{
  "id": "electroencephalography",
  "title": "Electroencephalography",
  "lang": "en",
  "source_lang": "en",
  "summary": "Electroencephalography records the brain’s electrical activity for clinical assessment, physiological monitoring, and neuroscience research.",
  "category": "health",
  "subcategory": "medicine",
  "type": "concept",
  "aliases": [
    "EEG",
    "electroencephalographic recording"
  ],
  "generated_at": "2026-10-22",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/electroencephalography/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/electroencephalography/",
      "json": "https://aiwiki.page/zh/electroencephalography.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Electroencephalography (EEG) is an electrophysiological method for recording electrical activity generated by the brain, usually through electrodes placed on the scalp. The resulting recording is called an electroencephalogram. Scalp EEG principally reflects the combined postsynaptic activity of populations of cortical neurons, rather than the electrical impulses of individual cells. Its fine temporal resolution makes it particularly useful for studying rapid changes in brain function. (ncbi.nlm.nih.gov)\n\nEEG has an established role in evaluating suspected seizures and epilepsy, characterizing sleep-related activity, and monitoring brain function in critically ill patients. Interpretation depends on the recording conditions, the person’s age and state of alertness, and the clinical context; artifacts and normal variants must be distinguished from abnormalities. Beyond clinical practice, EEG supports investigations of sensory processing, cognition, and brain–computer interfaces. Its uses range from conventional waveform interpretation to quantitative analysis, with standardized acquisition and reporting supporting reliable clinical assessment and reproducible research. (acns.org)"
    },
    {
      "heading": "Historical development",
      "text": "In 1875, the Liverpool physician Richard Caton reported electrical activity from the exposed brains of rabbits and monkeys in a short communication, “The Electric Currents of the Brain,” in the British Medical Journal. Using a sensitive galvanometer, he observed spontaneous fluctuations rather than merely responses to externally applied electrical stimulation. His expanded report in 1877 described experiments in a larger group of animals. These investigations established an experimental basis for recording the electrical activity of the brain, although they were not recordings through an intact human scalp. Subsequent animal experiments by Adolf Beck, reported in 1890, investigated spontaneous oscillations and their alteration by sensory stimulation. (pmc.ncbi.nlm.nih.gov)\n\nHans Berger, a psychiatrist in Jena, obtained his first human brain-electrical recording in 1924, initially working with patients who had skull defects; his first publication on the human electroencephalogram appeared in 1929. The two dates therefore identify different milestones: acquisition of the first recording and public presentation of the findings. Berger subsequently demonstrated that brain activity could be recorded through the intact skull, making a nonsurgical method possible. In 1934, Edgar Douglas Adrian and Bryan H. C. Matthews independently investigated the “Berger rhythm.” Their experiments documented an approximately 10-cycle-per-second rhythm, prominent with the eyes closed, and examined evidence distinguishing its cerebral origin from electrical activity generated by the eyes or cranial muscles. (pmc.ncbi.nlm.nih.gov)\n\nClinical EEG developed from limited recording arrangements into simultaneous multichannel examination, allowing activity over different head regions to be compared at the same moment. Electrode placement also became standardized: Herbert H. Jasper’s 1958 publication described the international 10–20 electrode system, subsequently incorporated into professional technical guidance. Standard positions made the spatial sampling of successive examinations and recordings from different laboratories more comparable. The importance of multichannel acquisition was explicitly codified in the American Clinical Neurophysiology Society’s (ACNS) 2016 minimum technical requirements, which specified at least 16 simultaneous recording channels to display the distribution of most normal and abnormal patterns. This was a minimum standard, not the date multichannel EEG was invented. (acns.org)\n\nThe replacement of analog paper traces by digital acquisition changed what could be done after an examination. A paper record preserved the waveform display produced by the selected electrode connections, amplification, and filters. Digital systems stored sampled signals that could subsequently be redisplayed with different gain, filtering, time scale, and montage—the arrangement of voltage differences shown as channels. Retrospective montage selection therefore allowed the same recorded event to be examined in several spatial configurations without repeating the examination. By its 2016 revision, the ACNS stated that digital equipment was used in most facilities and recommended digital recording rather than paper-based equipment. Its digital-media guideline specified a minimum acquisition rate of 256 samples per second and at least 16-bit digitization, illustrating how acquisition and display had become separate technical processes. (acns.org)\n\nDigital storage also facilitated synchronized video, networked access, and computer-assisted review. Time-linked video made it possible to compare movements and behavior with electrical events, rather than relying solely on annotations written beside a trace; the ACNS’s 2016 reporting guideline included time-locked video in its example technical description. Digital-media standards addressed event annotations, automated detection, postprocessing, and server-based storage for continuous video-EEG. These developments supported the review of lengthy recordings through searchable events and adjustable displays, while retaining the recorded signals for direct examination rather than substituting algorithmic detections for interpretation. (doi.org)"
    },
    {
      "heading": "Physiological origins and measurement principles",
      "text": "Scalp EEG primarily reflects the summed extracellular effects of postsynaptic currents, rather than the action potentials of individual neurons. At an excitatory synapse, entry of positively charged ions can create an extracellular current sink; return currents elsewhere along the cell membrane form associated sources. Inhibitory synaptic activity also contributes, with its effect depending on synaptic location and the electrochemical driving force. Cortical pyramidal neurons are especially important because their elongated apical dendrites are approximately parallel within a local cortical region and perpendicular to the cortical surface. This arrangement allows their electrical contributions to reinforce one another. Individual action potentials are comparatively brief and spatially localized, making them much less important than coordinated postsynaptic currents in conventional scalp recordings. (ncbi.nlm.nih.gov)\n\nDetectable scalp activity requires both spatial and temporal summation: many suitably oriented neurons must generate currents that overlap sufficiently in time. Contributions with opposing orientations or different timing can partially cancel. Consequently, EEG amplitude is not simply proportional to the number of active neurons or their average firing rate. A population receiving strongly coordinated synaptic input may produce a larger signal than a more actively firing but poorly coordinated population. Biophysical simulations demonstrate that distributing synaptic inputs broadly across pyramidal-cell dendrites can cause substantial cancellation, whereas concentrating inputs on apical or basal regions produces more consistent population signals. The location and coordination of input therefore matter alongside its strength. (ncbi.nlm.nih.gov)\n\nA spatially separated current sink and source can be represented approximately as an electrical dipole. Its orientation follows the geometry of the contributing cells, so cortical folding changes how a generator projects toward the scalp. Through volume conduction, the resulting potential field extends through brain tissue, cerebrospinal fluid, skull, and scalp. These tissues have different conductive properties: cerebrospinal fluid provides a relatively conductive pathway, whereas the skull’s high resistivity substantially attenuates and spatially smooths scalp potentials. One cortical generator can therefore influence several widely separated electrodes, and an electrode does not record only the cortex immediately beneath it. Clinical recordings near skull