EEG and TCD non-convulsive seizure detection

1. Three patient vignettes#

Vignette A. The canonical sedated TBI#

Aiden, 8 years old, 28 kg, severe TBI day 3 after a bicycle-versus-car collision. Intubated, sedated on midazolam infusion 0.1 mg/kg/h plus fentanyl. ICP monitor in place at 12 mmHg, CPP 65. Bedside aEEG shows a baseline narrow trace (sedation effect) with intermittent envelope rises every 3 to 5 minutes lasting 30 to 90 seconds; synchronous TCD shows MFV briefly rising from 70 to 110 cm/s in the right MCA. The bedside nurse asks: is this non-convulsive status, or sedation bursting through? Full-montage cEEG is hooked up and shows rhythmic 1.5 Hz spike-and-wave in the right hemisphere during the envelope rises, evolving in frequency and morphology. This is NCSE. Levetiracetam load is started. 1 2

Vignette B. The 6-month-old post-arrest#

Ravi, 6 months old, 8 kg, post-cardiac-arrest day 2 after a witnessed cardiac arrest secondary to a long-QT episode. Hypothermia 33 C complete, now at 36 C. Sedated on midazolam infusion for shivering. aEEG envelope shows a discontinuous pattern with sharp clusters every 2 to 4 minutes; bilateral TCD shows MFV at 60 cm/s baseline with brief rises to 95. cEEG confirms bilateral periodic discharges at 1 to 2 Hz with no clear evolution, lateralised epileptiform discharges with morphology consistent with the ictal-interictal continuum. The team must decide: is this electrographic seizure that needs treatment, or post-anoxic burst-suppression that needs time? The Hirsch 2021 ACNS nomenclature is the bridge: the pattern is GPDs (generalised periodic discharges), at the high-risk end of the IIC. Treatment is appropriate. 3 4

Vignette C. The misleading sedation burst#

Maya, 11 years old, 38 kg, post-cardiac arrest after a near-drowning, day 3, on a midazolam plus pentobarbital infusion targeted to burst-suppression with BSR 70%. aEEG shows the classic burst-suppression envelope with peaks every 4 to 6 seconds; TCD shows synchronous MFV rises 60 to 95 cm/s. The covering night-shift senior wonders: should we treat this as NCSE? Full-montage cEEG shows non-rhythmic polymorphic theta-delta bursts on a flat background, exactly as expected for the prescribed BSR target. This is the desired sedation pattern, not NCSE. The TCD-MFV rises are the metabolic signature of the bursts (transient flow-metabolism coupling during the active phase). No action taken; continue at the current target. 1 5

2. The clinical question#

For the sedated PICU patient with synchronous aEEG and TCD intermittent rises, how do we distinguish sedation-induced burst activity from non-convulsive status epilepticus, and what role does TCD play beyond cEEG?

3. Pathophysiology refresher#

The cerebral metabolic rate of oxygen (CMRO2) and the cerebral blood flow (CBF) are normally tightly coupled (flow-metabolism coupling). When CMRO2 rises, CBF rises a few seconds later through a metabolic vasodilator cascade (local lactate, adenosine, K+, nitric oxide, prostaglandins). When CMRO2 falls, CBF falls. This is the physiological substrate for the TCD-aEEG relationship in this scenario.

During a seizure, neuronal firing increases roughly threefold in the active focus, lifting regional CMRO2 by 50 to 100%. CBF rises in parallel (in autoregulated brain) or fails to rise (in disautoregulated brain, producing post-ictal hypoperfusion). The TCD MFV picks up the rise as a transient MFV peak, often seen 1 to 5 seconds after the EEG ictal onset.

During a sedation burst, the burst is itself a brief period of neuronal activity emerging from a quiescent suppressed background. Polymorphic theta-delta bursts of high amplitude consume metabolic substrate; CMRO2 rises; CBF rises briefly. The TCD MFV signature is similar in waveform shape to the ictal signature, but the magnitude is typically smaller (the burst recruits fewer neurons than a true ictal discharge), and the morphology is non-evolving (each burst looks like the previous one).

