Editing and Identifying Artifacts

Stage 1 sleep graphic © John S. Anderson. A two-channel display: raw EEG from AEEG1 and BEEG2 across a 30-second window, above waterfall spectrograms for each channel whose logarithmic frequency axis runs from about 3.7 Hz to 120 Hz. The high-power peaks cluster at the low end of that axis and the two channels track each other closely, showing the rise in theta amplitude that marks the drift into Stage 1 sleep. The reported mean frequency of 7.73 Hz sits at the theta-alpha boundary; read the peaks rather than the mean, because it is the shift of power down into theta, not a burst of alpha, that identifies this state.

This image contrasts the same EEG segment under two filter conditions: HFF 55 Hz on the left and HFF 15 Hz on the right. The higher filter setting preserves more high-frequency EMG activity, while the lower setting attenuates it. The longitudinal bipolar temporal EEG channels show prominent, irregular high-frequency activity consistent with myogenic artifact when the high-frequency filter is set to 55 Hz. Reducing the high-frequency filter to 15 Hz markedly attenuates and smooths the fast activity, demonstrating how filtering can suppress the visual appearance of muscle artifact without removing the underlying contamination from the recording.

Graphic adapted from © eegatlas-online.com. This EEG recording captures chewing artifact in a longitudinal bipolar EEG montage, showing rhythmic, high-amplitude EMG bursts from jaw muscle activity that obscure the underlying cerebral signal, especially across anterior and temporal leads.

Graphic adapted from © eegatlas-online.com. This recording illustrates muscle artifact in a double-banana EEG montage. This tracing demonstrates irregular, high-frequency EMG activity superimposed on scalp-recorded EEG, most prominent over the anterior and temporal derivations and relatively reduced along the midline. The artifact is asymmetric, intermittent, and sharply contoured in places, producing spike-like waveforms that may be mistaken for epileptiform discharges if interpreted without attention to frequency content, distribution, field, and temporal context. The preserved high-frequency components are emphasized by the 70 Hz high-frequency filter setting, while the red EKG channel provides cardiac timing for comparison. Overall, the pattern is consistent with myogenic contamination rather than primary cortical activity.

Graphic adapted from © eegatlas-online.com. This graphic illustrates focal EMG artifacts contaminating a longitudinal bipolar EEG montage. The red-outlined regions highlight irregular, high-frequency muscle activity superimposed on the EEG, most prominent in the left anterior temporal and left parasagittal derivations, including FP1–F7, F7–T3, T3–T5, FP1–F3, F3–C3, and C3–P3. The activity is asymmetric, anteriorly weighted, and relatively reduced in posterior and midline channels, a distribution consistent with focal myogenic contamination from scalp, facial, temporalis, or jaw muscle activity. The sharply contoured bursts and fast rhythmic components may obscure the underlying cerebral background and can be mistaken for pathologic fast activity or spike-like transients if interpreted without considering artifact morphology, spatial distribution, and lack of a consistent physiologic cortical field. Calibration markers indicate a 1-second time base and 140 µV amplitude scale.

Graphic adapted from © eegatlas-online.com. This graphic captures rapid blinking artifact in a longitudinal bipolar EEG montage. This tracing shows repeated, high-amplitude, sharply contoured anterior slow transients occurring in rapid succession, maximal in the frontal and frontopolar derivations and spreading posteriorly with decreasing amplitude. The pattern is broadly synchronous across anterior chains and is time-locked to repeated eyelid movements, consistent with rapid blinking artifact. Although eyelid and facial muscle activity may contribute high-frequency EMG components, the dominant waveform morphology reflects ocular/blink artifact rather than cerebral activity. The artifact obscures the underlying EEG background, especially in FP1–F7, FP1–F3, FP2–F4, and FP2–F8 channels, and could be mistaken for frontal rhythmic or sharply contoured activity if interpreted without attention to its stereotyped anterior distribution, repetitive blink timing, and lack of a physiologic cortical field. The EKG channel is displayed separately at the bottom, with calibration markers indicating 1 second and 70 µV.

Graphic adapted from © eegatlas-online.com. This recording captures rapid eye-blink artifact with superimposed muscle artifact in a longitudinal bipolar EEG montage. The tracing shows repetitive, high-amplitude, sharply contoured anterior transients produced by rapid blinking, maximal in the frontopolar and frontal derivations and diminishing posteriorly. These stereotyped blink waveforms recur in close succession and obscure the underlying EEG background, particularly across FP1–F7, FP1–F3, FP2–F4, and FP2–F8 chains. A separate band of irregular high-frequency activity labeled muscle artifact is visible near the lower portion of the recording, consistent with concurrent facial, scalp, or jaw EMG contamination. In contrast, the labeled posterior dominant rhythm is seen over posterior channels as a more regular occipital rhythm, helping distinguish physiologic cerebral activity from ocular and myogenic artifact. The EKG channel at the bottom provides cardiac timing and is not time-locked to the blink or muscle bursts.