defects illustrate this mechanism: reduced skull filtering can make cerebral discharges unusually sharp and spatially restricted. (ncbi.nlm.nih.gov)\n\nEach recording channel measures a difference in electric potential, not an absolute voltage: \\(V_{AB}(t)=\\phi_A(t)-\\phi_B(t)\\), where \\(\\phi_A\\) and \\(\\phi_B\\) are the potentials at two electrode sites. Polarity denotes the sign of that difference; exchanging the inputs reverses it. Changing either electrode can alter the waveform because both sites may contain cerebral activity. Amplitude expresses the size of the voltage excursion, commonly in microvolts, with \\(1\\,\\mu\\mathrm{V}=10^{-6}\\,\\mathrm{V}\\); peak-to-peak amplitude is the difference between a waveform’s maximum and minimum. Frequency is the number of cycles per second, measured in hertz, and phase specifies position within an oscillatory cycle. For an ideal sinusoid, \\(V(t)=A\\cos(2\\pi ft+\\theta)\\), these quantities are represented by amplitude \\(A\\), frequency \\(f\\), and phase offset \\(\\theta\\). Real EEG contains overlapping oscillatory and nonperiodic components rather than a single sinusoid. (ncbi.nlm.nih.gov)\n\nLocal cortical generators operate within interacting neural networks, including recurrent cortical connections and thalamocortical loops. These interactions can organize synaptic currents into rhythms, while volume conduction mixes contributions from multiple generators at each scalp site. Models incorporating cortical and thalamic feedback can reproduce spectral peaks near 10 Hz and show how rhythm frequency depends on physiological parameters. Such relationships do not establish a one-to-one correspondence between a waveform and a psychological state: the measured pattern depends jointly on network dynamics, cellular geometry, tissue conduction, and the electrode pair used to observe it. (pubmed.ncbi.nlm.nih.gov)"
    },
    {
      "heading": "Electrodes, instrumentation, and recording montages",
      "text": "Scalp electrodes commonly consist of gold or silver–silver-chloride cups attached with conductive paste or gel; electrode caps provide an alternative mounting arrangement. The conductive interface establishes electrical contact through the skin, while electrode–skin impedance helps characterize contact quality. The 2023 joint IFCN–ILAE standards recommend impedances below \\(5\\,\\mathrm{k\\Omega}\\), with values below \\(10\\,\\mathrm{k\\Omega}\\) considered acceptable. Balanced impedances matter because unequal contacts impair rejection of interference shared by amplifier inputs. Exceptionally low values, below approximately \\(100\\,\\Omega\\), can indicate a conductive bridge between electrodes rather than excellent contact. An electrode is a physical sensor; a recording channel represents a voltage difference between two inputs, so electrode count and displayed channel count are not interchangeable. (pmc.ncbi.nlm.nih.gov)\n\nThe international 10–20 system, documented by Herbert Jasper in 1958, places electrodes at proportional distances rather than fixed centimetre intervals. Measurements use the nasion, inion, and left and right preauricular landmarks; the name refers to spacings of 10% or 20% along measured head arcs. Letters identify frontopolar, frontal, central, parietal, occipital, and temporal positions—Fp, F, C, P, O, and T—while odd numbers indicate the left side, even numbers the right, and “z” the midline. Thus F3 and F4 are homologous frontal positions, whereas Cz is central and midline. The extended 10–10 system supplies intermediate sites and standardizes older T3/T4 and T5/T6 labels as T7/T8 and P7/P8. It supports denser arrays and targeted supplementary coverage. The 2023 standards suggest a 25-electrode array, adding six inferior temporal electrodes to the 19 scalp sites of the basic array, whenever feasible. (acns.org)\n\nA differential amplifier amplifies the difference between its inputs while suppressing their shared, or common-mode, signal. High input impedance limits loading of the electrode interface. An analog-to-digital converter then samples the amplified voltage and represents each sample numerically; sampling rate determines temporal spacing, whereas bit depth and input range determine voltage resolution and the risk of clipping. The American Clinical Neurophysiology Society’s 2016 digital-recording guideline specifies at least 256 samples per second and at least 16-bit digitization. Frequencies above half the sampling rate must be sufficiently attenuated before sampling to prevent aliasing. Calibration with a known signal checks amplitude scaling and channel responses; it does not establish that electrode contact or scalp coverage is adequate. (acns.org)\n\nAcquisition filters and display filters have different consequences. Analog filtering before digitization limits the frequency content retained in the recording; subsequent display adjustments cannot restore information already excluded. Digital review filters can be changed without altering preserved underlying samples. The 2023 standards suggest display cutoffs of 0.5 Hz for high-pass filtering and 70 Hz for low-pass filtering. Raising the high-pass cutoff attenuates slow activity; lowering the low-pass cutoff can blunt spikes and sharp waves. A mains-frequency notch filter can also distort waveforms. Changing display sensitivity alters apparent trace height, not the voltage originally acquired. (pmc.ncbi.nlm.nih.gov)\n\nA montage is an ordered arrangement of channel derivations. Referential montages display each electrode against a common reference, \\(V_i-V_R\\); activity at that reference consequently enters every channel. Bipolar montages subtract neighboring electrodes, such as Fp1–F3 and F3–C3. Average-reference montages subtract the mean of a selected electrode set, \\(V_i-\\frac{1}{N}\\sum_{j=1}^{N}V_j\\), making the result dependent on scalp coverage and which electrodes enter the average. Digital recordings permit retrospective comparison of these arrangements. A bipolar phase reversal occurs when adjacent derivations sharing an electrode deflect in opposite directions: a relatively negative potential at F3, for example, produces opposite signs in Fp1–F3 and F3–C3. This identifies a sampled scalp extremum, not necessarily the cortical generator directly beneath it. Reference choice, electrode spacing, generator orientation, and volume conduction influence the displayed field; broad activity can also cancel between neighboring electrodes. (acns.org)"
    },
    {
      "heading": "Examination procedures and recording formats",
      "text": "A clinical EEG begins with a defined question, such as whether recurrent episodes of impaired awareness might be seizures. The recording history documents the events and their frequency, the date of the most recent episode, relevant medication—including any sleep-inducing medication—and the patient’s behavioral state. A technologist applies scalp electrodes, checks contact quality, and obtains an interpretable baseline while observing the patient. Time-linked annotations identify movements, changes in alertness, activation procedures, and clinical events, allowing subsequent review to relate waveform changes to what occurred during acquisition. These requirements are specified in the American Clinical Neurophysiology Society’s 2016 technical guideline. (acns.org)\n\nEye opening and closure test background reactivity and help distinguish cerebral rhythms from eye-related artifacts. For example, the posterior waking rhythm normally attenuates with eye opening. Hyperventilation and intermittent photic stimulation are supervised activation procedures intended to reveal abnormalities not evident at rest: hyperventilation is particularly useful in evaluating suspected absence seizures, whereas flashing light tests for photosensitive responses. Both can provoke seizures. The 2023 joint IFCN–ILAE standards require advance explanation, contraindication screening, and continuous observation; hyperventilation contraindications include sickle cell disease or trait and severe cardiopulmonary disorders. Photic stimulation is stopped when generalized epileptiform discharges appear. (acns.org)\n\nRoutine awake EEG chiefly samples wakefulness, sometimes including drowsiness; sleep EEG deliberately samples sleep. Sleep-deprived EEG describes the preparation used to facilitate sleep, not a separate recording technology. Sleep can expose epileptiform discharges absent during wakefulness. The 2023 standards conditionally recommend 20 minutes for routine EEG and 30 minutes for sleep EEG, excluding preparation. These are minimum recording durations, not guarantees of abnormality detection. Longer or repeated examinations offer additional opportunities to sample intermittent discharges and relevant states; sleep deprivation is selected clinically because