LEARNING OBJECTIVES

Introduction

Amplitude-integrated electroencephalography (aEEG) is a trend algorithm used to examine the variance in amplitude of the electroencephalogram (EEG) signal over a compressed period of time. The limited number of channels as compared to conventional EEG simplifies the learning process for healthcare providers with little to no experience in neurophysiology. As with conventional EEG, background activity, behavioral states, and maturation of the developing brain can be assessed. In this chapter, we review the basic principles of aEEG and how it can be used in the newborn to identify background patterns and seizures.

Role and Utility of aEEG in the Neonatal Intensive Care Unit

The continuous and real-time monitoring of vital functions in the intensive care unit (ICU) is the fundamental basis of early intervention to reduce morbidity and mortality in the critically ill patient. The multimodal monitoring of brain function and its integration with other neurocritical care techniques has resulted in improved outcomes for critically ill patients with neurologic injury.^1–3 The neurologic examination is an important component of the routine assessment of the critically ill patient, including in newborns; however, the examination provides an assessment at only a single moment in time and can be limited in a comatose or sedated patient. Continuous EEG monitoring (cEEG) provides an objective measure of cerebral function, including the degree of brain maturation, severity of brain injury, presence of seizures, acute response to therapeutic interventions, and neurologic prognosis.

In the mid-twentieth century, given significant limitations to bringing EEG equipment into intensive care units, including limited bedside space, interference from adjacent monitors and equipment, and the need for highly trained personnel, Douglas Maynard, Pamela Prior, and D. F. Scott developed the Cerebral Function Monitor (CFM), the first aEEG device, to monitor brain function in adult patients after cardiac surgery. The CFM device filtered and compressed the amplitude of a single EEG channel into what is now known as an aEEG recording.⁴ In the ensuing years, its clinical applications were expanded to include neurologic monitoring in the setting of general anesthesia, cardiac arrest, status epilepticus, and cardiac surgery.^5–8

Beginning in the 1980s, the CFM device was applied to newborns for the identification of brain maturation patterns, seizures, and brain injury assessment. Verma et al.⁹ and Maynard et al.¹⁰ first defined the normal patterns of aEEG for gestational age in 1984. Lena Hellström-Westas, Ingmar Rosén, and Linda S. de Vries expanded its clinical applications to normal and critically ill newborns and published the first aEEG atlas for newborns in 2003.^11,12

aEEG has evolved significantly in the 21st century. Multiple aEEG devices have been developed for health care providers inexperienced in or with limited access to experts in neurophysiology. These devices offer one and/or two aEEG channels (Olympic CFM and BrainZ BRM3, Natus Medical, San Carlos, CA). Meanwhile, various companies have begun offering trend algorithms to enhance the review and analysis of continuous EEG recordings, and these include aEEG trends. Examples include Nihon-Kohden (Tokyo, Japan), NicoletOne (Natus Medical, San Carlos, CA), Persyst (Persyst Development Corporation, Prescott, AZ) and the Component Neuromonitoring System (CNS Monitor, Day One Medical, Ambler, PA).¹³ Although the precise details of the aEEG algorithm for a given device or software tends to be proprietary, the general visual aEEG patterns and voltage criteria described and delineated in newborns with the original CFM device remain consistent.^13–16

Introduction to Basic Principles of aEEG

How the Trace Is Generated

The EEG signal goes through a process of amplification, filtering, and time compression to create a trend display of brain activity known as the aEEG. The raw signal passes through a voltage amplifier and then an asymmetric 290band-pass filter with the purpose of filtering fluctuations in very slow (below 2 Hz) and fast frequencies (higher than 15–20 Hz), as these are most commonly associated with artifacts produced by sweating, muscle activity, and external electric currents, among others. For frequencies between 2 and 15–20 Hz, the band-pass filter has a positive gradient to give equal weight to the lower amplitudes of higher frequency activity and the higher amplitudes of lower frequency activity.¹⁷ Together this allows for amplification of the electroencephalographic signal between 2 and 20 Hz, highlighting the typical physiologic frequencies seen in recordings from mature as well as developing brains,¹⁷ and stabilizes the activity of lower frequencies while accentuating the activity of higher (more physiologic) frequencies (Figure 8.1).

