Magnetoencephalography for the pediatric population, indications, acquisition and interpretation for the clinician

Abstract

Magnetoencephalography (MEG) is an imaging technique that enables the assessment of cortical activity via direct measures of neurophysiology. It is a non-invasive and passive technique that is completely painless. MEG has gained increasing prominence in the field of pediatric neuroimaging. This dedicated review article for the pediatric population summarizes the fundamental technical and clinical aspects of MEG for the clinician. We discuss methods tailored for children to improve data quality, including child-friendly MEG facility environments and strategies to mitigate motion artifacts. We provide an in-depth overview on accurate localization of neural sources and different analysis methods, as well as data interpretation. The contemporary platforms and approaches of two quaternary pediatric referral centers are illustrated, shedding light on practical implementations in clinical settings. Finally, we describe the expanding clinical applications of MEG, including its pivotal role in presurgical evaluation of epilepsy patients, presurgical mapping of eloquent cortices (somatosensory and motor cortices, visual and auditory cortices, lateralization of language), its emerging relevance in autism spectrum disorder research and potential future clinical applications, and its utility in assessing mild traumatic brain injury. In conclusion, this review serves as a comprehensive resource of clinicians as well as researchers, offering insights into the evolving landscape of pediatric MEG. It discusses the importance of technical advancements, data acquisition strategies, and expanding clinical applications in harnessing the full potential of MEG to study neurological conditions in the pediatric population.

Introduction

Clinical overview

Magnetoencephalography (MEG) is a non-invasive, passive, and painless imaging technique that enables cortical activity assessment via direct measures of neurophysiology. Since its introduction, MEG has undergone significant hardware and accompanying analytical software advances that improved the spatial and temporal resolution of findings generated by source reconstruction methods. ¹ The contemporary MEG offers rapid multichannel acquisition of electrophysiological data from the whole cortical surface with full-head coverage.

Due to technical advances, the role of MEG in diagnosis and disease management is constantly growing, with U.S. Food and Drug Administration (FDA) approval for epilepsy^1–11 and presurgical functional brain mapping.^12–24 Neurologists should be aware of MEG’s expanding application for conditions including autism spectrum disorder (ASD),^25–30 mild traumatic brain injury (mTBI),^31–33 post-traumatic stress disorder (PTSD),^34,35 extreme prematurity, ³⁶ cerebral palsy,^37–39 schizophrenia, ⁴⁰ Williams and Landau-Kleffner syndromes, ⁴¹ Alzheimer’s disease,^42,43 depression, ⁴⁴ attention deficit hyperactivity disorder,^45,46 and mapping of eloquent cortex in brain tumor patients. ⁴⁷

The objective of this review is to orient the clinician to MEG’s indications, fundamental technical aspects, data acquisition and analysis, and interpretation.^34,35 Pediatric clinical domains highlighted include identification of the irritative zone in epilepsy, presurgical delineation of eloquent cortices, as well as translational applications in autism, and mild traumatic brain injury (mTBI).

Technical overview

The crux of most clinical MEG applications is the ability to perform magnetic source imaging (MSI): cortical or subcortical source estimation of a signal of interest. Directional current flowing through a depolarizing neuron generates a magnetic field, detectable with MEG. ⁴⁸ In contrast to all other functional imaging techniques, the data received are direct measures of post-synaptic potentials and are not reference based. With the current sub-millisecond temporal and millimeter spatial resolution, MEG can localize focal dipolar responses throughout the entire head, as opposed to subdural electroencephalography (EEG)’s regional detection. ⁴⁹

To perform a MEG scan, a patient’s head is placed in a helmet that contains an array of sensors. The cortical and subcortical activities of the brain generate magnetic fields. Magnetic fields are detected with superconducting coils called superconducting quantum interference devices (SQUID).^11,48 Methods such as signal space projection (SSP) ⁵⁰ and synthetic gradiometer ⁵¹ are effective noise cancellation methods used during data preprocessing for removing non-cerebral magnetic noise. Importantly, MEG’s source localization can also be co-registered with structural magnetic resonance imaging (MRI) to identify specific brain areas involved in function or pathology.

