MEG: Measuring Magnetic Fields Produced by Brain Activity
Author: Ukrainian Psychological Hub · Published: October 12, 2026 · Editorial Policy
Magnetoencephalography (MEG) is a noninvasive method for recording extremely weak magnetic fields generated by electrical currents in the human brain. Sensitive sensors positioned around the head capture changes in these fields while a person rests, perceives stimuli, speaks, moves, or performs a cognitive task. MEG lets researchers follow neural activity on a millisecond timescale and, with computational modeling, estimate which populations of neurons contributed to an observed signal. It is used in cognitive neuroscience and in selected clinical evaluations, especially epilepsy surgery planning and functional brain mapping. National Institute of Mental Health MEG facility; Kim and Davis, 2021.
The word 'scan' can create the wrong mental picture. An MEG system records time-varying magnetic signals outside the scalp. Brain maps shown after an examination are reconstructions inferred from these measurements, often aligned to a structural MRI. The magnetic recording is real and directly related to neural electrical activity; a colored map of a particular cortical source is a model-based estimate whose accuracy depends on anatomy, measurement geometry, noise, and analytical assumptions. Gross et al., 2013; Overview of MEG.
What is MEG, and what does an MEG brain scan measure?
MEG stands for magnetoencephalography. It detects magnetic fields produced by intracellular electrical currents in populations of neurons, predominantly synchronous postsynaptic currents in suitably oriented cortical pyramidal cells. These fields are extraordinarily small: measurements near the head are often on the scale of femtoteslas (10^-15 tesla) to picoteslas (10^-12 tesla), depending on the activity and recording conditions. Sensitive magnetic sensors and environmental shielding are necessary because nearby electrical equipment, moving metal, and Earth-related background fields can overwhelm brain signals. Kim and Davis, 2021; NIMH.
A MEG recording therefore provides information about when neural populations generate a detectable magnetic field and about the patterns and frequencies of that activity. It does not measure a person's thoughts, neurotransmitter concentrations, individual neuron firing throughout the brain, cerebral blood flow, or the literal location of every electrical event. Signals from many neurons add together, while signals with unfavorable orientation or weak synchrony may cancel or fall below the sensor's sensitivity. The distinction between a physical measurement and an inferred neural interpretation governs every scientifically responsible MEG result.
How does brain activity produce a magnetic field?
Neurons exchange information through electrical and chemical processes. During synaptic input, charged particles move across neuronal membranes and generate intracellular and return currents. According to electromagnetism, currents produce magnetic fields. In a sheet of similarly oriented cortical neurons, synaptic activity that is sufficiently coordinated can produce a net field detectable outside the head. Much of the measurable signal comes from aggregated postsynaptic currents in the dendrites of pyramidal neurons, rather than a one-to-one measurement of axonal action potentials. Hari and Salmelin, 2012; Kim and Davis, 2021.
Cortical geometry matters. Many cortical pyramidal cells are aligned approximately perpendicular to the local cortical sheet. Where the sheet folds into a sulcus, the associated current can be oriented tangentially relative to the head and produce a field well detected by conventional MEG. A radial source, oriented approximately toward or away from the scalp, can contribute little to conventional MEG under simplified spherical models. Real head geometry, extended sources, sensor orientation, and alternative measurement configurations complicate this textbook distinction, so radial invisibility should not be treated as an absolute property of all MEG systems. Overview of MEG; Kim and Davis, 2021.
The skull and scalp affect magnetic fields differently from electrical potentials recorded by EEG. Conductivity boundaries can strongly shape scalp voltage patterns, whereas brain-generated magnetic fields undergo comparatively little distortion by these tissues. That advantage can help source modeling, but it does not eliminate the magnetic inverse problem or guarantee exact source positions. Deep generators, cancellation of oppositely oriented currents, sensor-to-head distance, and environmental interference remain important.