To account for the wide variation in amplitudes seen in cerebral activity (isoelectric to seizures), the amplitude of this filtered signal is subjected to a semilogarithmic compression that is weighted by the frequency of the filtered signal.¹⁷ The amplitudes are compressed for higher peak-to-peak (p-p) amplitudes but not for p-p amplitudes in standard ranges. This provides a weighted-gain effect of the amplitudes of the electroencephalographic signal representing physiologic cerebral activity.

This filtered and weighted signal is then rectified, smoothed (Figure 8.2), and displayed on a semilogarithmic y-axis (linear scale between 1 and 10 µV; logarithmic scale between 10 to 100 µV) over a compressed time scale (x-axis in hours) (Figure 8.3). The semilogarithmic y-axis highlights the range of p-p amplitudes (0–25 μV) of physiologic interest. By tradition, the x-axis is compressed to a ratio of 1 mm/minute or 6 cm/hour, but in clinical practice can be modified to discern brief seizures or state changes.

The aEEG tracing appears as a thick band due to the compressed time x-axis. The lower margin of the tracing represents the lowest p-p amplitudes and thereby represents the overall background state of the EEG recording. The upper margin of the trace represents the highest p-p amplitudes. The bandwidth (difference between the upper and lower margins) represents the moment to moment variability of the p-p amplitudes or variability of the EEG signal.

aEEG Settings

ELECTRODE MONTAGE

The original CFMs had only a single channel (two electrodes) straddling the vertex (P3-P4) to capture watershed insults in adult critical care patients after cardiac surgery while minimizing temporalis muscle or movement artifacts.¹² Today stand-alone aEEG devices can support one- and/or two- channel aEEG recordings. EEG review software, in contrast, can display multiple aEEG channels as per the EEG reader. In newborns, the ideal aEEG locations to capture seizures are central and parietal, as hypoxic-ischemic events affecting watershed vascular territories are the most common type of injury.^18,19 Therefore, the American Clinical Neurophysiology Society (ACNS) recommends placing four electrodes to allow for interhemispheric comparison: C3-P3, C4-P4, C3-C4 and P3-P4¹⁸ (Figure 8.4).

PARAMETERS

As noted above, the standard aEEG tracing is displayed on a semilogarithmic y-axis of p-p amplitudes and on a compressed x-axis of time. The difference between the upper and lower margins of the aEEG represents the variance of the EEG amplitude and normally ranges between 10 and 40 µV, depending on the behavioral state of the patient. Time is represented on the x-axis on a compressed scale so as to comfortably display 6 hours per screen. The x-axis can be adjusted to display less time per screen (i.e., 2-hour screen display) to enable detection of brief events (e.g., neonatal seizures, which typically range from 20 to 140 seconds) or more time per screen (i.e., 8–12-hour screen display) to evaluate state changes (e.g., sleep-wake cycling, diurnal changes, improvement or deterioration of the background).¹³

The width of the aEEG band represents the maximum and minimum p-p amplitude (variance) over a 15-second period of the EEG recording.^5,12,20 The lower border indicates the lowest p-p amplitude while the upper border denotes the highest p-p amplitude during the same time period (Figure 8.5).

Indications for Use

Brain function is dynamic and associated with continuous fluctuations in perfusion, metabolism, and auto-regulation.²¹ Continuous EEG (cEEG) is an invaluable bedside tool enabling real-time assessment of brain function in patients with acute neurologic injury.^1–3,22 Newborns are at high risk for an unfavorable neurodevelopmental outcome after brain injury, further supporting the importance of neuromonitoring in neonates;²³ however, most neonatal ICUs (NICUs) have limited to no access to cEEG.

aEEG can be used to monitor high-risk newborns with potential brain injury in limited-resource NICU settings.^18,24,25 The ACNS provided clinical guidelines for cEEG monitoring of newborns in 2011,¹⁸ which can be extended to aEEG monitoring (Table 8.1), with the caveat that any abnormal aEEG finding should lead to neurologic evaluation and/or conventional EEG monitoring.