This recording captures eye blinks and lateral eye movements. This longitudinal bipolar EEG graphic contrasts three common ocular artifacts. The upper panel shows repeated eye-blink artifacts, appearing as large, stereotyped, frontally maximal slow deflections that are most prominent in FP1–F7, FP2–F8, FP1–F3, and FP2–F4, with attenuation posteriorly. The middle panel demonstrates sharp lateral eye movements, with asymmetric frontal-temporal deflections corresponding to leftward and rightward gaze shifts. The lower panel shows slow lateral eye movement, producing broader, lower-frequency drifting potentials across anterior derivations. Although eyelid and periocular muscle activity may add small fast components, the dominant pattern is ocular rather than primary EMG artifact, reflecting corneoretinal and eyelid-movement potentials that can obscure or mimic frontal cerebral activity. Calibration markers indicate 75 µV amplitude scaling.

This recording captures rapid blinking artifact in a bipolar EEG montage. This tracing shows a burst of repetitive, high-amplitude anterior deflections maximal in the frontopolar and frontal derivations, including FP2–F4, FP1–F3, FP2–F8, and FP1–F7. The waveforms are stereotyped, rhythmic, and sharply contoured, with clear attenuation in central, parietal, temporal-posterior, and occipital channels. The morphology and distribution are most consistent with rapid eye-blink artifact, reflecting ocular potentials with possible superimposed periocular muscle activity, rather than primary cortical activity. Because the repeated frontal transients are large and sharply contoured, they may obscure the underlying EEG and could be misread as frontal rhythmic slowing or epileptiform-appearing activity unless their anterior predominance, blink-like repetition, and lack of a physiologic cortical field are recognized.

Graphic adapted from © eegatlas-online.com. This EEG segment is displayed in a longitudinal bipolar montage with a low-frequency filter of 1.0 Hz, a high-frequency filter of 70.0 Hz, the notch filter turned on, and a simultaneous EKG channel recorded at the bottom. Historically, fast beta activity has been understood in clinical EEG as a low-amplitude, high-frequency background feature that is often most visible over frontal and central scalp regions. It is generally interpreted cautiously because it is highly sensitive to patient state, medication effects, muscle activity, and recording conditions. Cardiac artifact has an equally long-standing role in EEG interpretation as an extracerebral signal that can project into scalp channels and mimic sharply contoured cerebral transients if the EKG channel is not reviewed carefully.

The dominant feature in this tracing is diffuse beta activity. The fast activity is low in amplitude, broadly distributed, and present across multiple bilateral derivations rather than being confined to a single focal region. It appears most conspicuous in frontal, frontotemporal, and central chains, but it is not limited to one hemisphere or one electrode pair. The activity does not show a focal phase reversal, does not organize into an evolving rhythmic discharge, and does not have the morphology of a definite epileptiform pattern in this isolated sample. In clinical terms, this would be described as diffuse excessive beta or diffuse fast activity, while recognizing that the finding is nonspecific.

We adapted an ECG artifact graphic by Garces et al. (2007). The upper trace is labeled as an EEG signal containing ECG activity. Its amplitude is much smaller than the ECG trace, ranging approximately from −0.1 to +0.1 mV, which corresponds to about −100 to +100 µV. Across the 0.5-to-3.0 second epoch, the signal shows an irregular low-amplitude EEG background with superimposed deflections. The red arrows identify small waveform components in the EEG trace that occur at the same times as cardiac events in the lower ECG trace. These deflections are not large compared with the background, but their repeated temporal alignment with the ECG complexes makes them suspicious for cardiac artifact rather than independent cerebral activity.

The lower trace is the ECG signal itself. It has a larger amplitude range, extending roughly from −0.5 to +1.0 mV, and shows three prominent cardiac complexes over the displayed interval, occurring at approximately 0.95, 1.82, and 2.73 seconds. Each complex has a steep, high-amplitude QRS morphology, followed by slower recovery components. The red arrows point to earlier portions of the cardiac cycle immediately preceding the large QRS peaks. These same time points correspond to subtle deflections in the upper EEG trace, demonstrating that components of the cardiac signal are being transmitted into, or recorded by, the EEG channel.

