What Is MEG Actually Used For?

A Machine That Listens to Whispers From Inside Your Skull

There's a machine that can tell which part of your brain activated, and when, with the precision of a few millimeters and a few milliseconds. It can map the exact cortical tissue responsible for moving your right index finger. It can pinpoint, within centimeters, the spot in your brain where a seizure begins. It can watch language processing unfold across your temporal lobe in real time, resolving events that happen faster than you can blink.

That machine is called a magnetoencephalograph. And there's a good chance you've never been in the same building as one.

Despite being one of the most powerful brain measurement technologies ever invented, magnetoencephalography (MEG) is astonishingly rare. Fewer than 200 systems exist on the entire planet. For context, there are more particle accelerators in the world than MEG machines. There are more MRI scanners in the city of Tokyo alone than MEG systems on Earth.

This isn't because MEG doesn't work. It works phenomenally well. The problem is everything else: the cost, the infrastructure, the liquid helium, the magnetically shielded room that has to block out the Earth's own magnetic field. MEG is the Concorde of neuroscience. Brilliant engineering, extraordinary performance, and a price tag that keeps it locked away in a handful of elite institutions.

So what, exactly, do those institutions use it for? And what does MEG reveal about the brain that nothing else can?

A Quick Primer: What MEG Actually Measures

Before we get into applications, let's build the trunk of this knowledge tree.

Every time neurons in your brain fire, electrical current flows through them. And whenever electrical current flows, it generates a magnetic field. This is basic electromagnetism, the same physics that makes electric motors work. The magnetic field wraps around the current like an invisible sleeve.

A single neuron's magnetic field is laughably weak. We're talking femtoteslas, a unit so small that the Earth's magnetic field is roughly a billion times stronger. But when tens of thousands of cortical pyramidal neurons fire synchronously, their tiny magnetic fields add up into something just barely detectable.

"Just barely" is doing a lot of work in that sentence. To detect these fields, traditional MEG uses an array of superconducting sensors called SQUIDs (Superconducting Quantum Interference Devices). These sensors are so sensitive that they can detect magnetic fields a billion times weaker than what it takes to stick a magnet to your refrigerator. But to achieve that sensitivity, the SQUIDs must be cooled to -269 degrees Celsius, just four degrees above absolute zero, using liquid helium.

That's why MEG machines look the way they do. Picture a giant helmet mounted on a mechanical arm, with the patient sitting beneath it, head nestled inside a rigid sensor array. The whole thing sits inside a room whose walls are lined with layers of mu-metal and aluminum to shield out every stray magnetic field, from the building's electrical wiring to passing cars to the Earth itself.

It's an absurd amount of engineering. And it produces data that nothing else on Earth can match.

Clinical Application 1: Pre-Surgical Brain Mapping

This is MEG's most established clinical use, and it's the one that saves lives.

When a neurosurgeon needs to remove a brain tumor, they face a terrifying question: what's next to the tumor? Specifically, is the tissue they're about to cut responsible for something critical, like moving the patient's hand, understanding speech, or producing language?

Cut into eloquent cortex (the neuroscience term for brain tissue that does something irreplaceable) and the patient wakes up unable to speak, or with a paralyzed limb, or unable to understand what anyone is saying to them. The tumor is gone, but so is a piece of who they are.

This is where MEG comes in. Before surgery, the patient sits in the MEG system and performs a series of tasks. They move individual fingers. They listen to words. They name objects in pictures. They read sentences. During each task, MEG records the precise spatiotemporal pattern of brain activation, mapping which cortical tissue lights up, when, and in what sequence.

The neurosurgeon gets a functional map of the patient's brain overlaid on their structural MRI. Tumor here. Language cortex there. Motor cortex for the right hand right here, three centimeters from the tumor margin.

This isn't theoretical. MEG-based pre-surgical mapping has been an established clinical procedure since the early 2000s. The American Academy of Neurology recognized it as a valid clinical tool for presurgical functional mapping, and it's covered by Medicare and most major insurers in the United States for this purpose.

Why MEG and not fMRI? Both can do functional brain mapping. But MEG has two advantages in the surgical context. First, it provides millisecond timing information that helps distinguish primary cortical responses from later, less critical processing. When you move your finger, the primary motor cortex activates first, within about 20 milliseconds. Secondary motor areas activate later. MEG can separate those. fMRI blurs them together. Second, MEG is less affected by signal distortions near tumor tissue. Tumors often change the local blood flow patterns that fMRI depends on, creating misleading maps. MEG measures magnetic fields from neural currents directly, sidestepping this problem.