Event-related field (ERF) recording is one of the more well-established methods of MEG analysis. Similar to EEG event-related potential (ERP) analysis, ERFs are measured in response to repetitive stimuli (auditory, visual, tactile, electrical, etc.). Data are then measured from onset of stimulus to neuromagnetic peaks that correlate to specific sensory or cognitive processes. The neural response(s) specific to this stimulus is represented as a waveform and termed the “evoked response.” The variability in latencies and amplitude between patients or participants can be interrogated for their physiologic relevance. ²⁹

Neurophysiological activity can also be evaluated through cortical rhythms. To this end, graph theory and advanced computational methods can be applied to investigate the interactions of cortical and subcortical brain areas. While structural connectivity, white matter anatomical connections between different brain areas, can be measured with imaging techniques such as diffusion MRI, functional and effective connectivity require high temporal resolution, which MEG can provide. Functional connectivity is defined as statistical dependencies among remote neurophysiological events; effective connectivity refers to the effect that one neural system exerts on another, either at a synaptic or population level. ⁵²

Functional connectivity examines brain region interconnections using time series of cerebral activity. ²⁵ Power of induced or evoked oscillations, along with their frequencies, reflects processes like working memory, semantic processing, and sensory-motor integration.^36,53 High-frequency oscillations (HFOs) (beta: 15–30 Hz, gamma: >30 Hz) are associated with event-related desynchronization (ERD) and are impacted in pathological conditions.^34,54

Source estimation

Source estimation is the process of mapping the magnetic fields recorded by MEG, into a three-dimensional model of the brain. The model can be an anatomical MR image or a geometric approximation. Identifying underlying electrical currents from the induced magnetic fields does not have a unique solution since there is an infinite number of possible current dipole configurations. Transforming the sensor data into activity in specific structures of the brain requires physiologically reasonable assumptions to be made. There are several methods of source estimation, each with its own advantages and limitations.

The equivalent current dipole (ECD) method of source localization is a recursive algorithm that iterates the magnitude, strength, position, and orientation of an assumed or pre-specified number of current dipoles, until the modeled magnetic fields are a best fit to the measured magnetic fields (i.e., low fit error or high goodness-of-fit). This method is best suited for a focal source or a small number of focal sources; however, errors can result when solving for an unknown number of possibly distributed sources. ⁴⁸ Localization of interictal spikes for clinical epilepsy cases are mostly performed with the ECD method.

Beamforming is another method of source localization which attempts to isolate a specific focal region by using a spatial filter generated from a set of reconstructed recordings from a known source within a subject’s brain (Figure 1). This set of known values is called the leadfield. There are a number of beamforming techniques in open and closed source libraries, each having its own traits affecting accuracy. This method has been used in several recent MEG studies to improve the signal-to-noise ratio of particularly low amplitude signals of interest in the epileptic brain, such as HFOs.^55–57

Figure 1.

Electrophysiological basis of MEG signals and example of dipole and beamformer source localization results. (a) Shows an action potential generated from a neuron along with a corresponding electric and magnetic field that are produced by the movement of negative charge along the outer membrane of the axon. A magnified simplified view of the measurement of electric (post-synaptic potentials) and magnetic (intracellular current) fields generated. The two EEG electrodes measure the difference in voltage from two groups of neurons along the neuronal axon, while the MEG coil measures unreferenced magnetic flux produced tangentially from a single group of neurons. This can lead to differences in field strength between MEG and EEG in both gyri and sulci, for this type of activity. Figure created using BioRender software. (b) Example of CTF’s dipole source algorithm interface. The observed waveform with overlaid sensors can be seen at the bottom, with the red vertical line indicating the selected time point. The top three views show the computed dipole location. Information regarding the exact time point, goodness-of-fit/error, and dipole coordinates are also provided. Topographic maps looking down on the head with the plot for the acquired data at the top, the reconstructed data from the forward solution in the middle, and the difference at the bottom. Example of the source localization applied using a beamforming algorithm.

Acquisition of MEG data in pediatric population

Methodology in pediatric patients

A MEG study is a non-invasive and painless procedure performed in a room shielded from external electromagnetic fields, operated by a trained technologist. A MEG session can be divided into parts with a couple of minutes break in between and depending on the study, in total can last up to an hour or more. Simultaneous EEG is occasionally used. Depending on the clinical question, the data are collected with or without stimuli, with some tasks requiring a response from the patient. The resultant data are interpreted by clinicians and scientists with MEG expertise.

In our institutions, MEG studies have been incorporated into clinical and/or experimental protocols of patients who presented with neurological or neuropsychiatric symptoms for whom supplementation of exam, EEG, or anatomical imaging was felt to be helpful. For the studies performed at the Hospital for Sick Children (Toronto, Canada), patients and subjects were scanned supine or upright in a whole‐head 151‐channel MEG (CTF Omega, MISL, Coquitlam, Canada) and data were recorded continuously with a 600‐Hz sampling rate, DC‐100 bandpass, and the CTF Omega-specific third‐order spatial gradient noise cancellation. For those performed at the Boston Children’s Hospital (Boston, MA, USA), patients were scanned supine in the whole-head 375-channel pediatric MEG System (MagView Biomagnetometer, also known as “BabyMEG,” Tristan Technologies, San Diego, CA, USA) ⁵⁸ with 1024 Hz sampling rate, at Boston Children’s Hospital, Longwood, or scanned supine in the 76-axial gradiometer pediatric MEG system ⁵⁹ (“BabySQUID,” Tristan Technologies, San Diego, CA, USA) at Boston Children’s Hospital, Waltham, where MEG data were recorded continuously with a 1024 Hz sampling rate (Figure 2).

