Continuous EEG (cEEG)

1. Bedside vignettes: why this matters in the PICU#

Vignette A. Post-arrest day 1, IIC or seizure?#

A 6-year-old post out-of-hospital VF arrest, ROSC at 28 minutes, targeted temperature 33°C. Day 1: GCS 3 sedated; aEEG shows discontinuous pattern with intermittent bursts. cEEG (10–20 system) shows lateralised periodic discharges at 1.5 Hz with sharp morphology over the right hemisphere, lasting 30 seconds and recurring every 3 minutes. The bedside team asks: is this NCSE (treat) or the ictal-interictal continuum (less clear-cut)? Per ACNS 2021, frequency > 2.5 Hz with evolution is electrographic seizure; this is IIC. The team trials a midazolam bolus, the LPDs resolve, and the team interprets the resolution as functional evidence the LPDs were ictal: continue antiseizure therapy. 1 2 3

Vignette B. Refractory status epilepticus in a 5-year-old, the drug ladder#

A 5-year-old with prolonged febrile seizure that has continued for 30 minutes despite two doses of midazolam IM and IV lorazepam. cEEG confirms ongoing generalised electrographic seizure. Per ESETT, second-line is levetiracetam 60 mg/kg IV, fosphenytoin 20 mg PE/kg IV, or valproate 40 mg/kg IV (all three equivalent in efficacy). The team chooses levetiracetam (least sedating). The clinical seizure stops; cEEG shows resolution of the electrographic seizure 10 minutes later. The team continues cEEG for 24 hours and detects two further brief electrographic seizures without clinical correlate, which they treat. Without cEEG, those NCSE episodes would be missed. 4 5 6

Vignette C. SAH day 6, qEEG alpha-delta ratio falls 6 hours before clinical signs of DCI#

A 16-year-old with aneurysmal SAH post-coiling, day 6. cEEG running since day 2. Quantitative trends show the alpha-delta ratio (ADR) falling from 0.55 to 0.35 over 6 hours, which is statistically significant per the centre's algorithm. TCD MFV is rising in parallel but Lindegaard ratio is only 2.8 (borderline). The team escalates haemodynamics, repeats angiography (which confirms moderate vasospasm in the right MCA), and the patient is treated successfully. The clinical examination remained unchanged during the 6-hour ADR decline; qEEG was the earliest signal. 7 8 9

2. What EEG is, and what it is not#

EEG records the summated post-synaptic potentials of cortical pyramidal neurons, measured at the scalp via differential amplification between electrode pairs. The signal amplitude at the scalp is ~10–100 μV, attenuated by the skull and scalp by a factor of ~100 from the underlying cortical surface.

Five things follow.

EEG samples cortex only. Deep structures (basal ganglia, thalamus, brainstem) are not directly seen on scalp EEG. Subcortical seizures (e.g., from a deep mesial focus) may show only as secondary cortical propagation.

EEG is highly state-dependent. Sedation, hypothermia, neuromuscular blockade, hypoglycaemia, and metabolic encephalopathy all change the baseline. Reading EEG requires knowing the clinical state.

EEG distinguishes NCSE from other altered-consciousness causes. This is its most uniquely valuable contribution: NCSE causes ~20% of comatose ICU patients' altered consciousness and is invisible without continuous monitoring.

EEG-based SE management has trial evidence. ESETT established second-line drug equivalence; third-line (continuous-infusion) endpoints remain protocol-driven. Endpoint definitions (electrographic seizure resolution, burst suppression) are EEG-based.

EEG is not seizure-detector only. Background continuity, reactivity, and pattern (e.g., GPD, LPD) carry independent prognostic and diagnostic information.

3. Montage and electrode placement#

Electrode placement. The 10-20 international system places electrodes at proportional distances (10% or 20%) along the head circumference. Standard adult ICU montage uses 19 electrodes plus a reference. Pediatric NICU/PICU montages often reduce to 9 or 11 electrodes (less coverage but faster placement, less coverage but better fit on small heads).