Graphic adapted from © eegatlas-online.com. This EEG segment is displayed in a longitudinal bipolar montage with a low-frequency filter of 0.5 Hz, a high-frequency filter of 70 Hz, the notch filter turned on, and a simultaneous EKG channel shown in red at the bottom. The calibration marker indicates 200 µV, so the prominent activity in the upper left parasagittal channels is relatively high in amplitude compared with much of the remaining background. Historically, this is the type of recording in which careful montage-based reasoning is essential, because a localized electrode artifact can produce a striking apparent abnormality unless one asks whether the waveform is generated by cortex or by a single contaminated electrode.

The most conspicuous abnormal-looking activity is the rhythmic waveform labeled pulse artifact (C3). It is confined almost entirely to the two bipolar derivations that share the C3 electrode, namely F3-C3 and C3-P3. The waveform has a repetitive, rounded, mechanical quality rather than the morphology of a cerebral rhythm. It recurs at a fairly regular interval and appears in opposite polarity across the two adjacent channels, creating a phase-reversal-like pattern at C3. In a true cerebral discharge, a phase reversal can sometimes help localize a cortical voltage maximum, but in this case the localization to a single electrode shared by two channels, the regular pulse-like repetition, and the absence of a physiologic field into neighboring left parasagittal or homologous right-sided channels strongly favor artifact.

This EEG segment demonstrates a bridging artifact involving the left anterior frontal electrodes, most clearly seen in the highlighted Fp1-F3 derivation. Historically, bridging artifact has been recognized as an important technical pitfall in scalp EEG because it can make two nearby electrodes behave as though they are electrically shorted together. When conductive paste, sweat, saline, gel spread, or another low-resistance pathway connects two electrodes, the voltage difference between them becomes artificially reduced. In a bipolar montage, this can produce an abnormally flat or attenuated channel between the bridged electrodes, even while surrounding channels continue to show normal or artifact-contaminated EEG activity.

In this tracing, the highlighted Fp1-F3 channel is strikingly low in amplitude compared with adjacent and homologous derivations. It appears relatively flat and featureless across the displayed epoch, while the neighboring F3-C3, C3-P3, and P3-O1 channels show substantially larger mixed-frequency activity. This pattern is not physiologically plausible as an isolated absence of cerebral activity only between Fp1 and F3. Instead, it suggests that Fp1 and F3 are recording nearly the same electrical potential, leaving little voltage difference for the bipolar amplifier to display in the Fp1-F3 channel.

Graphic adapted from © eegatlas-online.com. This EEG segment shows a physiologic drowsy state in a longitudinal montage, with an EKG channel recorded at the bottom and a calibration of approximately 70 µV with a 1-second time marker. In the historical development of clinical EEG interpretation, drowsiness became important because the transition from relaxed wakefulness into stage N1 sleep changes the background in predictable ways: the posterior dominant rhythm loses persistence, eye blinks diminish or disappear, muscle activity often decreases, and the tracing becomes lower in sustained alpha organization with more mixed low-amplitude activity.

The most important feature in this recording is the attenuation of the posterior dominant rhythm. In a fully awake, relaxed, eyes-closed adult, one would expect a more sustained posterior alpha rhythm, usually maximal over the posterior head regions. In this segment, the posterior rhythm is mostly attenuated, especially across the posterior temporal and occipital derivations, consistent with the transition away from relaxed wakefulness. The tracing is not flat or suppressed; rather, it shows a low-amplitude mixed background in which the organized posterior alpha rhythm is no longer prominent.

This EEG recording is most consistent with a normal drowsy state. The tracing is shown in a longitudinal bipolar montage with a simultaneous EKG channel at the bottom, and the overall appearance is that of a low-amplitude, relatively symmetric background during the transition from relaxed wakefulness toward early sleep. In clinical EEG interpretation, drowsiness is recognized not as a pathologic state but as a physiologic change in background organization. As the patient becomes drowsy, the posterior dominant rhythm that is usually most evident during relaxed eyes-closed wakefulness becomes less sustained or attenuates, eye blinks diminish or disappear, and muscle activity often decreases as facial and scalp tone relax.

In this segment, the posterior dominant rhythm is mostly attenuated. The posterior derivations do not show a robust, continuous alpha rhythm, and instead the tracing contains lower-amplitude mixed-frequency activity distributed across the scalp. This is an expected feature of drowsiness, especially during the transition from wakefulness into stage N1 sleep. The background does not appear globally suppressed; rather, it has lost the organized posterior alpha pattern typical of fully alert relaxed wakefulness.

