03 Artifacts

Every abnormality you will learn in the later modules has an artifact that imitates it. A sharp transient can be an electrode pop. Rhythmic activity can be a ventilator. Slow waves can be sweat. Before a waveform is called cerebral, the reader has to have excluded the non-cerebral explanations, and in a child that exclusion is harder than in a still adult, because children blink, chew, cry, suck, and are patted, rocked and fed through their recordings.

When experts were asked which EEG findings residents must know, the artifacts they rated most important were ECG, eye blinks, electrode pop and muscle (Nascimento 2022). Those four are the commonest, the most often over-read, and the easiest to prove once you know what to look for. This module adds the artifacts specific to infancy and to the intensive care unit, where rhythmic artifact is the main imitator of seizures.

The single most useful rule

A cerebral discharge has a plausible field: a maximum at one electrode with smaller deflections at its neighbours, falling off with distance in a way that makes anatomical sense. An artifact usually does not. This is the sixth of the IFCN criteria for an epileptiform discharge, the scalp distribution that implies a source in the brain, and Module 5 applies it to the question of over-reading. It is worth learning here first, because it resolves most artifact questions on its own. A waveform confined to a single electrode, or one that appears in electrodes that are not neighbours, or one whose distribution matches a piece of equipment rather than a piece of cortex, is not cerebral until proven otherwise.

The second rule is physiology: an artifact is generated by something, and that something can usually be found. The heart leaves a trace on the ECG channel. The eyes can be seen on video. A ventilator has a set rate. Finding the generator is the proof, and the artifact table in this module lists where to look for each one.

Most artifact in a child comes from the child. Each physiologic artifact has a recognizable appearance and a way to confirm it.

Eye blink. The cornea is positive relative to the retina, and the eye rotates upward with a blink, so the positive pole moves toward the frontal electrodes. In a bipolar montage this produces a symmetric downward deflection at Fp1 and Fp2. The field is frontal only, with nothing posterior, and the eyes can be watched on video.

Lateral eye movement. A horizontal saccade moves the positive cornea toward one side, producing opposite polarity at F7 and F8, a mirror image across the hemispheres. Brief lateral rectus spikes at the start of each saccade can imitate frontotemporal sharp waves; their mirror-image distribution gives them away.

Muscle. Frontalis and temporalis contraction produces very fast, spiky activity, often temporal, during chewing, crying or grimacing. It disappears in sleep or with relaxation. Filtered muscle is the classic imitator of spikes, as Module 1 warned: lowering the high-frequency filter rounds it into sharply contoured waves.

ECG. The heart's electrical field reaches the scalp, producing a sharp transient time-locked to the QRS, often prominent in infants because the heart is close and the head is small. It is proved by lining the transient up with the ECG channel, which is one reason the neonatal guideline asks for that channel.

Pulse. An electrode over a scalp artery rocks with each pulse, producing slow, rhythmic waves at one electrode, following each QRS with a short delay. Moving the electrode removes it.

Sweat. A warm child sweats, and the changing skin potential produces very slow, high-amplitude drift, usually frontal or temporal. It is slower than any cerebral rhythm and it follows the temperature of the room.

Glossokinetic. The tongue is a dipole too, and its movement with talking, crying or feeding produces slow frontal and temporal waves. It occurs with the activity that generates it and stops when the child is quiet.

The second group of artifacts comes from the recording system and the room, and each has a mechanical signature.

Electrode pop. A poor contact or a sudden change in the electrode-gel-skin interface produces an abrupt, very steep deflection confined to one electrode. It is the sharpest thing on the page, which is why it gets called a spike, and it is the easiest artifact to prove: there is no field in the neighbouring electrodes. A cerebral spike at T4 would produce smaller deflections at F8, C4 and T6; a pop at T4 produces nothing there at all. Because the electrode is input to two channels in a bipolar chain, a pop can even produce a false phase reversal, so the absence of a field, not the presence of a reversal, is the test.

60 Hz line noise. Mains interference produces a thick, monotonous band in one or more channels. When it affects one channel, that electrode's impedance is high and the fix is at the scalp. When it affects all channels, the reference or ground is the problem. The notch filter hides it but does not fix it, as Module 1 noted, and a channel full of line noise may be hiding a real finding underneath.

High-frequency oscillatory ventilation. In the NICU, an oscillating ventilator shakes the whole baby at its set rate and produces rhythmic activity at the ventilator rate across many channels. It is monotonous, it does not evolve, and it matches the ventilator setting exactly. The next section returns to it as the most important rhythmic imitator of a neonatal seizure.

The general principle

Equipment artifact is regular in a way that brains are not. Line noise has a fixed frequency; a ventilator has a set rate; a pop has no field. When a waveform is too regular, too confined, or too exactly matched to a machine, look at the machine. Most of these artifacts are also visible to the technologist at the time of recording, so the technologist's notes are part of the read: an impedance problem noted at T4 explains a T4 pop without further work.

The artifacts that cause the most harm in pediatric practice are rhythmic, because rhythmic activity is what a seizure looks like, and a rhythmic artifact in a neonate or a ventilated child can lead to an antiseizure drug the child did not need.

Sucking, patting and rocking. An infant who is fed, patted or rocked moves rhythmically, and so do the electrodes. The result is rhythmic activity at the rate of the movement, which can be in the delta or theta range and can look like a rhythmic discharge. The proof is the video: the pattern begins with the movement and stops when the movement stops. A nurse's hand on the baby's back is a common generator of "rhythmic delta" in the NICU.

High-frequency oscillatory ventilation. The oscillator shakes the infant at its set frequency, producing rhythmic activity across many channels at that rate. It runs for as long as the ventilator does, unchanged, and it matches the setting exactly.

Evolution is the discriminator

Rhythmic artifact does not evolve. A seizure changes over its course in frequency, morphology or location, as Module 10 defines, and the neonatal seizure definition in Module 7 rests on the same idea. Patting stays at the rate of patting; a ventilator stays at the rate of the ventilator; a seizure speeds up, slows down, changes shape and spreads. Ten minutes of perfectly regular rhythmic activity in every channel is a machine, not a brain.

Because the stakes are a drug, the rule is absolute: always check the video before calling a neonatal or ICU seizure. If there is no video, say so in the report and describe the pattern as rhythmic activity of uncertain origin rather than as a seizure. And look for the generator: the ventilator's set rate, the nurse's hand, the feeding. When the generator is found, the report can say what the rhythmic activity was, which is more useful to the team than an equivocal description.

Put the rules together into a routine you can apply to any doubtful waveform.

The table collects the artifacts in this module with their appearance and their proof.

The sign-off for this module is ten artifacts correctly labelled on real records, at least three of them from infants or the ICU. Label each one with its proof, not just its name: "ECG artifact, time-locked to the QRS" is a labelled artifact; "artifact" is a guess.