Inside the MEG system: sensors, shielding, and recording
Conventional SQUID-MEG
Conventional whole-head MEG systems use arrays of superconducting quantum interference devices (SQUIDs) or related superconducting detectors housed in a helmet-shaped cryogenic assembly. The detectors sense extraordinarily weak changes in magnetic flux; cooling is needed for the superconducting technology, while the participant remains outside the cryogenic apparatus. Some facilities use magnetometers, others gradiometers, and many systems incorporate both sensor classes or reference channels to reduce interference. The precise number and configuration of sensors vary by model. NIMH MEG Core Facility; Hari and Salmelin, 2012.
Recording usually takes place in a magnetically shielded room. Head-localization coils, digitized anatomical landmarks, and sometimes high-resolution MRI information allow the recording to be registered with a person's head anatomy. Microphones, screens, projectors, response buttons, touch-free movement measurements, and simultaneous EEG may be incorporated depending on the protocol. These auxiliary measurements clarify the timing of stimuli, responses, and artifacts.
Optically pumped magnetometer MEG
Optically pumped magnetometers (OPMs) use atomic vapor and light to measure magnetic fields. Some OPM sensors can be mounted closer to a participant's scalp in lightweight, adjustable headgear instead of relying on a fixed-size cryogenic helmet. This offers an important research direction for people whose head shape, size, or movement makes conventional MEG difficult, including children. OPM systems still need sophisticated field control, shielding or active compensation, sensor calibration, and control of movement-related interference. Laboratory demonstrations should not be mistaken for universal clinical availability. Brookes and colleagues, 2022; Brickwedde et al., 2024.
What happens during an MEG examination?
An appointment starts with screening, an explanation of the procedure, and instructions tailored to the study. Participants may be asked to remove jewelry, watches, phones, or other magnetic or metallic objects. The team reviews implants, medical devices, and any factors that might interfere with the recording; whether a specific device is acceptable depends on the center and the imaging protocol. MEG itself passively records fields and uses no ionizing radiation. A separate MRI examination, if scheduled, has its own screening requirements and should not be conflated with MEG. {{RadiologyInfo: MEG::radiology}}.
During recording, the participant sits or lies with the head positioned in or near the sensor helmet. They may be asked to remain reasonably still, look at pictures, listen to sounds, move a hand or finger, read or name words, or rest with eyes open or closed. In an epilepsy assessment, the center may record spontaneous activity to detect interictal epileptiform discharges. The study team may simultaneously record scalp EEG, eye movements, heartbeat, muscle activity, or stimulus timing to aid interpretation. A typical protocol includes preparation and recording stages; total appointment length varies substantially by center and clinical question.
The procedure is generally painless, with no injected tracer or electrical stimulation from the recording itself. Holding still, the room environment, or repeated tasks may be uncomfortable for some people. The physical sensor assembly is close to the head, but a typical MEG recording is not a conventional MRI tunnel. People with movement, sensory, or communication differences should discuss accommodations with the center before the appointment. RadiologyInfo.
How quickly can MEG follow the brain? Temporal resolution versus localization
MEG is especially useful when the scientific question concerns timing. Data are often sampled hundreds or thousands of times per second, making it possible to analyze changes on a millisecond scale. A task might produce an early sensory response tens of milliseconds after a stimulus, followed by later processing and network interactions. The recorded time course is much closer to the electrical event than an fMRI BOLD response, which reflects a slower vascular process. Sampling rate, filtering, signal-to-noise ratio, and the duration of the neural event still constrain what can reliably be distinguished. Hari and Salmelin, 2012; Gross et al., 2013.
Spatial resolution is a different question. MEG has dense spatial sensor sampling and can sometimes localize relatively focal cortical activity with accuracy of several millimeters under favorable conditions. That number is not a uniform voxel size or an unconditional guarantee for an individual person. Source depth, geometry, the strength and extent of current sources, head motion, anatomy, noise, and the chosen inverse model all influence the resulting estimate. Multiple regions can produce similar sensor patterns; weak or highly correlated sources may be hard to disentangle. NIMH; Overview of MEG.