Figure 2.

Contemporary magnetoencephalography platforms. (a) Photograph of the CTF MEG system at the Hospital for Sick Children (Toronto, ON, Canada). (b) Photograph of the pediatric BabyMEG system at the Boston Children’s Hospital (Boston, MA, USA). (c) Photograph of the BabySQUID system at the Boston Children’s Hospital (Waltham, MA, USA).

The BabyMEG system, which accommodates up to 52 cm in head circumference, has better spatial resolution compared to most commercial adult systems due to its dense sensor array, small coils and short distance between each coil. ⁵⁸ It also provides a better signal-to-noise ratio as there is significantly less distance between the sensors and the scalp, minimizing magnetic field weakening with distance, particularly for pediatric patients. On the other hand, adult MEG systems can accommodate head sizes that pediatric MEG systems cannot. Additionally, for children whose head circumferences approach 52 cm, collecting simultaneous MEG and EEG data may not be possible in the smaller BabyMEG system.

Data from all MEG systems are processed offline, including functional connectivity maps. Head movement is monitored to ensure <5 mm movement during runs. Monitoring the patient’s brain position relative to MEG sensors is crucial. Head Position Indicator (HPI) coils are placed at four points on the subject’s head, digitally marked for reference. Anatomical fiducial points (nasion, left and right preauricular points) are also marked digitally to locate the head inside the helmet and guide MRI co-registration (digitization process). ² Adding more points improves co-registration accuracy but increases the time required.

Practicalities of magnetoencephalography signal capture in children

The quality of MEG data is highly sensitive to head movement. As the sensors are fixed, head movements change the relation of the sensors with neurological activity and can present a challenge for accurate localization. As some children may find staying still for the duration of the 15–45-min scan time uncomfortable or impossible, practical challenges can be solved with multiple techniques including behavioral interventions such as playing videos to keep them entertained, as well as recording during natural sleep or with anesthesia. For some young children, skipping a nap before the MEG recording is helpful for them to fall asleep easily during data acquisition. Some studies prefer to use sedation or anesthesia during MEG recordings, however, these agents can have an effect on the brain activity recorded with MEG.^60–62 Also, techniques for continuous monitoring of patient head movement and compensation (e.g., signal space separation) have been developed to reduce the error generated by head movement during a MEG scan. ⁶³ Having a child-friendly environment in the MEG facility and the shielded room, such as wall decals, age appropriate toys, and coloring books, are also helpful for creating a less intimidating atmosphere for the pediatric populations.

Clinical use of magnetoencephalography in pediatric population

Use of magnetoencephalography in epilepsy

Role of magnetoencephalography in epilepsy presurgical evaluation

Epilepsy is a disorder defined by recurrent seizures due to ictal discharges. These represent abnormal electrical discharges in the brain, and the time between seizures is thus termed the interictal period. The first-line treatment of epilepsy is anti-seizure medication, but only 60%–70% of patients achieve remission.^34,64 In appropriately selected candidates, the treatment of drug-resistant epilepsy (DRE) involves surgical resection of the epileptogenic zone (EZ), defined as the area of the brain responsible for generating clinical seizures.

The preoperative workup of epilepsy surgery involves the estimation of the EZ through a wide variety of modalities, including EEG, MRI, invasive electrophysiological monitoring, subdural electrodes, intracarotid amobarbital procedure (IAP or Wada), PET, SPECT, and MEG. ¹ MEG has a role in aiding the identification of the EZ and in the prediction of clinical outcomes following surgery, as further detailed below. ⁶⁵

Most commonly, estimating the EZ with MEG means delineating the irritative zone, the area generating interictal epileptiform discharges (IEDs or spikes). This zone is one of the estimators of the epileptogenic zone. Delineating the irritative zone with MEG involves dipole source modeling of the magnetic field deflections caused by IEDs. The field measured by the MEG is compared with a simulated field modeled by a computer to determine the location of a point source dipole capable of generating the discharge. This information is then overlaid on an MRI of the patient’s brain, to estimate the structural location of IEDs’ generation—the irritative zone (Figure 3).

Figure 3.

