Application of an Amplitude-integrated EEG Monitor (Cerebral Function Monitor) to Neonates

aEEG was originally developed as a bedside monitor for adult intensive care¹. The first publications detailing its use in neonates date back to the late 1980s^2,3. In early years, its clinical use was mainly for the detection of cerebral seizure activity, surveillance of antiepileptic drug treatment⁴, and prediction of cerebral outcome after birth asphyxia^5,6,7,8,9. Infants with birth asphyxia who did not show severe suppression of background activity and seizure activity had a more favorable outcome if they were cooled⁸, but research on this topic is still ongoing^10,11,12. Over the past 30 years, cerebral function monitoring in neonates has become more widespread in NICUs¹³. Nowadays, it is increasingly being used in the preterm infant population. aEEG has been proven to be a safe method for cerebral function monitoring, even in extremely preterm infants, and is generally well-accepted by NICU staff¹⁴. Several studies showed a correlation between early aEEG recordings and neurodevelopmental outcomes in preterm infants^{15,16,17,18,19,20}.

aEEG is based upon conventional electroencephalography that is recorded with two or four scalp electrodes, depicting the amplitude of the raw EEG on a time-compressed semi-logarithmic scale¹. The signal from two or four electrodes placed in the C3, P3, C4, and P4 positions of the international 10-20 system is passed through a bandpass filter, which enhances frequencies between 2 and 15 Hz. Frequencies under 2 Hz and above 15 Hz are attenuated in order to eliminate artefacts, such as sweating, movement, muscle activity, and electrical interference, as much as possible^1,4. Further processing includes filtering, rectification, smoothing, semi-logarithmic amplitude compression, and time compression. Amplitudes <10 µV are displayed on a linear scale and amplitudes >10 µV on a logarithmic scale²¹. The lowest-detected amplitude is shown as the lower border, and the highest amplitude is shown as the upper border²¹. By this means, even small changes in the lower amplitude remain visible, while an overloading of the display at high amplitudes is avoided²¹ (Figure 1). Due to time compression, 5 – 6 cm on the time scale represents 1 h, thus making the review of brain activity for hours and even days possible^1,4,13.

The visible information in the aEEG tracing is limited to changes of the amplitude. Modern devices offer the possibility of viewing the raw EEG, so the frequency and morphology of the raw EEG curve can also be considered for interpretation. This helps to distinguish between artefacts and real seizure activity during suspicious sections of the aEEG band⁴. Some aEEG devices can record a simultaneous video of the patient for even better identification of seizures and artifacts. Electrode impedance is monitored during the entire recording²¹. In two-channel aEEG devices that use four electrodes, the investigator can switch between two intraparietal curves or one transcerebral curve (Figure 2). Depending upon the manufacturer, software offers additional features, like seizure detection, burst rate analysis, electromyography, etc. It is also possible to derive an aEEG from a full-channel EEG device that offers video recording, electromyography, electrooculography, electrocardiography, etc.

The reduction of the electrophysiological information and time compression makes continuous monitoring and bedside interpretation possible without requiring specific knowledge about EEG. Because of the long recording time, even subclinical seizure activity can be detected, which would otherwise remain unnoticed^4,22 because conventional EEG monitoring for very long time periods is not available to date in NICUs. It must be considered, though, that not all pathological alterations, like seizures, are found due to the small area of brain surface covered by the recording¹³. Therefore, aEEG is not meant to replace conventional EEG, but to complement it¹³.

Electrocortical activity, as reflected by the aEEG background pattern, changes according to the infant's gestational age^4,23,24,25. In term infants and late preterm infants, the background pattern is mainly continuous with a lower amplitude above 5 µV⁴. During quiet sleep, the background pattern becomes more discontinuous²⁶. In very preterm infants, the dominating background pattern is discontinuous: episodes of high activity (i.e., high-amplitude bursts) alternate with episodes of low-amplitude activity²⁷. This physiological pattern must be distinguished from a burst suppression pattern, which is pathological²⁷. With increasing gestational age, aEEG and background patterns become more continuous, and the duration of continuous activity increases^27,28,29. Developing and existing pathological conditions can also be displayed by the aEEG tracing (e.g., the development of an intraventricular hemorrhage and periventricular leukomalacia is associated with acute disturbances in the background activity)^30,31. Severe meningitis can cause a flat trace.

The qualitative interpretation of aEEG generally includes three categories: classification of the background pattern, sleep-wake cycling, and the presence of seizures. Several authors have made suggestions for classifications and scores that describe brain maturation^16,21,24,25. The quantitative analysis of aEEG is less common, even though it is possible in modern devices, and few research groups made use of this approach^32,33,34. We would like to briefly present 3 different approaches to the qualitative and semi-quantitative assessment of aEEG tracings:

Hellström-Westas:²¹

The assessment of the tracing is solely qualitative, and the results are not transformed into a score. The classification allows for the description of pathological conditions. Normative values for gestational ages have been published to help interpret whether a pattern is adequate for the age²¹: (1) background patterns: continuous normal voltage (physiological), discontinuous normal voltage (physiological in preterm infants), burst suppression pattern (pathological), continuous low voltage (pathological), and flat trace (pathological); (2) sleep-wake cycling: none, imminent, mature (physiological/pathological, depending on the infant's age); and (3) seizure activity: none, single seizures, repetitive seizures, and status epilepticus.

Burdjalov:²⁵

The approach of this classification is the qualitative assessment of the tracing and its transformation into a score. The score rises with gestational age, and normative score values for each corresponding gestational age have been published: (1) 0 – 2 points for continuity, (2) 0 – 5 points for sleep-wake cycling, (3) 0 – 2 points for the amplitude of the lower border, (4) 0 – 4 points for the bandwidth, and (5) 0 – 13 points for the total score.

Olischar/Klebermass:^16,24

Percentiles regarding the percent duration of background patterns (i.e., discontinuous normal voltage, discontinuous low voltage, and continuous normal voltage) and burst rate were developed for gestational ages. Tracings are evaluated for an age-adequate background pattern, the presence of sleep-wake cycling, and the presence of seizure activity (i.e., repetitive seizures or status epilepticus). Then, the tracings are classified into a graded score as follows: (1) normal aEEG (all categories normal), (2) moderately abnormal (1 out of 3 categories classified as abnormal), and (3) severely abnormal (2 or 3 out of 3 categories classified as abnormal). This score has been shown to have a predictive value for neurodevelopmental outcome at 3 years of corrected age.

Changes in the aEEG tracing are caused by numerous extracortical factors, such as changes in cerebral blood flow, medication (e.g., opiates, sedatives, and caffeine), acidosis, changes in carbon dioxide tension, clinical conditions (e.g., hypogylcemia, sepsis, meningitis, and patent ductus arteriosus), etc.^{21,32,35,36,37,38}. The aEEG band itself is rather insensitive to changes of impedance, but significant changes are observed in terms of electrode distance and localization³⁹. Artifacts may pose a problem for interpretation: absolute values of the amplitude change as a consequence of scalp edema or interelectrode distance^39,40. Interferences caused by ECG, high-frequency oscillation ventilation, muscle activity, infant movement, or handling may result in a lower border increase⁴⁰. In modern devices, this can be partially avoided by the simultaneous recording of raw EEG and aEEG and by marking the beginning and end of handling. Liquids (e.g., sweat or ultrasound gel) can lead to connections between electrodes, feigning a flat trace pattern. Approximately 12% of the recording time in long-term aEEGs are altered due to artifacts, 55% caused by electrical interferences and 45% being movement artifacts⁴¹.