Why source localization is an inverse problem
The forward problem asks: given an assumed distribution of neural current sources, a head model, and sensor geometry, what magnetic fields should the sensors record? Physics and numerical modeling can predict that field distribution. The inverse problem asks the opposite: given external magnetic measurements, which internal sources produced them? The inverse problem is generally ill-posed because many distinct combinations of current sources can produce indistinguishable fields at the available sensors. Additional anatomical, statistical, and physiological assumptions are needed to choose a plausible solution. Kim and Davis, 2021; Gramfort et al., 2014.
Equivalent current dipole modeling represents an activity pattern with one or more fitted dipoles. This approach is especially familiar in clinical localization of interictal epileptiform discharges when the observed field is compatible with a focal generator. Minimum-norm estimates distribute currents over an assumed source space and favor particular solutions through regularization; the chosen constraint shapes the map. Beamformers construct spatial filters intended to estimate activity at candidate source locations while suppressing other contributions. These methods answer related but different questions and can produce meaningfully different maps from the same raw recording. Kim and Davis, 2021.
Structural MRI is often used to build an individualized source space and head model and to align the estimated activity with cortex and other anatomy. Magnetic source imaging (MSI) commonly refers to combining MEG source estimates with anatomical MRI visualization. The MRI supplies anatomical detail; MEG supplies time-resolved magnetic measurements. An anatomical overlay does not convert an inferred MEG source into directly observed tissue activity. Investigators should report model choices, registration quality, uncertainty, and appropriate validation where possible.
How researchers turn raw MEG signals into results
Recording and artifact identification
A continuous MEG dataset includes desired neural signals and unwanted contributions. Researchers inspect sensor channels, timing triggers, movements, eye blinks, cardiac activity, jaw or neck muscle activation, electrical interference, and residual environmental fields. No single filtering step can cleanly separate all neural signals from artifacts. A noise-removal method may suppress useful neural activity if its assumptions are violated; a plausible looking waveform is not enough to demonstrate a valid neural source. A 2025 systematic review of noise and artifact suppression specifically distinguishes environmental, physiological, and technical problems in both SQUID and wearable OPM-MEG. Wang et al., 2025; Gross et al., 2013.
Evoked responses and event-related fields
For controlled tasks, the signal can be divided into epochs around repeated events and averaged to estimate an event-related magnetic field (ERF). Repetition can make activity consistently time-locked to the event easier to detect while reducing unrelated fluctuations. A clean average reveals a reproducible time-locked pattern in the selected conditions, but not necessarily a unique cognitive operation or an isolated anatomical module. Different perceptual, attentional, and motor processes may overlap in both time and space.
Brain rhythms and time-frequency analysis
MEG can also characterize changes in oscillatory activity, such as alpha-, beta-, theta-, and gamma-range power, without equating each frequency band with a fixed mental state. Spectral power measures how strongly particular frequencies appear in a signal; phase and phase consistency can provide additional timing information. Event-related synchronization and desynchronization reflect changes relative to a baseline, not an on/off switch for a single function. Rhythms arise in distributed circuits and depend on behavioral state, task, anatomy, and analysis decisions. Gross et al., 2013; Kim and Davis, 2021.
Functional connectivity and statistical testing
Connectivity analyses ask whether signals from different sensors or inferred sources vary together in time, phase, or amplitude. Correlation, coherence, phase-based metrics, and envelope coupling quantify particular relationships; they do not by themselves establish a direct anatomical connection, causal direction, or communication between two isolated regions. Common sources, volume conduction or field spread, reference choices, signal leakage, and head motion can create spurious associations. Multiple comparisons across sensors, source locations, times, and frequencies also require an appropriate statistical strategy. Transparent preprocessing, preregistered questions where feasible, data-quality reporting, and independent replication strengthen conclusions. Gross et al., 2013; Gramfort et al., 2014.