it can cause distress and has safety implications. (mknft.hu)\n\nAmbulatory EEG uses a portable multichannel recorder in the patient’s ordinary environment, extending observation beyond a laboratory appointment and permitting sampling of everyday activities and overnight sleep. It may include synchronized video, but “ambulatory” does not itself imply that video is available. The distinction matters when an electrical change must be compared with visible behavior: an EEG-only recording cannot provide the same direct behavioral documentation. The ACNS published dedicated minimum technical requirements for ambulatory EEG in 2022. (pubmed.ncbi.nlm.nih.gov)\n\nInpatient long-term video-EEG monitoring combines continuous EEG with synchronized audiovisual recording in a staffed monitoring unit. During captured events, personnel can assess responsiveness, language, movement, and subsequent recall, relating these observations to electrical onset and evolution. Compared with home recording, the inpatient setting permits closer surveillance, clinician-directed medication adjustment when indicated, and rapid intervention for prolonged or repeated seizures. Duration is individualized according to the clinical question and the number and types of habitual events captured rather than a universal fixed interval. Capturing a representative episode is especially important in the differential diagnosis of recurrent attacks, although some focal epileptic seizures lack a detectable scalp EEG correlate. The joint IFCN–ILAE inpatient monitoring guideline appeared in 2022. (neurosoft.com)"
    },
    {
      "heading": "Normal activity across wakefulness, sleep, and development",
      "text": "EEG frequency bands describe oscillation rates rather than distinct mental states. The conventions adopted here follow the 2017 International Federation of Clinical Neurophysiology glossary by Nick Kane and colleagues: delta below 4 Hz, theta 4–8 Hz, alpha 8–13 Hz, beta 14–30 Hz, and gamma above 30 through 80 Hz. These are descriptive ranges, not sharp physiological boundaries; clinical and research definitions differ, particularly at band edges and in the upper limit of gamma. Frequency alone does not establish whether activity is normal: its distribution, organization, reactivity, and relationship to age and behavioral state also matter. (pmc.ncbi.nlm.nih.gov)\n\nDuring relaxed, eyes-closed wakefulness, the adult posterior dominant rhythm is usually an 8–12 Hz alpha rhythm, strongest over the occipital regions. It attenuates with eye opening and returns with eye closure. Normal background organization includes broadly similar frequencies over the two hemispheres, although amplitudes need not be identical, and lower-amplitude, faster activity anteriorly. Beta activity commonly appears over frontocentral and temporal regions. The sensorimotor mu rhythm also occupies approximately 8–12 Hz but has a central rather than occipital maximum and an arch-shaped waveform. Unlike posterior alpha, mu attenuates with movement or imagined movement rather than simply with eye opening. Thus, two rhythms within the same frequency band can differ in anatomical distribution and physiological reactivity. (ncbi.nlm.nih.gov)\n\nDrowsiness brings attenuation and fragmentation of posterior alpha, increasing theta activity, and slow rolling eye movements. In light non-rapid-eye-movement sleep, particularly N1, vertex waves appear as sharply contoured transients maximal near the scalp vertex; their sharp appearance is physiological in this setting. N2 is identified by sleep spindles and K-complexes. Spindles are waxing-and-waning trains of 11–16 Hz activity, usually 12–14 Hz, lasting at least 0.5 seconds and generally most prominent centrally. K-complexes consist of a well-defined negative sharp component followed by a positive component, together lasting at least 0.5 seconds, with a frontocentral maximum. (ncbi.nlm.nih.gov)\n\nDeeper N3 sleep contains prominent high-amplitude slow-wave activity. In adult sleep scoring, N3 requires qualifying 0.5–2 Hz slow waves, at least 75 microvolts peak-to-peak over frontal regions, to occupy at least 20% of a 30-second epoch; this operational definition is narrower than the general delta band. Rapid-eye-movement sleep instead shows low-amplitude, mixed-frequency EEG activity that can resemble wakefulness. Rapid eye movements and low chin-muscle tone provide essential accompanying evidence, illustrating why sleep staging uses EEG together with eye-movement and muscle recordings. (pmc.ncbi.nlm.nih.gov)\n\nNeonatal recordings require interpretation against postmenstrual age, combining gestational age at birth with elapsed postnatal age. Discontinuity—bursts of activity separated by quieter intervals—can be normal in premature infants and becomes less pronounced with maturation. In healthy term neonates, wakefulness and active sleep generally have continuous backgrounds, whereas quiet sleep may show tracé alternant, alternating higher- and lower-amplitude activity. These age- and state-dependent patterns are not equivalent to pathological suppression; continuity, interburst activity, symmetry, and sleep–wake organization must be assessed together. (ncbi.nlm.nih.gov)\n\nDuring infancy, the posterior dominant rhythm emerges at approximately 3–4 Hz around two months, accelerates to about 5–7 Hz by twelve months, and reaches approximately 8 Hz by three years in many children. Sleep spindles typically appear by two to three months and can initially be asynchronous between hemispheres. Childhood backgrounds retain more theta and delta activity than adult backgrounds; high-amplitude rhythmic slowing during drowsiness, termed hypnagogic hypersynchrony, is particularly common around two to four years. Posterior slow waves of youth can persist into adolescence or young adulthood. Consequently, a slow component that is age-appropriate in a child cannot be interpreted using adult waking-background expectations alone. (ncbi.nlm.nih.gov)"
    },
    {
      "heading": "Artifacts, normal variants, and interpretive safeguards",
      "text": "Artifacts are recorded signals arising outside the brain, whereas normal variants are genuine cerebral patterns that can resemble abnormalities. Physiological artifacts include ocular, muscular, cardiac, sweat-related, and movement-related activity. The eye behaves as an electrical dipole, with the cornea positive relative to the retina: blinking usually produces large, bilaterally synchronous frontopolar deflections that diminish posteriorly, while lateral eye movements produce opposing polarities near the two outer canthi. Muscle activity from chewing, jaw tension, or forehead contraction produces irregular, fast, sharply contoured activity, often strongest over frontal and temporal regions. Its relationship to visible movement and reduction with relaxation help distinguish it from cerebral activity. (onlinelibrary.wiley.com)\n\nCardiac electrical contamination produces recurring sharp deflections time-locked to the electrocardiogram (ECG). Mechanical pulse artifact is different: pulsation beneath an electrode can generate a slower waveform after each cardiac contraction. Sweating alters skin potentials and the electrode–skin interface, producing slow baseline shifts that can resemble cerebral slowing. Body movement may simultaneously generate muscle potentials, displace electrodes, and move cables; consequently, its appearance ranges from localized transients to widespread irregular disturbance. ECG channels and synchronized video provide temporal evidence for these extracerebral origins rather than relying on waveform appearance alone. (pmc.ncbi.nlm.nih.gov)\n\nTechnical artifacts include abrupt electrode “pops,” unstable contact, cable movement, mains interference, and signals from nearby equipment. A disturbance confined to one electrode appears in every displayed channel containing that electrode and can create an apparent phase reversal in a bipolar montage, mimicking a focal spike. Poor contact and increased electrode–skin impedance can increase susceptibility to interference at 50 or 60 Hz. Auxiliary recordings help establish signal origin: the 2023 joint IFCN–ILAE minimum standards specify an ECG channel, recommend additional electromyography channels when motor events are suspected, and describe electrooculography for distinguishing eye movements from cerebral slow waves. Time-synchronized video documents behavior and movement at the instant an apparent discharge occurs. (onlinelibrary.wiley.com)\n\nWicket waves illustrate why a pointed contour is not sufficient evidence of epileptiform activity. These temporal, arch-shaped waves occur singly or in short trains, commonly during drowsiness or light sleep in adults. Their frequency is approximately 6–11 Hz and their amplitude can reach 60–200 μV. Unlike typical