Epilepsy Monitoring Unit workflow — continuous video-EEG recording, AED taper, ictal capture, and electrographic onset analysis to localize the seizure onset zone
Long-term video-EEG monitoring (LTM) is the diagnostic gold standard for localizing drug-resistant epilepsy. A patient is admitted to a dedicated Epilepsy Monitoring Unit where 21+ scalp electrodes are affixed according to the International 10-20 system, and both EEG signal and synchronized video are recorded continuously — often for 3 to 7 days — until enough habitual seizures are captured to answer the clinical question.
Electrode placement follows the 10-20 system, first standardized in 1958, which positions electrodes at intervals of 10% or 20% of measured skull landmarks (nasion to inion, and left to right preauricular points). This guarantees that "T3" or "Fp2" refer to the same anatomical location on every patient, enabling reproducible comparison across recordings, centers, and decades of literature.
Electrode naming logic: • Fp = frontopolar, F = frontal, C = central, P = parietal, O = occipital, T = temporal • Odd numbers = left hemisphere, even numbers = right hemisphere, z (zero) = midline • Higher-density 10-10 montages add intermediate electrodes for finer spatial resolution when a scalp focus needs to be pinpointed more precisely
Each electrode is filled with conductive gel and secured with collodion adhesive or an electrode cap, then impedance-tested (<5 kΩ target) before recording begins — a poor contact anywhere can masquerade as, or obscure, an epileptiform discharge.
A comfortable, secure electrode array must survive days of normal activity — sleeping, eating, walking to the bathroom with a telemetry backpack — without losing contact, which is why EMU nursing staff re-check impedances and re-glue electrodes daily.
EEG alone tells you when electrical activity changes; video tells you what the patient actually experienced and did. Time-locked video-EEG is essential because:
• Scalp EEG can be misleadingly normal during focal seizures arising from deep or mesial structures (the signal must travel through skull and tissue that acts as a low-pass filter) • Many "seizure-like" events are actually psychogenic non-epileptic seizures (PNES) — video captures characteristic semiology (asynchronous limb movement, pelvic thrusting, eye closure) that differs from epileptic seizures despite a normal-looking or artifact-obscured EEG • Semiology itself is localizing: which body part twitches first, whether speech is preserved, direction of head/eye deviation — all correlate statistically with the region of cortex involved
A dedicated ceiling-mounted infrared camera, synchronized audio, and a bedside event button (pressed by patient, family, or nursing staff) all timestamp against the same EEG clock so that every second of the record can be replayed as a unified audiovisual-electrographic timeline.
Before any seizure occurs, days of baseline "interictal" recording establish what this patient's normal and abnormal-but-non-ictal EEG looks like:
• Interictal epileptiform discharges (IEDs) — spikes and sharp waves seen between seizures — often cluster over the same region that later generates seizures, giving an early localizing clue • Sleep and wake states, hyperventilation, and photic stimulation are all sampled, since some discharges only appear during specific physiological states (many focal epilepsies activate in drowsiness and light sleep) • Establishing a clean, artifact-characterized baseline lets the team distinguish true ictal onset later from muscle, movement, or electrode artifact that can mimic it
This stage sets the stage, quite literally, for everything that follows: without a well-instrumented and validated baseline, the ictal recording captured days later cannot be interpreted with confidence.
The entire point of EMU admission is to record a habitual seizure — one that resembles the patient's typical clinical events. Because anti-epileptic drugs (AEDs) suppress seizures by design, they are gradually reduced or fully withheld under close, monitored conditions to provoke an event within the admission window, while safety precautions scale up in parallel.
AEDs raise the seizure threshold by stabilizing neuronal membranes, enhancing inhibitory (GABAergic) tone, or dampening excitatory (glutamatergic) transmission. Reducing or stopping them predictably increases seizure likelihood — which is exactly the diagnostic information needed, but it must be done safely:
• Taper schedules are individualized to seizure frequency, drug half-life, and any history of status epilepticus • Rescue medication (IV benzodiazepines) and full resuscitation equipment stay at bedside throughout • Seizure precautions escalate: padded rails, oxygen saturation monitoring, staff proximity, and sometimes temporary restriction of ambulation • The taper is titrated day-by-day against EEG background changes and any early clinical signs, and can be slowed or reversed if events cluster dangerously
The taper is a controlled provocation, not an uncontrolled withdrawal — the clinical team balances "record enough seizures to localize the focus" against "never let a seizure cluster into status epilepticus," adjusting rate and rescue thresholds continuously.
Beyond AED reduction, several time-tested provocation techniques increase yield within a limited admission window:
• Sleep deprivation: staying awake overnight measurably lowers seizure threshold in many focal and generalized epilepsies • Hyperventilation: 3–5 minutes of forced deep breathing can activate absence and some focal discharges via cerebral vasoconstriction and pH shift • Photic stimulation: strobe light at varying frequencies can trigger photoparoxysmal responses in photosensitive epilepsies • Stress and routine disruption of hospitalization itself can be mildly activating for some patients
Each activation method is logged precisely against the EEG/video timeline, since it matters for interpretation whether a discharge occurred spontaneously or was provoked — provoked interictal discharges are still useful localizing data but are weighed differently than a fully unprovoked habitual seizure.
