Polysomnographic artifacts are extraneous signals in a recording channel that obscure or provide false data. Because we simultaneously record multiple different biophysiological signals for long hours, artifacts in polysomnography (PSG) are ubiquitous and pervasive. Artifacts in PSG can be good, bad, or ugly. Examples may be easily identified. Artifact is good when muscle or movement artifacts herald arousals, shifts in body position, or parasomnias. Bad when false theta activity due to digital EEG aliasing is misinterpreted as EEG slowing (1). Ugly when an implanted deep brain stimulator (DBS) totally obliterates all our alternating current (AC) recorded signals (2). In this chapter, we discuss common and uncommon artifacts in PSG, their salient features, causes, clinical significance, and strategies for fixing them when needed.
UNDERSTANDING WHICH PHYSIOLOGICAL SIGNALS ARE RECORDED IN PSG
PSG is a diagnostic tool in sleep medicine during which multiple different biophysiological signals are continuously and simultaneously recorded across a sleep period to characterize sleep and identify sleep disorders. The name is derived from the Greek root πολύς (polus for “many, much”) and the Latin root, somnus (“sleep”), and Greek γράφειν (graphein, “to write”). The resulting study is a polysomnogram (also abbreviated as PSG).
In our sleep laboratory, we typically record 11 to 16 (or more) different biophysiological variables for 6 to 8 hours when performing a comprehensive overnight in-laboratory (level 1) PSG. Given the number of different physiological signals and time duration recorded in PSG, the number and types of artifacts encountered often seem exponential.
Identifying artifacts in PSG begins by knowing which biological signals are recorded and the normal technical settings, waveforms, and patterns for each of these (3). We can score sleep/wake states using only three physiological parameters: EEG, eye movements (electrooculography [EOG]), and submentalis electromyography (chin EMG). However, far more physiological parameters are typically recorded to identify or exclude sleep pathologies and disorders.
Most comprehensive overnight in-laboratory (level 1) PSGs are recorded to identify sleep disordered breathing, most often obstructive sleep apnea (OSA). Figure 11.1 shows a 120-second epoch record in a 60-year-old man with severe OSA. Almost all of the respiratory parameters recorded in PSG provide only qualitative data. To compensate for this, multiple different respiratory signals are recorded (nasal pressure [NP] airflow, oronasal thermal sensors, thoracic and abdominal respiratory effort, electrocardiogram, and pulse oximetry); redundancy is useful when one or more fail. We additionally monitor carbon dioxide (CO2) using end-tidal and/or transcutaneous CO2 with capnometry when recording PSG in adults with suspected sleep-related hypoventilation and all children. Figure 11.2 shows a 120-second epoch of PSG recorded on a child with severe OSA and Down syndrome showing how we also record CO2. When introducing and titrating positive airway pressure (PAP) to assess its effect on sleep disordered breathing, we additionally record PAP flow (Figure 11.3) and sometimes PAP pressure and leak.
Sleep laboratories are certified by the American Academy of Sleep Medicine (AASM). We record, score, and interpret PSGs using the AASM 260Manual for the Scoring of Sleep and Associated Events (4). It provides rules for scoring sleep stages, arousals, and respiratory, cardiac, movements, and parasomnias during sleep. It also specifies standard montages, electrode placements, and digitization parameters. Table 11.1 summarizes standard low-frequency (LFF) and high-frequency filter (HFF) settings for each of the PSG signals.
Following the AASM Scoring Manual, we preferentially record right frontal, central, and occipital EEG referenced to left mastoid (F4-M1, C4-M1, and O2-M1) but also record F3-M2, C3-M2, and O1-M2 as backup (Figure 11.4). However, we record video-PSG with an expanded EEG montage (16–19 or more channels) when sleep-related seizures are suspected (Figure 11.5). We routinely monitor lower extremity limb EMG from the left and right anterior tibialis muscles. However, when rapid eye movement (REM) sleep behavior (REM behavior disorder [RBD]) is suspected, we also record independently left and right flexor digitorum superficialis and extensor digitorum communis muscles to detect REM sleep without atonia (RWA) because RWA and RBD events more often involve the upper extremities (Figure 11.6).
EEG and EOG
Oronasal thermal airflow and thoracoabdominal respiratory belts
DC or ≤0.03
PAP device flow
AASM, American Academy of Sleep Medicine; DC, direct current; EMG, electromyography; EOG, electrooculography; HFF, high-frequency filter; LFF, low-frequency filter; PAP, positive airway pressure.
Recognizing and reducing unwanted artifacts in PSG begins and ends with biocalibration to verify proper functioning of the digital polygraph, electrodes, and sensors. The AASM Scoring Manual recommends systematically performing biocalibration at the start and end of the PSG recording. Biocalibration provides a wealth of information for scoring events and artifacts in the PSG. Table 11.2 summarizes the protocol and rationale for biocalibration recommended by the AASM Scoring Manual.
When forced to classify PSG artifacts into types, we favor first dividing them into two groups: physiological and nonphysiological. Physiological artifacts originate from the patient (e.g., cardiac, muscle, respiration, sweat, and movements of eyes, tongue, body, and limbs). Nonphysiological artifacts can be further subdivided into two types: instrumental (artifacts that originate from the instruments and equipment used to record the PSG) and environmental (artifacts that arise from nearby surrounding environment). Table 11.3 summarizes these.
ARTIFACT IDENTIFICATION STRATEGIES
Identifying artifacts in PSG begins by knowing which biological signals are recorded and the normal ranges and patterns for each of these (3). Many artifacts in PSG can be recognized by their characteristic appearance and distribution but others only identified if the technologist identifies them at the time of the recording. Identifying artifacts in PSG begins by knowing which biological signals are recorded, and the normal ranges and patterns for each (3). Observing unexpected or unusual patterns in the PSG should prompt the questions shown in Figure 11.7.
Many PSG artifacts can be reduced or eliminated by (a) proper application of recording electrodes to obtain low and nearly equal impedances, (b) electrodes and sensors placed in the proper locations and securely attached, (c) properly grounded equipment; and (d) reducing electrical and electromagnetic interference (EMI) in the surrounding recording environment. The sleep technologist identifying an artifact needs to document its presence in the comment log, determine what is needed to fix/eliminate it, and finally consider whether it affects patient safety, or prevents interpretation of the PSG. Sometimes, letting the patient sleep is the best choice.