Identifying the irritative zone (Example 1). A pediatric patient (younger than 10 years of age) with intractable epilepsy. Spontaneous data were collected with the whole-head pediatric MEG system at Boston Children’s Hospital while the patient was awake. Data were bandpass filtered between 4 and 70 Hz with a notch filter at 60 Hz, bad channels and segments of the data with motion artifacts were removed from the analysis. Data was analyzed with the Brainstorm software. T1-weighted MRI data of the patient were processed through the Freesurfer software and registered to the patient’s brain for source modeling to be performed using the Boundary Element Method. Equivalent current dipoles were thresholded at 70%. No ictal activity was observed in the data. Most of the interictal spikes had a tight cluster of dipoles localized in the right parietal lobe where the center of the cluster surrounds the intraparietal sulcus extending both to the superior and inferior parietal lobule. The anterior aspect of the intraparietal sulcus had a suspected focal cortical dysplasia. There were also scattered dipoles localized to the right frontal lobe, primarily in the parasagittal region.

While due to the time limit of MEG sessions (not to exceed 1–2 h as per IFCN-endorsed practical guidelines) and head movement control challenges, most studies investigating MEG in epilepsy presurgical planning are focused on characterizing interictal discharges, ⁶⁶ in certain settings MEG can be also used to investigate ictal discharges.^6–8 Studies with adult epilepsy patients have shown that MEG has shown success in discovering interictal discharges and employment of successful dipole source modeling providing additional useful information for presurgical evaluation.^3–5 Furthermore, in another study of 121 children with temporal lobe and temporal lobe plus epilepsy, MSI localization data showed significant predictive value. ⁶⁷ Further demonstrating the importance of multi-modal evaluation, three recent studies on pediatric patients demonstrated and quantified the presurgical accuracy of both MSI and electric source imaging (ESI) in localizing IEDs, highlighting the relative superiority of MSI, the complementary role of the two techniques, and the benefit of applying complementary clustering analysis.^55,66,68

Studies that successfully recorded ictal MEG discharges have small sample sizes. Among these, there is a general agreement that ictal MEG is superior to interictal MEG in terms of sensitivity and localization.^6–8 Finally, an increasing number of recent studies have been reporting the potential of MEG in detecting and localizing epileptic HFOs, which are relatively novel interictal epilepsy biomarkers. ⁴⁹ The recent studies of Tamilia and colleagues^55,69 demonstrated that MEG can detect and localize interictal ripples in children with DRE, consistently with the ripple-zone defined by intracranial EEG, and is more likely to predict seizure freedom in children when ripple-zones are resected. Ripples overlapping on spikes and ripples at onset of a propagating ripple activity are more often linked to the EZ. ⁶⁹ Migliorelli and colleagues also demonstrated in their cohort of nine patients (8 children) that interictal ripples on MEG can help localize the EZ and correlated with HFOs on intracranial EEG. ⁷⁰

Non-surgical uses of magnetoencephalography in epilepsy

While the predominant focus of MEG has been in presurgical evaluation, research has also shown possible burgeoning roles in non-surgical epilepsy patients in the understanding of underlying mechanisms and prognostication. For example, connectivity studies in childhood absence epilepsy show that strength of HFOs in MEG reflects the severity of absence seizures ⁷¹ and ictal HFOs can be used to localize seizure onset zones such as those at the thalami, cingulate gyri, and anterior cerebellar regions.^72,73

For childhood epilepsy with centrotemporal spikes (CECTS), a recent study by Kai Niu and colleagues correlated improved cognitive outcomes of patients who showed normalization of connectivity networks after 1 year of anti-seizure medication. ⁷⁴ Another study of 28 pediatric patients with CECTS compared to 14 control subjects suggested that abnormal network topology might explain underlying cognitive impairments. ⁷⁵

In summary, both ictal and interictal MEG measurements, as well as connectivity measurements, have been shown to be helpful in presurgical evaluation of epilepsy patients, both in terms of localization of the epileptogenic zone and predicting post-surgical seizure freedom (Figure 4). Though ictal MEG discharges tend to be more accurate, they are more difficult to obtain due to limited MEG session length and difficulty in minimizing head movements. Moreover, MEG should be used in conjunction with other imaging modalities such as EEG, MRI, PET, and/or SPECT for a more comprehensive assessment.

Figure 4.

Identifying the irritative zone (Example 2). A pediatric patient (younger than 5 years of age) with tuberous sclerosis complex, developmental delay, infantile spasms, refractory complex partial seizures, and status epilepticus. Simultaneous spontaneous data during natural sleep were collected using the whole-head pediatric MEG system at Boston Children’s Hospital. Data were bandpass filtered between 4 and 60 Hz with a notch filter at 60 Hz, bad channels and data segments were removed from the analysis. The data was analyzed with the Brainstorm software. Patient’s T1-weighted MRI data was processed with the Freesurfer software and co-registered to the patient’s brain for source modeling to be performed by the Boundary Element Method. Equivalent current dipoles were thresholded at 85%. No ictal activity was observed. Interictal activity was observed over the right hemisphere where dipole localization involved the inferior right precentral gyrus, adjacent anterior inferior right parietal lobe, posterior right insula, and posterior aspect of the right superior temporal gyrus. There is a tuber in the precentral gyrus just superior to the dipole cluster in the right inferior precentral gyrus.