epileptiform spikes, wickets lack a following slow wave and do not disrupt the background. An isolated fragment can look more suspicious than the longer train from which it arises, making inspection of surrounding recording essential. (ncbi.nlm.nih.gov)\n\nSmall sharp spikes, historically called benign epileptiform transients of sleep, are usually shorter than 50 ms and lower than 50 μV. They occur chiefly during drowsiness and light sleep, often in temporal leads, without background disruption or a prominent following slow wave. Rhythmic midtemporal theta of drowsiness consists of sharply contoured, sometimes notched temporal theta bursts that may be unilateral or bilateral. Its stable frequency and distribution, without progressive evolution or spread, distinguish it from many ictal rhythms. Despite their historical names and sharp appearance, these characteristic variants do not themselves establish epilepsy. (pmc.ncbi.nlm.nih.gov)\n\nInterpretation therefore combines morphology with the scalp electrical field, background organization, vigilance state, temporal evolution, auxiliary signals, and clinical history. A phase reversal identifies a voltage relationship within a montage, not necessarily an epileptiform generator; likewise, rhythmicity alone does not identify a seizure. In a 2022 prospective study of 3,050 routine recordings using standardized SCORE reporting, rhythmic temporal theta of drowsiness occurred in 6.03% and small sharp spikes in 1.05%. Such variants also appeared in recordings containing separate abnormalities, so recognizing a benign pattern neither establishes epilepsy nor makes the entire recording normal. (onlinelibrary.wiley.com)"
    },
    {
      "heading": "Abnormal patterns and epilepsy assessment",
      "text": "Focal slowing, generalized slowing, and background disorganization primarily indicate disturbed cerebral function rather than an enduring susceptibility to seizures. Persistent focal theta or delta activity in an awake patient suggests dysfunction in the corresponding region; possible causes include stroke, a tumor, or transient postictal dysfunction. Generalized slowing indicates more widespread dysfunction and may accompany metabolic or toxic disturbances or sedative effects. Background disorganization includes loss of the expected age- and state-appropriate rhythm, symmetry, or reactivity. These findings are distinct from interictal epileptiform discharges—abnormal transients occurring between seizures that support seizure susceptibility. Neither nonspecific slowing nor an isolated epileptiform discharge, without the clinical context, establishes epilepsy. (ncbi.nlm.nih.gov)\n\nSpikes and sharp waves are distinguished partly by duration: a spike lasts 20 to less than 70 milliseconds, whereas a sharp wave lasts 70–200 milliseconds under the American Clinical Neurophysiology Society’s 2021 terminology. An epileptiform interpretation also depends on the transient’s electrical field, distinction from the background, and associated disruption or following slow wave, not simply its pointed appearance. A spike-and-wave complex combines a brief spike with a slower component; repeated complexes may form generalized rhythmic discharges. Interictal spikes are not themselves equivalent to clinical seizures. (pmc.ncbi.nlm.nih.gov)\n\nIctal recordings are assessed for a sustained pattern and its evolution in frequency, morphology, and spatial distribution. A focal seizure may produce rhythmic activity initially confined to one region, followed by changes in frequency and spread to neighboring or bilateral regions. Temporal seizures, for example, can show evolving localized rhythmic theta or delta activity rather than a succession of isolated spikes. In childhood absence epilepsy, a representative generalized ictal pattern is approximately 3-Hz spike-and-wave activity, with a recognized range of 2.5–4 Hz, accompanying the absence episode. Brief generalized interictal fragments may resemble portions of this pattern, so behavioral correlation helps distinguish an electrographic discharge from a clinical absence seizure. (ncbi.nlm.nih.gov)\n\nEEG contributes to seizure classification and epilepsy-syndrome assessment through the distribution of discharges, background organization, sleep dependence, and recorded ictal changes. Its findings are interpreted alongside the event history, age at onset, neurological examination, and magnetic resonance imaging, which supplies structural information rather than electrical recordings. A normal interictal EEG does not exclude epilepsy: epileptiform abnormalities are intermittent, and even prolonged recordings can lack them. Likewise, a captured habitual event must be assessed through synchronized behavior and EEG, rather than through the tracing alone. Video-EEG can distinguish epileptic events from nonepileptic episodes, but absence of a scalp ictal discharge is not sufficient evidence of a nonepileptic mechanism; spatially restricted focal seizures may have no detectable scalp correlate, and movement artifacts can obscure ictal activity. (ncbi.nlm.nih.gov)\n\nPresurgical assessment distinguishes the region generating interictal spikes from the recorded seizure-onset zone and the tissue whose treatment is necessary for seizure control; these regions need not coincide. Scalp ictal recordings help formulate a localization hypothesis, but rapid propagation can obscure or falsely localize the onset. When noninvasive findings are insufficient or discordant, surgically implanted electrodes can test specific anatomical hypotheses. Electrocorticography records from cortical-surface contacts, commonly subdural grids or strips, providing contiguous surface coverage. Stereoelectroencephalography uses stereotactically implanted multicontact depth electrodes to sample selected superficial and deep structures, including sulcal cortex and mesial temporal regions. Both differ fundamentally from routine scalp EEG: they require invasive implantation and record only the tissue sampled by the chosen contacts. Their complementary coverage supports localization, but neither constitutes a complete survey of the brain. (neurosoft.com)"
    },
    {
      "heading": "Critical care, anesthesia, and other clinical monitoring",
      "text": "Continuous EEG in intensive care detects seizures that may have no visible motor signs, particularly in patients with persistently altered consciousness after convulsions, acute brain injury, or otherwise unexplained coma. It distinguishes continuing electrographic seizures from postictal impairment and medication-related sedation, and documents whether treatment suppresses seizure activity. The American Clinical Neurophysiology Society’s (ACNS) 2015 consensus statement proposed at least 24 hours of recording for most indications, with duration individualized to clinical and EEG findings. During continuous intravenous antiseizure treatment, recording can reveal persistent or recurrent seizures despite apparent clinical control; acquisition commonly continues through treatment withdrawal. “Continuous” describes acquisition, not necessarily uninterrupted real-time interpretation. (acns.org)\n\nThe ACNS’s 2021 terminology, developed by Lawrence J. Hirsch and colleagues, describes background frequency, symmetry, continuity, voltage, and reactivity to stimulation. Suppression denotes activity below \\(10\\,\\mu\\mathrm{V}\\); burst suppression consists of higher-voltage bursts alternating with suppression occupying 50–99% of the record. Periodic discharges recur at approximately regular intervals with discernible gaps, whereas rhythmic delta activity consists of repeated slow waves without intervening gaps. Descriptors specify generalized or lateralized distribution, frequency, evolution, and associated features. Some patterns occupy an ictal–interictal continuum: they do not meet definite electrographic seizure criteria but may be associated with impaired function or neuronal injury. This designation is descriptive, not itself a diagnosis of status epilepticus. (acns.org)\n\nDuring anesthesia, EEG provides information about drug-related changes in cortical activity. In a 2013 study, Patrick L. Purdon and colleagues associated propofol-induced loss of responsiveness with increased slow activity and coherent frontal alpha oscillations, together with loss of coherent posterior alpha activity. (pmc.ncbi.nlm.nih.gov) Processed monitors reduce selected EEG features to numerical indices: the bispectral index, for example, uses a 0–100 scale. Such an index is a calculated output, distinct from the underlying waveform; its stated role is to assist assessment of certain anesthetic effects. (accessdata.fda.gov) In selected vascular operations, including carotid endarterectomy, multichannel EEG is also used to identify changes associated with reduced cerebral perfusion