Why aEEG alone cannot distinguish. aEEG compresses 10 to 15 minutes of cortical activity into a one-screen envelope. It captures the amplitude envelope but loses the morphology and the rhythmicity. An ictal saw-tooth envelope is morphologically very similar to a burst-suppression envelope at low temporal resolution. The ACNS pediatric cEEG indications are explicit: aEEG is a screening tool; full-montage cEEG is the diagnostic standard for NCSE. 2 3

Why TCD adds value. TCD adds three pieces of information aEEG does not provide: (i) timing of the metabolic-flow response (how delayed; whether the flow peak follows or precedes the envelope peak); (ii) localisation (which hemisphere has the larger MFV rise; useful when cEEG montage is reduced); (iii) autoregulation status (Mx-style correlation of MFV with MAP across the ictal cycle informs whether the brain is still autoregulating). 1 6

Why NIRS adds further value. NIRS rSO2 falls in the hemisphere that is metabolising more than its flow can supply, and the asymmetry timing relative to EEG and TCD informs the autoregulation question. 7

4. The multimodal picture#

5. Decision tree#

flowchart TD
  Obs[Synchronous aEEG and TCD peaks] --> Sed{Sedation depth target?}
  Sed -->|At burst-suppression target| Conf[Confirm with cEEG morphology]
  Sed -->|Below target| NotBSR[Not the burst-suppression state]
  Conf --> Morph{Morphology}
  Morph -->|Polymorphic bursts, non-evolving| Burst[Sedation burst; no action]
  Morph -->|Rhythmic spike-and-wave, evolving| NCSE[NCSE; treat]
  NotBSR --> NCSE
  NCSE --> Load[ESETT second-line load]
  Load --> Persist{Persists?}
  Persist -->|Yes| Infusion[Anaesthetic infusion to cEEG endpoint]
  Persist -->|No| Watch[Continue cEEG, watch breakthrough]
  Burst --> ReduceQ{Need to reduce sedation?}
  ReduceQ -->|Yes| Slow[Slow wean under cEEG]
  ReduceQ -->|No| Continue[Continue current target]

6. Step-by-step bedside actions#

For Aiden (8 y, 28 kg, sedated TBI with intermittent aEEG narrowing). Times are from the first reported envelope change.

1. 0 to 10 min: gather the trace, hold action. Do not adjust sedation or treat empirically before the full-montage cEEG is reviewed. Bedside aEEG alone is not a diagnostic standard. Page neurology and the on-call cEEG technician.
2. 10 to 30 min: get full-montage cEEG on. Hook up the full 21-electrode montage; the morphology snapshot is the diagnostic standard. Print or screenshot the most recent envelope-narrowing event for the neurology read.
3. 30 to 60 min: review morphology. Rhythmic spike-and-wave with evolution in frequency, amplitude, or location = NCSE. Polymorphic theta-delta bursts on a flat background, non-rhythmic, non-evolving = sedation burst. The Hirsch 2021 ACNS nomenclature is the reference. 3
4. NCSE confirmed: treat per the SE pathway. Levetiracetam 60 mg/kg IV over 10 min as ESETT-validated second-line (assuming benzodiazepine load already given as part of sedation). For Aiden's 28 kg: 1680 mg, max 4500 mg adult ceiling. 8 9
5. Sedation burst confirmed: no action; continue target. If the target is burst-suppression with BSR 50 to 90%, the pattern is expected. Document and continue.
6. Synchronously check the TCD trace. The TCD MFV rise magnitude (>30%) and the post-ictal undershoot (drop below baseline for 30 to 60 seconds) support NCSE; small symmetric rises with no undershoot support burst.
7. NIRS asymmetry check. If rSO2 falls >5% on the affected hemisphere during the rise and undershoots after, NCSE is more likely.
8. ICP trend. Cumulative ICP rise across multiple events (e.g., baseline 12 climbing to 18 over 30 minutes of events) supports NCSE-driven secondary injury.
9. NPi (pupillometry) at the next 2-hour neurocheck. Falling NPi (especially asymmetric) suggests evolving structural change beyond pure NCSE; consider repeat imaging.
10. Reassess at 4 to 6 hours. Persistent rhythmic ictal discharges despite second-line: escalate to anaesthetic infusion (midazolam 0.2 mg/kg bolus then 0.1 to 2 mg/kg/h), targeting cEEG seizure cessation.

7. Management ladder and endpoints#

Success looks like: cEEG morphology resolves (no more evolving rhythmic discharges); aEEG envelope normalises; TCD MFV peaks return to expected sedation baseline; NIRS symmetry restored; ICP stable; haemodynamics stable on titrated infusion.

