01 Signals, montages and technical settings

The scalp EEG is a record of voltage differences, not of neurons firing. What the electrode sees is the summed postsynaptic current of many thousands of cortical pyramidal cells whose apical dendrites lie parallel to one another and perpendicular to the cortical surface. When those cells are depolarized together, each behaves as a tiny dipole with a negative pole and a positive pole. Only when the dipoles are aligned and synchronous does their field reach the scalp; single action potentials are too brief and too scattered to contribute. This is why the EEG reflects the state of a cortical population, and why a small, deep or tangentially oriented generator can produce a large discharge at depth and nothing on the surface.

The skull and scalp smear the field. A generator under one electrode also raises the voltage at its neighbours, falling off with distance, which is what an electroencephalographer means by a discharge's field. A genuine cerebral event therefore has a plausible field: a maximum at one electrode, smaller deflections at adjacent ones, and a gradient that makes anatomical sense. Something confined to a single electrode with no field at all is far more likely to be artifact, a rule you will meet again in Module 3 and as one of the six IFCN criteria in Module 5.

The display convention

Every channel displays the voltage at input 1 minus input 2. By convention, a negative difference is drawn as an upward deflection. Most epileptiform discharges are surface-negative, so they point up, which is convenient, but it also means the direction of a pen deflection tells you nothing by itself. What matters is the sign of the difference between two electrodes, and that depends on the montage. The same discharge will look different, and even point in opposite directions, in different montages, without any change in the brain. Reading polarity correctly is the skill on which every later localization rests.

A bipolar montage links electrodes in chains, each channel comparing one electrode with the next: Fp1-F3, F3-C3, C3-P3, P3-O1 in the longitudinal ("double banana") arrangement. Because each electrode is input 2 of one channel and input 1 of the next, a negative maximum at a shared electrode produces deflections that point toward each other in the two adjacent channels. That is a negative phase reversal, and it marks the electrode of maximum negativity. A positive maximum produces the mirror image, deflections pointing away from each other, which you will see with some artifacts and with the positive end of a horizontal dipole.

Use the figure above to convince yourself of this. Move the source along the chain and watch which channel pair reverses. Then widen the field and notice that the phase reversal stays at the maximum even as the neighbouring channels grow. Localization by phase reversal is arithmetic on differences: the montage is a choice of subtraction, and the brain does not change when you change it.

The end-of-chain problem

A phase reversal needs a channel on each side of the maximum. An electrode at the end of a chain, such as Fp1, Fp2, O1 or O2 in the longitudinal montage, is input to only one channel, so a maximum there cannot reverse. The discharge simply appears in the last channel, largest there, with no reversal to prove where the peak lies. When a sharp transient sits at the end of a chain with no reversal, do not guess. Read it in a referential montage, where the maximum shows as the largest amplitude, or in a transverse chain, where the same electrode may sit in the middle and can reverse.

Bipolar montages have one more property to keep in mind: they subtract what two neighbouring electrodes share. A broad, diffuse field looks small in a bipolar chain because adjacent electrodes see nearly the same voltage, and the difference between them is close to zero. A focal discharge, by contrast, is emphasized. That makes bipolar montages good for finding focal maxima and poor for judging the true amplitude or extent of widespread activity.

A referential montage compares every electrode with the same reference, so the channel with the largest deflection is the electrode closest to the maximum. There is no phase reversal to look for; amplitude does the work. This makes referential montages the right place to confirm an end-of-chain maximum, to judge the true amplitude of a discharge, and to appreciate a broad field that a bipolar chain would flatten.

The price is that everything depends on the reference being quiet. If the reference electrode itself picks up the discharge, that voltage is subtracted from every channel, and the discharge appears everywhere, sometimes with reversed polarity in channels far from the true source. An ear or mastoid reference is close to the temporal lobe, so a temporal discharge can contaminate it. A common average reference sums all electrodes and so dilutes any one source, but a large discharge still leaks into the average and shows up as a small, opposite-polarity deflection across the other channels. When a referential montage shows a discharge in an implausible distribution, or the same waveform at every electrode, suspect the reference before you suspect the brain.