during arterial clamping, often alongside other neurophysiological measurements. (pmc.ncbi.nlm.nih.gov)\n\nIn encephalopathy, EEG characterizes cerebral dysfunction and its evolution rather than uniquely identifying its cause. Sedatives, antiseizure medication, hypothermia, and the interval since brain injury influence background continuity and reactivity, so identical-looking patterns can have different implications in different circumstances. (acns.org) The 2023 guidelines for neuroprognostication after adult cardiac arrest classified suppression or burst suppression as a moderately reliable predictor of poor functional outcome only when assessed at least 72 hours after return of spontaneous circulation—or after rewarming when therapeutic hypothermia was used—and without sedation or other relevant confounders. They specified multimodal assessment, incorporating neurological examination, imaging, and other physiological tests rather than an isolated EEG conclusion. (pmc.ncbi.nlm.nih.gov)\n\nElectrocerebral inactivity is a recording finding, not equivalent to death by neurological criteria. EEG assesses predominantly cerebral cortical activity and does not establish loss of brainstem function. The 2023 United States consensus guideline issued jointly by the American Academy of Neurology, American Academy of Pediatrics, Child Neurology Society, and Society of Critical Care Medicine therefore states that EEG must not be used as an ancillary test for this determination. An electrically inactive scalp recording cannot substitute for the required clinical assessment, including brainstem reflexes and apnea testing, or for an accepted ancillary investigation when indicated. (pmc.ncbi.nlm.nih.gov)"
    },
    {
      "heading": "Quantitative analysis and source estimation",
      "text": "Digital signal processing begins with identifying defective channels, selecting usable intervals, filtering, and rereferencing. Average rereferencing subtracts the selected channels’ mean voltage from each channel; consequently, a contaminated electrode can affect otherwise usable recordings. Artifact rejection discards contaminated intervals, whereas independent component analysis (ICA) decomposes multichannel data into statistically independent component time courses and their scalp projections. Components associated with blinking, cardiac activity, or muscle activity can then be excluded during reconstruction. These operations are not physiologically neutral: filtering can alter waveform amplitude and timing, and rejecting a component containing both artifact and cerebral activity removes both. ICA components are mathematical estimates, not automatically individual anatomical generators. These dependencies are addressed in the 2020 COBIDAS MEEG recommendations by Cyril Pernet and colleagues. (eeglab.org)\n\nSpectral analysis describes how signal variance is distributed across frequency. A Fourier transform, often calculated using a fast Fourier transform, converts a finite voltage segment into frequency coefficients; appropriately normalized squared magnitudes provide power estimates. Power spectral density can be expressed in \\(\\mu\\mathrm{V}^{2}/\\mathrm{Hz}\\), and integrating it across a frequency band gives absolute band power in \\(\\mu\\mathrm{V}^{2}\\). Relative band power divides that quantity by power over a specified broader frequency range; it can therefore increase because another band decreases, without an absolute increase in the band of interest. Moving-window Fourier analysis and wavelet methods produce time–frequency representations, with spectrograms displaying power by time, frequency, and color. Longer windows improve frequency discrimination but blur brief changes; shorter windows provide better temporal localization. Changes relative to a baseline, including percentage or decibel changes, are distinct from a band’s proportion of total power. (mne.tools)\n\nFunctional connectivity measures statistical relationships between signals rather than anatomical connections. Magnitude-squared coherence is \\(C_{xy}(f)=|S_{xy}(f)|^{2}/[S_{xx}(f)S_{yy}(f)]\\), where \\(S_{xy}\\) is the cross-spectrum and \\(S_{xx}\\) and \\(S_{yy}\\) are the individual power spectra. Other measures assess phase consistency or amplitude-envelope correlation. A shared reference can introduce common fluctuations, while volume conduction allows one generator to appear simultaneously at several electrodes, producing apparent connectivity without interaction between separate neural populations. Imaginary coherency and phase-lag measures reduce sensitivity to instantaneous mixing, but do not establish causation or eliminate every confound. Estimates also depend on segment duration, sample size, frequency smoothing, and channel- or source-selection choices. (mne.tools)\n\nSource estimation reverses a forward model that predicts electrode voltages from hypothetical intracranial currents. The model incorporates electrode positions, tissue boundaries and conductivities, and candidate source locations and orientations; individual magnetic resonance images can supply anatomical geometry. In the linear expression \\(\\mathbf v=\\mathbf G\\mathbf j+\\boldsymbol\\epsilon\\), \\(\\mathbf G\\) is the lead-field matrix, \\(\\mathbf j\\) represents source strengths, and \\(\\boldsymbol\\epsilon\\) represents noise. The inverse problem is nonunique: different current configurations can explain the same scalp measurements. Regularization stabilizes estimation by balancing data fit against assumptions such as small current magnitude, smoothness, or sparsity. Minimum-norm estimation favors low overall source strength; dipole fitting instead assumes a limited number of generators. The resulting anatomical map depends on these assumptions and on errors in geometry, conductivity, electrode alignment, and noise estimation. (mne.tools)\n\nQuantitative trend displays compress prolonged recordings into spectral-power maps, band-power ratios, suppression measures, or automated event markers. They help reviewers identify intervals requiring closer examination rather than independently determining their clinical meaning. The American Clinical Neurophysiology Society’s 2015 consensus statement, led by Susan T. Herman, explicitly treats quantitative trends as adjuncts to continuous EEG review and notes that synchronized video helps distinguish artifacts and state changes. Automated seizure detections likewise require checking against the underlying waveforms and clinical context; algorithmic labels are not independently validated diagnoses, and performance established in one recording population does not by itself establish performance in another. (acns.org)"
    },
    {
      "heading": "Experimental neuroscience and brain–computer interfaces",
      "text": "Evoked potentials are voltage changes associated with sensory stimulation, such as a flash, sound, or peripheral nerve stimulus. The broader term event-related potential (ERP) includes activity associated with perception, preparation, decisions, and responses. Unlike spontaneous background activity, ERPs are characterized relative to a defined event: stimulus-locked recordings align trials to stimulus onset, whereas response-locked recordings align them to an action, such as a button press. These alignments address different questions because reaction time varies between trials; activity consistently preceding a response can become blurred when averaged relative to stimulus onset. (pages.ucsd.edu)\n\nResearchers divide continuous recordings into event-centered epochs, establish a voltage baseline, exclude contaminated trials, and average corresponding time points. Averaging preserves activity consistently time-locked to the event while reducing unrelated background fluctuations; it does not isolate every neural process occurring during the task. Latency describes a component’s timing relative to the event, usually in milliseconds, while amplitude describes its voltage, commonly measured as a peak or mean within a specified window in microvolts. Peak timing and amplitude depend partly on noise and measurement choices. The 2021 ERP CORE resource, developed by Emily Kappenman and colleagues, provides standardized experimental tasks, recordings, and analysis procedures for seven commonly investigated ERP components, including response-related and stimulus-related measures. (pages.ucsd.edu)\n\nIn sensory and attention research, experimental contrasts can separate changes caused by stimulus properties from those associated with task demands. For example, comparing responses to physically identical stimuli when attended and unattended reveals attentional modulation without changing the sensory input itself; visual attention studies have documented modulation of early P1 and N1 components. Sleep and memory experiments also combine