Use of magnetoencephalography for mapping of eloquent cortices

Prior to neurosurgical procedures, neurosurgical landmarks, such as somatosensory, motor, visual, and auditory cortices, must be identified to prevent inadvertent loss of brain function (Figure 5). Currently, the gold standard for identification of these eloquent cortices is invasive cortical stimulation and the IAP for language lateralization.^34,35 These invasive procedures have inherent risks of morbidity. ³⁵ The use of MEG for mapping eloquent cortices has been extensively studied in literature, including with pediatric epilepsy patients.^76,77 The American Clinical Magnetoencephalography Society (ACMEGS) has released two position statements and a clinical practice guideline to support the use of MEG for presurgical mapping.^35,78

Figure 5.

Mapping of eloquent cortices. Example of typical averaged waveforms with overlaid sensors (first column) for somatosensory, auditory, visual, and language paradigms. The thin vertical lines (for sensory, auditory, and visual) indicate the time point of interest for the dipole fit, and the yellow box (for language) indicates the time points for a moving dipole fit. The middle column shows topographic plots at the selected time points. The projection by convention is looking down on the head with the nose at the top and the ears at the sides. The last column shows the dipole location obtained from the source reconstruction algorithm, and then overlaid onto individual MRIs after co-registration with fiducials.

Somatosensory and motor cortex

The mapping of the somatosensory cortex is often required prior to tumor resection surgeries, because tumors have been known to shift the position and shape of the central sulcus. ¹² MEG has been validated against fMRI and invasive techniques for this purpose.^12,35 Electrical or tactile stimulations of the median nerve or the fingers, respectively, generate a somatosensory evoked field (SEF) from the primary somatosensory cortex.^12,35 The mapping of both primary and secondary somatosensory cortices was shown to be reliable in patients under anesthesia and in coma.^12,35 MEG has also been used to localize the somatosensory cortex in typically developing children and pediatric epilepsy patients.^79,80

MEG is superior to other imaging modalities in identifying the motor cortex due to its greater temporal resolution. ¹² All movements (active, passive, or imagined) can be utilized to map the motor cortex and its associated areas. ³⁵ Two methods have been suggested to map the motor cortex, either via repeated motor movements to produce an average evoked field or by using EMG (electromyography) to detect movement onset for correlation with MEG signals.^81,82 A study by Gaetz and colleagues successfully localized the hand region primary motor cortex, as a result of right and left index finger movements, in pediatric patients undergoing presurgical evaluation. ⁸³

Visual and auditory cortices

The localization of the visual cortex is difficult due to the crossing of the visual pathway at the optic chiasm resulting in simultaneous activation of multiple occipital areas within the brain, which may result in signal cancellations. ¹² However, it has been established that visual evoked fields (VEFs) in each brain hemisphere can be stimulated with a hemifield reversing checkerboard pattern and localized with the ECD approach ^12,35; this mapping has been validated against fMRI. ¹⁵ VEF analysis is part of clinical practice during the presurgical evaluation of patients with epilepsy or brain tumors, especially if posterior brain regions are involved. Future clinical use of VEFs includes early identification of disorders such as multiple sclerosis, Parkinson’s disease, and stroke. ⁸⁴

The localization of the auditory cortex and cortices responsible for language function are also being explored with MEG. Cheour and colleagues were able to use MEG successfully to study auditory discrimination in infants. ⁸⁵ Moreover, peripheral and central lesions of auditory pathways can affect auditory evoked fields, localized to Heschl’s gyri and the supratemporal auditory cortices. ⁸⁶ Language pathways can be re-routed in patients suffering from long-term epileptic activity, tumors, or stroke ⁸⁶ ; therefore, mapping prior to surgery is important to prevent deficits. Some studies have attempted to localize the frontal cortical area responsible for expressive language, ⁸⁷ while others focused on localizing the receptive area within the temporal lobe or tempo-parietal junction.^23,24 However, the protocols are not standardized, and more experimental research remains to be performed on this topic. Thus, auditory cortex mapping is currently not frequently used in a clinical setting alone, but more often applied in conjunction with the lateralization of language. ³⁵