In the minutes to hours before an electrographic seizure, background EEG activity can shift subtly: increased amplitude, slight frequency drift, or intermittent rhythmic slowing localized to the eventual onset region. These changes are frequently too subtle to flag prospectively but become visible retrospectively once the ictal onset is known.
Automated seizure-detection algorithms (amplitude and frequency-based, and increasingly deep-learning classifiers) continuously scan the incoming signal to alert staff the moment a possible event begins, since the diagnostic value of the recording depends entirely on catching the earliest seconds of the discharge — not just the obvious later stages.
When the seizure finally occurs, every second matters. The EEG stream and the synchronized video feed jointly capture the full arc of the event — earliest electrographic change, evolving rhythmic discharge, clinical semiology as it unfolds on camera, and the post-ictal recovery period — creating the single most information-dense artifact of the entire admission.
A well-captured ictal recording typically shows several distinguishable phases in sequence:
1. Electrographic onset — the earliest rhythmic, evolving discharge, often subtle low-voltage fast activity or rhythmic theta/alpha, confined to a small number of channels 2. Propagation — the discharge recruits neighboring channels and often changes frequency and morphology as it spreads, sometimes generalizing to the whole scalp 3. Clinical onset — the first visible behavioral change on video (aura report, automatism, focal motor sign), which may lag, coincide with, or occasionally precede the scalp electrographic onset 4. Ictal semiology evolution — the sequence of signs (e.g., oral automatisms → dystonic limb posturing → head version) is itself a localizing and lateralizing signature refined over decades of seizure semiology research 5. Electrographic termination and post-ictal state — background suppression, slowing, or asymmetry immediately after the seizure ends can add further lateralizing information
Post-ictal Todd's paresis (transient focal weakness after a seizure) and post-ictal EEG suppression that is more pronounced on one side are both classic lateralizing clues — the weaker/more suppressed side usually corresponds to the seizure-onset hemisphere.
The moment an event is flagged (by automated detection, patient event button, or bedside observation), EMU protocol activates:
• A "push-button" annotation is placed on the EEG record instantly, and staff verbally narrate observations on the synchronized audio track ("eyes deviate right," "right arm clonic jerking begins now") — narration becomes part of the permanent clinical record • Staff perform a structured bedside seizure exam: test responsiveness, language, and lateralized motor/sensory function *during* the event when safely possible, since these bedside findings are themselves localizing • Safety-first: protecting the airway, cushioning the patient, and timing the event are prioritized over any exam maneuver that would put the patient at risk • Rescue medication is administered per protocol if the seizure exceeds a pre-set duration threshold (commonly 5 minutes), converting the recording from a diagnostic capture into an emergency intervention if needed
One of the EMU's most important diagnostic functions is capturing events that turn out NOT to be epileptic — most commonly psychogenic non-epileptic seizures (PNES), but also syncope, movement disorders, parasomnias, or panic attacks. A recorded event with a completely normal, artifact-free EEG throughout a clinical spell strongly supports a non-epileptic etiology, which fundamentally changes management away from AEDs and surgery toward psychiatric or cardiology referral.
Distinguishing a subtle focal seizure with poor scalp signal from a genuinely non-epileptic event is one of the most challenging judgment calls in epileptology, and is precisely why capturing several habitual, video-documented events — not just one — is the goal before any localization conclusion is finalized.
After capture, epileptologists perform detailed channel-by-channel review of the ictal EEG, often frame-by-frame at slow paper speed, to pinpoint the earliest channels showing a clear rhythmic, evolving ictal discharge — distinct from artifact, muscle activity, or simple propagation from elsewhere. This "onset zone" on the scalp is the anchor for every subsequent localization step.
No single display reveals everything, so epileptologists systematically re-review the same seizure in several montages:
• Bipolar (longitudinal / transverse "banana") montages compare adjacent electrode pairs, which is excellent for localizing a discrete focal source via phase reversal — the classic sign where the deflection points oppositely in two channels sharing the source electrode • Referential montages (each electrode vs. a common or average reference) are better for judging true amplitude and field extent, since bipolar chains can "cancel" a broadly distributed discharge • Source-derivation / Laplacian montages sharpen spatial resolution further, emphasizing local maxima and reducing the smearing caused by volume conduction through skull and scalp
Slowing the paper speed and expanding sensitivity around the presumed onset window lets reviewers separate genuine rhythmic evolution (increasing/decreasing frequency, recruiting channels in a specific spatial sequence) from artifact or simple asymmetric muscle tension.
Phase reversal in a bipolar chain is one of the most classic and teachable EEG localization signs: when the same discharge appears "up" in one channel and "down" in the adjacent channel sharing an electrode, that shared electrode sits closest to the underlying source.