MEG versus EEG, MRI, fMRI, and PET
MEG and electroencephalography (EEG) record different physical components of related neural electrical activity. EEG measures voltage differences at scalp electrodes; MEG records magnetic fields. Both can follow fast neural events, and their sensitivity to source orientation, source depth, noise, and head tissues differs. Their measurements are complementary: simultaneous EEG-MEG can provide more information about a candidate generator than either alone. MEG is often more costly and less widely available; EEG is portable and accessible in many healthcare settings. Neither method is universally superior for locating every source or assessing every neurological problem. Kim and Davis, 2021.
Structural MRI characterizes tissue anatomy and lesions at high spatial detail but is not a direct millisecond-scale recording of neural current. Functional MRI commonly measures blood-oxygen-level-dependent (BOLD) signals linked through neurovascular physiology to neural activity; it offers different spatial coverage and slower temporal information. PET uses radioactive tracers to measure tracer-dependent processes such as metabolism or molecular targets, so PET involves ionizing radiation and has different questions and timescales. These tools often converge in multimodal research and clinical evaluation. Our guide to MRI, fMRI, PET, and EEG measurements explains their comparative measurement principles. Kim and Davis, 2021; RadiologyInfo.
The clinically relevant choice is driven by the decision to be made: whether a lesion is visible, whether a seizure-related electrical source can be estimated, whether a language-related cortical response is present, or whether a metabolic pattern adds information. An impressive-looking brain map should never replace a clearly defined question and suitable reference standard.
What is MEG used for in clinical practice?
Presurgical evaluation for epilepsy
MEG has an established role as an adjunct in the presurgical evaluation of selected patients with focal epilepsy, particularly when the location of a potential seizure-related network is uncertain. Epileptic activity recorded between seizures, called interictal epileptiform discharges or spikes, can be modeled to estimate probable generators. These estimates can help clinicians form or refine hypotheses for further testing, including intracranial EEG when indicated. A location associated with interictal spikes is not automatically the precise seizure onset zone, the full epileptogenic network, or the complete tissue that should be removed. Geller et al., 2024; ACMEGS practice guidance.
Clinical interpretation integrates symptoms, seizure semiology, scalp video-EEG, structural MRI, and sometimes PET, other investigations, or invasive recordings. A 2025 qualitative systematic review of nine focal-epilepsy studies involving 354 participants found that MEG offered complementary localizing information and that multimodal combinations were often more informative than MEG alone. The evidence differed across patient groups, reference standards, combinations, and study designs; individual performance percentages from selected series should not be advertised as a universal diagnostic accuracy or a guarantee of successful surgery. Kim et al., 2025 systematic review.
Guidelines from the American Clinical Magnetoencephalography Society specify clinical recording, functional mapping, reporting, and professional standards. These standards concern how appropriately trained teams acquire and interpret MEG data; they do not imply that an MEG result alone establishes a diagnosis or selects surgical treatment. Whether MEG is useful for a particular patient is a specialist decision within a comprehensive epilepsy evaluation. ACMEGS clinical practice guidelines.
Mapping sensory, motor, and language functions
MEG can record time-locked responses during stimulation or tasks engaging somatosensory, auditory, visual, motor, and some language-related processes. When estimated sources are aligned to anatomy, this may contribute to presurgical functional mapping, for example by helping identify responsive cortex near tissue under consideration for surgery. The test protocol must be matched to the function of interest, the person's ability to complete it, and the surgical question. A response in a cortical location is evidence of task-related physiological engagement, not proof that the whole psychological faculty resides exclusively there. ACMEGS functional mapping guideline; RadiologyInfo.