Clinical Application 2: Epilepsy Localization

Here's where MEG arguably has its biggest clinical impact.

About one-third of people with epilepsy don't respond to medication. For these patients, surgery to remove the seizure focus, the specific patch of brain tissue where seizures originate, can be life-changing. Some patients go from having dozens of seizures a day to being seizure-free.

But you have to find the focus first. And the brain doesn't come with labels.

Epilepsy surgery is only as good as the localization. Remove the wrong tissue and the seizures continue. Remove too much and you risk neurological deficits. The entire clinical challenge comes down to one question: where, exactly, do this patient's seizures start?

MEG is exceptionally good at answering this question. Between seizures, epileptic brain tissue produces brief, abnormal magnetic spikes called interictal epileptiform discharges. These are the electrical tantrums of irritable neurons, little bursts of synchronized abnormal activity that happen hundreds or thousands of times per day, even when the patient isn't having a seizure.

MEG detects these spikes and, using mathematical source localization algorithms, traces them back to their point of origin in the brain. The result is a cluster of dipoles, a constellation of points on the patient's MRI showing where the abnormal activity is coming from.

Multiple studies have shown that MEG-guided epilepsy surgery leads to better seizure outcomes. A landmark 2013 study published in Brain found that patients whose surgical resection included the MEG-identified focus had significantly higher rates of seizure freedom than those whose resection missed it. MEG doesn't replace other diagnostic tools. It adds a layer of precision that can mean the difference between a successful surgery and one that falls short.

In many epilepsy centers, the clinical pathway now looks like this: long-term EEG monitoring identifies that the patient has focal epilepsy, structural MRI looks for visible lesions, and MEG narrows down the functional source of the abnormal activity. If all three converge on the same location, the surgical team can proceed with higher confidence. If they disagree, that disagreement itself is clinically valuable, because it tells the team they need more information before cutting.

Clinical Application 3: Language Lateralization

Here's a question that comes up before almost every brain surgery near language areas: which hemisphere of this patient's brain controls language?

For roughly 95% of right-handed people and about 70% of left-handed people, language is lateralized to the left hemisphere. But "roughly" and "about" are not words you want to hear when someone is about to operate on your brain.

The traditional method for determining language lateralization is the Wada test, developed in the 1960s. During a Wada test, a barbiturate is injected into one of the carotid arteries, temporarily anesthetizing one entire hemisphere. While half the brain is asleep, the neurologist tests whether the patient can still speak, understand language, and remember new information. Then they do the other side.

The Wada test works, but it's invasive, carries risks (including stroke), stresses the patient, and provides only a binary answer: left or right. It can't tell you which specific areas within a hemisphere are doing the heavy lifting.

MEG can. During a MEG language mapping session, the patient performs language tasks while the system records the spatiotemporal cascade of activation across both hemispheres. The result isn't just "left or right." It's a detailed map showing which specific cortical regions in which hemisphere activate during receptive language (understanding), expressive language (producing speech), and naming.

This has made MEG an increasingly accepted alternative to the Wada test, particularly for patients where the Wada test carries elevated risk or where more detailed mapping is needed. The American Clinical MEG Society has published guidelines supporting MEG as a non-invasive alternative for language lateralization, and a growing number of epilepsy surgery programs use it routinely.

Research Frontier 1: Auditory Processing

Now let's step out of the clinic and into the research lab, where MEG is revealing things about the brain that would be invisible to any other technology.

Auditory neuroscience may be the field where MEG has had its most profound impact. The reason is timing.

When a sound hits your ear, the auditory system processes it in a cascade of neural events that unfold over hundreds of milliseconds. The brainstem responds within 5 to 10 milliseconds. The primary auditory cortex in the temporal lobe fires at around 20 to 50 milliseconds. Higher-order processing, where your brain figures out what the sound means, where it came from, whether it's a word or a noise, unfolds from 50 to 500 milliseconds.

This entire cascade happens faster than a heartbeat. And MEG can see every stage of it.

The Neurosity Crown gives you real-time access to your own brainwave data across 8 EEG channels at 256Hz, with on-device processing and open SDKs.

See the Crown

The most famous auditory MEG signal is the M100 (also called the N100m), a strong magnetic response that peaks about 100 milliseconds after a sound onset. The M100 is generated in the auditory cortex on the superior temporal plane, and its properties have taught us enormous amounts about how the brain processes sound.

