Neonatal seizure recognition and bedside amplitude-integrated EEG in the Australian NICU

Neonatal seizures remain one of the most challenging clinical events in any NICU, particularly in preterm infants whose fragile physiology can mask what would otherwise be obvious neurological signs in older children. Across Australian perinatal centres from Brisbane to Perth, bedside clinicians frequently encounter situations where a baby simply "looks off" yet the standard observation chart offers few objective clues. Electrographic seizures in this age group are commonly clinically silent or present only with subtle signs such as eye deviation, lip smacking, or autonomic fluctuation, which can easily be attributed to non-convulsive movement or simply prematurity itself.

Continuous conventional EEG remains the diagnostic gold standard, yet access to dedicated paediatric neurophysiologists and the wiring required for full montage recordings is concentrated in major tertiary centres. This gap has driven the widespread uptake of amplitude-integrated EEG, often shortened to aEEG, as a pragmatic bedside tool for seizure screening, treatment titration, and brain monitoring over hours or days. Modern devices pair two or four scalp electrodes with software that compresses raw EEG into a narrow impedance-based tracing that even a busy junior doctor can interpret at a glance.

In Australia, geography shapes the conversation. With vast distances between regional birthing centres and tertiary NICUs, retrieval services such as NETS NSW and the Victorian Neonatal Retrieval Service routinely transport encephalopathic newborns after initial resuscitation. Bedside aEEG performed during stabilisation, and again on arrival, helps receiving teams stratify risk before formal MRI and long-term video EEG can be arranged. For clinicians working in Darwin, Townsville, or rural Western Australia, a compact aEEG channel can transform triage decisions that previously depended on intermittent clinical checks by exhausted registrars.

Within national training frameworks and professional bodies like PSANZ and the Australian College of Neonatal Nurses, competency in aEEG interpretation is increasingly expected of neonatal fellows and senior nurses managing high-acuity cots. Yet interpretation still varies between units, particularly when traces fall outside the usual teaching repertoire. The following sections walk through how seizures hide in plain sight, how the aEEG trace is constructed, how to read common seizure patterns, and where Australian teams can sharpen their day-to-day practice.

Why neonatal seizures are easy to miss clinically

In the first 28 days of life, the immature brain rarely produces the generalised tonic-clonic activity seen in older children or adults. Seizures instead manifest as fragments: brief automatisms, cycling movements, brief apnoeas, or sudden changes in heart rate and blood pressure that overlap with common neonatal physiology. A term infant with hypoxic-ischaemic encephalopathy on therapeutic cooling at a Sydney quaternary centre may have exclusively subclinical electrographic seizures, meaning that without a brain tracing the treating team would never know a seizure was occurring.

Under-recognition is particularly acute in preterm cohorts cared for in level 6 nurseries. Background movement, ventilator artefact, and sedation all obscure the clinical examination. Studies consistently show that nurses report only a minority of electrographic events, with sensitivity often below 30 percent. For Australian retrieval teams crossing long distances with a sick neonate, missing seizures during transit is a recognised concern, especially when staff are simultaneously managing airway, ventilation, and inotropes.

This is where amplitude-integrated EEG earns its place. By running continuously from a small electrode pair, the device captures what clinical observation cannot, providing a real-time window into cortical activity that the bedside team can act upon within minutes rather than waiting hours for a formal reading.

How aEEG works at the bedside

A standard two-channel aEEG relies on two pairs of needle or paste electrodes placed in a cross-cerebral configuration, often P3-P4 and C3-C4 referenced to the contralateral shoulder or mastoid. The raw signal is filtered, typically between 2 and 15 Hz, then rectified and time-compressed so that each vertical pixel of the display represents a short interval of the underlying voltage. The result is a narrow band tracing, usually plotted over several hours on a single screen, with the upper and lower margins representing the smoothed amplitude envelope.

Modern devices also display the raw EEG channel beneath the compressed trace, which is essential for distinguishing true seizure activity from artefact. Most contemporary machines used in Australian NICUs incorporate automated seizure detection algorithms that flag suspected events, but these remain adjuncts rather than stand-alone diagnostics. Interpretation still depends on the clinician's ability to confirm the band-like background, recognise the abrupt rise in lower margin that characterises an electrographic seizure, and exclude common confounders such as patting, sucking, or high-frequency ventilation.

In practice, electrode placement takes around five minutes once the team is familiar, and electrodes can stay in place through most routine nursing care. This portability has made aEEG a fixture in cooling protocols at the Royal Women's Hospital in Melbourne, the Mater Mother's in Brisbane, and similar tertiary hubs, where hourly assessment of background activity feeds directly into prognosis discussions with families.

Reading common seizure patterns

Classic neonatal seizures on aEEG show a sudden rise in the lower border of the trace accompanied by a corresponding narrowing of the upper border, producing a transient saw-tooth or pedestal elevation that lasts at least ten seconds. Status epilepticus can be suspected when this pattern persists or recurs with gaps shorter than the seizure-free interval defined in local guidelines. Single, brief events in an otherwise normal background carry a different implication from recurrent seizures arising from a severely suppressed trace.

Equally important is the background pattern between events. A normally structured background with sleep-wake cycling is reassuring and supports a more conservative approach. Continuous low voltage, burst-suppression, or a flat trace suggest a more ominous substrate where any seizure heralds significant encephalopathy. Australian teams commonly use background classification schemes such as the Toet or Hellström-Westas criteria, with local adaptation to account for the specific populations seen in units like King Edward Memorial or the Royal Brisbane and Women's Hospital.