Other neurological and psychiatric research
Researchers have used MEG to examine brain dynamics in dementia, movement disorders, pain, psychosis, autism, ADHD, and other conditions. Some studies report group-level differences in rhythmic activity or functional connectivity. These are active areas of investigation, with varying sample sizes, diagnostic heterogeneity, medication and movement confounds, and substantial differences in preprocessing and models. A reproducible group association does not automatically yield a sensitive, specific, independently validated test for an individual patient. MEG should not be presented as a routine standalone diagnostic test for a psychiatric disorder. Brickwedde et al., 2024; 2024 methods and clinical review.
What can MEG reveal about thinking, perception, and behavior?
In cognitive neuroscience, MEG is especially valuable for the order and timing of processing. Researchers can examine how sensory responses evolve after a sound or picture; how attention changes the amplitude or phase of population-level activity; when signals related to an upcoming movement emerge; and how activity changes during memory encoding, retrieval, speech perception, or working memory tasks. MEG can also test hypotheses about coordination among distributed neural populations as experimental demands change. The method is a bridge between the temporal dynamics of neural signals and observable performance. Hari and Salmelin, 2012; Gross et al., 2013.
Such findings should be interpreted alongside controlled task design and behavioral measures. For example, an observed response 200 milliseconds after a word is not a direct measurement of a specific thought, and an attention-related change in alpha activity does not identify a universal marker of concentration. Reaction time, task accuracy, language context, previous experience, and other experimental variables help determine which interpretation is justified. For a broader account of how investigators connect measurements to cognitive questions, see cognitive neuroscience and its research scope; for the circuit-level framework, see how neurons, circuits, and networks shape behavior.
Key scientific limitations and sources of uncertainty
The main limitation is incomplete observability. Sensors detect the combined fields of multiple generators, with stronger sensitivity to some cortical source orientations than others. Deep structures are usually harder to measure and separate reliably. A lack of detectable MEG signal from a region does not demonstrate that the region is inactive. Similarly, an apparent focal map can be produced by different plausible generator configurations. Overview of MEG; Kim and Davis, 2021.
Source reconstruction introduces assumptions about cortical geometry, conductivity, dipole orientation, regularization, noise covariance, and source extent. A different head model or inverse method can shift an estimate or change the apparent number of active regions. Head movement and imperfect MRI coregistration further affect accuracy. Differences between research groups may reflect acquisition hardware, choice of frequency range, artifact handling, task design, sample composition, or statistical thresholds. Gramfort et al., 2014; Gross et al., 2013.
Connectivity estimates add another uncertainty layer. Two inferred cortical time series may correlate because of leakage from a common source; an apparent temporal lead can also depend on filtering and noise. Evidence for functional coordination is valuable, but demonstrating directional causal influence requires stronger assumptions or interventions and appropriate convergent evidence. MEG is an observational recording method: it does not directly manipulate the brain as stimulation methods do.
Real-world access is also a limitation. Conventional MEG requires specialized equipment, trained staff, expensive infrastructure, and carefully controlled magnetic conditions. OPM technology changes some engineering constraints but introduces others, including sensor drift, field compensation, motion artifacts, and cross-system standardization. The 2025 systematic review of SQUID and OPM artifacts underscores that improvements in sensor technology do not remove the need for careful measurement practice. Wang et al., 2025.
Recent developments: wearable MEG and the evidence frontier
Wearable OPM-MEG can bring sensors close to the scalp and adapt to different head sizes. Research groups have demonstrated dynamic recordings with movement allowances that are difficult for conventional fixed-helmet systems, prompting interest in developmental neuroscience, more natural interactions, and selected neurological populations. Improvements in magnetic shielding, active field compensation, whole-head sensor coverage, and source reconstruction are the subjects of ongoing work. Brookes and colleagues, 2022; Brickwedde et al., 2024.
