The electroencephalogram (EEG) is a recording from the surface of the scalp of the time-varying electric fields generated by summated excitatory and inhibitory post-synaptic potentials generated in dendrites of layer I of the gyral crests of the cerebral cortex.

Let’s approach this in pieces.

The recording is from the surface of the scalp. Electrodes are placed in a standardized grid pattern on the scalp. The “brain waves” are generated in the brain but propagate through the various layers: CSF, meninges, calvarium, galea, skin, and finally to the elecrodes. As a result the signals are somewhat depleted and small. Most EEG signals are of the order of 1E-5 volts—compare to ECG signals at 1e-3 volts. The EEG can be overwhelmed by artefactual signal from electrical interference, movement, muscle activity, and other things.

Moreover, although the scalp presents a more-or-less smooth convex surface, the underlying brain is thrown into myriad convolutions. Signals are directed more or less at right angles to the local cortical surface, meaning EEG signals from a sulcal wall can interfere with a distant electrode that is also detecting signal from directly underneath it.

There are also therefore broad areas of cortex that are inaccessible by scalp recording. The mesial surfaces of the frontal, parietal, temporal, and occipital lobes are deep to the scalp and oriented parallel to the overlying skull surface. The insula is several cm away from the scalp surface and is overlain by other cortices. The basal temporal and frontal areas are chock-a-block against the basal skull far away from the skin surface.

And many signals of interest (such as interictal spikes) are only manifest if created by a synchronous region of cortex some 6–10 cm² in area. Recording directly from brain surface (as in electrocorticography or ECoG) often reveals many more spikes than are detectable at the surface because of that technique’s superior sensitivity.

Scalp recording therefore means that roughly but 40% of the cortex is sampled; the some areas can be electrically silent; and that the absence of abnormal activity cannot with certainty be correlated with the absence of epileptic disease.

The time-varying fields are detected as voltages, which means a potential difference. Voltage is only meaningful as a difference of electrical potential between points—you can only ever measure voltage with respect to another point. In EEG therefore, the signal display is often constructed to show differences between adjacent electrodes or between electrodes and a common reference point.

Post-synaptic potentials (PSPs) are voltage changes that occur when an electrical synapse either partially depolarizes or hyperpolarizes the post-synaptic neuron through the release of neurotransmitters. Large pyramidal cells in the cortex have large dendrite trees that extend all the way up to layer I and are connected to by thousands of other neurons.

In any given patch of cortex there may be more or less random excitations and inhibitions of the pyramidal cells. In this case the summed voltage over the patch is going to be nearly neutral, or close to zero, with variations over time in a small range of voltage. When patches of cortex become synchronously active all together they can generate larger negative or positive deflections in voltage.

Gyral crests are the preferential location for detection of signal because their signals are oriented directly outward toward the scalp where electrodes can more easily detect them. Within sulcal walls, the signals are oriented at angles to the scalp and are more depleted in strength because of the travel through more tissue. Therefore epileptiform discharges and epileptic seizures arising in sulcal walls or deep cortices may not be detected by scalp EEG until or unless they spread to include dorsolateral gyral crest.

The interictal discharge


EEG is useful in three major ways.

First, EEG can detect IEDs. The presence of IEDs is a marker of increased risk for focal seizures. They are not diagnostic of epilepsy, however. Seizure-free persons, including those who are in remission and those who have never had a seizure, may have spikes and sharp waves on their EEG. Given estimates of 1% prevalence of IEDs in patients without epilepsy (and this varies to up to 6% or so), as well as a prevalence of epilepsy of 0.5% of the population and an expected sensitivity of EEG of 85% (over repeated studies), one can calculate that the positive predictive value of IEDs in random EEGs done in the general population is 25% or less. It is therefore imperative that 1) EEG be ordered only in cases suspected to be epileptic; and 2) that EEG results be interpreted alongside the clinical history.

EEG can detect seizures. Routine EEGs of 30–60 min are unlikely to capture seizures in any but the most burdened patients. (They do, nevertheless, occasionaly capture them.) Longer-term EEGs of several days, especially when coupled with provocative measures in a video-EEG monitoring unit, are good at detecting epileptic seizures. The detection of seizures can ipso facto make a diagnosis of epilepsy (absent cases wherein critically-ill patients have symptomatic seizures, etc).

With standard EEG electrode densities the seizure onset zone can be estimated at best down to the sublobar level. Frequently only a lobar onset can be inferred and not uncommonly seizure onset is so diffuse as to confound attempts at localization.

Sometimes this may be due to seizure onset and initial propagation within sulci or within inaccessible cortices (mesial brain, insula, basal frontal or temporal) which only manifests on EEG once the seizure has spread (often diffusely) into areas more detectable at the scalp.

Some seizures, usually sensory auras, may be undetectable by EEG due to the small extent of cortical activation and/or its occurrence entirely within sulci or deep cortex. Moreover, many frontal lobe seizures have very abrupt and rapid spread resulting in almost no EEG seizure before the record is entirely contaminated with motion artefact.

In cases such as these, it is up to the clinical neurophysiologist or epileptologist to use video and other evidence to infer whether the event is an epileptic seizure or not. This comes with knowledge and experience.

Lastly EEG can rule out epileptic seizure as a cause for paroxysmal clinical events. In cases such as conversion disorder, syncope, movement disorders, and other “imitators of epilepsy” the recording of a patient’s symptomatic event coupled with knowledge of semiology and a negative EEG tracing can definitively rule out epilepsy as a cause.

The main caution here is that, as discussed above, some seizures do not manifest on EEG. And some seizures have fairly bizarre semiologies which inexperienced practitioners may be too quick to label as impossibly epileptic.