Subtle artefacts deserve explicit attention. Patting for comfort, endotracheal suction, and even phototherapy lights can produce sharp transients that mimic seizures, which is why the raw channel should always be reviewed before treatment escalation. Conversely, brief seizures that fall below the display resolution can be missed, reinforcing that aEEG is a screening tool rather than a replacement for full montage EEG when diagnostic certainty is required.

Integrating aEEG with clinical assessment

A useful bedside workflow combines structured clinical assessment with continuous aEEG, allowing team members in Perth, Hobart, or suburban Sydney to escalate findings consistently. A practical sequence begins with confirming electrode integrity and signal quality, then assessing background pattern and sleep-wake cycling, and only then focusing on event detection. Documenting each of these elements hourly creates a longitudinal record that consultants can review at handover or during teleconference rounds with regional referral centres.

Clinical context still leads the picture. A sudden rise in upper margin accompanied by confirmed desaturation, tachycardia, and limb stiffening on nursing observation is far more compelling than a similar trace change in a stable, spontaneously breathing infant. Australian guidelines increasingly recommend dual review, with one team member interpreting the trace while another reviews the infant, particularly when loading phenobarbitone or considering second-line anticonvulsants such as levetiracetam or a midazolam infusion.

Communication with families is also strengthened by aEEG. Showing parents a six-hour trace where seizures resolved and background normalised can transform what feels like abstract neurology into visible recovery, an approach several neonatal units across the country now embed in family update meetings.

Training and workforce considerations

Workforce training is the limiting factor in many Australian centres. While paediatric neurology input is concentrated at the Royal Children's, the Children's Hospital at Westmead, and equivalent paediatric hubs, most level 5 and 6 NICUs rely on neonatologists and senior nurses to provide out-of-hours aEEG interpretation. Even though fellowship programs expose trainees to the technology, ongoing bedside exposure varies considerably, particularly for nurses rotating across general paediatrics and neonates.

Practical solutions include regular multidisciplinary case reviews, joint tele-interpretation sessions with tertiary neurophysiologists, and short simulation modules focused on recognising common patterns and artefacts. Several state-based retrieval services have built mandatory aEEG competencies into their induction packages, recognising that accurate interpretation during retrieval directly affects downstream management decisions in the receiving unit. Sustained pattern recognition under high cognitive load also leads to fatigue during overnight shifts, which is why hospital wellbeing programs sometimes circulate guidance on post-holiday recovery steps to help clinical staff return alert and attentive after extended leave.

Limitations and common pitfalls

Even experienced teams encounter pitfalls. Electrode displacement during handling or during retrieval transit produces high-impedance periods that interrupt the trace, sometimes coinciding with seizure activity that therefore goes unrecorded. Sedation, particularly with morphine or midazolam infusions used during cooling, suppresses background activity and can hide both seizures and emerging abnormalities, complicating prognostication at 24 and 48 hours.

Automated seizure detection algorithms are improving rapidly, yet false positive rates remain clinically significant, especially when EMG artefact from facial muscles or ventilator-driven chest movement crosses into the signal band. Treating every flagged event as a seizure leads to unnecessary anticonvulsant exposure, with implications for blood pressure, respiratory drive, and later neurodevelopment. The discipline of confirming each event on the raw channel, and aligning it with documented clinical change, cannot be substituted by software alone.

Another pitfall is reading the trace in isolation. Background recovery, timing relative to insult, and adjunctive imaging, particularly MRI on day 5 to 14, contribute far more to counselling families about long-term outcome than any single aEEG snapshot taken in isolation.

Future directions and Australian research priorities

Research priorities in Australia echo international efforts but are tailored to local realities. A key theme is optimising aEEG use during neonatal retrieval, with recent collaborations between NETS NSW and tertiary neurophysiology teams exploring lightweight single-channel devices suitable for road and fixed-wing transport. Standardised protocols for electrode placement, signal quality assurance, and interpretation hand-off are being piloted, with the goal of producing nationally consistent documentation that any receiving team can trust.

Another active area is the integration of aEEG with quantitative metrics such as spectral power and connectivity indices, which may help detect subtle injury patterns missed on visual review. Multi-centre registry work coordinated through PSANZ is exploring how early aEEG trajectories correlate with two-year neurodevelopmental outcomes, building a stronger evidence base for parent counselling.

The equipment itself is also evolving. Newer devices combine aEEG with multi-channel conventional EEG capability, opening the possibility of upgrading from screening to definitive diagnosis without changing the electrode set. For remote sites such as Cairns or Alice Springs, such hybrid capability may eventually reduce the diagnostic gap that currently relies on transferring the infant, or even the recordings, across vast distances to specialist review.

Comparison of common aEEG devices available in Australia

These figures are indicative only and vary with contracts, consumables, and service packages negotiated by individual hospitals. For procurement decisions, clinicians in Adelaide, Perth, or regional NSW often pilot devices in parallel for a quarter before standardising across the unit.

Looking ahead, neonatal teams across Australia have a practical opportunity to embed structured aEEG interpretation into daily rounds, retrieval handovers, and family communication. Building local expertise, sharing case material through PSANZ-affiliated networks, and contributing data to national registries will sharpen recognition of the seizures that currently hide in plain sight. With deliberate practice, the bedside trace becomes not just a monitor but a shared language between neonatologists, nurses, retrieval staff, and the families whose babies depend on every team member catching the silent signals early. Related discussion on progesterone and preterm birth and the considerations raised around perinatal palliative care sit within the same continuum of perinatal medicine, where early neurological recognition overlaps with decisions about prognosis and family-centred goals.