The current evidence supports substantial technical promise while leaving important translational questions open. Rigorous head-to-head comparisons, reliable calibration, longitudinal reproducibility, robust clinical outcomes, access and cost-effectiveness in different health systems, and adequate data-processing standards are needed before broad claims of clinical equivalence or superiority. More detailed sensory signals or sharper source estimates should not be advertised as the ability to diagnose a broad set of psychiatric disorders. Brickwedde et al., 2024; Wang et al., 2025.
Common misconceptions about MEG
Myth: MEG directly photographs the mind. Reality: it measures external magnetic fields associated with neural electrical currents; any source image is reconstructed and task interpretation requires behavioral and experimental context.
Myth: millisecond temporal resolution means millimeter-perfect localization. Reality: temporal sampling and source localization are different properties, and source accuracy varies with geometry, noise, and modeling.
Myth: MEG finds every active brain area. Reality: sensitivity depends on current orientation, synchrony, depth, and cancellation. Undetected activity may still be physiologically important.
Myth: a MEG abnormality identifies a psychiatric diagnosis. Reality: most condition-related MEG signatures are research findings, and individual clinical interpretation requires validated criteria and convergent assessment.
Myth: MEG and MRI are interchangeable. Reality: MEG measures fast magnetic dynamics, while structural MRI describes anatomy; joint MEG-MRI source maps combine their different information.
Frequently asked questions
Is an MEG scan safe?
A standard MEG recording is passive and noninvasive and does not use ionizing radiation. Preparation still matters because metal objects, implants, or devices can interfere with the magnetic measurement; an accompanying MRI is a separate procedure with different safety screening. The center should assess the individual setup. RadiologyInfo.
Does MEG involve magnetic stimulation or change brain activity?
MEG records fields emitted by neural electrical currents. It is distinct from transcranial magnetic stimulation (TMS), which applies magnetic pulses to induce electric fields in brain tissue. A conventional MEG recording itself is not a brain stimulation treatment.
What is the difference between MEG and EEG?
EEG measures scalp voltage differences, whereas MEG measures the magnetic field produced by overlapping neural currents. Both capture rapid activity; they have distinct sensitivity profiles and sources of noise and are often complementary in clinical and research contexts. Kim and Davis, 2021.
Can MEG locate the origin of seizures?
MEG can help estimate the sources of some epileptiform discharges and contribute to presurgical hypotheses. In selected focal epilepsy cases, its strongest clinical role is complementary localization with EEG, MRI, and other tests. Interictal spike sources, seizure onset, and the entire tissue required for surgery are not necessarily identical. Geller et al., 2024.
Can MEG show emotions, memories, or specific thoughts?
It can detect time-varying neural patterns during tasks that involve emotion, memory, or language, and researchers can statistically relate features to task conditions. It does not provide a universal readout of an individual's private mental content. Predictions are conditional on training data, the task, noise, and the model used.
Does MEG measure dopamine, serotonin, or blood flow?
MEG measures neuromagnetic signals related to population electrical activity. It does not directly quantify dopamine, serotonin, receptors, metabolism, or perfusion. Other specialized methods may address those variables under specific protocols.
Is MEG better than fMRI?
MEG offers much finer temporal tracking of electrical dynamics, while fMRI provides a different spatially resolved measure related to hemodynamic activity. Which technique is more informative depends on the experiment or clinical question. Combining methods can clarify the relationship between rapid signaling and slower vascular responses. Kim and Davis, 2021.
Why are MEG pictures often placed on top of MRI images?
The MEG sensors record magnetic data outside the head. Anatomical MRI provides a structural coordinate system for modeling and displaying estimated generators. The resulting image is a spatial estimate, and its precision depends on how accurately the datasets were aligned.
How long does the exam take, and must I remain still?
Preparation and acquisition time depend on the center and the purpose of the test, from short research recordings to more extensive presurgical protocols. Conventional systems generally benefit from minimized head motion; some newer wearable OPM systems allow more movement, with technical restrictions. Ask the imaging team for the specific appointment instructions.