Trend analysis, including aEEG, can be performed using any channel(s) and can be used concurrently with cEEG.¹⁷ In general, aEEG is helpful to quickly scan a prolonged recording for normal and abnormal events, to assess the EEG background, to identify changes in state, to monitor for improvement or deterioration of a patient’s brain function, and to provide prognostic information. Amplitude-integrated EEG can be used in patients with acute encephalopathy, congenital heart disease, metabolic disorders, ischemic and hemorrhagic lesions, those supported by extracorporeal membrane oxygenation (ECMO) and others.¹²

The same features of aEEG that allow for visualization of global trends limit its application to resolve fine details of the EEG recording. aEEG cannot be used to assess reactivity, synchrony, or specific graphoelements (e.g., delta brushes, frontal sharp transients) that can aid in the evaluation of the patient with encephalopathy.²⁶ As such, the aEEG should be considered only as a screening tool to determine who may need a cEEG at those centers without access or with only limited access to continuous long-term monitoring.

Background

Several aEEG classification systems have been proposed, with the most widely used classifications based on voltage and pattern recognition.

Voltage Classification

A quantifiable classification system based on voltage criteria was proposed by al Naqeeb et al. in 1999. They provided standardized upper and lower margin p-p amplitudes in term newborns, grouping the background activity as shown in Figure 8.6.²⁷ They did not address sleep states. Their classification was highly predictive of neurodevelopmental outcome in newborns with acute encephalopathy and had good correlation with the prognostic information derived from conventional EEG and MRI.

There are limitations to using voltage as the sole criteria to grade an aEEG. Voltage can be affected by impedance, distance between electrodes, scalp edema or hemorrhage, and high-frequency oscillatory ventilation. These can lead to false elevation of the lower margin of the aEEG, especially in tracings with a depressed background.¹⁶

Pattern Classification

In 2006, Hellström-Westas, de Vries, and Greisen²⁶ proposed a new aEEG classification during a period when aEEG and conventional EEG were being used simultaneously. They observed that the pattern of the aEEG tracing (bandwidth) changed with the degree of activity and continuity on the conventional EEG recording regardless of gestational age, thereby reflecting EEG background activity and state changes. The names given to each pattern of the aEEG tracing reflected the activity of the conventional EEG background: continuous (C), discontinuous (DC), burst suppression (BS), low voltage (LV), or inactive/flat (FT). Within this classification, there is a subclassification of burst suppression focused on the density of the bursts. Burst suppression (+) consists of dense bursts (≥100 bursts/hour), while BS (−) consists of sparse bursts (<100 bursts/hour). They also described normative aEEG patterns for gestational age, providing maturational standards (Figure 8.7), and a classification for sleep-wake cycling (i.e., no sleep-wake cycling, imminent/immature sleep-wake cycling, and developed sleep-wake cycling).

aEEG in the Preterm Infant

Though less well characterized than in the term infant, normative aEEG patterns have been defined in the preterm infant. Normal aEEG changes parallel the conventional EEG changes that occur in wakefulness, active sleep, and quiet sleep with gestational age. Therefore, it is important to understand the normal development of EEG features in this age group to relate them to the expected aEEG features for gestational age.

The ACNS has standardized the descriptions of behavioral states and sleep-wake cycling in the newborn.²⁶ The conventional EEG in the extremely preterm infant is primarily discontinuous (tracé discontinu), with brief periods 292of continuity emerging at 28 weeks postmenstrual age (PMA). The conventional EEG becomes more continuous with increasing PMA. With increasing gestational age, the duration of the interburst interval (IBI) decreases and the bursts are characterized by a greater degree of organization, higher frequencies, and lower amplitudes. In keeping with this, the aEEG bandwidth narrows with gestational age (i.e., becomes more continuous with a smaller difference in p-p amplitude)^14,28,29 (Figures 8.8 and 8.9). Normative IBI amplitude and duration for gestational age have been described.^28,30,31 The emergence of quiet sleep and the cycling between quiet and active sleep is represented on the aEEG tracing as cyclic variations of the lower margin and bandwidth^16,29,32,33 (Figure 8.10).

