Environmental & Technical Artifacts

Where physiologic artifacts originate inside the patient, technical artifacts originate in the apparatus and its surroundings: the electrodes, the leads, the amplifier, the environment of electrical machinery, and the mechanical interface between metal and skin. They are, in principle, fully preventable, because they are problems of engineering rather than of biology. The mark of a skilled technologist is not the ability to filter these artifacts after the fact but the ability to recognize their electronic signatures in real time and eliminate them at the source. The reader who understands the circuit understands the artifact, and the recurring theme of this module is that almost every technical artifact is a symptom of a failure somewhere in the chain that runs from scalp to display - a chain comprising the electrode-electrolyte-skin junction, the lead wire, the amplifier input, the analog-to-digital converter, the reference, and the montage reconstruction.

Modern digital systems have changed the texture of this chain without changing its physics. Because acquisition is now referential and montages are reconstructed in software, a single bad electrode contaminates every derivation in which it participates, and the same epoch can be re-montaged instantly to test a hypothesis - a diagnostic luxury the paper era lacked. At the same time, digital systems introduce their own failure modes: aliasing if the anti-aliasing filter is inadequate relative to the sampling rate, quantization and saturation when an input exceeds the converter's dynamic range and clips into a flat-topped square wave, and digital filtering artifacts such as the ringing that a steep filter can ring around a sharp transient. The contemporary reader must therefore reason about both the analog interface and the digital pipeline.

Line Noise and the Failure of Common-Mode Rejection

Mains-frequency interference - 60 Hz in North America, 50 Hz across most of Europe, Asia, and Africa - is the most familiar environmental artifact, appearing as a perfectly regular sinusoidal contamination at the line frequency. To understand why it appears, and more importantly why it sometimes does not, one must understand the differential amplifier. Every EEG channel measures the voltage difference between two inputs, and it does so with a differential amplifier whose great virtue is common-mode rejection: any signal that appears identically on both inputs, such as the ambient electromagnetic field radiating from building wiring, is subtracted away and ideally vanishes. The ambient 50 or 60 Hz field is very nearly a common-mode signal because it bathes both electrodes almost equally. In a healthy recording it is therefore almost completely rejected, and a clean EEG shows little or no line frequency at all. The figure of merit, the common-mode rejection ratio, is extraordinarily high in modern amplifiers, which is precisely why a well-prepped study needs no notch filter.

Line noise appears prominently only when common-mode rejection fails, and the dominant cause of that failure is impedance imbalance between the two electrodes of a derivation. The common-mode rejection of the amplifier depends on the two input impedances being matched; when one electrode has high or unstable contact impedance, the shared ambient field is no longer divided equally between the two inputs, the common-mode signal is partially converted into a differential signal, and 50 or 60 Hz blooms in that channel. This yields the diagnostic insight that line noise is rarely a problem of the room and almost always a problem of an electrode. A single channel buzzing with line frequency points to a bad electrode in that derivation, not to a need for more filtering; line noise across many channels points to a poor common ground, a disconnected ground or reference, or a powerful nearby source such as an infusion pump, a dialysis machine, or a poorly shielded bed. The correct response is to identify the offending electrode - it will be the common member of the noisy derivations - and to re-prep and re-gel it to lower and balance its impedance, restoring the amplifier's ability to reject the field.

This is precisely why the notch filter is a crutch rather than a remedy. A notch filter is a narrow band-stop filter centered on the line frequency that surgically removes energy in a narrow band around 50 or 60 Hz. It can render an otherwise unreadable trace legible, and in that narrow sense it is useful, but it carries hazards the expert never forgets. First, it does not fix the underlying problem; an electrode whose impedance is high enough to break common-mode rejection is also an electrode whose signal fidelity is compromised across the whole spectrum, so applying a notch masks the symptom while the diseased electrode continues to distort the data in ways the notch cannot touch. Second, the notch removes genuine cerebral and pathologic information that lives at the line frequency. Real EEG has spectral content at 50 and 60 Hz, and so do clinically meaningful signals, most notably high-frequency oscillations - ripples in the 80 to 250 Hz band and fast ripples in the 250 to 500 Hz band - that are of intense and growing interest as biomarkers of the epileptogenic zone. A naive notch, or worse a cascade of notches at the line frequency and its harmonics, can blindly excise or distort exactly the band a surgical workup may depend upon. Third, a steep notch can introduce ringing around sharp transients, fabricating small oscillatory wavelets that did not exist in the data.

Electrode Pops and the Mechanics of the Skin Interface

An electrode pop is the signature of a momentary failure at the electrode-electrolyte-skin junction. It appears as an abrupt, often dramatic vertical deflection - a step or a sharp transient, sometimes with an exponential return to baseline - that is strictly confined to the single electrode involved. The mechanism is a sudden change in the electrochemical contact: a drying gel bridge, a momentary loss and re-establishment of contact, a hair trapped under the disc, or a static discharge across the junction, any of which transiently alters the half-cell potential of that one electrode. The defining feature is the field, or rather the absence of one. Because a pop arises in a single electrode rather than in a region of cortex, it has no physiologic field whatsoever. In a bipolar montage it appears in exactly the two derivations that share the affected electrode, and with equal amplitude and opposite polarity in those two derivations - a pattern that superficially resembles a phase reversal but violates the smooth amplitude gradient of a true cerebral phase reversal, where the maximum is graded across several contacts. In a referential montage it appears in the single channel of that electrode.