Check, Observe, and If Needed Fix
Check and document impedances for all the EEG, EOG, and EMG electrodes; fix if needed.
Ensure impedances for EEG, EOG, and EMG derivations are ≤5 kΩ and relatively equal
Wake EEG recorded for 30 seconds with patient lying quietly with eyes closed.
Identify presence of dominant posterior rhythm for scoring wake and sleep onset
Look up, down, left, right, and blink without moving head.
Identify presence and direction of horizontal and vertical eye movements; appreciate more subtle eyeblinking.
Grit teeth or chew.
Assess display of chin EMG; may see glossokinetic artifacts (reproduced by asking patient to say Tom Thumb or Lilt-Lilt-Lilt)
Snore or hum.
Check snoring microphone channel
Breathe in and out.
Check for sufficient sensitivity and proper polarity of airflow and respiratory effort channels
Hold breath for 10 seconds.
Ability to record an apnea
Hold breath for 10 seconds.
Ability to record an apnea
Breathe only through nose for 10 seconds, then only through mouth for 10 seconds.
Nasal pressure transducer and oronasal thermal sensors to detect air flow are functioning properly, have sufficient sensitivity and appropriate polarity
Take a deep breath and exhale slowly for 10 seconds (prolonged expiration)
Identify prolonged expiration seen with sleep-related expiratory groaning
Flex left foot/raise toes on left foot (×5) then flex right foot/raise toes on left foot (×5).
262Optimal functioning of anterior tibialis leg EMG
Flex fingers on right hand (×5) and then left hand without bending at the two distal joints, then extend fingers on left and then right hand without moving the wrist if upper extremity EMG is recorded.
Optimal functioning and display of flexor digitorum superficialis and extensor digitorum communis when recording upper limb EMG muscles for suspected REM behavior disorder
EMG, electromyography; EOG, electrooculography; REM, rapid eye movement.
SEMs, REMs, eye flutter, eyeblinks, nystagmus, asymmetric EOG
Muscle and movement
Muscle; whole body movements, position shifts; sucking, chewing, eating; HT; ALMA; RMD; tremor; myokymia; hemifacial spasm; RBD events; NREM arousal parasomnias
Sweat; galvanic sympathetic skin response
Tongue (glossokinetic); palatal myoclonus, glottis click, inspiratory stridor, snore vibration
ECG, pulse, ballistocardiographic
263Electrode pop, high impedance, imbalanced electrode pairs, ground recording artifact
Channel inversion, improper settings
Jack box, amplifiers
60-Hz artifact, electromagnetic, electrostatic
Smartphone, cellphone, laptop, tablets
Oscillating beds, fans, ventilators, IV pumps, sequential compression devices
Hypoglossal nerve stimulator, deep brain stimulator, vagal nerve stimulator, cardiac pacemaker, cardiac defibrillator
Chest percussion, patting, rocking, feeding, CPR.
ALMA, alternating leg muscle activation; EOG, electrooculography; CPR, cardiopulmonary resuscitation; HT, hypnagogic tremor; IV, intravenous; NREM, nonrapid eye movement sleep; RBD, REM behavior disorder; REM, rapid eye movement; RMD, rhythmic movement disorder; SEM, slow eye movement.
The ability to re-reference electrodes can be invaluable for identifying and ameliorating a questionable or spurious PSG signal (3). If, after re-referencing, an artifact appears in another channel, the amplifier is not the problem. If an amplifier channel goes bad, the input can be moved to a spare channel. If the mastoid reference is bad on one side, the electrodes can be re-referenced to the ipsilateral rather than contralateral side (makes scoring the amplitude of slow wave activity in nonrapid eye movement sleep [NREM] 3 a bit off but tolerable).
Some artifacts have a stereotypical appearance, so pattern recognition of these is often helpful. When observing medium-to-high amplitude atypical waveforms in only one EEG derivation, this should warrant suspicion that these are artifact because EEG typically distributes a maximal potential which gradually drops off with increasing distance across the scalp. Most in-laboratory PSG records have time-locked video using ultraviolet light. Video can be invaluable in confirming PSG artifacts and diagnosing sleep disorders (benefits of which are summarized in Table 11.4).
• Identify the artifact, movement, or parasomnia
• Was that a snore, snort, sigh, gasp, wheeze, or stridor?
• Patient out of bed, technologist or parent in room
• Who really is snoring?
• What caused the probe to malfunction?
• Who or what caused the arousal?
The AASM Manual specifies backup electrodes which are to be recorded and deployed if those fail (4). For example, the recommended EEG derivations are F4-M1, C4-M1, and O2-M1; but if M1 malfunctions and cannot be fixed, the backup montage is F3-M2, C3-M2, and O1-M2. Three chin EMG electrodes are mandated, such that changing pairs can eliminate the signal emanating from the bad electrode. When recording respiratory signals the preferred sensor for identifying an apnea during a diagnostic study when the oronasal thermal airflow sensor is not functioning or the signal unreliable is to use: NP sensor or respiratory inductance plethysmography (RIP) flow or sum signals. The preferred sensor for scoring hypopnea is the NP transducer sensor; if malfunctioning or unreliable, we use oronasal thermal sensors, RIP sum or flow, or dual thoracoabdominal RIP belts.
Physiological artifacts are those which originate from the patient. Given the long duration of overnight PSG, these are diverse and pervasive. Some are welcome, helping us to recognize diagnostic sleep pathologies. Others so obscure the recording as to render parts of the recording unreadable. One of the technologist’s crucial roles is to recognize artifacts and, when possible, fix or eliminate them. The most common are muscle, muscle, sweat, respiration, cardiac, and movements of the eyes, tongue, and limbs.
Muscle Artifact (Electromyography)
EMG muscle artifact is one of the most common physiological artifacts in PSG, observed when muscle groups beneath the particular electrode site contract (3). Muscle artifacts often intrude into EEG, EOG, and EMG signals. They are often of much higher amplitude (≥100 µV), two to five times higher than the EEG.
Most scalp-recorded muscle artifact in PSG emanates from frontalis, temporalis, and occipitalis muscles. Using expanded EEG montages, muscle artifact is most prominent in the frontal (F3, F4), midtemporal (T3, T4), and parietal (P3, P4) electrodes reflecting frontalis, temporalis, and masseter muscles. Occipital electrodes (O1, O2) may pick up EMG artifact from the occipitalis, trapezius, and supraspinal muscles. Like EEG tracings, muscle artifact tends to be least over the midline channels (Fz, Cz, and Pz). We use this to our advantage; when recording patients suspected of sleep-related hypermotor epilepsy, electrographic seizures may be best seen at the Cz electrode derivation (Figure 11.5).