event timing with measurements of ongoing rhythms. In a 2024 high-density EEG study, Andrea Sánchez-Corzo and colleagues presented memory-associated odor cues during sleep and observed increases in spindle rate and duration, together with enhanced slow oscillations and spindle coupling to their up states. Such results concern differences between controlled conditions and candidate mechanisms of memory processing, not direct readings of remembered content or thoughts. (pmc.ncbi.nlm.nih.gov)\n\nAn EEG-based brain–computer interface (BCI) converts selected signal features into device commands, with feedback allowing the user to learn control. In 2004, Jonathan R. Wolpaw and Dennis J. McFarland demonstrated two-dimensional cursor control in four participants using sensorimotor rhythms. Weighted combinations of mu-band activity at 8–12 Hz or beta-band activity at 18–26 Hz over left and right sensorimotor cortex controlled horizontal and vertical movement. The cursor updated every 50 ms, and an adaptive algorithm adjusted feature weights after each trial. Users commonly employed imagined movements early in training; successful control depended on learned rhythm modulation rather than decoding arbitrary intentions. (neurotechcenter.org)\n\nThe participants completed 22–68 training sessions. In their final three sessions, individual target-acquisition rates ranged from 70% to 92% within a 10-second limit; mean movement times on successful trials ranged from 1.9 to 3.9 seconds. These results establish feasibility in trained individuals, not population-wide usability. Practical evaluation separately considers between-person variability, electrode setup, recording stability across days, calibration, fatigue, and assistance requirements. Home-use studies therefore assess sustained operation and user experience rather than laboratory accuracy alone. Independent evaluation must also distinguish performance on previously unseen recordings and users from performance on data used to select features or tune the decoder; held-out testing addresses a different question from continued improvement within an adaptive training session. (neurotechcenter.org)"
    },
    {
      "heading": "Capabilities, limitations, and complementary methods",
      "text": "EEG can follow changes in population electrical activity on a millisecond timescale, making it suitable for distinguishing the timing and sequence of events that slower measurements may merge. Scalp equipment can be used at the bedside or in ambulatory recordings, and passive acquisition without ionizing radiation permits prolonged and repeated observations. Technical sampling, however, is not equivalent to physiological certainty: sampling at \\(1{,}000\\) Hz produces one voltage measurement every millisecond, not a guaranteed identification of every neural event within that interval. Detectability also depends on signal amplitude, noise, filtering, and whether the active tissue generates a measurable scalp field. (camelice.org)\n\nSpatial specificity is limited because each electrode receives contributions from multiple generators after currents spread through brain tissue, cerebrospinal fluid, skull, and scalp. Differences in tissue conductivity reshape and broaden the surface potential distribution; the largest scalp voltage therefore need not lie directly above its generator. Source strength, depth, orientation, and the extent and synchrony of active cortex all affect sensitivity. Contributions from differently oriented cortical surfaces may partially cancel, while deep or spatially restricted activity may be obscured by stronger superficial activity. Increasing electrode density improves spatial sampling but does not eliminate the nonuniqueness of the inverse problem: different source configurations can produce similar scalp measurements. (pmc.ncbi.nlm.nih.gov)\n\nA simultaneous scalp and stereotactic depth-electrode study published in 2022 illustrates the distinction between recording continuously and detecting all events. Among 27 patients undergoing evaluation for drug-resistant epilepsy, depth electrodes recorded 172 seizures, but only 100 had an identifiable scalp ictal pattern. Scalp EEG detected 33% of focal aware seizures and 33% of subclinical seizures, compared with 97% of focal impaired-awareness seizures and 100% of focal-to-bilateral tonic–clonic seizures. These figures describe a selected surgical-evaluation population, not universal sensitivity estimates. Nevertheless, they demonstrate why absence of a scalp correlate cannot by itself establish absence of a seizure, especially when activity remains restricted. (pmc.ncbi.nlm.nih.gov)\n\nMagnetoencephalography (MEG) measures magnetic fields generated by neural currents rather than scalp voltage differences, while retaining millisecond-scale temporal sensitivity. Its source sensitivity differs from EEG: MEG is generally much less sensitive to approximately radial generators than to tangential generators, whereas EEG detects both orientations. In a 2010 realistic-head modeling study, Seppo Ahlfors and colleagues found median ratios between the least- and most-sensitive source orientations of 0.06 for MEG and 0.63 for EEG. Thus, the same neuronal activity can have substantially different visibility in the two modalities; combining them adds information rather than simply duplicating a measurement. Neither modality directly supplies an anatomical image, and localization still requires source models. (pmc.ncbi.nlm.nih.gov)\n\nStructural magnetic resonance imaging uses magnetic fields and radiofrequency excitation to obtain tissue-dependent signals, depicting anatomy and structural lesions rather than the rapid electrical fluctuations measured by EEG. Functional MRI commonly measures blood-oxygen-level-dependent contrast, an indirect vascular response whose evolution over seconds differs from electrical activity on millisecond timescales. The methods can therefore address different aspects of the same event. In 2002, Christian Bénar and colleagues studied four patients with focal epilepsy during 80-minute simultaneous EEG–fMRI sessions, using EEG discharge timing to examine associated BOLD responses. Such correspondence links electrical events to regional hemodynamic changes; it does not make either signal a comprehensive record of all neural activity. In clinical recording, diagnostic yield likewise depends on the question, electrode coverage, recording duration, and whether relevant states or habitual events are sampled; the 2023 IFCN–ILAE standards distinguish routine and sleep recordings for these reasons. (nibib.nih.gov)"
    },
    {
      "heading": "Standards, safety, and research reproducibility",
      "text": "The International Federation of Clinical Neurophysiology (IFCN) and International League Against Epilepsy (ILAE) jointly develop international recording standards. Their 2023 routine and sleep EEG guideline contains 16 consensus-based recommendations covering indications, technical requirements, recording duration, sleep induction, and activation procedures. Its evidence review used formal quality-assessment methods, with modified Delphi consensus where strong evidence was lacking. The American Clinical Neurophysiology Society (ACNS) publishes complementary technical standards for clinical acquisition, electrode nomenclature, montages, digital recording, and reporting, together with setting-specific terminology, including its 2021 critical-care EEG terminology. These documents standardize how recordings are obtained and described rather than replace clinical judgment. (pubmed.ncbi.nlm.nih.gov)\n\nACNS’s 2016 reporting guideline, authored by William O. Tatum and colleagues, specifies five sections: history, technical description, EEG description, impression, and clinical correlation. The history identifies the clinical question and relevant medications. Technical documentation includes recording duration, behavioral states sampled, electrode arrangements, activation procedures, and limitations affecting interpretability. The descriptive section records background organization, symmetry, reactivity, sleep features, abnormalities, and captured events without conflating observation with diagnosis. The impression identifies the principal findings and their significance; clinical correlation relates them to the presenting question. Documenting artifact or incomplete sampling makes the evidential limits of a report explicit. (acns.org)\n\nOrdinary scalp recording measures voltage differences through electrodes; it does not deliver electrical stimulation to the brain. Local redness or skin irritation can occur at electrode sites. Activation procedures have distinct risks: hyperventilation may produce light-headedness, while hyperventilation and flashing lights can provoke seizures in susceptible individuals, requiring