Montages. Two principal display modes:

• Bipolar (longitudinal "Banana"): each channel shows the difference between two adjacent electrodes. Useful for localising focal patterns; "phase reversal" in bipolar identifies the seizure focus.
• Referential (average reference): each channel shows one electrode against an average of all others. Useful for global pattern recognition (burst suppression, generalised discharges).

Bedside teams often run both simultaneously and switch view depending on the question.

Pediatric considerations. Neonatal EEG uses a reduced montage (typically Fp1, Fp2, C3, C4, T3, T4, O1, O2, Cz; 9 electrodes) per ACNS 2013 neonatal guidelines. Head size, scalp thinness, and the need to balance coverage against placement time guide the choice. 13 10

4. The signal: bands, rhythms, reactivity, continuity#

A bedside EEG reading covers four dimensions:

Frequency bands (the basic vocabulary):

Rhythms: posterior dominant rhythm (PDR) is the canonical awake-resting alpha over occipital cortex; loss of PDR is a non-specific marker of encephalopathy.

Reactivity: does the background change when the patient is stimulated (loud noise, painful stimulus)? Reactive background is favourable; unreactive is concerning regardless of the resting frequency.

Continuity: is the background continuous (always present) or discontinuous (bursts separated by silence)? Severely discontinuous patterns (suppression, burst-suppression) reflect deep encephalopathy or anaesthesia.

5. ACNS 2021 terminology: GPD, LPD, RDA, BIRDS, IIC#

The ACNS 2021 standardised terminology is the modern reference for ICU EEG reading. The key categories:

Seizure criteria (ACNS 2021): rhythmic activity > 2.5 Hz with evolution in frequency, location, or morphology lasting ≥ 10 seconds; OR rhythmic activity ≤ 2.5 Hz with clinical correlate; OR repeated rhythmic activity meeting threshold. 1 6

The IIC dilemma: many patterns (LPDs at 1–2 Hz, GPDs at 1.5 Hz, rhythmic delta) are below the seizure cutoff but may still drive metabolic distress. The bedside approach: trial of antiseizure medication; resolution of the pattern and clinical/electrographic improvement supports an ictal interpretation. Non-resolution suggests interictal or post-ictal pattern.

6. Pattern library: what to recognise on cEEG#

7. Try it: interactive widgets#

8. Status epilepticus management section#

The modern SE ladder is based on trial evidence (ESETT, RAMPART). EEG is used at three points: (1) confirming initial diagnosis (especially NCSE), (2) titrating second- and third-line therapy, (3) defining endpoints (electrographic seizure resolution; burst suppression for super-refractory).

8.1 First-line: benzodiazepines#

• Midazolam 0.2 mg/kg IM (max 10 mg) as first-line therapy (AES 2016 guideline; the efficacy of the IM route was established by RAMPART, which used fixed autoinjector doses of 10 mg if > 40 kg or 5 mg if 13 to 40 kg), or lorazepam 0.1 mg/kg IV (max 4 mg). 4 14 15
• Repeat once if seizure continues at 5 minutes.

8.2 Second-line: ESETT-equivalent#

ESETT (Kapur 2019) compared levetiracetam 60 mg/kg, fosphenytoin 20 mg PE/kg, valproate 40 mg/kg. All three were equivalent at 60 minutes. Choose by patient context:

• Levetiracetam: least sedating, no hepatic interactions; often first choice.
• Fosphenytoin: established, but cardiac monitoring required.
• Valproate: avoid in hepatic disease, suspected metabolic disease, age < 2 years.

8.3 Third-line (refractory SE)#

Continuous IV anaesthetic infusion:

• Midazolam 0.2 mg/kg bolus, then 0.05–2 mg/kg/h infusion.
• Propofol 1–2 mg/kg bolus, then 1–4 mg/kg/h (caution in children, risk of PRIS).
• Pentobarbital 5–15 mg/kg load, then 1–5 mg/kg/h (super-refractory; reserved when above fails).

