Artifact Recognition
Artifact Recognition
John Giang, MD
Purpose: separate non-cerebral signals from brain activity so that artifact is not reported as spikes, seizures, or attenuation, and recognize when the recording itself is not good enough to interpret.
Why it matters
- Artifact is the leading cause of EEG over-reading; the physiological patterns it mimics (spike-and-wave, rhythmic ictal evolution, electrodecrement) have already been described in earlier sections, so the task here is knowing the look-alikes.
- Children generate more artifact than adults: crying, sucking, patting, rocking, chewing, hiccups, and poor cooperation, and in the NICU the environment adds ventilators, pumps, warmers, and ECG.
- A technically poor study (high impedance, wrong filters, too short) lowers sensitivity and specificity at the same time, so technical review comes before pattern review.
Key points
Rule of thumb for any suspicious waveform
- Does it have a physiologic field (gradient across neighboring electrodes), or is it confined to one electrode or spread identically everywhere?
- Is it time-locked to something non-cerebral (QRS, respiration, a movement on video, a blink)?
- Does it respect the background (physiologic discharges disrupt it; artifact sits on top)?
- Can the technologist reproduce or abolish it (ask the child to move the tongue, re-gel the electrode, pause the pump)?
Physiologic artifacts
|
Artifact |
Appearance |
Pitfall it mimics |
Confirmation |
|---|---|---|---|
|
Eye blink / eye closure |
High-amplitude positive deflection at Fp1-Fp2 (cornea positive, Bell's phenomenon); eye opening gives a bifrontal negative deflection; eye flutter gives repetitive frontal positivity |
Frontal slow waves or spike-and-wave; blinks have no preceding spike, no posterior field, and do not disrupt the background |
Frontal-only field, synchronous with video blink |
|
Lateral eye movement |
Opposite polarity at F7 and F8; the side looked toward becomes positive, so bipolar chains show a positive phase reversal on that side and a negative one contralaterally |
Bitemporal sharp waves; slow undulating frontotemporal waves in drowsiness mimic temporal slowing |
Opposite polarities at F7/F8 on the same wave |
|
Muscle (frontalis, temporalis) |
Very fast, often low-amplitude activity riding on the background, frontal and lateral temporal, minimal at the vertex |
Beta activity, polyspikes, or GPFA, which is slower (beta range), lower amplitude, and generalized |
Disappears with relaxation or jaw opening; faster than any cortical rhythm scalp EEG can resolve |
|
Chewing / sucking |
Sudden bursts of generalized very fast temporalis EMG, rhythmic at chew or suck rate, often with slower glossokinetic delta |
Generalized paroxysmal fast activity; in neonates, rhythmic sucking artifact imitates a focal rhythmic seizure |
Video; the rhythm matches the mouth movement |
|
Glossokinetic (tongue) |
Diffuse, synchronous delta, maximal frontally and in the temporal chains (tongue tip is negative) |
2.5 Hz bifrontal spike-and-wave; the real discharge has a small spike before each slow wave and tapers posteriorly |
Reproduce with "la-la-la" |
|
ECG |
Small regular sharp transients time-locked to the QRS, usually more prominent or only present on the left, especially in referential montages and in infants with short necks |
Posterior epileptiform discharges or POSTS, which are not QRS-locked; periodic discharges |
Overlay with the ECG channel |
|
Pulse (cardioballistic) |
Slow, periodic wave at one electrode lying over an artery, lagging the QRS by about 200–300 ms |
Focal delta slowing |
Single-electrode field, fixed delay from QRS |
|
Respiration, hiccups, patting, rocking |
Slow, monotonous rhythmic waves at breathing rate (ventilator or chest wall) or at the rate of the caregiver's hand; hiccups give brief generalized movement bursts |
Rhythmic delta mimicking a neonatal seizure or periodic pattern; never evolves in frequency or morphology |
Respiratory belt, nursing notes, video |
|
Sweat and salt bridge |
Very slow (<0.5 Hz) drifting baseline in sweat; a gel bridge between two electrodes makes their bipolar derivation nearly flat |
Diffuse delta; focal attenuation or "suppressed" channel |
Drift respects no field; a flat single channel that is normal in referential montage is a bridge |
Instrumental and environmental artifacts
|
Artifact |
Appearance |
Pitfall it mimics |
Confirmation |
|---|---|---|---|
|
Electrode pop |
Abrupt, steep, single-electrode transient with instant return, no field |
Focal spike or sharp wave |
Phase reversal with no spread to neighbors; recurs at the same electrode; fixed by re-gelling |
|
Lead movement / high impedance |
Irregular slow waves or noise confined to one electrode; polymorphic, often with 60 Hz riding on it |
Focal slowing, focal attenuation if the channel is dead |
Impedance check; one electrode only |
|
60 Hz (50 Hz) mains |
Monotonous fast activity, uniform across channels sharing a bad electrode or ground |
Muscle or beta; obscures low-amplitude fast activity |
Remove the source first; apply the notch only when nothing else works, because it can attenuate spikes |
|
Pumps, ventilators, ECMO, warmers, pacifiers, phones |
Regular, stereotyped waveforms at the device rate, often with a sharp component, identical in all channels or limited to the electrodes nearest the device |
Periodic discharges, rhythmic seizure-like runs, or a "monotonous" background |
Rhythm identical to the device cycle; unaffected by patient state |
Study quality checks before interpreting
- Impedances no higher than 5 kΩ, rechecked whenever a pattern could be artifact.
- Low-frequency filter no higher than 1 Hz and high-frequency filter no lower than 70 Hz for routine review; raising the LFF to hide sweat or lowering the HFF to hide muscle sharpens muscle potentials into spike look-alikes and smooths real spikes.
- Sensitivity 5–10 µV/mm (7 µV/mm standard); over-amplified traces look "slow and disorganized", under-amplified traces look "attenuated".
- At least 20 minutes of technically satisfactory recording, with calibration at start and end and a dedicated ECG channel, which is the single most useful artifact-detection tool.
- Technologist annotations of state, movement, stimuli, and nursing care are part of the data; an unannotated rhythmic run in a neonate should be assumed to be artifact until video or notes say otherwise.
Common pitfalls
- Calling an electrode pop a spike because it "phase reverses": a true discharge has a field across at least two or three adjacent electrodes; a pop lives in one.
- Reading rhythmic patting, sucking, or ventilator artifact as a neonatal seizure: artifact keeps a fixed frequency and shape, while an electrographic seizure evolves.
- Mistaking a flat channel for focal attenuation when the cause is a salt bridge or a disconnected lead; check the referential montage and the impedance log.
- Interpreting ECG transients in referential montages as periodic or posterior discharges, particularly in infants and in left-sided channels.
- Switching on the 60 Hz notch or lowering the HFF to "clean" the record, which blunts genuine spikes and converts muscle into spike-like transients.
- Reporting a drowsy child's slow eye movements as bitemporal delta slowing; the opposite F7/F8 polarities give it away.
- Over-reading a technically inadequate study instead of stating that it is non-diagnostic and should be repeated.