional EEG, critical elements of neonatal aEEG interpretation include background assessment, detection of seizures, and identification of artifacts. In general, there is good consistency between aEEG background patterns and their correlates on cEEG [6]. There is not one universally accepted method for describing or classifying aEEG patterns; the approach described here has been adapted in various forms.
One approach to aEEG background interpretation focuses on quantitative assessment of the aEEG activity band, with some pattern recognition (Table 1) (reviewed in detail elsewhere [8]). This system classifies aEEG into five categories: continuous, discontinuous, low voltage, burst suppression, or inactive. In a term neonate, only a continuous background is normal. In preterms, depending on gestational age, a discontinuous background pattern may also be normal. Low-voltage, burst-suppression, and inactive patterns are always abnormal. In normal aEEG recording from a term newborn, the lower margin of the activity band is greater than 5 μV, and the upper margin is greater than 10 μV and often greater than 25 μV (Fig. 5). This reflects a raw EEG pattern with consistently normal amplitudes. In contrast, a discontinuous EEG pattern will have some periods of higher amplitude alternating with very low amplitude; this wider range of amplitudes is reflected in a wider activity band, with the lower margin sometimes below 5 μV (Fig. 6). In these patterns, normal variability in amplitudes is reflected in variability in the margins of the activity band. This is in contrast with abnormal patterns such as burst suppression. In burst suppression, the lower margin is also below 5 μV and the upper margin greater than 25 μV, but the lower margin has no variability – it is a near flat line, reflecting the true suppression of the background between bursts on the corresponding EEG. This suppression alternating with short, high-amplitude bursts results in a “comb-like” pattern. A severely abnormal aEEG background pattern may be described as low voltage when the upper margin is below 10 μV or inactive when the upper margin is below 5 μV and the lower margin below 2 μV and invariant (Fig. 7).


Table 1
Assessment of aEEG background
Background category | Upper margin of activity band (μV) | Lower margin of activity band (μV) | Variability: upper margin | Variability: lower margin |
---|---|---|---|---|
Continuous | 10–50 | >5 | Present | Present |
Discontinuous | >10 | <5 | Present | Present |
Low voltage | <10 | <5 | Present | Present |
Burst suppression | Bursts >25 | 0–2 | Widely variable (bursts) | Absent |
Inactive | <5 | 0–2 | Absent | Absent |

Fig. 5
Example of continuous aEEG. The lower margin of the activity band is above 5 uV, while the upper margin is above 10 uV

Fig. 6
Example of discontinuous aEEG. The lower margin of the activity band is below 5 uV, while the upper margin is above 10 uV