EEG endpoint: resolution of electrographic seizures for refractory; burst suppression (typically 60% suppression ratio for 24 hours) for super-refractory.

9. Clinical contexts: cEEG across acute brain injuries#

9.1 Convulsive and non-convulsive status epilepticus#

The canonical indication. NCSE causes ~20% of comatose ICU patients' altered consciousness. cEEG is the only way to detect it. ESETT defines the modern second-line drug ladder. 4 5 16 17 7

9.2 HIE and post-cardiac arrest neuroprognosis#

Post-arrest cEEG patterns carry strong prognostic information:

• Continuous, reactive background within 24–48 h: favourable.
• Discontinuous, unreactive, low-amplitude: unfavourable.
• Burst suppression with identical bursts: highly unfavourable.
• Status epilepticus, especially myoclonic SE: unfavourable but not invariably fatal.
• Isoelectric at > 72 h: very poor.

The AHA 2020 pediatric post-arrest guidelines and Topjian 2019 pediatric statement endorse cEEG as part of routine post-arrest neuroprognostication. 2 18 3 19

9.3 SAH and DCI (qEEG)#

Quantitative EEG trends, particularly the alpha-delta ratio (ADR), detect DCI hours before clinical signs in adult SAH cohorts. Pediatric SAH data are sparse but the principle applies. The Foreman 2012 review summarises the qEEG-for-DCI evidence. 20 8

9.4 Bacterial meningitis with seizures#

cEEG is recommended when seizures are suspected (subtle motor signs, fluctuating consciousness) or in severely encephalopathic patients per European meningitis and IDSA encephalitis guidelines. 21 22

9.5 Sepsis-associated encephalopathy#

Sepsis-associated encephalopathy presents with delirium, altered consciousness, and a spectrum of EEG findings from diffuse slowing to GPD and NCSE. cEEG detects NCSE in 10–20% of comatose septic patients. 1 17

9.6 Severe TBI: subclinical seizures#

cEEG detects subclinical / non-convulsive seizures in 15–25% of severe TBI patients in the first 7 days, particularly in those with intra-axial haemorrhage or cortical contusion. PBTF 4 pediatric guidelines recommend cEEG for severe TBI with GCS ≤ 8 or for refractory ICP suggestive of seizure-driven elevation. 23 24 6

9.7 Neonatal seizures#

Neonatal seizures are often subtle or invisible clinically. The 2017 ILAE neonatal seizure classification (Pressler 2021) uses cEEG as the diagnostic standard. Sansevere 2019 Pediatric Neurology cEEG review summarises modern neonatal cEEG practice. aEEG remains the bedside trend tool, but cEEG (10–20 system, ideally video) is the diagnostic standard. 10 11 13

9.8 ECMO#

cEEG on ECMO detects seizures (~20% of pediatric ECMO patients) and provides background continuity / reactivity for ongoing prognostication. Movement artefact from circuit and ventilator is a recurrent challenge. Pisano 2024 (a review of ECMO neurologic complications) supports cEEG on neuro-ECMO indications. 25 26

10. Multimodal integration: EEG in the MMM/MNM stack#

The qEEG alpha-delta ratio as an early marker of delayed cerebral ischaemia in SAH is established by Claassen 2004 and extended by Foreman 2012. 7 8

11. Setup and technique: a step-by-step#

11.1 Electrode placement (10-20 system)#

Measure scalp landmarks (nasion, inion, preauricular points). Place electrodes at proportional 10% / 20% distances. Standard adult ICU montage uses 19 electrodes (Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T3, T4, T5, T6, Fz, Cz, Pz) plus reference and ground.

Pediatric reduction (9 electrodes): Fp1, Fp2, C3, C4, T3, T4, O1, O2, Cz. Trades coverage for placement time and fit on small heads.

11.2 Skin preparation#

• Mark the position, abrade lightly with prep gel.
• Apply conductive paste or gel.
• Place the electrode, secure with collodion or adhesive disk.
• Verify impedance < 5 kΩ ideally (< 10 kΩ acceptable).

