Routine EEG Interpretation

The single most important habit in electroencephalography is that the interpretation precedes the abnormality. A novice scans the page for a spike; an expert first asks what the brain's resting electrical state actually is, because almost every meaningful judgment in EEG is relative to the background. A sharp transient is benign or epileptiform only in the context of the rhythms around it; a focal slow wave is trivial or sinister depending on whether it disrupts an otherwise normal field. For this reason the disciplined reader imposes the same fixed sequence on every study, regardless of the clinical question, so that nothing is skipped under the pressure of an obvious finding. The pull toward the obvious is precisely the trap: an unmistakable left temporal spike train invites the eye to declare a focus and stop, while a contralateral subtle asymmetry, a unreactive posterior rhythm, or a second independent discharge goes unrecorded. The sequence exists to defeat that premature closure. This module builds that sequence, then connects it to the three interpretive acts physicians most want from EEG: saying where a problem is, separating real epileptiform activity from its many imitators, and saying what it all means in language a referring clinician can act on.

It helps to hold the whole enterprise in a probabilistic frame from the outset. An EEG does not return a diagnosis; it returns evidence that shifts the prior probability of a clinical hypothesis. A normal study lowers, but rarely abolishes, the probability of epilepsy; a focal epileptiform discharge raises it substantially and localizes it; a triphasic-morphology periodic pattern raises the probability of a diffuse encephalopathy while leaving its cause open. Every sentence in a competent report is, implicitly, a likelihood statement, and the reader's discipline is to calibrate those likelihoods honestly - neither overcalling a benign variant into an epileptiform abnormality nor dismissing a genuine discharge as artifact. The systematic review is the machinery that produces calibrated evidence; localization tells you where; the report communicates the posterior.

The systematic review: background before abnormality

Begin with state and continuity. Is the patient awake, drowsy, or asleep, and is the tracing continuous or interrupted by attenuations? State is the denominator for everything that follows: a 6 Hz rhythm is pathological slowing in a vigilant adult but is unremarkable in drowsiness, and a posterior rhythm that has merely slowed because the patient is dozing must not be read as encephalopathic. Establishing state first - from eye movements, muscle tone, the presence or absence of blinking, and the response to alerting - is therefore not a formality but the act that calibrates the entire read. Next characterize the posterior dominant rhythm (PDR): its frequency, its reactivity to eye opening, its amplitude, and crucially its symmetry. In an awake, cooperative adult the PDR should sit at roughly 8.5-12 Hz, attenuate briskly with eye opening, and be reasonably symmetric between the two occipital regions. A frequency below about 8 to 8.5 Hz in a fully alert, properly alerted adult is abnormal and is the most common quantitative sign of a diffuse encephalopathy - but only once drowsiness has been excluded, because the most frequent reason a PDR looks slow is simply that the patient was not fully awake.

Two quantitative facts about the PDR are worth committing to memory because they anchor real decisions. First, the developmental trajectory: the PDR is roughly 6 Hz at age one, reaches about 8 Hz by age three, and climbs into the mature 8.5-12 Hz adult range by around twelve to thirteen years - so what counts as a normal frequency is itself age-dependent, and a value normal for a toddler would be abnormal in an adult. Second, the asymmetry rule: the right occipital amplitude is normally slightly higher than the left (the skull is, in most people, marginally thicker on the left), so a modest right-greater-than-left amplitude difference is physiologic. The conventional thresholds for an abnormal asymmetry are an interhemispheric amplitude difference of greater than about 50 percent or a frequency difference of greater than 1 Hz, with the abnormality residing on the lower-amplitude or slower side. Because right-predominance is the physiologic baseline, a left-predominant amplitude asymmetry deserves more suspicion than its mirror image and should prompt a careful search for a structural explanation.

Then assess organization and gradient: a normal adult record shows an anterior-to-posterior gradient with faster, lower-amplitude activity frontally and slower, higher-amplitude alpha posteriorly. Loss of this gradient, or intrusion of frontal delta, signals diffuse dysfunction. Evaluate symmetry systematically by comparing homologous chains - the two temporal chains against each other, the two parasagittal chains against each other - for amplitude, frequency, and reactivity. Persistent focal asymmetry of frequency or amplitude is one of the most reliable scalp indicators of a structural lesion. Finally, test reactivity and variability: a normal brain changes. It responds to eye opening, to alerting, to drowsiness; a record that is invariant - the same pattern minute after minute regardless of stimulation - is itself an abnormal finding, particularly in the comatose patient, where reactivity becomes one of the most prognostically loaded features a reader can document. Only after all six axes - state, continuity, PDR, organization, symmetry, reactivity - have been characterized has the reader earned the right to call anything abnormal.