Can MEG diagnose ADHD, autism, depression, or schizophrenia?
MEG research has investigated neural dynamics and population differences associated with these conditions. Such findings do not constitute validated standalone diagnostic tests for an individual. Diagnosis, when relevant, follows established clinical criteria and comprehensive assessment rather than a MEG map. Brickwedde et al., 2024.
Scientific conclusion
MEG is a powerful method for measuring rapid magnetic signatures of collective neuronal electrical activity. Its principal contribution is temporal precision combined with carefully modeled spatial inference. It supports research on sensory, motor, language, and cognitive dynamics and provides complementary information in selected clinical settings, particularly epilepsy evaluation and presurgical functional mapping. Its results become stronger when acquisition quality, source modeling, behavior, anatomy, alternative explanations, and clinical context are all examined together. Gross et al., 2013; ACMEGS guidelines.
Related Articles
References
American Clinical Magnetoencephalography Society. (2011; current guideline listing). Clinical Practice Guidelines 1–4 for MEG recording, presurgical functional mapping, reporting, and personnel. Official clinical guidelines.
Brickwedde, M., et al. (2024). Applications of OPM-MEG for translational neuroscience: A perspective. Translational Psychiatry, 14, 341. Journal article.
Brookes, M. J., Leggett, J., Rea, M., Hill, R. M., Holmes, N., Boto, E., & Bowtell, R. (2022). Magnetoencephalography with optically pumped magnetometers (OPM-MEG): The next generation of functional neuroimaging. Trends in Neurosciences, 45(8), 621–634. https://doi.org/10.1016/j.tins.2022.05.008
Geller, A. S., Teale, P., Kronberg, E., & Ebersole, J. S. (2024). Magnetoencephalography for epilepsy presurgical evaluation. Current Neurology and Neuroscience Reports, 24, 35–46. . https://doi.org/10.1007/s11910-023-01328-5
Gramfort, A., et al. (2014). MNE software for processing MEG and EEG data. NeuroImage, 86, 446–460. . https://doi.org/10.1016/j.neuroimage.2013.10.027
Gross, J., et al. (2013). Good practice for conducting and reporting MEG research. NeuroImage, 65, 349–363. . https://doi.org/10.1016/j.neuroimage.2012.10.001
Hari, R., & Salmelin, R. (2012). Magnetoencephalography: From SQUIDs to neuroscience. NeuroImage, 61(2), 386–396. . https://doi.org/10.1016/j.neuroimage.2011.11.074
Kim, H.-S., Lee, J., Jeong, C.-H., & Koo, Y. S. (2025). Clinical utility of magnetoencephalography in epilepsy evaluation: A qualitative systematic review. Journal of Epilepsy Research, 15(2), 83–92. . https://doi.org/10.14581/jer.25010
Kim, J. A., & Davis, K. D. (2021). Magnetoencephalography: Physics, techniques, and applications in the basic and clinical neurosciences. Journal of Neurophysiology, 125(3). . https://doi.org/10.1152/jn.00530.2020
National Institute of Mental Health. Magnetoencephalography (MEG) Core Facility: Procedures, superconducting sensors, and source localization. Official NIMH resource.
Radiological Society of North America and American College of Radiology. MEG (Magnetoencephalography): Clinical uses, preparation, and procedure. RadiologyInfo patient resource.
Wang, R., et al. (2025). Noise and artifact suppression in SQUID and wearable OPM-MEG: A systematic review of background, physiological, and technical interference. NeuroImage, 318, 121403. . https://doi.org/10.1016/j.neuroimage.2025.121403
Yang, Y., Luo, S., Wang, W., Gao, X., Yao, X., & Wu, T. (2024). From bench to bedside: Overview of magnetoencephalography in basic principle, signal processing, source localization and clinical applications. NeuroImage: Clinical, 42, 103608. https://doi.org/10.1016/j.nicl.2024.103608