This EEG segment shows a normal sleep-transition pattern with clear features of drowsiness progressing into stage N2 sleep. The recording is displayed in a longitudinal bipolar montage, with bilateral anterior-posterior chains that allow comparison of frontal, central, temporal, parietal, and occipital regions. Historically, drowsiness in EEG was recognized by attenuation of the posterior dominant rhythm and the emergence of low-amplitude mixed-frequency activity, while stage N2 sleep became defined by the appearance of characteristic graphoelements such as K-complexes and sleep spindles. This image contains both the subdued, mixed background expected with reduced wakefulness and more specific stage N2 features.

The most conspicuous labeled event is a K-complex near the middle of the displayed epoch, around 03:27. It appears as a relatively high-amplitude, sharply contoured slow complex with a broad field, most evident over the frontocentral derivations. Its morphology is consistent with a normal sleep-related K-complex rather than an epileptiform discharge, because it is broad, state-dependent, not followed by an evolving ictal rhythm, and occurs in a background that otherwise shows sleep architecture. The field is not restricted to one electrode or one channel, which argues against a focal electrode artifact.

Later in the segment, beginning around 03:30 and continuing toward 03:32, there are prominent rhythmic waxing-and-waning bursts consistent with sleep spindles. These are most apparent over central and parasagittal derivations, with bilateral expression and a frequency visually compatible with the sigma range, typically around 12–14 Hz. Their morphology is regular and spindle-like, with gradual buildup and decline rather than abrupt onset and termination. This is a normal physiologic sleep pattern and supports classification as stage N2 sleep rather than simple quiet wakefulness.

Graphics © John S. Anderson. The two images above each contain raw EEG tracings from two scalp electrodes: A (EEG1) shows P3–A1, and B (EEG2) shows P4–A2. The raw tracings are displayed with a 0–60 Hz bandpass and a 50 µV vertical scale. Below each set of raw tracings are compressed spectral arrays (CSA) presenting the same information as a three-dimensional image: frequency on the x-axis (here 0–40 Hz), power on the y-axis (0–12.0 in units of µV², which this software labels picowatts), and time on the z-axis (10 seconds).

Note that the top image, recorded during the initial eyes-closed period, shows an active posterior rhythm of about 10 Hz; the spectral display clearly renders bright yellow (higher-amplitude) peaks at 10 Hz. The lower image shows a much slower pattern in the raw tracing, with clear lateral eye-movement artifacts along with some theta and delta activity. The CSA reflects the same change, including the loss of the 10 Hz posterior rhythm. Together, these two images detail the transition from an awake, eyes-closed EEG to a drowsy, Stage 1 pattern.

Graphic adapted from © eegatlas-online.com. This EEG segment shows a prominent movement artifact in a Cz-referential montage, with the most obvious contaminated interval highlighted near the right side of the tracing. The recording shows multiple scalp derivations referenced to Cz, with a simultaneous EKG channel at the bottom. Before the highlighted event, the EEG background consists of relatively lower-amplitude mixed-frequency activity with intermittent slower fluctuations. During the highlighted interval, however, there is an abrupt, large-amplitude, irregular disturbance that appears nearly simultaneously across many EEG channels.

The artifact has the typical appearance of movement contamination: the waveforms are large, abrupt, jagged, and nonphysiologic, with sudden baseline shifts and superimposed irregular faster components. The activity does not resemble a normal cerebral rhythm, a focal epileptiform discharge, or an evolving seizure pattern. Instead, it appears as a broad mechanical disruption of the recording. The morphology varies from channel to channel, but the timing is shared across much of the montage, which suggests movement of the patient, electrodes, leads, or reference pathway rather than a localized cortical generator.

The simultaneous disturbance in the EKG channel is especially important. The red EKG tracing at the bottom shows a clear disruption during the same interval as the EEG artifact, with large noncardiac deflections superimposed on or replacing the usual cardiac rhythm. When a large EEG disturbance is also evident in the EKG channel, the finding strongly supports a non-cerebral source, because true cortical activity would not be expected to produce a simultaneous large artifact in the cardiac lead. This temporal correspondence indicates that the event likely reflects body movement, cable movement, electrode displacement, or generalized mechanical disturbance affecting multiple recording channels.

In the left panel, labeled “Head movement O2,” the most conspicuous abnormality is in the P8-O2 derivation. This channel shows a repetitive, relatively large-amplitude, slow rhythmic waveform that is not seen with comparable amplitude in the neighboring right temporal or left-sided channels. Because the affected channel includes O2, the pattern suggests movement or mechanical instability involving the right occipital electrode region. The waveform has a smooth, repetitive, mechanical quality rather than the spatially distributed field expected from a physiologic posterior rhythm. Its confinement to a posterior derivation and its regular movement-like morphology favor head movement or electrode motion near O2 rather than a focal occipital cerebral discharge.