Lateralization of language

Clinical MEG studies of language mapping have focused on the lateralization of language function: determining the dominant language hemisphere and the intrahemispheric localization of language function prior to surgery. Language mapping is particularly important for epilepsy patients because they show more variability in hemispheric dominance compared to neurologically typical individuals.^88,89 The standard procedure for determining language dominance has been the Wada test for years ⁹⁰ ; however, this test is invasive, and it does work for intrahemispheric localization of language. ⁹¹ For now, fMRI might be the non-invasive test of choice in the lateralization of language, due to its greater standardization of technology and protocol.^34,92 However, MEG is a promising technique yielding relatively high concordance rates with both IAP and fMRI. For example, a study of 85 surgical candidates (ages: 8–56 years) for drug-resistant epilepsy (DRE) showed 87% concordance between MEG and IAP. ¹⁶

MEG is used to assess the localization of receptive and expressive language function in epilepsy patients through a variety of tasks (for a review see ⁹¹ ). One of the most commonly used receptive language mapping tasks is passive listening tasks where patients listen to words, tones, or vowels. Such tasks are the easiest for pediatric populations since they do not require any active participation from the patient. Other tasks may require patients to actively respond via button presses such as semantic decision tasks where the patient decides if a given word is abstract (“freedom”) or concrete (e.g., “table”) ⁹³ or word recognition tasks that require patients to indicate when a previously presented target word is among a list of words presented ¹⁶ but these active tasks are not very commonly used for pediatric populations. According to a review study, word recognition task helps with the intrahemispheric localization of language and has a high concordance rate between MEG and IAP and has been used in both children and adult populations. ⁹¹

For the mapping of expressive language, tasks such as picture naming, word generation, phonemic fluency, and verb generation have been used. ⁹¹ One of the most commonly used tasks is the verb generation task where participants are asked to generate a verb associated with a given written noun or picture of an object (e.g., “BALL”—“throw”/“catch”). ⁹⁴ This method had high concordance rates with other methods, such as cortical stimulation mapping and fMRI, for localization of language in 89% of the pediatric epilepsy patients. ⁹⁴ Another expressive language task, especially easier for younger populations, is the verbal fluency task where patients are asked to produce words that start with a letter they are given. ⁹¹

New evidence suggests that atypical language lateralization may be common in pediatric patients with a variety of neurological disorders.^20,21,30,95 Thus, such patients require a multi-modal approach to ensure the most accurate lateralization of language prior to surgery. This can be especially true when neurovascular decoupling compromises the utility of fMRI alone, such as arteriovenous malformations or vasogenic edema.

In summary, the mapping of eloquent cortices is an important step before neurosurgical procedures. A growing body of literature has supported the use of MEG in the identification of somatosensory, motor, visual, and auditory cortices as well as in the lateralization of language. ACMEGS recommends the routine clinical use of non-invasive MEG/MSI in the presurgical evaluation of patients with brain lesions and epilepsy. ³⁵

Use of magnetoencephalography in autism spectrum disorder

ASD is a neurodevelopmental condition characterized by challenges in social-emotional reciprocity and the presence of restricted or repetitive behaviors or interests. ⁹⁶ Though it is a clinical diagnosis, attempts have been made to study the atypical neuronal processing using fMRI,^97,98 EEG,^99,100 and MEG.^25–29,101 MEG ASD studies have been done in both adults and children.^26,28 As MEG is quieter than MRI and provokes less claustrophobia, it confers an advantage in children with ASD.

A recent systematic review summarized studies prior to 2015 that utilized MEG to investigate brain function in ASD. ²⁵ Out of 38 studies that met inclusion criteria, 24 focused on responses to visual, auditory, or somatosensory stimuli. The rest investigated cognitive tasks or spontaneous activity. Studies using auditory stimuli (Figure 6) generally showed a significant difference between the evoked potentials and spontaneous activity between children with ASD and neurotypical controls.^{86,95,102,103} However, the review noted great variability in the specific differences discovered among the studies, which the authors attributed to differences in stimulation used, heterogeneity of the participants, and heterogeneity of methods. ²⁶ MEG research to date has been conducted at the group level, future studies are needed to examine whether MEG can predict ASD presence, type, and severity at the individual level, especially helping identify individuals with auditory and language impairments who might need intervention.^104–107

Figure 6.