For example, MEG research on the M100 revealed that the brain has a tonotopic map in the auditory cortex, an organized spatial layout where different sound frequencies are processed by different patches of tissue, much like keys arranged along a piano. High frequencies activate one end. Low frequencies activate the other. MEG was able to resolve this spatial organization because of its millimeter-level source localization, something that scalp EEG alone couldn't achieve.

MEG has also been instrumental in studying the "mismatch negativity" response, the brain's automatic error signal when a sound deviates from an expected pattern. Play someone a series of identical tones and then throw in an oddball, and the brain produces a distinctive magnetic response about 150 to 250 milliseconds after the deviant sound. This response is generated without the person even paying attention to the sounds. It's the brain's built-in novelty detector, and it's impaired in conditions like schizophrenia and autism, making it a potential biomarker that MEG can measure with precision.

Research Frontier 2: Developmental Neuroscience

Here's the "I had no idea" part of this guide.

Traditional MEG has a major limitation for studying children, especially young children and infants. The sensor helmet is built for adult heads. It's rigid, fixed in size, and requires the patient to sit very still. If you've ever met a toddler, you know that "sit very still inside a giant machine" is not in their behavioral repertoire.

This means that for decades, one of the most powerful brain measurement tools was essentially useless for studying the population neuroscientists most wanted to understand: developing brains.

The developing brain is where the action is. Between birth and age five, the human brain forms roughly 700 new synaptic connections every second. Neural circuits for language, social cognition, emotional regulation, and sensory processing are being wired and rewired at a pace that will never be matched again in that person's lifetime. Understanding how these circuits form, when they form, and what happens when they form incorrectly is arguably the most important question in all of neuroscience.

And until recently, MEG couldn't contribute.

That changed with OPM-MEG (optically pumped magnetometer MEG), a new generation of sensors that we'll discuss in detail later. But even with traditional systems, creative researchers have found ways to study older children and adolescents. MEG studies of children aged 5 and up have revealed striking differences in how the developing brain processes language compared to adults.

For instance, MEG research has shown that children's brains process grammar in different cortical regions than adults, and that the adult-like pattern emerges gradually through adolescence. The spatiotemporal cascade of language processing, which in adults flows smoothly from auditory cortex to Wernicke's area to Broca's area, is more distributed and less organized in children, becoming increasingly streamlined with age.

These findings have direct relevance for understanding developmental language disorders, dyslexia, and autism. If you know the typical developmental trajectory of language networks, you can identify when and how a child's trajectory diverges.

Why Is MEG So Rare?

We need to talk about the elephant in the magnetically shielded room.

If MEG is this good, why are there fewer than 200 systems worldwide? Why doesn't every hospital have one? The answer comes down to three brutal constraints.

Constraint 1: Cost. A traditional MEG system runs $2 million to $3 million for the hardware alone. The magnetically shielded room (MSR) adds another $500,000 to $1 million. Installation, calibration, and commissioning add more. You're looking at $3 million to $5 million before you record a single brain. And that's just the purchase price.

Constraint 2: Liquid helium. The SQUID sensors at the heart of traditional MEG must operate at superconducting temperatures, roughly -269 degrees Celsius. Maintaining this temperature requires a constant supply of liquid helium, which boils off slowly over time and must be replenished regularly. Depending on the system and local helium prices, this costs $50,000 to $100,000 per year. And helium is a non-renewable resource. It's produced by radioactive decay deep in the Earth's crust, and once it escapes into the atmosphere, it's gone. Global helium shortages have caused prices to spike multiple times in the past decade, making MEG operating costs unpredictable.

Constraint 3: Infrastructure. You can't just plug a MEG into a wall outlet. You need the magnetically shielded room. You need a building with low magnetic interference (don't put it next to an elevator or a subway line). You need specialized technical staff who know how to operate and maintain the system, handle liquid helium, and run the source localization software. Most hospitals and universities simply can't justify this level of investment for a single neuroimaging modality.

Compare this to MRI, which also costs millions but serves dozens of clinical indications every day, from knee injuries to chest pain to stroke. An MRI scanner at a busy hospital might run 20 to 30 scans per day. A MEG system might run 3 to 5. The per-scan economics are brutal.

This is why MEG has remained a niche technology despite decades of proven clinical utility. It's not a question of capability. It's a question of infrastructure economics.

The OPM Revolution: MEG Without the Liquid Helium

And now for the part that might change everything.

In the past few years, a new type of magnetic sensor has emerged that could fundamentally reshape the MEG landscape. These sensors are called optically pumped magnetometers, or OPMs.