Scalp EEG has real physical limitations that make onset determination genuinely difficult:
• Volume conduction spreads any deep or broad cortical generator across many electrodes simultaneously, blurring the true point of origin • A significant fraction of focal seizures — especially from mesial temporal, orbitofrontal, or insular cortex — can be scalp-EEG-silent or show only subtle, late, diffuse change at onset because the generating cortex is far from the scalp surface or oriented unfavorably • Muscle and movement artifact accompanying the seizure's earliest motor signs can obscure or mimic true ictal rhythms in nearby channels • Very rapid propagation (a discharge can spread across the cortical surface and reach distant scalp regions within a second or two) means the first channel to show a change is not always the true anatomical origin
This is exactly why the onset call weighs the *pattern* of evolution — a discharge that clearly starts in one or two channels and progressively recruits neighbors — rather than the single earliest millisecond alone, and why concordance across multiple montages and multiple captured seizures raises confidence.
Certain scalp ictal patterns recur often enough to carry pattern-recognition value, though none is perfectly specific:
• Rhythmic theta/alpha build-up over a focal region: classic mesial or lateral temporal onset pattern • Low-voltage fast activity (beta/gamma range) at onset: often associated with neocortical or extratemporal foci, and correlates with better surgical outcome when clearly focal • Repetitive spiking that gradually organizes into rhythmic activity: another common focal-onset signature • Bilateral, diffuse onset without clear focality: raises suspicion for a deep midline generator, rapid bilateral synchronization, or a primary generalized process rather than a resectable focal lesion
Each recorded seizure is scored independently, and localization confidence grows substantially when multiple habitual seizures show a concordant onset region rather than resting on a single, possibly atypical, event.
The final and most consequential step fuses two independent data streams — what the video showed the patient doing (semiology) and where the EEG showed the discharge beginning (electrographic onset) — into a single lateralization and localization hypothesis. This synthesized conclusion, alongside MRI, PET, and neuropsychological data, ultimately determines whether a patient becomes a surgical candidate.
No single data source is treated as definitive on its own. The epilepsy multidisciplinary team (epileptologist, neurosurgeon, neuropsychologist, neuroradiologist) weighs and cross-checks:
• Electrographic onset zone — the scalp region(s) where the ictal discharge began, from Stage 4 review • Ictal semiology — lateralizing and localizing behavioral signs captured on video (e.g., unilateral dystonic posturing, versive head turning, preserved vs. impaired language during the event) • Interictal epileptiform discharges — where spikes cluster between seizures, historically • Structural imaging (MRI) — an identifiable lesion (hippocampal sclerosis, focal cortical dysplasia, tumor) concordant with the electrographic zone strongly increases surgical confidence • Functional imaging (ictal/interictal SPECT, PET) and neuropsychological testing, which can lateralize even when structural MRI is normal
When video semiology, EEG onset, and imaging all point to the same region, the case is termed "concordant" and confidence in a focal, resectable epilepsy is high. Discordant findings trigger further workup — often intracranial EEG (stereo-EEG or subdural grids) to map the zone with much higher spatial precision than the scalp ever could.
A classic teaching example: right-hand automatisms with a dystonic (stiffly postured) left arm strongly lateralizes to a right temporal onset — the automatisms occur ipsilateral to the seizure focus while the dystonic limb is typically contralateral, a semiology pattern validated across thousands of surgical cases.
The EMU stay culminates in a formal report summarizing every captured event, montage-by-montage onset analysis, correlated semiology, and a stated localization/lateralization conclusion with an explicit confidence level. This document becomes the backbone of the epilepsy surgery conference discussion, where the full multidisciplinary team decides among several paths:
• Proceed directly to resective surgery if scalp data plus imaging are strongly concordant • Proceed to intracranial monitoring (stereo-EEG electrodes or subdural grid/strip placement) to refine an ambiguous or discordant localization before committing to resection • Consider neuromodulation (responsive neurostimulation, vagus nerve stimulation, deep brain stimulation) for patients who are not resection candidates, e.g. bilateral or poorly localized onset • Continue medical management and re-attempt monitoring later if no habitual seizures were captured or the record remains inconclusive
Scalp video-EEG localizes to a region of the scalp, not a precise anatomical structure — it is inherently a screening and hypothesis-generating tool rather than a millimeter-accurate map. Its core limitations include volume conduction blurring, poor sensitivity to deep or small-surface-area generators, and the physical impossibility of sampling every square millimeter of cortex with ~21–25 electrodes.
When the surgical stakes are high and scalp data leaves ambiguity, intracranial EEG (stereo-EEG depth electrodes or subdural grids placed directly on or within the brain) provides orders-of-magnitude better spatial and temporal resolution, at the cost of an invasive procedure guided by exactly the localization hypothesis this scalp video-EEG workup produced. In this sense, the entire Video-EEG Localization workflow is not the final answer for many patients — it is the essential first map that makes every subsequent, more invasive step possible and safer.