Auditory evoked fields. Auditory evoked fields generated by the auditory presentation of beep sounds (∼220 epochs) from the left hemispheres of a typically developing (TD) child (aged between 7 and 12 years old) and a child with an autism spectrum disorder (ASD) (aged between 7 and 12 years old). Data were collected with the BabySQUID system, at Boston Children’s Hospital, Waltham. BabySQUID (Tristan Technologies, Inc., San Diego, CA, USA) is a pediatric MEG system (143) (BCH, Waltham), with a partial head coverage sensor array (coverage area ∼265 cm²) consisting of 76-axial gradiometers (10 mm pickup coil diameter, 30 mm baseline, and 12–14 mm coil center-to-coil center spacing, gap of 7–10 mm between each coil and scalp). During the recording, children lay on the BabySQUID bed, resting their head on the helmet. Minimum norm estimate (in pico ampere meter) solutions overlaid on the participant’s own cortical surface for the TD participant and on an age matched template brain for the child with ASD at the peak of M100 which was delayed in the child with ASD compared to the TD child as shown in previous studies [103].

Studies suggest atypical development of the primary auditory cortex in ASD and have implications for clinical markers for the diagnosis of ASD. Latencies in M50, M100, and mismatch field (MMF), and auditory event-related fields^104,107 have shown significant findings in ASD, including both language impairment and overall level of support needed.^106,108 One study also showed that frontal and temporal residual M100 latencies were found to be longer in children with ASD with auditory sensitivity. ¹⁰⁹ M100 latencies, in another study, were additionally associated with the level of support needed. ¹⁰⁶ Berman and colleagues showed that the latency of M100 response to be related to ASD level of support needed and the MMF delay to language impairment. ⁹⁵ Matsuzaki and colleagues showed that latency delays in ASD are associated with poorer language ability persisting into adulthood. ¹⁰² Roberts et al. (2019) showed M50 and M100 auditory responses of children with ASD with significant language disability had significantly delayed latencies compared to controls. ¹⁰³

MEG has also been used to study functional connectivity in children with ASD supporting the fMRI and immunocytochemistry derived hypothesis of long-range under-connectivity and local over-connectivity. A MEG study conducted during awake state when children were watching a video, found reduced long-range functional connectivity in the theta band in children with ASD compared to TD children between the left-anterior and right-posterior brain regions, where the reduction was correlated with clinical severity in the right-handed children with ASD. ²⁶ Another MEG study also found decreased connectivity in the theta frequency in children with ASD compared to TD controls in a memory task. ¹¹⁰ A MEG study of face response found reduced functional connectivity between the fusiform face area and the inferior frontal gyrus in children with ASD that correlated with more severe symptomatology. ¹¹¹

There is a high comorbidity between epileptic activity and ASD; however, it is not known if epileptiform activity, in the absence of seizures, contributes to ASD symptoms. ¹¹² A study reported that MEG successfully identified abnormal discharges in 82% of children with ASD in contrast to EEG which identified in only 68% ²⁷ (for a review on EEG studies of epilepsy and ASD both in children and adults please see ¹¹² ). MEG studies have shown subclinical epileptiform activity especially in the perisylvian regions in children with ASD.^27,113

Face processing in children with ASD has also been studied using MEG. A study found altered functional connectivity, increased alpha band phase synchronization, and happy face processing in ASD compared to controls. ¹¹⁴ Abnormal reduction in the local and long-range functional connectivity in children with ASD for inverted faces has been shown in a MEG study. ¹¹⁵ Another study showed that adolescents with ASD had atypical neural activity in insula, anterior and posterior cingulate, and temporal and orbitofrontal regions for angry and happy face processing. ¹¹⁶

In summary, MEG has been used to study a variety of neuronal processes in children with ASD. It has been used to understand the neurobiology of ASD and how it functions as compared to neurotypical individuals. MEG can add information in the characterization of neural processing in ASD in the hopes of future clinical applications in diagnosis, stratification, and treatment response.

Use of magnetoencephalography in mild traumatic brain injury

Traumatic brain injury (TBI) is a major public health problem among children and adolescents. ¹¹⁷ 70%–90% of pediatric TBIs are classified as mild traumatic brain injury (mTBI), commonly referred to as a concussion.^117–119 mTBI is responsible for a substantial number of deaths, hospitalizations, and emergency department visits during childhood. ¹²⁰

mTBI occurs when the head experiences acceleration and deceleration forces from a direct impact or rotational forces that affect the brain. ¹²¹ Injury symptoms spontaneously resolve in around 70%–80% of cases within 3 months.^122,123 However, at least 20% of injured individuals continue to experience chronic physical, cognitive, and emotional impairments, known as persistent post-concussive symptoms.^122–124 These chronic effects of mTBI in childhood and adolescence have been associated with reduced quality of life, difficulties in school, and poorer long-term psychological and behavioral outcomes.^125,126