Tracé discontinu of the preterm infant can be distinguished from burst suppression on aEEG by differences in the lower margin and bandwidth of the trace and by the presence or absence of state change. Suppression is defined as an IBI with voltage less than 2 µV, while the lower margin of a tracé discontinu tracing on aEEG is 0–5 µV, allowing for differentiation of these two patterns based on voltage criteria. Moreover, the tracé discontinu pattern should demonstrate variability, while burst suppression should be invariant.

Seizures

As recent EEG-based neonatal seizure studies have shown, seizures in critically ill newborns are common and are associated with increased mortality and worsened neurodevelopmental outcomes.^34–38 The majority of seizures in neonates are acute symptomatic (i.e., due to hypoxic-ischemic encephalopathy, intracranial hemorrhage, perinatal arterial ischemic stroke, and congenital or perinatal CNS infections). Seizure onset typically occurs within the first 3 days of life and reflects the underlying mechanism of injury. In patients with postnatal CNS infections, inborn errors of metabolism, metabolic disturbances, CNS malformations, and genetic conditions, seizure onset is often after 3 to 5 days of life.

Symptomatic seizures in the newborn period have been associated with higher rates of mortality, later epilepsy, and cognitive and behavioral disability.^39–42 Continuous EEG monitoring may allow for early seizure detection and treatment,^43–45 reduced exposure to antiseizure medication by ensuring that nonseizure movements are not treated as seizure,^46,47 and potentially improved outcomes.

The ACNS classifies neonatal seizures as clinical-only, electroclinical, or electrographic-only. Electrographic seizures in the newborn are defined as a “sudden, abnormal EEG event defined by a repetitive and evolving pattern with a minimum 2 µV p-p voltage and duration of at least 10 seconds.”²⁶ The EEG evolution can be in frequency, voltage, morphology, and/or location. There is no minimum frequency requirement, in contrast to older children and adults.

On aEEG, the electrographic seizure is typically seen as an abrupt upward deflection or rise in the lower more than the upper margin of the trace, with concomitant narrowing of the bandwidth (less variance in p-p amplitudes), which is sustained for a brief period of time before gradually returning to baseline (Figure 8.11). Multiple, repetitive seizures can give the aEEG tracing a saw-tooth appearance.⁴⁸ Hellström-Westas et al. proposed characterizing seizure burden as single, repetitive (occurring more frequently than every 30 minutes), and status epilepticus (continuous seizures >30 minutes).²⁶ However, the authors would recommend following one of the ACNS-proposed definitions of seizure burden: frequency per hour, percent of the record, and/or status epilepticus defined as a single seizure >30 minutes or seizures occurring over greater than 50% of a 1-hour epoch.

Sensitivity and Specificity for Seizure Detection

The limitations of aEEG for seizure detection are important to note as a guide to when a continuous EEG should be obtained. Obvious limitations include the number and location of electrodes placed.^48,49 The sensitivity of a single-channel aEEG for seizure detection has been reported to be below 50%,^41,50,51 increasing with additional electrodes to as high as 92%. However, this sensitivity is as low as 38% if the user has less than 1 year of experience with aEEG.^19,41,52,53

Patterns of seizure localization and propagation, amplitude, duration, and frequency can also affect the sensitivity and specificity of aEEG for seizure detection. Electrographic seizures with low amplitudes similar to background (common in newborns)^49,50,53 or short durations will not produce a discernable deflection of the aEEG tracing (Figure 8.12). The average newborn seizure duration of 100 seconds corresponds to a 1.4-mm deflection on a standard aEEG display of 6 cm/hour. Therefore, electrographic seizures less than 40 seconds cannot be distinguished by standard aEEG.^51,54,55 Moreover, electrographic seizures overlapping in time (multifocal independent seizures) or with brief time intervals in between will appear as a single prolonged seizure on aEEG.

Motion artifact can also mimic the aEEG deflection seen with electrographic seizures. Patting artifact, for example, can be abrupt in onset and evolve on EEG. Patting artifact is often stereotypical to the person doing the patting and recurrent. However, without a synchronized video for review, patting artifacts are indistinguishable from seizures on aEEG.

293Conventional EEG remains the gold standard for seizure detection and characterization (number, duration, localization, propagation) in newborns. aEEG is a useful screening device; however, cEEG is necessary to accurately capture seizure burden and to guide seizure treatment.