The power spectrum of contracting striated muscle measured by surface EMG has a broad bandwidth of 20 to 300 Hz, but most of the power is between 15 and 100 Hz (5). The peak frequency of masseter muscles when chewing is 50 to 60 Hz, 30 to 40 Hz for the frontalis muscles when wrinkling the brow, and 40 to 80 Hz for the temporalis muscles when jaw clenching. The posterior head muscles have higher peak frequencies around 100 Hz although they may vary depending on the direction, force of contraction, and the patient’s sex. The peak frequency can extend up to 600 Hz in some facial muscles due to their smaller size and higher innervation ratio (5).
If muscle artifact is not too prominent during wakefulness, we tolerate it, cognizant that interference will decrease with sleep onset. Skeletal EMG activity typically shows a progressive decrease with deepening NREM sleep and reaches its lowest levels in REM sleep. One of the most important biomarkers of REM sleep is a low-amplitude chin EMG. A poorly attached chin EMG can prevent optimal scoring of chin muscle atonia in REM sleep (Figure 11.8). The sudden appearance of prominent muscle artifact from NREM 3 sleep, coupled with a diffusely slowed EEG and sinus tachycardia, prompts review of the video to identify a NREM arousal parasomnia (Figure 11.9). Excessive amounts of muscle activity/movement artifact during REM sleep (accompanied by other biomarkers of REM sleep such as rapid eye movements and irregular respiration) warrants review of the video (to identify RBD events) and consideration of a diagnosis of REM sleep without atonia associated with RBD (Figure 11.10).
Humans at sleep onset assume a posture of relative repose. The motor system reduces its level of activity and tone even before sleep onset. Subsequent movements during sleep typically occur during (a) postural shifts, 265(b) arousals or awakenings, (c) sleep stage shifts, or (d) following paroxysmal events (hyperpneic phase of a respiratory event, a periodic limb movement, seizure, or parasomnia). The frequency of body movements is a fairly reliable measure of sleep and sleep stages, decreasing with increasing depth of sleep (wakefulness > NREM 1 > REM > NREM 2 and least during NREM 3 sleep) (6).
What else affects sleep-related movement? Older adults move far less in sleep than younger. One classic study found the mean hourly frequency of body position shifts (postural shifts) decreases with increasing age: from 4.7 per hour of sleep between ages 8 and 12 years to 2.1 per hour between ages 65 and 80 years (7). Progressively more intrusive stimuli lead to recruitment of progressively more extensive motor sequences. Real words or calling a sleeper’s name cause a differential responsiveness. The brain appears to perform a complex set of discriminations which can lead to movement, arousal, or waking when needed. Bedpartners can influence how we move (sleeping well or poorly with each other). And last, those who are depressed and/or have suffered child abuse often exhibit increased motor activity at night (as though sleeping with one eye open).
Body movements in prolonged PSG recordings can cause electrode pops, signal blocking, and muscle artifact in EEG and ECG. Major body movements can affect every channel (Figure 11.11). Repetitive facial movements such as chewing, blinking, hemifacial spasm, or tremor can produce combinations of fast muscle and slow movement artifacts which resemble cerebral discharges, especially if the combinations repeat with similar shape (8). Many sleep-related movements cause stereotypical movement and muscle artifacts, which sleep specialists and technologists learn to recognize. These include chin scratching (Figure 11.12), mother rocking baby (Figure 11.13), sleep bruxism (Figure 11.14), rhythmic movement disorder (Figure 11.15), hypnagogic foot tremor (Figure 11.16), alternating foot tremor (Figure 11.17), excessive fragmentary myoclonus (Figure 11.18), and REM sleep with atonia (Figure 11.10). Observing these prototypic motor patterns when recording or reviewing a PSG should prompt video review to confirm them. It is important to understand that many sleep-related movements have no clinical consequences (do not disturb the patient’s sleep nor impact upon their daytime performance) and are not considered sleep disorders. For example, rhythmic movement disorder (9), hypnagogic foot tremor, alternating leg muscle activation, and excessive fragmentary myoclonus (Figures 11.16–11.18, respectively) are typically asymptomatic (10).
Contrary to popular opinion and traditional teaching, motor symptoms of most diurnal movement disorders do not disappear with sleep (11). Most actually recur (and some persist) during sleep. For example, the rest tremor of Parkinson’s disease (PD) tends to disappear with sleep onset and is rarely seen in deep NREM 3 sleep. Compared to wakefulness, the bursts of tremor during sleep are typically brief (<15 seconds) and most often occur after awakenings; NREM 1 sleep; and during, or in transitions to and from, REM sleep. Tremor artifact is typically rhythmic 4 to 6 Hz, begins abruptly, emanates from the nuchal muscles, and seems isolated from the rest of the EEG background, often maximal in the occipital regions (Figure 11.19). PD tremor is often less or even absent the first 1 to 2 hours after awakening from sleep (so-called sleep benefit). Examples of palatal myoclonus artifact (periodic, very brief, low-amplitude 100–200 millisecond bursts of pharyngeal muscle activity), frontal myogenic artifact (may appear as unilateral periodic bursts of 30–70 Hz muscle activity lasting 200–400 milliseconds, bursts recurring depending upon the individual every 1 to 5 seconds), and hemifacial spasm (prolonged high-frequency bursts lateralized to one side which last as long as the spasm).
Sweat artifact represents long-duration, high-amplitude slow undulating shifts in the electrical potential between the skin and the recording electrodes caused by perspiration from sweat glands. Sweat contains high amounts of sodium chloride and lactic acid which react with the exposed metal of the electrode, generating action potentials which combine with skin and sweat gland potentials (12). Sweat artifact typically (a) occurs at a frequency of 0.5 Hz or slower, (b) is synchronous (but can be lateralized or asymmetric), (c) is often most prominent in frontal electrodes, and (d) can mimic slow wave activity of NREM 3 sleep (Figure 11.20).
