07 Neonatal EEG II: abnormal background, seizures and NICU monitoring

The neonatal background is the single most informative part of the record, and the ACNS standardized terminology gives it a vocabulary that every reader in the unit can share (Tsuchida 2013). The categories are defined by voltage and by the length of the interburst interval, which is why the maturation tables from Module 6 and the sensitivity setting from Module 1 both have to be known before a grade can be given.

Reading the table

The first two rows are descriptions, not verdicts. A discontinuous background is normal at many postmenstrual ages and in quiet sleep at term, as tracé alternant. It becomes excessively discontinuous only when the interburst intervals are too long or too low for the age, and the terminology deliberately sets no numeric cutoff, because the cutoff is the age-specific limit from the maturation tables. The bursts in an excessively discontinuous record still contain some normal features.

Burst suppression is different in kind, not degree. The bursts are abnormally composed, the interburst voltage is below 5 µV, and, most tellingly, the pattern is invariant: it does not change with state and does not react to stimulation. Reactivity, which Module 6 showed becomes reliable by 35 to 36 weeks, is the feature that most reliably separates burst suppression from tracé alternant in a term baby. The two low-voltage categories describe a continuous record whose amplitude is too low: borderline low voltage keeps some normal features at 10 to under 25 µV, and low voltage suppressed has none, with a baseline below 10 µV.

Grade the background in every state that was recorded, note whether it reacts, and state the postmenstrual age against which the grade was made. The grade is the part of the report that the NICU team will act on most, and the sign-off asks the resident to grade five neonatal backgrounds in agreement with the faculty read.

A seizure on EEG is recognized by evolution, and in the neonate the ACNS terminology turns that into a definition with two numbers. A neonatal electrographic seizure is a sudden, repetitive, evolving pattern of at least 2 µV lasting at least 10 s (Tsuchida 2013). An evolving rhythmic pattern that lasts under 10 s is a brief rhythmic discharge (BRD), named separately so that it is neither ignored nor counted as a seizure. Neonatal status epilepticus is seizures summing to at least 50% of any 1-hour epoch.

How this differs from older patients

The pediatric and adult definitions in Module 10 are stricter about what qualifies and looser about how much is needed for status. An electrographic seizure in an older patient needs either discharges averaging more than 2.5 Hz or definite evolution, for at least 10 s; the neonatal definition asks only for evolution and 10 s. Status in the older patient is 10 continuous minutes or 20% of an hour; in the neonate the threshold is 50% of the hour. A resident who reads both populations must keep the two sets of numbers apart, and the commonest error is to apply the 20% rule to a neonate and call status where the terminology does not.

Reading evolution

Evolution means a change in frequency, morphology or location across the run, not merely rhythmicity. Rhythmic activity that begins and ends without changing is the pattern Module 3 warned about: the ventilator, patting, sucking. A neonatal seizure characteristically begins as rhythmic activity at one location, slows or speeds, changes shape, and spreads to a neighbouring region before it stops. Because the ACNS definition needs no clinical accompaniment, an evolving run with nothing visible on video is a seizure in full, an electrographic-only seizure, and the next sections explain why these are the majority in some populations.

Measure the run against the time marks, in the state in which it occurred, and describe onset, evolution and offset. A 40-second run at C3 that speeds from 1 Hz to 2 Hz and spreads to T3 is a seizure whether or not the baby moved.

The EEG definition tells you that a seizure happened; the ILAE classification tells you what to call it. The 2021 modification of the ILAE framework for the neonate makes four points that change how a reader works (Pressler 2021).

Video-EEG is the gold standard for diagnosis. The classification is built on the recording, not on what a bedside observer saw, and the next section shows why.

Neonatal seizures are treated as focal in onset. There is no focal-versus-generalized split in the neonatal scheme. Every neonatal seizure is described from its site of onset on the EEG, which is another reason the localization skills from Module 1 are needed in the NICU.

Seizures are typed by their predominant clinical feature. The categories are motor (automatisms, clonic, epileptic spasms, myoclonic, tonic), non-motor (autonomic, behaviour arrest) or sequential, when several features follow one another. The type is read from the video that runs alongside the EEG.

Electrographic-only seizures are a category of their own. This is the point with the largest clinical weight. The paper cites 50% to 80% of seizures being electrographic-only in selected populations with hypoxic-ischemic encephalopathy. In other words, in the babies most likely to be monitored, most seizures have no visible expression at all.

What changes at the desk

The classification means that a neonatal seizure report has two parts: the EEG description (onset, evolution, offset, duration, and burden over the hour) and the clinical type from the video, which may be "electrographic-only". It also means that the absence of clinical signs, far from arguing against seizures, is the expected finding in encephalopathic neonates. The reader who waits for the baby to move before calling a seizure will miss most of them.

The case for monitoring rests on one study that every resident should be able to quote. In a study of 51 monitored neonates, 526 electrographic seizures occurred in 9 babies. Only 34% of those seizures had any clinical sign on video, and staff recognized 9%. The error ran both ways: of 177 events that staff suspected were seizures, only 27% had an EEG correlate (Murray 2008).

Reading those numbers

Two-thirds of the seizures had no visible expression at all, which matches the ILAE's 50% to 80% electrographic-only figure in encephalopathic populations (Pressler 2021). Of the third that did show something, most were still not recognized by experienced staff, so fewer than one seizure in ten was identified at the bedside. And nearly three-quarters of the movements that staff did call seizures were not seizures. Bedside observation therefore both undercounts and overcounts, and a treatment plan based on it will give antiseizure medication to babies who are not seizing while leaving most of the seizure burden untreated in babies who are.

What follows for the reader

The EEG is not confirming a clinical diagnosis; it is the diagnosis. That has three consequences. First, the reader must review the whole record, not the epochs around the nurses' event marks, because the seizures are mostly elsewhere. Second, the video must be checked for every EEG seizure, to type it and to exclude the rhythmic artifacts of Module 3, but the absence of movement never argues against a seizure. Third, the reader owes the team a seizure burden, expressed as minutes per hour, because that is the number the definition of status depends on and the number that tracks the response to treatment. The sign-off asks the resident to identify seizures on two continuous EEG studies in agreement with the faculty read; the Murray data are the reason that skill cannot be delegated to the bedside.

Because most neonatal seizures are invisible at the bedside, the question is not whether to record but whom and for how long. For high-risk neonates, the ACNS guideline recommends 24 hours of conventional EEG to screen for seizures, at least 1 hour to assess the background, and, after a seizure, monitoring until 24 hours seizure-free (Shellhaas 2011). A 2025 ACNS guideline updates the indications; its recommendations are conditional, because the evidence is of very low certainty (Wusthoff 2025).

The three numbers, and what each is for

Using the definitions together

Over a monitored day, the reader grades the background in each state, identifies every run that meets the seizure definition, sums the seizure time in each hour, and reports whether any hour reached the 50% threshold for neonatal status epilepticus (Tsuchida 2013). A team reading that report can see both the trajectory of the background, which carries prognostic weight, and the seizure burden, which guides treatment. The ACNS terminology PDFs linked below are short; read the neonatal one in full before the first NICU read, and keep the definitions table beside the screen.

Amplitude-integrated EEG

Many units run amplitude-integrated EEG (aEEG) at the bedside. It is a compressed trend, useful for spotting a change in background or a candidate seizure over hours, but a trend is not a diagnosis. Every event seen on a trend must be confirmed on the raw EEG, a rule Module 11 applies to all quantitative EEG. The Calgary course linked below has lessons on both the seizure read and aEEG.