Failure looks like: breakthrough electrographic seizures within 6 hours; NIRS asymmetry persisting or worsening; rising NPi asymmetry; rising ICP; new focal deficit when sedation lightened.

When to escalate:

• Persistent NCSE despite second-line, start midazolam infusion to cEEG endpoint.
• Persistent NCSE despite midazolam at 1 mg/kg/h, add ketamine (1 to 3 mg/kg bolus then 1 to 5 mg/kg/h) or switch to pentobarbital.
• Persistent NCSE more than 24 h, declare SRSE; broaden workup; consider immunotherapy and ketogenic diet.

When to de-escalate:

• cEEG seizure-free for 24 h at the target BSR; slow wean (20 to 30% per 6 to 12 h).
• Reversible drivers (fever, electrolytes, infection) addressed.
• Family goals-of-care conversation current.

8. Variant subsections#

8.1 Post-cardiac-arrest NCSE in the cooled child#

Up to 30% of comatose post-arrest children have electrographic seizures on cEEG; many are non-convulsive. The cooled child cannot generate the autonomic surges that hint at seizure clinically. TCD adds value by detecting the MFV-CMRO2 couple even when the cortex is suppressed by hypothermia. NIRS asymmetry is similarly informative. Treatment improves seizure burden; whether it improves outcome remains uncertain. 4 10

8.2 Ictal-interictal continuum patterns#

The 2021 ACNS standardised nomenclature defines patterns that lie between definite seizure and definite background (GPDs, LPDs, BIPDs, GRDA, LRDA). Whether to treat is a clinical decision that integrates the cEEG pattern with the clinical state, TCD MFV behaviour, NIRS asymmetry, and ICP trend. Treat aggressively when (a) the pattern is rhythmic and high-frequency, (b) the TCD-NIRS signature supports active metabolism-flow demand, or (c) there is a treatable downstream injury (rising ICP, new deficit). 3

8.3 Sedation burst with paradoxical TCD swings#

A child at deep sedation with burst-suppression target may show TCD MFV swings of 30 to 50% with each burst. This is not pathological; it is the metabolic signature of the burst itself. The discriminator is the cEEG morphology and the non-evolving nature of the bursts. Continue at target.

8.4 NCSE in NORSE/FIRES#

In febrile-prodrome refractory SE, the cEEG often shows multifocal evolving discharges that wax and wane over hours. The TCD MFV rises are often bilateral and shifting (the focus migrates). aEEG envelope is wide. Treatment is per the RSE/SRSE pathway with early ketogenic diet and immunotherapy. 11

8.5 NCSE after meningitis or encephalitis#

Particularly common in HSV encephalitis and bacterial meningitis with cortical involvement. The TCD-NIRS pair adds vasculitic vasospasm detection (rising MFV without flow-metabolism coupling correlation). The treatment ladder is identical to standard NCSE; the workup must include the infectious driver. 12 13

8.6 NCSE on ECMO#

VA-ECMO produces a non-pulsatile baseline TCD; intermittent MFV peaks during sedation bursts or NCSE may be difficult to distinguish from circuit-related flow swings. cEEG morphology and NIRS asymmetry are the discriminators. The HITS-detection capability of TCD on ECMO is unrelated but a useful concurrent finding. 14 15

9. Multimodal integration matrix#

10. Worked alternative scenarios#

10.1 What if the aEEG narrowing is actually electrode artefact?#

An obese 14-year-old with sweating and movement artefact on aEEG. The narrowing event is repetitive and synchronous with TCD MFV peaks, but the morphology on full-montage cEEG shows muscle artefact and electrode-pop, not rhythmic activity. Reapply electrodes, dry the scalp, sedate further if movement is driving the artefact; the TCD MFV peaks are likely cardiac variability captured by the bedside trend.

10.2 What if the TCD rises but the cEEG is silent?#

A 7-year-old on burst-suppression for pentobarbital coma. aEEG envelope is narrow and flat; cEEG morphology is isoelectric; yet TCD MFV is showing 30% rises every 4 to 5 minutes. The most likely explanation is cardiac variability captured at low temporal resolution, an autonomic surge from light anaesthesia, or systemic vasoactive boluses (e.g., nurse-administered noradrenaline pulse). Check the haemodynamic trace concurrently; the TCD swings should correlate with MAP swings.