To reduce/eliminate sweat artifact: (a) try cooling the patient by lowering the room temperature to less than 68°F (20°C) by either uncovering the patient or turning on air conditioning; (b) place a towel beneath the patient’s head; (c) fan the patient’s head using a sheet of paper but do so gently to avoid 60 Hz or movement artifact; (d) reassure the patient to reduce test anxiety; and (e) if persistent, set LFF in the affected channels to 1, 3, or 5 Hz. Realize that changing LFF to 1, 3, or 5 Hz tends to remove delta EEG frequencies important for recognizing and scoring slow wave sleep and eye movements (Figure 11.21). A PSG pearl: sudden onset of generalized slowing in the EEG with prominent sweat artifact in a patient with diabetes mellitus warrants consideration of nocturnal hypoglycemia (3).
Eye Movement Artifacts
To record EOG in PSG we set the LFF at 0.3 Hz and the HFF at 35 Hz, the same bandpass we use to record EEG. EOG in PSG is generated when the 266eyeball rotates about its axis, causing large shifts in ACs detected by electrodes placed near the eyes. The cornea is 100 mV positive compared to the retina (−100 mV). When the electropositive cornea moves toward an EOG electrode referenced to the contralateral mastoid, it results in a downward deflection. When the eyes close or blink, both eyeballs move up (Bell’s phenomenon) which results in a maximum deflection in frontopolar and frontal EEG derivations. When the eyes are open, downward eye movements cause negative potentials in the frontopolar and frontal EEG derivations. Of note, when the eyeballs are in a fixed position, no change is observed in the EOG.
We use eye movements recorded in EOG and often detected in frontal EEG derivations to identify sleep/wake states. Slow eye movements appear at sleep onset (Figure 11.22). Slow eye movements (0.2–1 Hz) heralding the onset of NREM sleep are typically horizontal and out of phase: surface positive when positively charged cornea moves toward it, surface negative in the opposite eye. Rapid eye movements coupled with a loss of chin muscle tone and irregular respiration characterizes REM sleep (Figure 11.4). Of note, rapid eye movements in REM sleep are most often oblique, less often vertical, occasionally horizontal, and sometimes dysconjugate (13). When we record EOG using FPz as a common reference, the EOG also records frontopolar EEG K-complexes, and slow wave activity is typically of highest amplitude in the frontal and frontopolar derivations. K-complexes tend to cause in-phase deflections in the EOG (Figure 11.23A), whereas rapid eye movements cause out-of-phase deflections in recommended EOG during REM sleep (Figure 11.23B). EOG in PSG is typically detected in frontal and central EEG derivations (Figure 11.23B).
The AASM Scoring Manual permit recording eye movements in a PSG using either recommended or acceptable derivations (Figure 11.24A). Using the recommended EOG derivations, vertical or lateral conjugate eye movements result in out-of-phase deflections. EEG activity or artifacts in recommended EOG derivations typically appear as out-of-phase deflections. When the deflection in F4-M1 is downward, it means F4 is positive relative to M1. When the eyes move to the right, they cause a downward deflection in E2-M2. The acceptable derivations allow determination of the direction of eye movements. Vertical eye movements will show in-phase deflections and horizontal eye movements out-of-phase deflections. We prefer the acceptable method for recording EOG because it shows all directions of eye movements and also detects low-amplitude eye movements, such as blinking (Figure 11.24B).
Using AASM Scoring Manual recommended EOG derivations (Figures 11.24A and B), vertical or lateral conjugate eye movements result in out-of-phase deflections in both EOG derivations when one eye moves toward and the other away from one EOG electrode. EEG activity or artifacts in recommended EOG derivations typically appear as out-of-phase deflections. When the deflection in F4-M1 is downward, it means F4 is positive relative to M1. When the eyes move to the right, they cause a downward deflection in E2-M2. We prefer the alternative method for recording EOG because it shows all directions of eye movements and also detects low-amplitude eye movements, such as eyeblinking. Eyeblinks are typically brief, high-amplitude electropositive–electronegative waveforms in frontal EEG and EOG as the cornea moves upward (Figure 11.25).
• Ipsilateral eye absent or prosthetic
• Ipsilateral retinal disease
• Ipsilateral paresis of vertical gaze
• EOG electrodes asymmetrically placed
• Ipsilateral frontal skull defect
• Low-frequency filter settings set wrong
Eye flutter is characterized by rhythmic 3–6 Hz activity and confined to EOG and frontal EEG derivations (Figure 11.26). Eye flutter more often occurs in tense or anxious patients, and can be interrupted by asking the patient to open their eyes and fixate gaze (14). A small but interesting study recorded EEG and eye movements in two healthy adults using magnetic search coil technique (15). They found that spontaneous blinks produced small eye movements directed down and inward, whereas slow or forced blinks were associated with delayed upward eye rotations (i.e., Bell’s phenomenon). Both types of blinks caused positive EEG potentials with bifrontal distribution maximum at Fp1 and Fp2.
Given our dependence on EOG to score sleep, we are more concerned when encountering an asymmetric EOG in a PSG (Figure 27). Table 11.5 summarizes considerations when observing decreased or absent EOG in one eye in a PSG. Two PSG eye movement artifact pearls: (a) a prosthetic eye made of glass has no electric potential but one made of metal is capable of generating a slow potential with the eye closed or open and (b) in the presence of a skull defect, the EOG will be of lower amplitude ipsilateral to the bone defect (the opposite of a breach rhythm in EEG).
Tongue and Other Oropharyngeal Artifacts
Sucking (Figure 11.28), swallowing, and talking (Figure 11.29) are common during sleep (most likely an expression of central pattern generators), 267involve tongue movements, and often cause artifacts which invade EEG, EOG, chin EMG, and airflow channels (16). When the tongue moves during these, it functions as a dipole (the tip relatively electronegative compared to the base). Tongue movements can produce so-called glossokinetic artifacts (Figure 11.30). When the tongue moves (particularly when touching the roof of the pharynx) the change in the electrical field spreads to involve the scalp. This results in single or brief runs of rhythmic delta activity usually maximal frontal. Glossokinetic potentials cause diffuse synchronous delta frequency slow waves that are often maximal in the frontal derivations. Glossokinetic artifacts can resemble eye movement artifacts or frontal delta activity. They can be confirmed by asking the patient when awake to say words that contain lingual consonants (“Tom Thumb” or “Lilt-lilt-lilt”).