With the background described, the search for abnormalities proceeds along two axes: epileptiform versus non-epileptiform, and focal versus generalized. Epileptiform discharges - spikes, sharp waves, spike-and-wave complexes, and their rhythmic ictal equivalents - are defined morphologically by a paroxysmal waveform that stands out from the background, typically with an asymmetric, steep ascending limb, a duration in the spike (under 70 ms) or sharp-wave (70-200 ms) range, a following slow wave, a disruption of the surrounding rhythm, and a physiological electrical field that spans more than a single electrode. Non-epileptiform abnormalities are dominated by slowing: focal slowing, especially focal polymorphic delta, implicates local cortical or subcortical dysfunction, while generalized slowing implicates a diffuse process. The reader must also recognize the catalog of benign variants and artifacts that mimic pathology, because over-reading a normal variant is a far more common and more consequential error than missing genuine epileptiform activity - a false-positive read can commit a patient to years of unnecessary antiseizure medication and a driving prohibition built on a wicket spike.

Localization and lateralization

Localization is the geometry of EEG. In a bipolar montage, each channel displays the voltage difference between two adjacent electrodes, and the cardinal localizing sign is the phase reversal: when a discharge is maximal beneath one electrode, the channels on either side of it deflect in opposite directions, pointing toward the electrode of maximal involvement. A negative phase reversal at, say, F7 in a longitudinal bipolar (double banana) chain points to a generator near that electrode. Two conceptual sharpenings keep this from becoming a source of error. First, the phase reversal is purely localizing, not a marker of abnormality - normal rhythms, artifacts, and benign variants all phase-reverse, so a phase reversal tells you only where the field maximum is, never whether it is pathological. Second, the field maximum coincides with the cortical source only for a radially oriented generator; for a tangentially oriented source the peak scalp negativity can sit over a different region entirely than the generator, which is one reason no localization is trusted until it is confirmed across montages.

A referential montage complements bipolar recording by showing the absolute voltage at each electrode against a (relatively) inactive reference, allowing the reader to map the amplitude field and identify the single electrode of maximal negativity. The discipline is to reconcile the two views: the bipolar phase reversal and the referential amplitude maximum should converge on the same generator. Discordance usually means the reference itself is contaminated by the discharge - an active reference injects its own activity, with reversed polarity, into every channel and is the classic cause of spurious or smeared localization, particularly with vertex and midtemporal sources. This is precisely why modern source-oriented practice favors an average reference over linked-ears. Lateralization - simply which hemisphere - is usually easier than precise localization and is anchored by consistent asymmetry of background, by the side of maximal slowing, and by the side of the epileptiform discharges. The interactive model below lets you place a generator and watch how its polarity and field project onto the scalp chains, which is the fastest way to build the intuition that links a waveform on the page to a location in the head.

A correction that separates the careful reader from the careless one concerns what the standard electrode labels actually overlie. The classic 10-20 array has no electrode over the anterior and basal temporal lobe - the region that is, inconveniently, the single most common source of epileptogenic activity. The familiar teaching that F7 and F8 are the anterior temporal electrodes is an entrenched approximation; in the modern IFCN framing, F7 and F8 actually overlie the inferior frontal gyrus, not the basal temporal cortex. Apparent anterior temporal spikes at F7 or F8 may in truth be inferior frontal in origin. True coverage of the anterior and inferior temporal region requires the inferior chain - F9 and F10, FT9 and FT10, T9 and T10 - which sit roughly 10 percent below the standard line and which supersede the older, non-standard T1 and T2 (Silverman) electrodes once used for the same purpose. The same lesson applies in reverse elsewhere: the electrodes formerly labeled T5 and T6, now P7 and P8, actually overlie the posterior temporal lobe despite the parietal letter. The honest stance is that the one-to-one mapping of electrode letter to brain lobe should be de-emphasized, and that a scalp localization of a temporal focus is inherently limited by where the electrodes are not.

Benign variants: the imitators of epilepsy

Because over-reading is the dominant error, the expert keeps an explicit catalog of benign variants - paroxysmal-looking patterns that are normal and carry no association with epilepsy. The unifying discriminator across nearly all of them is the absence of the features that define a true epileptiform discharge: they lack a convincing after-going slow wave, they do not disrupt the surrounding background, and where they are rhythmic they characteristically fail to evolve in frequency, amplitude, and field over their course, whereas a true electrographic seizure evolves. Wicket spikes are the single most over-read pattern: arciform, mu-like temporal waves of drowsiness with no after-going slow wave and no background disruption. Small sharp spikes (SSS), also called benign epileptiform transients of sleep (BETS), are low-amplitude, brief biphasic transients of light non-REM sleep that, by consensus, have no value in seizure evaluation. Rhythmic temporal theta of drowsiness (RTTD) - the pattern once unhelpfully named the 'psychomotor variant' precisely because the old name wrongly implied a seizure link - is notched, rhythmic 5-7 Hz temporal theta that does not evolve, distinguishing it from a temporal ictal rhythm.