10.3 What if the TCD MFV peaks but PI does not fall?#

In a true ictal flow-metabolism couple, PI typically falls (vasodilation) during the discharge and recovers afterward. If MFV rises but PI is stable or rising, the pattern is more consistent with proximal vessel stenosis, vasospasm, or hyperaemia from a different driver (fever, sympathetic surge), not NCSE.

11. Outcome data#

• Claassen 2004: in a series of adults with coma post-arrest or post-injury, electrographic seizures were detected in 19% within the first 24 h of cEEG, with 8% of those being NCSE. Almost all required cEEG to detect; clinical exam was negative. 1
• Foreman 2012, 2022 reviews: cEEG yield in modern PICUs is 10 to 40% depending on indication. Highest in suspected SE; intermediate in post-cardiac arrest; lower in routine neuroprognostication. 6 5
• Hirsch 2021 ACNS nomenclature: improved inter-rater reliability for periodic and rhythmic patterns (kappa values 0.6 to 0.8 for most common patterns versus 0.3 to 0.5 in earlier classifications). This matters because pattern classification drives treatment decisions. 3
• Topjian 2021 pediatric AHA: cEEG is recommended for the comatose post-arrest child; treatment of electrographic seizures is recommended; the magnitude of outcome benefit is uncertain. 4
• Naim 2023 PCCM: seizure burden after pediatric cardiac arrest correlates with 12-month neurological outcome; the relationship is strongest for status, weaker for non-status seizures. 10
• ESETT (Kapur 2019; Glauser 2016): second-line AED equivalence holds whether the seizing is convulsive or non-convulsive. Treatment ladder is the same. 9 8

12. Pitfalls#

• Treating aEEG narrowing as diagnostic. The aEEG envelope cannot reliably distinguish sedation burst from ictal activity. Always confirm with full-montage cEEG before action.
• Treating sedation bursts as NCSE. A patient at a target BSR will show TCD MFV swings; do not over-treat the desired state.
• Lightening sedation before cEEG. Lightening sedation when NCSE is suspected risks both a clinical seizure and the misinterpretation of the lightened-sedation EEG as the baseline.
• Hyperventilating to manage suspected ICP rises. NCSE drives small ICP rises by raising CBV; treat the seizure, not the CO2.
• Forgetting the reversible drivers. Fever, hypoglycaemia, hyponatraemia, sepsis, and AED non-adherence (or extravasation) all trigger or sustain NCSE.
• Believing a normal NPi rules out NCSE. NCSE without focal mass effect or oedema produces little change in NPi; preserved NPi is consistent with NCSE.
• Using PI as ICP. During NCSE, PI behaviour is dominated by vasodilation, not by raised intracranial pressure; do not invert the Bellner regression in this physiology. 16
• Ignoring the ictal-interictal continuum patterns. GPDs, LPDs, and BIPDs at the high-risk end of the IIC (high frequency, evolving, with mass effect or clinical change) deserve treatment; not all are benign. 3

13. Pediatric considerations#

14. Combine with#

• EEG / aEEG modality page: full montage versus reduced, morphology, ACNS nomenclature.
• TCD / TCCD modality page: MFV, PI, the flow-metabolism couple.
• NIRS modality page: rSO2 asymmetry as a localising tool.
• Pupillometry / NPi page: the focal-injury sentinel that survives paralysis.
• Integration: Refractory status epilepticus: the full RSE/SRSE pathway.
• Integration: HIE post-arrest: the cooled child with electrographic seizures.
• Foundations: flow-metabolism coupling: why CMRO2 and CBF move together (and when they do not).

15. Evidence summary#

16. Recent literature (2022 to 2025)#

• Foreman 2022 review updates the case for cEEG in the modern PICU. Reduced-channel continuous trace is the bridge tool; full montage remains the diagnostic standard. 5
• Hirsch 2021 ACNS nomenclature is now the standard for IIC pattern reporting; treatment decisions hinge on the pattern classification. 3
• Topjian 2021 AHA pediatric post-arrest standardises cEEG monitoring expectations across centres. 4
• Naim 2023 PCCM quantifies the seizure-burden to outcome relationship in pediatric cardiac arrest. 10
• TCD-EEG synchronous monitoring is increasingly available with combined acquisition platforms; the published case series are small but growing.
• NIRS-EEG localising studies in pediatric SE show that hemispheric rSO2 asymmetry correlates with the cEEG focus in about 70% of focal cases. 7

17. Self-check#