supervised protocols. These risks are separate from those of intracranial recording. Implanting surface grids, strips, or depth electrodes requires neurosurgery and introduces risks including infection and intracranial bleeding, occasionally requiring further surgery. Electrical stimulation through implanted contacts is another procedure, not an intrinsic component of passive recording. (nhs.uk)\n\nResearch reproducibility requires documentation of decisions throughout acquisition and analysis. The 2020 OHBM COBIDAS MEEG recommendations, presented by Cyril Pernet and colleagues, address experimental design, acquisition, analysis, reporting, and sharing. Relevant details include participant selection, trial structure and timing, recording reference, sampling and filtering, rejected channels and epochs, artifact-removal criteria, and software versions. Statistical reporting identifies outcomes, analysis windows, uncertainty estimates, and treatment of multiple testing across electrodes, times, or frequencies. Sharing analysis code and preserving processing parameters allow another group to reconstruct the workflow rather than merely repeat its verbal description. (cibm.ch)\n\nAutomated-model evaluation additionally requires separation between model development and independent testing. For claims about previously unseen patients, segments from one participant cannot be distributed across training and test partitions: models may recognize participant-specific recording characteristics rather than the target condition. A 2024 translational EEG study demonstrated this data leakage mechanism; in its seizure-classification experiment, accuracy fell from 79.1% with segment-based splitting to 65.1% with participant-based splitting. Feature selection, learned normalization, and hyperparameter tuning likewise belong within training procedures, with an untouched test set reserved for final evaluation. The partitioning unit therefore determines what kind of generalization the reported performance actually tests. (pmc.ncbi.nlm.nih.gov)\n\nEEG-BIDS and iEEG-BIDS, published in 2019, provide standardized research-data organization rather than anonymization guarantees. They combine naming conventions with machine-readable metadata describing channels, recording settings, events, and electrode coordinates where applicable; iEEG-BIDS also supports intracranial contact localization and coordinate-system documentation. Informed consent and privacy protection remain separate responsibilities. Identifying information can persist in file headers, clinical annotations, images, or synchronized video and audio. Data sharing therefore requires attention to consent scope, removal of identifiers, and appropriate access controls; assigning participant codes or passing a format validator does not itself establish anonymity. (pure.ed.ac.uk)"
    }
  ],
  "links": [
    "brain",
    "electrode",
    "neuron",
    "epilepsy",
    "electric-current",
    "action-potential",
    "synapse",
    "electric-dipole",
    "electric-potential",
    "electrical-impedance",
    "calibration",
    "diagnosis",
    "magnetic-resonance-imaging",
    "consciousness",
    "anesthesia",
    "brain-death",
    "signal-processing",
    "fourier-transform",
    "fast-fourier-transform",
    "correlation",
    "inverse-problem",
    "regularization",
    "experimental-design",
    "test-set",
    "neurosurgery",
    "reproducibility",
    "multiple-testing",
    "data-leakage",
    "generalization",
    "informed-consent",
    "data-privacy"
  ],
  "sources": [
    {
      "title": "Adolf Beck: A pioneer in electroencephalography in between Richard Caton and Hans Berger",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC3902832/"
    },
    {
      "title": "The Berger Rhythm: Potential Changes from the Occipital Lobes in Man",
      "url": "https://docs.backyardbrains.com/assets/files/AdrianMathews-1934-EEG-recordings-4f74d8f2f457d02d53d5c8c9bd21092b.pdf"
    },
    {
      "title": "American Clinical Neurophysiology Society Guideline 5: Minimum Technical Standards for Pediatric EEG",
      "url": "https://www.acns.org/UserFiles/file/Guideline5-MinimumTechnicalStandardsforPediatricEEG.pdf"
    },
    {
      "title": "American Clinical Neurophysiology Society Guideline 1: Minimum Technical Requirements for Performing Clinical Electroencephalography",
      "url": "https://www.acns.org/UserFiles/file/EEGGuideline1Tech.pdf"
    },
    {
      "title": "EEG Guidelines Introduction",
      "url": "https://www.acns.org/UserFiles/file/EEGGuidelineintro.pdf"
    },
    {
      "title": "American Clinical Neurophysiology Society Guideline 4: Recording Clinical EEG on Digital Media",
      "url": "https://www.acns.org/UserFiles/file/Guideline4-RecordingClinicalEEGonDigitalMedia_v1.pdf"
    },
    {
      "title": "American Clinical Neurophysiology Society Guideline 7: Guidelines for EEG Reporting",
      "url": "https://doi.org/10.1097/WNP.0000000000000319"
    },
    {
      "title": "Basic Mechanisms Underlying Seizures and Epilepsy",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK2510/"
    },
    {
      "title": "On the validity of electric brain signal predictions based on population firing rates",
      "url": "https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1012303"
    },
    {
      "title": "Focal Epileptiform Discharges Can Mimic Electrode Artifacts When Recorded on the Scalp Near a Skull Defect",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6104203/"
    },
    {
      "title": "Introduction — Electroencephalography (EEG): An Introductory Text and Atlas of Normal and Abnormal Findings in Adults, Children, and Infants",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390346/"
    },
    {
      "title": "Fine spatiotemporal structure of phase in human intracranial EEG",
      "url": "https://pubmed.ncbi.nlm.nih.gov/16737849/"
    },
    {
      "title": "Prediction of electroencephalographic spectra from neurophysiology",
      "url": "https://pubmed.ncbi.nlm.nih.gov/11308514/"
    },
    {
      "title": "Routine and sleep EEG: minimum recording standards of the International Federation of Clinical Neurophysiology and the International League Against Epilepsy",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10006292/"
    },
    {
      "title": "American Clinical Neurophysiology Society Guideline 2: Guidelines for Standard Electrode Position Nomenclature",
      "url": "https://www.acns.org/UserFiles/file/Guideline2-GuidelinesforStandardElectrodePositionNomenclature_v1.pdf"
    },
    {
      "title": "Guideline 3: A Proposal for Standard Montages to Be Used in Clinical EEG",
      "url": "https://www.acns.org/UserFiles/file/EEGGuideline3Montage_final20160323revclean_v1.pdf"
    },
    {
      "title": "Guideline 4: Recording Clinical EEG on Digital Media",
      "url": "https://www.acns.org/UserFiles/file/EEGGuideline4Digital.pdf"
    },
    {
      "title": "Appendix 2. Principles of Digital EEG",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390353/?report=printable"
    },
    {
      "title": "American Clinical Neurophysiology Society Guideline 1: Minimum Technical Requirements for Performing Clinical Electroencephalography",
      "url": "https://www.acns.org/UserFiles/file/EEGGuideline1Tech_finalrev20160411clean_v1.pdf"
    },
    {
      "title": "Routine and sleep EEG: Minimum recording standards of the International Federation of Clinical Neurophysiology and the International League Against Epilepsy",
      "url": "https://www.mknft.hu/wp-content/uploads/2023/03/2023.-Peltola_Routine-and-sleep-EEG_minimum-standards-of-the-IFCN-and-ILAE.pdf"
    },
    {
      "title": "Minimum Technical Requirements for Performing Ambulatory EEG",
      "url": "https://pubmed.ncbi.nlm.nih.gov/35916885/"
    },
    {
      "title": "Minimum standards for inpatient long-term video-EEG monitoring: A clinical practice guideline of the international league against epilepsy and international federation of clinical neurophysiology",
      "url": "https://neurosoft.com/files/catalog/catalog/1677/en/files/ILAE-and-IFCN-LTM-Guidelines.pdf"
    },
    {
      "title": "A revised glossary of terms most commonly used by clinical electroencephalographers and updated proposal for the report format of the EEG findings. Revision 2017",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6123891/"
    },
    {
      "title": "The Normal EEG",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390343/"
    },
    {
      "title": "ATS Core Curriculum 2016: Part I. Adult Sleep Medicine",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC5461994/"
    },
    {
      "title": "The AASM Manual for the Scoring of Sleep and Associated Events: Summary of Updates in Version 2.0",
      "url": "https://aasm.org/wp-content/uploads/2017/11/Summary-of-Updates-in-v2.0-FINAL.pdf"