The middle panel, labeled “Respiration,” shows slower rhythmic baseline fluctuations that correspond to breathing-related movement. The arrows mark respiratory cycles, and several channels show broad, slow deflections that rise and fall in a pattern compatible with chest, neck, head, or cable movement during respiration. This type of artifact may be more prominent in temporal or referential derivations, depending on electrode placement, wire tension, body position, and reference configuration. Unlike cerebral rhythmic activity, respiratory artifact tends to have a slow, periodic cadence tied to the breathing cycle, often with broad baseline shifts rather than organized cortical waveforms.

The right panel, labeled “Tongue – F8,” demonstrates a more irregular movement and muscle-related artifact, maximal in the F8-T4 derivation. The highlighted channel contains high-amplitude, jagged, irregular activity, with additional contamination in adjacent right anterior and temporal channels such as Fp2-F8 and T4-T6. This distribution suggests a source near the right anterior temporal/frontotemporal region, compatible with tongue movement, jaw activation, or nearby facial muscle activity affecting the F8 electrode region. The waveform is not sinusoidal or physiologically organized; instead, it is abrupt, uneven, and sharply irregular, which is typical of orofacial movement or electromyographic contamination.

The 60-Hz artifact graphic © John S. Anderson. This EEG segment is shown in a referential montage in which the scalp electrodes are referenced primarily to A1, with a sensitivity of 15 µV/cm. The recording is heavily contaminated by a widespread, regular, high-frequency signal consistent with 60-Hz line-frequency artifact. The artifact appears as a fine, tightly spaced, nearly sinusoidal oscillation riding on top of the slower EEG background across many channels. Its uniformity and persistence distinguish it from physiologic beta activity, which is usually less mechanically regular, more variable over time, and more dependent on state, medication, or regional scalp muscle activity.

The contamination is broadly distributed across the montage rather than limited to one physiologic field. Channels from frontal, central, temporal, parietal, and occipital regions all show varying degrees of this fast regular activity. Because the montage is referential, a problem involving the reference, ground, electrode impedance balance, environmental electrical interference, or amplifier shielding can project the artifact widely across the recording. The A1-referenced layout is important because contamination affecting the reference pathway can make a technical problem appear generalized.

Graphic adapted from © eegatlas-online.com. This EEG segment demonstrates radiofrequency contamination, labeled here as a telephone artifact, in a longitudinal bipolar montage. The recording uses filters from 1.0 to 70.0 Hz with the notch filter turned on, and includes simultaneous EKG and photic channels. The artifact is most clearly visible in the boxed interval between approximately 4 and 6 seconds, where a rhythmic, tightly packed, high-frequency oscillation appears abruptly in several channels. Its sudden onset, nearly mechanical regularity, and abrupt disappearance strongly favor contamination from an external electronic source rather than physiologic cerebral activity.

Graphic adapted from © eegatlas-online.com. The recording is displayed in a longitudinal bipolar arrangement with an EKG channel at the bottom. The red EKG tracing is regular and does not account for the large irregular activity in the highlighted EEG channels. The key abnormal-looking activity appears in the P3-C3 derivation and the Pz-Cz derivation, where the waveforms are much larger, more irregular, and more sharply unstable than the surrounding EEG background. These discharges fluctuate abruptly, with jagged high-amplitude deflections and intermittent vertical transients. Their morphology is not consistent with a normal cerebral rhythm, an evolving seizure pattern, or a reproducible epileptiform field.

Graphic © John S. Anderson. This EEG segment shows an electrode-pop artifact localized to the T3 electrode, displayed in a referential montage with most channels referenced to A1. The circled event occurs around 06:55 to 06:56 and is most conspicuous in the T3-A1 channel, where there is an abrupt, high-amplitude, sharply contoured voltage excursion followed by slower baseline recovery. Historically, this type of artifact has been recognized as a common technical pitfall in EEG because sudden changes at the electrode–scalp interface can create waveforms that look strikingly sharp or paroxysmal, yet are not generated by cerebral cortex.

The morphology is typical of an electrode pop. The waveform begins suddenly, has a steep vertical component, and is followed by a slower, irregular return toward baseline. This combination of an abrupt transient and subsequent baseline drift suggests a sudden impedance change or intermittent loss and recovery of electrode contact. Common technical causes include drying conductive paste, a loose electrode, tugging on the lead wire, sweat or motion at the electrode site, or a brief mechanical shift at the scalp–electrode interface. Because the affected electrode is T3, the artifact appears most prominently in the T3-A1 derivation.