Traditional MEG uses SQUID sensors that require cryogenic cooling. OPMs use an entirely different physical principle. Inside each OPM sensor is a small glass cell filled with an alkali metal vapor (usually rubidium). A laser beam "pumps" the atoms into a specific quantum state. When an external magnetic field (like the one produced by your brain) hits the vapor, it perturbs the atoms' quantum state, and this perturbation is detected as a change in the laser light passing through the cell.

The critical difference: OPMs operate at room temperature. No liquid helium. No cryogenic dewar. No rigid helmet.

This has cascading consequences that go far beyond cost savings.

OPM-MEG is still maturing. Current commercial systems have fewer sensors than traditional SQUID arrays, and the magnetic shielding requirements remain significant. But the trajectory is clear: the technology is moving from physics labs into clinical validation studies, and several companies are developing commercial OPM-MEG systems for both research and clinical use.

For developmental neuroscience in particular, OPM-MEG is already producing data that was previously impossible to obtain. Studies at University College London and the University of Nottingham have recorded MEG-quality data from children as young as two years old wearing OPM sensor arrays, capturing the spatiotemporal dynamics of sensory processing in developing brains for the first time.

The Bigger Picture: MEG, EEG, and the Physics of Brain Measurement

Here's something worth stepping back to appreciate.

MEG and EEG are reading the same page of the same book. When pyramidal neurons in your cortex fire synchronously, they produce both electrical potentials and magnetic fields. EEG picks up the electrical side. MEG picks up the magnetic side. Same neurons, same moment, different physics.

This means that every insight MEG generates about where and when the brain activates is, at its core, information about the same neural signals that EEG detects. MEG achieves better spatial precision because magnetic fields pass through the skull without distortion, while electrical signals get smeared and spread. But the underlying source, populations of cortical pyramidal neurons firing in sync, is identical.

This is why EEG and MEG are considered complementary rather than competing technologies. And it's why advances in understanding the brain through MEG directly inform what we can learn from EEG.

When MEG research maps the tonotopic organization of auditory cortex or traces the millisecond-by-millisecond cascade of language processing, it's building a knowledge base that applies to EEG as well. The spatial precision of MEG helps researchers develop better source localization algorithms for EEG data. The functional maps generated by MEG help EEG researchers know what to look for in their own recordings.

Consumer EEG devices like the Neurosity Crown measure the electrical counterpart of what MEG measures magnetically. The Crown's 8 channels at 256Hz capture real-time cortical activity from the same pyramidal neuron populations that MEG detects, processed on-device through the N3 chipset. You can't replicate MEG's spatial precision with scalp EEG. But you can capture the temporal dynamics of brain activity, track brainwave patterns, and build applications that respond to your neural signals, all from a device that fits on your head and goes wherever you go. No shielded room. No liquid helium. No $3 million price tag.

That tradeoff, spatial precision for accessibility and portability, is the fundamental tension in brain measurement technology. And it's a tension that's gradually resolving as both sides improve.

What Comes Next

The story of MEG is, in many ways, a story about what happens when brilliant technology runs into the hard reality of infrastructure economics. MEG works. It works extraordinarily well. It saves lives in pre-surgical mapping. It pinpoints seizure foci with a precision that changes surgical outcomes. It reveals the millisecond dynamics of brain processing with a clarity that no other non-invasive tool can match.

But for forty years, that capability has been locked behind a wall of liquid helium, magnetic shielding, and multimillion-dollar price tags. The result is a technology that fewer than 200 institutions on Earth can access.

OPM-MEG is the first credible crack in that wall. Room-temperature sensors that sit on the scalp. Flexible arrays that fit any head size. Systems that tolerate natural movement. The physics is the same. The engineering is completely different. And the implications for who gets access to MEG-quality brain measurement are profound.

But here's the question that keeps the field up at night, and it's a question worth sitting with.

If the fundamental constraint on understanding the brain has been the gap between what our tools can measure and what the brain actually does, then what happens as that gap closes? What happens when millimeter-and-millisecond brain measurement is no longer locked in a shielded room? What happens when it's on your desk, or on your head, or woven into the fabric of everyday life?

We're not there yet. But the trajectory is clear. And every year, the tools get lighter, the sensors get more sensitive, and the distance between "what the brain does" and "what we can see" gets a little bit smaller.

The brain has been whispering its secrets in magnetic fields since long before we evolved the cleverness to listen. We're just now building the ears.