Respiratory and Vibration Artifacts
We encounter many artifacts in PSG related to respiratory effort. We preferentially record respiratory effort in PSG with LFF set to 0.1 Hz and HFF 15 Hz. Respiratory artifact often intrudes into the EEG derivations. Respiration signals are very slow and of high amplitude relative to EEG. It most often appears as a slow sway in the baseline EEG which is time-locked, occurring in-phase with phases of inspiration and expiration (Figure 11.31). Less often, respiratory effort artifact is a slow or sharp wave time-locked with inspiration and/or expiration but involves only some of the EEG electrodes (usually those on which the patient is lying) (Figure 11.32). Respiratory artifact is more often seen when the patient is lying in a manner that each breath causes a slight movement of EEG electrodes relative to the scalp. Once again, this is often most evident in electrodes which are loose or have high impedance.
To identify respiratory artifact, assess whether the artifact is time-locked to respiratory effort. Respiratory artifact is typically faster than sweat artifact and usually localized. Sweat artifact is more often generalized and involves several channels. To reduce or eliminate respiratory artifact when detecting during PSG collection: (a) check electrode impedances; if bad, regel or replace offending electrodes; (b) reposition the patient’s head; and (c) reroute electrode leads so patient not lying on them. A different respiration-related artifact is snoring-related vibration, which can cause high-frequency artifacts in other channels (Figure 11.33). This artifact cannot be removed but should be recognized as snoring-related.
The heart is a major source of artifacts in PSG (and EEG). It produces electrical and mechanical artifacts, both time-locked to contractions. Cardiac-related artifacts include those related to (a) ECG, (b) pulse, (c) ballistocardiographic movements, and (d) implanted cardiac pacemakers.
We record only one ECG channel (modified lead II) in PSG using a LFF of 0.3 Hz, HFF 70 Hz, and a minimum sampling rate of 200 Hz (500 Hz desirable). ECG tracings are typically recorded using a paper speed of 25 mm/sec. A 10-second time screen of PSG is roughly equivalent to a time screen of 30 mm/sec. Recording ECG in PSG using a modified level II, sinus rhythm is characterized by an upright P, R, and T waves. ECG recorded on the scalp is usually 1 to 2 µV in amplitude compared to EEG with typical amplitudes of 20 to 100 µV. ECG artifact can occur in any PSG channel, but especially in the AC-recorded EEG, EOG, and chin/limb EMG. The ECG signal has far greater amplitude (measured in mV) compared to the scalp-recorded EEG (μV).
ECG artifact in the EEG is usually diphasic (monophasic or triphasic in some). Small artifacts reflect primarily the R wave of the QRS. Larger ECG artifact may reflect other components; very large artifacts are often due to pacemakers. ECG artifact in EEG derivations is best seen in referential montages especially those to the left ear or wide interelectrode distances (all of which we typically record in PSG). The R wave is usually maximal over the left posterior head regions (Figure 11.34), lower over the right anterior head because the main cardiac vector produced a wave that is positive directed diagonally from right to left, anterior to posterior. If the head is turned, electrodes situated on left and posterior with regard to torso will still record maximum positivity. ECG artifact is enhanced in PSG when: (a) using referential montages because of their greater interelectrode distances and the field of the ECG is approximately equipotential across the head, (b) imbalanced electrode impedance between electrode pairs, (c) fatty deposits over the mastoid reference sites, and (d) individuals with short, thick necks or low-voltage scalp EEGs. ECG artifact is less noticeable in Cz or bipolar derivations.
ECG artifact usually cannot be eliminated from PSG because it is secondary to volume conduction of the ECG waveform. ECG is usually not due to a bad electrode. Given our preference for M1 and M2 as references, and that polygraphs do not easily permit linking M1 and M2, we are left with often prominent ECG artifact. Placing the ground electrode on the right side of the patient may reduce ECG artifact. Sometimes, turning the head relative to the trunk will change the field and lower the amplitude. The amplitude of the ECG artifact often changes in distribution and amplitude as the patient breathes. Breathing changes the position of the heart with respect to the head. Premature ventricular complexes are usually maximum over the posterior head regions, and are greater in amplitude and duration than the normal conducted heartbeat (Figure 11.35).
268ECG artifact from cardiac arrhythmias is more challenging to recognize at first glance (Figure 11.36). Always check whether each ECG artifact is time-locked to a QRS. The heart rate tends to increase slightly with inspiration and decrease with expiration and, when prominent, this is called sinus arrhythmia. Heart rate cutoffs for sinus tachycardia and bradycardia are different during sleep than when awake: sinus tachycardia in adults during sleep is scored when heart rate is greater than 90 beats per minute (bpm) for greater than 30 seconds and sinus bradycardia in sleep in individuals 6 years of age or older with a heart rate less than 40 bpm greater than 30 seconds. Premature ventricular complexes are usually maximal over posterior head regions, and are greater in amplitude and duration than normal conducted heartbeats.
Pulse artifact is caused by movement of an electrode when it rests over a pulsating artery (Figure 11.37). It appears as a smooth, periodic slow wave which follows the QRS ECG complex artifact by a slight delay (approximately 200 milliseconds); delayed by the time it takes the blood to travel from the heart to the pulsating blood vessel. Sometimes, the pulse artifact is triangular or sharply contoured. Pulse artifacts are more likely to occur in frontal, central, or temporal rather than occipital regions. To eliminate or reduce pulse artifact in PSG, first try moving the patient’s head slightly off the particular electrode; if the artifact persists, reapply the electrode a small distance away from the superficial artery.
Ballistocardiographic artifacts are related to pulse artifacts and caused by slight recoil movements of the patient’s head or body, which occur as the blood pumps from the heart. They typically occur as small “pulse waves” that correspond to cardiac contractions in the ECG tracing with a slight delay. They are cardiogenic oscillations that are most often observed in the PSG channels that record airflow or respiratory effort, but sometimes are observed in the EEG derivations. Figure 11.38 shows ballistocardiographic artifact in the abdominal respiratory effort channel and Figure 11.39 in the NP transducer and thoracic respiratory effort channel during central apneas. Some argue that this, when seen, suggests the upper airway is patent. There is considerable debate about their cause, including (a) recoil movements of the patient’s head or body as the blood pumps from the heart, (b) pulsatile force on the aortic arch from abrupt redirection of blood flow, and/or (c) changes in intrathoracic pressure produced with each heartbeat, resulting in the movement of air through the patient’s upper airway. There are ways to reduce or eliminate ballistocardiographic artifact. If the artifact is due to movement of the head on a pillow, try raising the head off of the bed with a roll of towels. However, this may introduce unwanted respiratory or motion artifacts. Once again, it is often seen when electrode impedances are high, so check and fix them if bad.