Status Epilepticus

Frequent repetitive seizures, whether meeting the electrographic definition of status epilepticus or not, will give a saw-tooth appearance (margin deflections) to the aEEG tracing⁴⁸ (Figure 8.13). Continual or near-continual electrographic seizure activity can be mistaken for a normal high-amplitude aEEG tracing, as the lower and upper margins do not visibly return to baseline.

Prediction of Outcome Based on Background Features

Acute Encephalopathy

Like conventional EEG, aEEG background has been shown to parallel the degree and course of acute encephalopathy, as well as predict outcome. In the case of hypoxic ischemic encephalopathy (HIE), early aEEG findings correlate with adverse outcomes, including death, cerebral palsy, and global developmental delay. Indeed, aEEG background within the first 24 hours of life can predict outcome at 12 months of age in over 90% of newborns.^26,56–58 In the precooling era, burst suppression, low-voltage continuous, or flat aEEG tracings within the first 3 hours of life predicted adverse outcome with a sensitivity of 85% and specificity of 77%.^24,57,59 Noncooled neonates with an abnormal aEEG have a worse outcome than infants who are cooled.^60,61 Normalization of aEEG background by 24 hours of life in noncooled and by 48 hours of life in cooled neonates is associated with lower rates of disability.^56,58 Moreover, the absence of sleep-wake cycling by 36 hours of life has been associated with adverse outcome in 82% of cooled neonates^12,62 (Figure 8.14).

In the postcooling era, EEG and aEEG background trends can inform the bedside user of the infant’s response to neuroprotective strategies. Mild to moderate hypothermia attenuates the amplitude of electrocerebral activity and shortens seizure duration but does not affect the correlation between background and outcome at 18–24 months of age.

Extracorporeal Membrane Oxygenation

ECMO may be used in neonates with respiratory or cardiac failure refractory to medical management. This requires cannulation of large neck vessels, blood contact with synthetic surfaces, and anticoagulation, resulting in an increased risk for neurologic complications, including hypoxic-ischemic injury, intracranial hemorrhage, and embolic infarcts.^63,64 Continuous neuromonitoring in the form of conventional EEG is suggested, but prolonged aEEG monitoring is often used instead by tertiary NICUs. The aEEG is noninvasive and portable⁶⁵ and has been shown to modify the course of ECMO treatment.⁶⁶ A burst suppressed aEEG background during ECMO is associated with a poor prognosis.^63,67 Electrographic seizures have been reported to occur in up to 18% and status epilepticus in up to 11% of newborns undergoing ECMO.^65,68,69

Congenital Heart Disease

Complex congenital heart disease, such as hypoplastic left heart syndrome, critical aortic stenosis, aortic coarctation, and transposition of the great arteries, is accompanied by a high risk for hypoxic-ischemic or anoxic brain injury given the abrupt changes in cardiac output, hypoxia, and hypotension these infants often experience before, during, and after their staged surgical corrections. Deep hypothermic circulatory arrest and cardiopulmonary bypass can also provoke cerebral blood flow fluctuations. An abnormal preoperative aEEG background is associated with a higher risk for brain atrophy on MRI.^70–72 During deep hypothermic circulatory arrest (DHCA), aEEG amplitude is suppressed and gradually increases during rewarming. Normalization of aEEG background occurs in 62% of infants by 24 hours after DHCA and in 74% by 48 hours. Infants with a persistently abnormal background 48 hours after DHCA have a 50% increased risk of mortality and worse cognitive, language, and motor outcomes, compared to a 14% risk of mortality in those whose background normalizes by 48 hours.^73,74

Seizures are the most common neurologic complication observed in the perioperative period. Between 11% and 18% of infants have a seizure(s) within the first 2 days after surgery, usually electrographic-only and refractory to multiple antiseizure medications.^73,75–78 Seizure occurrence and high seizure burden have been shown to correlate with new MRI lesions after surgery, increased mortality, and poor language, motor, and developmental outcomes.^74,79,80

Intraventricular Hemorrhage

IVH is one of the most common types of brain injury among preterm infants, especially prior to 28 weeks PMA. Abnormal aEEG background for PMA (prolonged discontinuity or poor sleep-wake cycling) or presence of seizures has been associated with severity of hemorrhage (Figure 8.15).^81–83 In contrast, cerebrospinal fluid drainage after posthemorrhagic hydrocephalus has been associated with an improvement in aEEG background.⁸⁴