    },
    {
      "title": "Technical standards for recording and interpretation of neonatal electroencephalogram in clinical practice",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC2811985/"
    },
    {
      "title": "The Developmental EEG: Premature, Neonatal, Infant, and Children",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390356/"
    },
    {
      "title": "Protocol for designing, conducting, and analyzing event-related potentials in human participants",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12151294/"
    },
    {
      "title": "The Effects of Electrode Impedance on Data Quality and Statistical Significance in ERP Recordings",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC2902592/"
    },
    {
      "title": "Methodological aspects of EEG and body dynamics measurements during motion",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC3970018/"
    },
    {
      "title": "Routine and sleep EEG: Minimum recording standards of the International Federation of Clinical Neurophysiology and the International League Against Epilepsy",
      "url": "https://camelice.org/wp-content/uploads/2023/04/Routine-and-sleep-EEG-Minimum-recording-standards-of-the-internationa-federation-of-clinical-neurophysiology-and-the-international-league-against-epilepsy.pdf"
    },
    {
      "title": "Benign Variants in the EEG",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390352/"
    },
    {
      "title": "EEG normal variants: A prospective study using the SCORE system",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9294211/"
    },
    {
      "title": "The Abnormal EEG",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390357/?report=printable"
    },
    {
      "title": "EEG in the Epilepsies",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK390347/"
    },
    {
      "title": "American Clinical Neurophysiology Society’s Standardized Critical Care EEG Terminology: 2021 Version",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8135051/"
    },
    {
      "title": "Childhood Absence Epilepsy (CAE): EEG",
      "url": "https://www.epilepsydiagnosis.org/syndrome/cae-eeg.html"
    },
    {
      "title": "Consensus Statement on Continuous EEG in Critically Ill Adults and Children, Part I: Indications",
      "url": "https://www.acns.org/UserFiles/file/Consensus_Statement_on_Continuous_EEG_in.1.pdf"
    },
    {
      "title": "American Clinical Neurophysiology Society’s Standardized Critical Care EEG Terminology: 2021 Version",
      "url": "https://www.acns.org/UserFiles/file/ACNSStandardizedCriticalCareEEGTerminology_rev2021.pdf"
    },
    {
      "title": "Electroencephalogram signatures of loss and recovery of consciousness from propofol",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC3607036/"
    },
    {
      "title": "BARM – Traditional 510(k)",
      "url": "https://www.accessdata.fda.gov/cdrh_docs/pdf21/K213273.pdf"
    },
    {
      "title": "Indications for use for BIS EEG Monitor System (VISTA Monitor and BISx4), and the BISx device",
      "url": "https://www.accessdata.fda.gov/cdrh_docs/pdf7/K072286.pdf"
    },
    {
      "title": "Carotid Endarterectomy Surgeries: A Multimodality Intraoperative Neurophysiological Monitoring Approach",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9348437/"
    },
    {
      "title": "Guidelines for Neuroprognostication in Comatose Adult Survivors of Cardiac Arrest",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/10241762/"
    },
    {
      "title": "Pediatric and Adult Brain Death/Death by Neurologic Criteria Consensus Guideline: Report of the AAN Guidelines Subcommittee, AAP, CNS, and SCCM",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10791061/"
    },
    {
      "title": "Independent Component Analysis for artifact removal",
      "url": "https://eeglab.org/tutorials/06_RejectArtifacts/RunICA.html"
    },
    {
      "title": "Issues and recommendations from the OHBM COBIDAS MEEG committee for reproducible EEG and MEG research",
      "url": "https://cibm.ch/wp-content/uploads/s41593-020-00709-0.pdf"
    },
    {
      "title": "Frequency and time-frequency sensor analysis",
      "url": "https://mne.tools/stable/auto_tutorials/time-freq/20_sensors_time_frequency.html"
    },
    {
      "title": "mne_connectivity.spectral_connectivity_epochs",
      "url": "https://mne.tools/mne-connectivity/stable/generated/mne_connectivity.spectral_connectivity_epochs.html"
    },
    {
      "title": "Algorithms and other implementation details",
      "url": "https://mne.tools/stable/documentation/implementation.html"
    },
    {
      "title": "Tutorials/SourceEstimation",
      "url": "https://neuroimage.usc.edu/brainstorm/Tutorials/SourceEstimation"
    },
    {
      "title": "Event-related brain potentials: an introduction",
      "url": "https://pages.ucsd.edu/~scoulson/cogs179/ColesRugg1995chpt1.pdf"
    },
    {
      "title": "ERP CORE: An open resource for human event-related potential research",
      "url": "https://www.emilykappenman.org/wp-content/uploads/2021/07/Kappenman-et-al_2021_NeuroImage_ERP-CORE-An-open-resource-for-human-event-related-potential-research.pdf"
    },
    {
      "title": "How does noise affect amplitude and latency measurement of event-related potentials (ERPs)? A methodological critique and simulation study",
      "url": "https://pubmed.ncbi.nlm.nih.gov/23216521/"
    },
    {
      "title": "Event-related brain potentials in the study of visual selective attention",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC33798/"
    },
    {
      "title": "Control of a two-dimensional movement signal by a noninvasive brain-computer interface in humans",
      "url": "https://www.neurotechcenter.org/sites/default/files/misc/Control%20of%20a%20two-dimensional%20movement%20signal%20by%20a%20noninvasive%20brain-computer%20interface%20in%20humans.pdf"
    },
    {
      "title": "Independent home use of a brain-computer interface by people with amyotrophic lateral sclerosis",
      "url": "https://www.neurotechcenter.org/sites/default/files/misc/Independent%20home%20use%20of%20a%20brain-computer%20interface%20by%20people%20with%20amyotrophic%20lateral%20sclerosis.pdf"
    },
    {
      "title": "Table 1, Main features to compare MEG and EEG",
      "url": "https://www.ncbi.nlm.nih.gov/books/NBK597471/table/ch9.Tab1/?report=objectonly"
    },
    {
      "title": "Sensitivity of MEG and EEG to Source Orientation",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC2914866/"
    },
    {
      "title": "The Sensitivity of Scalp EEG at Detecting Seizures—A Simultaneous Scalp and Stereo EEG Study",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8290181/"
    },
    {
      "title": "Magnetic Resonance Imaging (MRI)",
      "url": "https://www.nibib.nih.gov/science-education/science-topics/magnetic-resonance-imaging-mri"
    },
    {
      "title": "The BOLD response to interictal epileptiform discharges",
      "url": "https://pubmed.ncbi.nlm.nih.gov/12414258/"
    },
    {
      "title": "Routine and sleep EEG: Minimum recording standards of the International Federation of Clinical Neurophysiology and the International League Against Epilepsy",
      "url": "https://pubmed.ncbi.nlm.nih.gov/36762397/"
    },
    {
      "title": "Guidelines and Consensus Statements",
      "url": "https://www.acns.org/practice/guidelines"
    },
    {
      "title": "Guideline 7: Guidelines for EEG Reporting",
      "url": "https://www.acns.org/UserFiles/file/EEGguideline7Report.pdf"
    },
    {
      "title": "EEG (electroencephalogram)",
      "url": "https://www.nhs.uk/tests-and-treatments/electroencephalogram/"
    },
    {
      "title": "Routine EEG",
      "url": "https://www.cuh.nhs.uk/patient-information/routine-eeg/"
    },
    {
      "title": "Day case EEG",
      "url": "https://www.guysandstthomas.nhs.uk/health-information/day-case-eeg"
    },
    {
      "title": "Invasive EEG monitoring",
      "url": "https://www.gosh.nhs.uk/conditions-and-treatments/procedures-and-treatments/invasive-monitoring/"
    },
    {
      "title": "Data leakage in deep learning studies of translational EEG",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11099244/"
    },
    {
      "title": "Data leakage in deep learning studies of translational EEG",
      "url": "https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2024.1373515/pdf"
    },
    {
      "title": "Achieving Reproducibility in EEG-Based Machine Learning",
      "url": "https://facctconference.org/static/papers24/facct24-98.pdf"
    },
    {
      "title": "EEG-BIDS, an extension to the brain imaging data structure for electroencephalography",
      "url": "https://www.pure.ed.ac.uk/ws/portalfiles/portal/98143283/s41597_019_0104_8.pdf"
    }
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}