Cardiac Pacemaker Artifact
Pacemaker artifact appears as a high-amplitude fast spike when internal cardiac pacemaker delivers an electrical stimulus, which may be seen in the EEG and EOG channels, time-locked with ECG. The pacing spike is of much shorter duration than ECG artifact. Usually heart rate variability with respiratory events are not seen when heart rate is completely paced. This artifact cannot be eliminated. Figure 11.40 shows pacemaker artifact in a PSG tracing.
The presence of ventricular tachycardia (now called wide complex tachycardia) typically prompts concern when observed in overnight PSG. It needs to be distinguished from ventricular pacing in an individual with a pacemaker. Pacemakers with a single ventricular pacing lead in the right ventricle, when activated, result in a wide complex QRS, which converts to a normal sinus rhythm if the atrial rhythm falls below the set threshold or if an atrial beat is not conducted. Ventricular pacing artifact almost always occurs at a frequency less than 90 bpm, whereas ventricular tachycardia (wide complex tachycardia) is typically tachycardia.
Instrumental artifacts in PSG are generated by improper calibration, application, connection, incorrect settings, and/or malfunction of instruments and other equipment used to record the PSG. These include electrode popping, reversed polarity, incorrect filter settings, high impedance, imbalanced electrode pairs, overamplification, and ground recording artifact. Environmental artifacts are those which originate from surrounding environment.
Electrical Interference Artifacts
The most common environmental artifact is electrical noise generated by power lines and nearby electrical equipment (60 Hz in North America, 50 Hz in many other countries). A 60-Hz artifact usually appears as a thick line in the channel of interest (undulations at a frequency of 60 per second can be counted if the epoch speed is changed to 1 per second per screen) and affects AC recorded PSG channels (Figure 11.41). It is caused by a problem in continuity or grounding anywhere between the patient-to-digital polygraph circuitry.
269When 60-Hz artifact affects only one or a few AC-recorded channels, it often reflects high or imbalanced electrode impedance levels (Figure 11.42). When observing 60-Hz artifact, first check electrode impedances. Maximum electrode impedances for EEG and EOG in PSG need to be less than 5 kΩ and the difference between electrode pairs less than 1 kΩ. Unbalanced impedance between electrode pairs defeats the ability of the differential amplifier to cancel out the electrical noise. A 60-Hz artifact in one electrode may also be due to a broken or disconnected wire, poor electrode contact, and/or nearby electrical equipment.
If artifact involves a single electrode, press on the offending electrode, still present the regel, and if it persists, reapply. Further strategies to reduce 60-Hz artifact in AC-recorded PSG channels are as follows: (a) keep the patient’s head and connections to the polygraph device as far from power cables as possible; (b) electrode wires should be as short as possible, and bound to together in a single bundled cable; and (c) identify and minimize sway of the electrode wires or movement of tubing. If 60-Hz artifact involves all AC-recorded channels, it may be a loose ground electrode (reassess and replace if needed). If not, it could be due to electrical interference: (a) first untangle and straighten cables and wires; (b) try turning off or unplugging unnecessary (briefly if necessary) equipment, systematically assessing the effect (turning back on if no effect on it); and (c) if all this fails, run the study with the 60-Hz notch filter on.
Finally, 60-Hz artifact that appears in all AC-recorded channels of all recordings made in the same recording room is likely due to unshielded electromagnetic currents flowing through nearby power cables, transformers, or electromotors (8). If so, contact your biomedical department for identification and proper shielding. Proper shielding of the room and wiring of power cables can reduce environmental EMI.
Electrode Pop Artifact
Electrode pops are single or multiple spike or sharp waveforms due to an abrupt shift in impedance, which typically occur in a single electrode at a time, do not disrupt the ongoing EEG background activity, and have no electrical field. Changes in skin temperature, movement, dry electrodes, or air bubbles in the electrode-electrolyte solution trigger transient electrode pops. When the patient moves, the electrode can pull away from the skin (Figure 11.43). When the impedance change is less abrupt, electrode pops may appear as a low-amplitude arrhythmic slow wave. Electrode popping is more often seen with (a) poor or unbalanced electrode impedances; (b) moving, jerking, or touching the lead wire; or (c) a break in skin contact. To fix electrode pops: (a) identify the electrode, (b) check electrode impedances, and (c) press and secure the electrode; if it persists, regel, and if needed reapply or replace it.
Incorrect Filter and Sensitivity Settings
Sleep specialists need to understand how to best use and display the digital PSG by reducing or eliminating channel with artifact, amplifying critical signals, or viewing the study at varying time frames for each channel and signal (3). Modern digital polysomnographic equipment is designed to permit the user to change filter and sensitivity settings for each PSG channel during and after the recording. Filter settings for different PSG parameters recommended by the AASM are summarized in Table 11.2. EEG and EOG are typically displayed in PSG using a bandpass of 0.3 (LFF) to 35 Hz (HFF).
Technologists sometimes adjust the filter settings of particular channel(s), and when the file is opened, we may miss these, leading to interpretative errors. When the LFF is increased from 0.3 to 1.59 or 5.3 Hz by the technologist to reduce sweat sway artifact (Figure 11.21), the amount of NREM 3 present may be underestimated. As mentioned earlier, incorrect filter setting in EOG channel may eliminate the EOG signal in that channel (Figure 11.27). REM sleep is often unscored because of falsely elevated chin EMG due to poorly attached chin EMG electrodes (Figure 11.43). Overamplification of EEG may lead to missing sleep spindles crucial for scoring NREM 2 sleep (Figure 11.44). Overamplification is most often seen in children during NREM 3 sleep when the high-amplitude slow activity (often 200–400 µV) saturates the EEG channels; correct by decreasing the channel gains from 7 µV/mm to 10 or 15 µV/mm. A tracing showing all PSG signals are flat and most often appears when the patient is disconnected from the jack box, usually to use the bathroom (Figure 11.45) or when computer shuts down (Figure 11.46). It is important for the technologist to check all the connections between the patient and amplifier.