294Medication Effects

The use of sedatives, paralytics, antiseizure medications, and other CNS-acting medications can modify the alertness, reactivity, and sleep state of the neonate and, thereby, attenuate the EEG signal, affect the variance of the p-p amplitude, and alter natural sleep-wake cycling. Sedatives such as morphine, fentanyl, and benzodiazepines increase IBIs, resulting in a greater degree of discontinuity, with a more profound effect in preterm than in term infants. Midazolam infusions can provoke acute changes in the aEEG background, including induction of a burst-suppression pattern, and may induce a later onset of sleep-wake cycling.^85–92 Phenobarbital, lidocaine, and lorazepam can cause voltage suppression as well as altered sleep-wake cycling.^92–96 Surfactant therapy in preterm newborns has been shown to cause a profound decrease in brain activity on aEEG for approximately 10 minutes; this is associated with a sudden fall in mean arterial blood pressure in an immature CNS where cerebral autoregulation is impaired.^97,98 Depending on the type and dose, medications cause different changes in cortical brain activity affecting the aEEG background and the cycling between behavioral states. A through medical and medication history is required to effectively interpret the aEEG.

Artifacts

A modern, tertiary NICU is a harsh environment for recording neurophysiologic information given that adjacent medical equipment generates mechanical and electrical disturbances. Isolating the EEG electrodes (wrapping the head), amplifier box and cable, and power outlet, amplifying the EEG signal, and applying bandpass filters to eliminate very slow or fast frequencies are measures used to minimize artifact. Video synchronization is useful to identify some artifacts, especially those caused by movement of the patient and/or EEG wires, but is not available with stand-alone aEEG devices. Stand-alone aEEG devices do display an impedance map to assist the bedside user in recognizing when electrodes need to be repositioned or replaced. Despite these measures, artifacts persist and require expertise to prevent misinterpretation.⁹⁹ For simplicity, artifacts will be divided into nonphysiologic and physiologic artifacts.

Nonphysiologic artifacts arise from the environment. They can be mechanical artifacts from feeding and infusion pumps, bedside cares and other interventions, standard ventilators, high-frequency oscillators, dialysis, and ECMO. Conventional mechanical ventilation can generate a rhythmic, high- amplitude slow-wave artifact, altering the amplitude margins of the aEEG trace (Figure 8.16A).¹⁰⁰ High frequency oscillatory ventilation and movements or handling of the baby, like patting, may cause a sudden increase of the lower margin on the aEEG, which can be interpreted as an improvement in the neurologic state (oscillator) or a seizure (patting).¹⁰¹ NIRS cables, radiant warmers, 60 Hz outlets, internet cables in the walls and cellphones in the range of 60 Hz can also introduce electrical artifacts (Figure 8.16B).

Physiologic artifacts arise from the patient and include electrocardiographic (EKG) artifacts, myogenic artifacts originating from the temporalis and frontalis muscles when grimacing, from the masseter or tongue muscles during sucking or swallowing, or from the fine muscles of the globe in association with movement of the eyes.^100,102,103 These artifacts typically increase the lower margin of the aEEG.

Conclusion

Amplitude-integrated electroencephalography is a bedside tool that can assist in spot, serial, or prolonged monitoring of brain activity in the critically ill neonate. It involves a simple setup of two or four electrodes and has a relatively straightforward interface to check electrode impedance and review tracings. It can be used to assess the severity of encephalopathy and the maturational state of the brain. It can also be used for seizure detection and can aid in the prognosis of neurodevelopmental outcome. For these reasons, many neonatologists and neonatal neurologists consider the aEEG an invaluable adjunctive tool in the NICU.

Related posts:

Templates and Treatment Algorithms Cases Questions and Answers by Chapter Brief Rhythmic Discharges, Seizures, and Mimics in Critically Ill Neonates Seizures in Critically Ill Children Convulsive and Nonconvulsive Status Epilepticus in Critically Ill Children

Stay updated, free articles. Join our Telegram channel

Join