A working routine

Read a suspicious transient in more than one montage before you commit to a location. Find the phase reversal in a longitudinal bipolar chain, check that it holds in a transverse chain, then confirm the maximum by amplitude in a referential montage. If the three agree, the localization is secure. If they disagree, one of them is misleading you, and the usual culprits are an end-of-chain maximum in the bipolar montage or an active reference in the referential one. This routine is what the sign-off for this module tests: five discharges localized on real records in both bipolar and referential montages, with a faculty reader.

The tracing you read is shaped by four settings, and each can create or hide a finding.

The low-frequency filter (high-pass) attenuates slow activity. Raising it makes a record look cleaner by removing sweat and movement drift, but it also shrinks slow waves that are real. Focal delta over a lesion can become inconspicuous, and the after-going slow wave that helps identify a spike can be lost. When slowing is the question, read the record with the low-frequency filter as low as the record tolerates.

The high-frequency filter (low-pass) attenuates fast activity. Lowering it smooths muscle artifact, which is why it is tempting in a restless child, but the smoothing turns fast, spiky muscle potentials into rounded, sharply contoured waves that can look like spikes or sharp waves. A record that is "cleaned up" this way can acquire epileptiform-looking transients it did not have. If you must lower the high-frequency filter, note that you did, and go back to the unfiltered record before calling anything epileptiform.

The notch filter removes line frequency. It is useful when an electrode has poor contact and a thick band of line noise obscures the channel, but it is a patch. Persistent line noise in one channel means that electrode's impedance is high, and the fix is at the scalp, not in software.

Sensitivity scales voltage to display height. Too low a sensitivity setting (fewer microvolts per millimetre) exaggerates amplitude and can make ordinary activity look high-voltage; too high a setting flattens a low-voltage record. Judge amplitude by the number shown, not by the height of the trace.

Time base

The page duration sets how compressed the waveforms are. Neonatal records are often reviewed at a slower paper speed than routine records, which compresses each wave and makes activity look faster and spikier than it is. A frequency judged by eye on a compressed page will be wrong. Always check the page duration before judging frequency, and count cycles against the time marks rather than by impression.

Children are not small adults on the recording table, and the technical standards say so explicitly. For neonates, the American Clinical Neurophysiology Society recommends a 10-20 array modified for the small head, with a reduced electrode set, and it asks for ECG and respiration channels as part of the recording, because heart rate and breathing are needed to identify sleep state and to separate physiologic events from cerebral ones (Shellhaas 2011). Eye-movement and chin electromyography channels are often useful for staging neonatal sleep but are not universally required. Sleep staging matters more in a neonate than at any other age, because, as Module 6 shows, the same background can be normal in one state and abnormal in another.

For older infants and children, ACNS Guideline 5 sets the minimum technical standards for pediatric EEG (Kuratani 2016). Read it once, in full, so that you know what a technically adequate pediatric study looks like and can recognize when a study falls short: too few electrodes, no sleep obtained, activation procedures skipped, or no video when the question is an event. A report that interprets an inadequate study without saying so misleads the reader.

What changes at the bedside

A neonate's head is small, so electrodes are close together and the field of any discharge spans fewer of them. The record is often reviewed on a longer page, so check the time base before judging frequency, as in the previous section. Infants and toddlers move, feed and cry through their studies, which is why the recording needs video and why Module 3 spends so long on the artifacts of infancy. Hyperventilation cannot be obtained from a very young child and photic stimulation may be limited, so the report must state which activation procedures were done and how adequately. Finally, sedation and medications change the record: sedatives add fast activity and alter sleep, and antiseizure medications can suppress discharges and slow the background, so both belong in the technical description of every study, a point Module 2 turns into a reporting rule.