Channel Inversion Artifacts
The polarity of each PSG channel can also be adjusted individually as per AASM recommendations. We frequently encounter inverted waveforms in one or many recording channels. If negative inputs are placed in positive ports of the headbox, the displayed signal will be inverted. This most often happens in the ECG (Figure 11.47), easily fixed by inverting the channel. Inappropriate inversions in the respiratory signals (especially in the NP), thermal sensor, and respiratory effort channels can lead to significant misinterpretations (Figure 11.48). First review the biocalibration to determine if inversion of the signal was accidental and missed.
270Malfunctioning Respiratory Sensors
Most level 1 PSG recordings are to evaluate patients for sleep disordered breathing. Almost all respiratory sensors provide qualitative (not quantitative) data. Redundancy (multiple different respiratory sensors) is the name of the game when recording respiration in a PSG. Malfunctioning respiratory sensors can lead to a myriad of missing or false data. Respiratory artifacts most often appear as a low-frequency baseline sway that occurs in-phase with respiration (Figures 11.21 and 11.22); patients are often lying in a position where each breath causes a slight movement of electrode(s) relative to the scalp and a shift in the electrical potential. This is most evident in a loose or high-impedance electrode. Flattened respiratory effort channels (with or without notches), despite good respiratory effort confirmed by video, suggest the respiratory effort sensor is loose or poorly positioned.
To identify respiratory artifacts, check whether they coincide with the respiratory effort signals. Respiratory artifact is typically faster than sweat artifact and usually localized. Sweat artifact is slower and more often appears generalized. To reduce/eliminate respiratory artifact: (a) straighten electrode wires avoiding loops; (b) check electrode impedances; if the scalp contact is bad, regel or replace offending electrodes; (c) have patient change position; and/or (d) reroute electrode leads so patient is not lying on them.
NP Sensor Artifacts
NP transducer is the preferred sensor for identifying hypopneas or respiratory event-related arousals (RERAs) in PSG. However, they are particularly prone to malfunction (especially in children with profuse nasal secretions and a great dislike for the nasal cannula inserted in their nostrils) (Figure 11.49). We record NP using LFF of 0.03 Hz or direct current (DC) and the HFF set at 100 Hz. Figure 11.50 shows the NP sensor channel is inverted. The crest of the NP signal should appear sharp. If the bottom of the waveform appears sharp, flip the channels to properly identify flow limitation from obstructive hypopneas. Table 11.6 summarizes common causes for NP sensor malfunction in PSG.
Thermal Sensor Artifacts
Thermal sensors are the preferred method for identifying apneas in PSG. They measure changes in temperature induced by airflow: cooler, inspired aire causes an up signal and warmer, expired air a down signal. The signal is qualitative, not quantitative, and does not vary proportionately with airflow. Thermal sensors usually identify apneas but can overestimate ventilation and underestimate hypopneas. When the thermal sensor is dislodged or lying on the skin, it may not detect the changes in temperature between inspiration and expiration. If recognized, adjust by repositioning the thermal sensor properly. If the NP sensor is malfunctioning and forced to use the thermal sensor to assess airflow, it may fail to detect subtle changes in airway (respiratory effort-related arousals).
• Patient pulled the nasal cannula out of the nose
• Obligate or intermittent mouth breather
• Profuse nasal secretions
• Supplemental oxygen during PSG dilutes NP signal
• Too small an NP prong in large nostrils may produce a small NP signal; too large an NP prong in small nostrils may cause a large signal
• Nasal prongs which partially or completely obstruct one or both nares can lead to spuriously higher apnea and arousal indexes
NP, nasal pressure; PSG, polysomnography.
Pulse Oximeter Artifacts
We typically detect arterial oxyhemoglobin saturations during PSG using pulse oximeters. They calculate the oxyhemoglobin saturation (ratio of oxyhemoglobin to reduced hemoglobin in arterial blood) using spectrophotometry (the quantification of compounds by their light absorption characteristics) (17). The peak absorption frequency of reduced hemoglobin is 660 nm, and oxyhemoglobin 940 nm.
The pulse oximeter contains a photoreceptor that isolates the pulse signal from which the oxygen saturation is calculated and also generates a plethysmographic signal (the pulse waveform). Pulse waveforms are extraordinarily useful in confirming whether the pulse oximeter signal is “reliable.” Observing low peripheral capillary oxygen saturation (SpO2), we confirm the pulse waveform is preserved, if blunted or absent, the SpO2 signal deemed unreliable (Figure 11.50). Pulse oximetry can be falsely peripheral for many reasons, summarized in Table 11.6. To fix poor oximeter readings: (a) reapply or select another oximeter probe; (b) remove nail polish and avoid acrylic nails; (c) try a different location (earlobe and toe). The lag time in the maximal fall in the oxygen desaturation related to apneas or hypopneas in a PSG varies with the location of the recording probe: 10 to 30 seconds when placed on the tip of the finger over the nail bed versus 5 to 10 seconds when attached to the earlobe. Table 11.7 summarizes causes of pulse oximeter artifact or malfunction.
• Movement artifact
• Poor perfusion in patient who is hypovolemic, hypotensive, or cold
• Inadequate light transmission detected by device because of tissue edema, nail polish, acrylic nails, dark or thick skin, improper probe placement
• Excessive ambient room light
• Venous pulsations misinterpreted as arterial
• Significantly higher SpO2 values in heavy tobacco smokers with carboxyhemoglobin levels greater than 10%
SpO2, peripheral capillary oxygen saturation.
PAP Tubing Artifacts
Water in the NP airflow channel causes a fast, undulating signal artifact. It appears as an M-shaped waveform with each breath as the water moves back and forth in the tubing. It occurs during diagnostic PSGs if the moisture gets into the cannula or the tubing connecting it to the NP sensor. We see it most often during continuous positive airway pressure (CPAP) titration if there is a mismatch between room temperature and the humidification in the CPAP circuit (Figure 11.51).
Computer Monitor Display Artifacts
The number of pixels on the computer monitor screen places an absolute limit on the upper range of signal frequencies which can be displayed. The AASM recommends a minimum computer monitor resolution of 1,600 × 1,200 pixels. A high-resolution, digital PSG super VGA monitor often has a horizontal screen resolution of 1,600 pixels. When sampling PSG at 10 seconds per screen, this results in a resolution of only 150 pixels per second, 72 pixels per inch. The fastest EEG frequencies seen viewing a 10-second PSG sample are less than 75 Hz, less than 20 to 25 Hz when the EEG is displayed as a 30-second time screen, and 2.7 Hz or less when viewing a 180-second epoch. To see less than 75-Hz EEG frequencies on a 30-second screen, 3,000 pixels (which are not yet commercially available) are required. Analog EEG and PSG machines recording ink on paper (of which only a few of us now recall using) had 2 to 3 times this resolution.