The catalog continues with patterns that demand more nuance. SREDA (subclinical rhythmic electrographic discharge of adults) is the most alarming of the benign variants because it genuinely mimics an electrographic seizure - a widespread, often parietally predominant 5-7 Hz rhythm - and is distinguished mainly by its relative lack of evolution and the absence of clinical change. The breach rhythm is not a brain rhythm at all but the accentuation - higher amplitude, sharper contour, faster appearance - of otherwise normal rhythms over a skull defect, where the missing bone no longer filters and attenuates the signal; the trap here is the opposite of over-reading, because a breach should prompt the reader to be conservative rather than to ignore the region, since genuine spikes can coexist near a craniotomy. The 6 Hz phantom spike-and-wave earns the most important asterisk in the whole catalog and must not be taught as unconditionally benign. The benign form is summarized by the mnemonic FOLD - Female, Occipital, Low amplitude, Drowsiness - whereas the WHAM form - Waking, High amplitude, Anterior, Male - overlaps with genuinely pathological generalized spike-and-wave and cannot be dismissed.

Clinical correlation and the structured report

An EEG finding is not a diagnosis. The interpretive act the referring physician needs is the correlation - the translation of an electrographic pattern into a statement about the clinical situation, bounded by what EEG can and cannot establish. Here the Bayesian frame becomes concrete. Interictal epileptiform discharges raise the probability of epilepsy and help classify it (focal versus generalized) and localize it, but the sensitivity of a single routine study is modest. The classic operational data are worth holding precisely: across serial recordings, roughly half of patients with epilepsy show interictal discharges on the first EEG, with the cumulative yield rising to about 84 percent by the third study and around 92 percent by the fourth, after which additional routine studies add little. Sleep is the most powerful single maneuver - the classic figures contrast roughly a third of patients showing discharges awake against the majority in sleep - and sleep deprivation and the activation procedures add further yield. The corollary is unambiguous: a single normal routine EEG, especially one without sleep, never excludes epilepsy.

The same calibration applies to specificity, and to two distinctions that are routinely conflated. First, the yield of an EEG in a patient known to have epilepsy (about 50 percent on a first study) is a different quantity from the sensitivity of an EEG for predicting seizure recurrence after a single unprovoked seizure (closer to 17-19 percent in meta-analysis); using one number for the other will mislead. Second, interictal discharges are highly specific but not pathognomonic: incidental epileptiform discharges occur in people without epilepsy at a rate that, depending on the population and how strictly discharges are defined, runs from well under 1 percent in carefully screened adults to a few percent in children and in prolonged recordings. The practical upshot is to treat a discharge as strong but not infallible evidence, weighted by the pre-test probability the clinical history supplies - an unequivocal discharge in a patient with stereotyped focal events is nearly diagnostic, while the identical waveform reported as an incidental finding in an asymptomatic adult warrants caution before it becomes a label.

The last row deserves a dedicated correction because the old teaching is wrong in two ways. What were once called triphasic waves are, in the current ACNS critical-care terminology, no longer a standalone entity; 'triphasic' is a morphology modifier applied to a periodic pattern, usually generalized periodic discharges. Their canonical morphology, in the ACNS convention, is negative-positive-negative, dominated by a large surface-positive middle phase - a point of genuine confusion in the literature, where some sources number the phases differently; the substantive, agreed fact is a prominent surface-positive component flanked by smaller negative deflections, classically with an anterior predominance and a fronto-occipital time lag visible on a longitudinal bipolar montage. They are also not specific to hepatic encephalopathy - they are in fact at least as common in renal failure and with drugs such as cefepime - and, most importantly, they overlap with nonconvulsive status epilepticus. Inter-rater agreement for the descriptor itself is only fair, and a benzodiazepine trial together with assessment of reactivity and clinical context, rather than morphology alone, is what separates a metabolic pattern from an ictal one.

The structured report mirrors the review sequence and exists to be useful to a reader who will never see the raw tracing. The ACNS guideline specifies five sections, not four: a clinical History; a Technical Description (montage, electrodes, recording conditions, activation procedures, and limitations such as artifact or absence of sleep); an EEG Description of the background and any abnormalities in neutral, objective language; an Impression classifying the study as normal or abnormal and naming each abnormality (conventionally limited to no more than three or four ranked findings); and a Clinical Correlation relating the findings to the stated indication in terms a generalist can use. The cardinal discipline is the firewall between the Description and the Impression: the Description is to be as objective as possible, so that a future reader - or the writer, months later, in a medicolegal review - can re-derive the conclusions from the observations. A report that asserts the study is 'abnormal due to left temporal sharp waves' without ever describing those waves is unverifiable and, in the worst case, wrong in a way no one can audit.