Traditional neuroimaging methods such as computed tomography (CT) and MRI are commonly used to clinically assess mTBI. CT scan excludes hemorrhage and skull fracture, but isn’t reliable for detecting mTBI and visualizing certain brain structures. Structural MRI improves visualization but may lack sensitivity for detecting white matter microstructure changes. ¹²⁷ MEG can be more sensitive to brain injury as it can detect cellular activity not detected by macroscopic structural neuroimaging. Previous adult mTBI studies have shown that resting state MEG can detect brain abnormalities that go along with persistent behavioral changes when traditional imaging methods fail to do so.^128,129 Those abnormalities include a pathological increase in slow-wave oscillations in the delta and theta frequency bands in frontal and temporal regions in symptomatic adults after mTBI,^130–133 which are evidence of neurological damage. ¹³⁴

More recent studies also assessed the pediatric population with mTBI using MEG, revealing abnormalities in the electromagnetic signals from the children and adolescents with mTBI. Those abnormalities included pathological slow-wave oscillations,^124,135,136 similar to adult mTBI studies, but also showed dysregulated alpha activity^124,135 in pediatric patients with chronic mTBI. Huang et al. discovered increased low frequency activity in bilateral temporal lobes and hypoactivity in prefrontal cortices and precuneus areas in alpha bands using rs-MEG. They also observed hyperactivity in the inferior temporal lobe, parahippocampus, and insula. Additionally, the study found hyper- and hypoactivity in frontal, temporal, occipital, and cerebellar regions in beta and gamma bands, ¹³⁵ suggesting abnormal communication patterns in both remote and local networks. ¹³⁷ While the slow-wave results are comparable to what has been found in adult mTBI studies, the outcomes of other frequency bands may be distinct to pediatric mTBI.

In the alpha band, reduced alpha peak power also suggests damage to white matter tracts, consistent with axonal injury.^138,139 Thus, the hypoactivity findings may indicate a deficiency in thalamocortical interactions, which could cause a decline in functional suppression within the prefrontal cortices and precuneus cortical regions of children with mTBI. ¹⁴⁰ On the other hand, the hyperactivity in the alpha bands observed in the inferior temporal lobe, parahippocampus, and insula may be associated with increased functional inhibition in these regions. This inhibition may lead to deficits in memory function and sensory information integration.

To better assess the post-concussive symptoms and their recovery, recently Huang et al. found that MEG data combined with machine learning algorithms can accurately distinguish between children with mTBI and the control group, with a sensitivity of 83.7% and a specificity of 85.7%. The researchers also found that the MEG data can be used to track the recovery of children with mTBI over time, as the differences between the two groups became smaller as the children with mTBI recovered. ¹⁴¹

In summary, mTBI affects neural communication in the brain across different frequencies, with long-lasting consequences for children and adolescents. These effects are not limited to specific regions and persist for approximately 1 year after the injury. MEG can be a valuable tool for objectively assessing and monitoring pediatric mTBI, leading to better diagnosis and treatment for children.

Conclusions

MEG is a non-invasive technique that enables study of the brain’s neural activity in real time. Within our institutions, the use of MEG has contributed to identifying epileptogenic foci and eloquent cortices, as well as to further our understanding of the neuronal processes underlying ASD and mTBI. These processes represent only a few of the many uses of MEG to delineate neurological pathologies. As technology advances, we will likely see more routine clinical uses of MEG, either alone or in combination with other brain imaging techniques.

Author contributorship

A.A.D., MD MPH MSc: conception and design of the study, literature review, and drafting a significant portion of the manuscript and figures. A.H., MD: literature review and drafting a significant portion of the manuscript and figures. G.N., MD: literature review, drafting a significant portion of the manuscript and figures, and acquisition and analysis of data. E.T., PhD: literature review, drafting a significant portion of the manuscript and figures, and acquisition and analysis of data. M.P., BS: literature review, drafting a significant portion of the manuscript and figures, and acquisition and analysis of data. L.F. Be.: literature review and drafting a significant portion of the manuscript and figures. P.E.G., MD MSc: literature review and drafting a significant portion of the manuscript and figures. E.P., PhD: conception and design of the study and acquisition and analysis of data. B.A., DPhil MSc: literature review, drafting a significant portion of the manuscript and figures, and acquisition and analysis of data.

Ethical statement

Ethical approval

Data presented in this review paper have been collected at Hospital for Sick Children (Toronto, Canada) and Boston Children’s Hospital (Boston, MA, USA). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the SickKids Research Ethics Board (REB# 1000053377 and REB# 1000049940) and the Boston Children’s Hospital Institutional Review Board (IRB-P00000247 and IRB-P00023088).

Consent to participate

An informed written consent was obtained from all parents of children from whom data were collected.

ORCID iDs

Adam A. Dmytriw https://orcid.org/0000-0003-0131-5699

Banu Ahtam https://orcid.org/0000-0003-2036-092X