Frequency aliasing introduces display and potential interpretative errors in PSG and EEG. Frequency aliasing in old black and white television programs made the wagon wheels appear to be stationary or turning backward. Inadequate sampling rates can produce spurious representation of EEG activity. Epstein showed how spurious (false) theta activity in EEG channels on computer monitor display was due to aliasing EMG artifact. This was caused by undersampling of the EMG signal, resulting in errors in waveform display from exceeding Nyquist frequency on the computer screen display (1). The false theta disappeared when the HFF was turned down to 35 Hz. Such distortions can be lessened or avoided by viewing the EEG using a 10-second time screen with the HFF set at ≤40 Hz (we typically set our HFF at 35 Hz). Undersampling errors can also distort relationships between spike-and-slow waves. This is best viewed using 10-second or, even better, 5-second time screens.
Cell Phone Artifacts
Potential sources of PSG artifact change with time (environment and technology). Mobile telephones, computer tablets, and laptop computers are much more common sources of electrical interference artifact in the sleep laboratory. A cell phone ringing at the bedside can cause EMI (Figure 11.52). When the patient types on a laptop while we are recording PSG, it provokes intolerable artifact.
One case report described an EMI artifact produced by a smartphone, which was characterized by runs of 100-µV, 5- to 9-Hz sharply contoured waveforms intermixed with intermittent high-frequency (20–50 Hz) sinusoidal waves lasting for 1 to 20 seconds (18). These occurred throughout the night, and review of the audio recording revealed an intermittently rhythmic sound during the discharges. The patient had placed her mobile device 12 inches away from the recording jack box. Others have reported that the cellular telephone artifact can produce 11- to 16-Hz waveforms mimicking sleep spindles; high-amplitude short bursts of activity in chin EMG misinterpreted as REM sleep without atonia; ventricular tachycardia in the ECG channel; and unilateral saccadic eye movement artifact in EEG when repeatedly swiping to left; the ringing phone delays sleep onset, causes arousals and awakenings, and alters sleep architecture (19,20).
Cellphones are low-power radiofrequency transmitters that can cause narrowband EMI (21). They send and receive microwave frequencies from 272approximately 900 to 1800 Hz. Coupled with nearby electrical equipment, they can act as a radio receiver and interfere with their functioning. The fundamental principle of EMI is the level of exposure in a far field proportional to the square of the distance from the radiating source. Clinically relevant EMI at distances greater than 1 meter are negligible. The solution is to keep mobile devices greater than 3 feet (≥1 meter) from the medical device (e.g., mechanical ventilator, pacemaker, EEG, or PSG equipment). We recommend (a) electronic devices be placed at least 3 feet away from the PSG equipment, (b) preferably be turned off, (c) and the patient should not touch it when it is charging and we are recording.
Vagal Nerve Stimulation Artifacts
Artifacts due to periodic activations of implanted vagal nerve stimulators cause a periodic abrupt reduction in amplitude and a marked increase in the rate of the respiratory signals, which lasts as long as the stimulator fires (most often 30 seconds, but sometimes 17 seconds in “rapid cycling” modes) and recurs at the particular activation frequency (typically 5 minutes but sometimes 3 minutes). More often, vagus nerve stimulator (VNS) activation causes no arousal, awakening, or significant oxygen desaturation. However, case reports abound describing worsening of obstructive sleep disordered breathing (especially among those with OSA). This may be due to spasm in the upper airway muscles or laryngeal vocal cord dysfunction when the VNS activates, resulting in obstructive apneas. If symptomatic, reduce the current flow (lower VNS stimulus frequency from 30 to 20 Hz) or tape magnets across the stimulator to turn it off when the patient sleeps (22). Illustrative examples of this can be found in these references (23–25).
DBSs are used in the United States to treat PD, dystonia, neuropsychiatric disorders, and medically refractory epilepsy. DBS generates electrical artifact that affects every PSG signal recorded using AC. Bipolar DBS causes fewer artifacts than unipolar DBS. Unfortunately, turning off the DBS during the recording often results in emergence of tremor artifact in patients with PD. Bilateral monopolar DBS produces a high-frequency artifact which obscures most channels of EEG and is most pronounced in those with long interelectrode distances. Bipolar DBS causes fewer artifacts. Figure 11.53 shows an example of DBS artifact in PSG.
Artifacts Which Require Technologist or Video to Identify
It is best for the technologist recording the study to identify the artifact for the reader. Figure 11.54 shows a recurring electrographic interference artifact due to a heating pad repeatedly turning on. After the fact, we observed the heating blanket when the patient kicked off her blanket.
Because overnight PSGs are recorded for hours and many different biologic parameters recorded, artifacts are ubiquitous. Some artifacts in PSG are useful, helping us identify an arousal, body position change, respiratory event, or parasomnia. However, other artifacts may obscure significant signals in the PSG containing important diagnostic information. Some artifacts have a stereotypical appearance, so pattern recognition of these is often helpful. The ability to re-reference electrodes can be invaluable for identifying and ameliorating a questionable or spurious PSG signal. Failure to recognize artifacts may lead to false-positive reports that may contribute to errors in clinical management. The sleep technologist identifying an artifact needs to document its presence in the comment log, determine the source and perform what is needed to eliminate it from the recording, and finally consider whether it affects patient safety, or prevents interpretation of the PSG. Sometimes, letting the patient sleep even when artifact is present is the best choice.
Abd, abdominal respiratory effort; C4-M1, right central, right central-left mastoid; Chest, thoracic respiratory effort; ECG, modified lead II; F4-M1 right frontal EEG, right frontal-left mastoid; L EOG, left extraocular muscles; NP, nasal pressure sensor; O2-M1, right occipital EEG (right occipital-left mastoid); R EOG, right extraocular muscles; SaO2 (%), pulse oximetry signal; Snore Mic, snore microphone; Sum, summed signal of thoracic and abdominal respiratory effort signals; Therm, oronasal thermal sensor.