11.3 Filters and sampling#

• Sample rate: 200–512 Hz typical for ICU cEEG; > 256 Hz for high-frequency oscillation work.
• Low-frequency filter (LFF): 0.5–1 Hz (preserve delta).
• High-frequency filter (HFF): 70 Hz (reject muscle/EMG).
• Notch filter: 50 or 60 Hz mains.

11.4 Artefact reduction#

• Common artefacts: EMG (muscle), ECG (cardiac), eye movement, electrode pop, ventilator, sweat.
• EMG: paralyse only when essential; otherwise filter and accept some contamination.
• Eye movement: dominant in frontal channels; usually obvious morphology.
• Electrode pop: sudden high-amplitude spike on a single channel; re-secure the electrode.
• Ventilator: rhythmic at ~12–20/min; isolated by frequency.

11.5 Video EEG#

Continuous video alongside EEG is the standard for SE management and seizure semiology characterisation. The video stream captures eye deviation, automatisms, head turn, and post-ictal state, all of which add diagnostic information.

11.6 qEEG trend display#

Modern ICU EEG systems compress 24-hour traces into:

• Spectrogram: colour map of frequency power over time (per electrode).
• Alpha-delta ratio: ratio of 8–12 Hz to 1–4 Hz power; falls in DCI.
• Suppression ratio: percentage of time with EEG below an amplitude threshold; rises in deep anaesthesia.
• Asymmetry index: left-vs-right power; flags unilateral pathology.

12. Pitfalls and artefacts#

• Under-reading: missing GPD or NCSE under deep sedation; the encephalopathy can mask the pattern.
• Over-reading: calling RDA or GPD an electrographic seizure without meeting ACNS criteria.
• Inter-rater variability: even among trained epileptologists, agreement on IIC patterns is moderate.
• Sedation-induced patterns: propofol-induced burst suppression looks like post-anoxic burst suppression; clinical context is essential.
• Movement artefact: ventilator, ECMO circuit, CRRT.
• Montage choice obscuring focal patterns: bipolar may miss a true generalised pattern; always verify with referential.
• Electrode dislodgement: long ICU recordings; verify impedance every shift.
• Salt-bridges: conductive paste connecting adjacent electrodes; falsely suggests synchrony.
• Pseudo-seizure artefact: chewing, shivering, eye flutter; verify with video.
• Pediatric montage limits: 9-electrode neonatal montage will miss some focal patterns; accept the trade-off for placement feasibility.

13. Combine with…#

• qEEG: for the alpha-delta ratio, spectrogram, suppression ratio.
• aEEG: for the compressed bedside trend (NICU).
• BIS: for operating-room sedation depth.
• Evoked potentials: for subcortical and brainstem function.
• TCD: for the NCSE-vs-spasm discrimination in SAH.
• NIRS: for the rSO2 reactivity pair in HIE.
• ICP: for the seizure-driven ICP spike pair.

14. Evidence summary and recent literature#

14.1 Evidence summary#

14.2 Recent literature (2022–2025)#

• ACNS 2021 standardised terminology (Hirsch): the modern reference for ICU EEG reading; GPD, LPD, RDA, BIRDS, IIC defined. 1
• Topjian 2019 AHA pediatric post-arrest: cEEG endorsed as routine post-arrest neuroprognostication tool. 2
• Slovis 2024 pediatric post-arrest brain injury review: integrates cEEG into multimodal neurosurveillance. 3
• Sansevere 2019 neonatal cEEG: practical review of modern neonatal cEEG; reduced montages, aEEG-cEEG combination, video integration. 11
• Foreman 2012 qEEG DCI review: foundational synthesis of the qEEG-for-DCI evidence base; ADR as the workhorse trend. 8
• Benedetti 2023 pediatric qEEG review: contemporary qEEG metrics in children (the Part II noninvasive-neuromonitoring review). 31
• Laws 2022 (pediatric MMM review): cEEG described as essential pediatric neurocritical-care monitoring. 27

15. Self-check#