Electrode pops are among the most frequent false alarms in long-term monitoring precisely because their abrupt vertical morphology can superficially resemble a spike, and because automated spike-detection software is notoriously prone to flagging them. The discriminator is unfailing and worth stating as a rule: a transient that involves only one electrode, with no spread to its physical neighbors and no plausible cortical field, is an electrode artifact regardless of how sharp or how large it appears, because no cortical generator projects to one electrode and abruptly to none of its neighbors. When pops recur, the offending electrode should be re-prepped or replaced. Reviewing the same epoch in more than one montage is the fastest confirmation: the pop will follow the single electrode through every montage, appearing only in derivations containing that electrode, while a true discharge will redistribute according to the logic of each derivation chain and maintain a coherent field. A false phase reversal across a popping electrode is the classic trap, and re-montaging to a referential display, where the pop collapses onto its single channel, dispels it instantly.

Movement, Sweat, and Impedance: The Mechanical Domain

Movement artifact is the broad category produced when the patient, the leads, or the electrodes physically move. Mechanical motion changes electrode contact and induces voltages in the moving wires, producing irregular, often large, frequently slow deflections that involve multiple electrodes simultaneously and without any of the orderly field structure of cortical activity. Rhythmic movements deserve particular caution. Tremor, patting, chewing, sucking, rhythmic rocking, hiccups, or the periodic motion of a ventilator or an oscillating cooling device can imprint a rhythmic artifact on the EEG that, taken in isolation, can be mistaken for a rhythmic cerebral discharge or even an electrographic seizure. The protection against this error is twofold: the artifact will lack the smooth evolution in frequency, morphology, and field that defines a true seizure, and synchronized video will reveal the movement that generates it. This is the central justification for video-EEG; many rhythmic artifacts are diagnosable only by watching what the patient is doing while the rhythm appears, and a frequency that tracks a ventilator rate or a compression rate rather than evolving is artifact by definition.

Sweat artifact is a distinctive slow contamination produced when a film of perspiration, rich in electrolytes such as sodium chloride and lactate, forms across the scalp. The sweat alters electrode impedance and can physically bridge adjacent electrodes, and the autonomic and thermal fluctuations of sweating impress a very slow, undulating, large-amplitude wave on the affected channels, typically well below 1 Hz and often in the 0.1 to 0.5 Hz range. This slow rolling baseline can obscure the underlying EEG and, when prominent, can be mistaken for pathologic delta activity. The cues are its very low frequency - below the band of physiologic delta - its broad and shifting distribution following the sweat film rather than a vascular or cortical territory, and its prompt response to cooling the patient or the room. Engaging a more aggressive low-frequency (high-pass) filter will flatten the slow undulation and can rescue a swamped trace, but the same caution applies as with the notch: filtering treats the display, not the cause, and an over-aggressive high-pass filter distorts genuine slow activity and can even create artificial sharp transients at the edges of the filtered slow waves.

A related and more insidious phenomenon is the salt bridge, in which the conductive film of sweat or excess gel literally connects two neighboring electrodes, short-circuiting them so that their derivation reads near zero. A suspiciously flat, isoelectric channel between two electrodes that should differ is a classic sign of a salt bridge and is corrected by drying and separating the electrodes; it is dangerous because an artificially silent channel can be misread as focal attenuation or, in the wrong context, as evidence bearing on a brain-death evaluation. The lesson is that artifacts can falsely create activity and can also falsely abolish it, and the reader must be alert to both directions of error.

Underlying several of these artifacts is the single most important quality metric in EEG recording, electrode impedance. Impedance is the opposition to current flow at the electrode-skin interface, and the standard of good practice, codified by professional society guidelines, is to keep each electrode below 5 kilohms while keeping impedances reasonably balanced across electrodes; many laboratories also accept a small floor above which impedance should not fall, since a near-zero reading can itself signal a salt bridge. High impedance degrades signal quality directly by producing baseline drift and excess thermal noise, and indirectly by breaking the common-mode rejection that suppresses line noise; imbalanced impedance is in some respects worse than uniformly high impedance because it is the mismatch between inputs that converts common-mode interference into differential artifact. Many technical artifacts - line noise, drift, instability, susceptibility to movement - share a single root cause in poor or unbalanced impedance, which is why the impedance check at the start of every recording is not a formality but the foundation of a clean study.

A practical bedside sequence operationalizes this principle. When a noisy channel appears, the disciplined technologist first re-montages the epoch to localize the problem to a single electrode or a shared reference; then checks and corrects impedance at the implicated electrode; then inspects the lead and connector for mechanical fault; then looks for an environmental source and, if many channels are affected, for a ground or reference failure; and only after exhausting these source-level fixes considers a filter, documenting it explicitly. This ordering - re-montage, re-prep, re-secure, remove source, and only then filter - is the difference between a study that is clean and a study that merely looks clean while concealing a compromised electrode beneath a notch.