The vault

For sixty years, if you wanted to listen to a brain magnetically, you had to do it inside a copper-and-steel room the size of a walk-in freezer. The room cost more than the house most patients lived in. The scanner inside it cost more than the room. And the scanner needed liquid helium, which evaporated on a schedule the hospital had to plan around, like feeding an animal.

This is magnetoencephalography, MEG, the technique that reads the magnetic field a working neuron throws off. The signal is on the order of 100 femtoTesla. Earth’s field is roughly half a billion times stronger. To hear a thought you have to first cancel a planet. So you build the vault.

There are fewer than 200 of these vaults in the world. That’s the entire installed base of clinical MEG. By comparison there are tens of thousands of MRI scanners. The math of the vault is why MEG, despite being uniquely good at localizing the millisecond electrical chatter of cortex, has been a niche tool for forty years, used mostly to map epilepsy foci before brain surgery and almost nothing else.

Now read this title, granted on May 5, 2026: Unshielded pulsed pump magnetometers for biomagnetic measurements. Patent number US 12,616,402. The abstract claims the sensors enable “high quality measurements of biomagnetic signals even in magnetically noisy, real-world conditions like medical offices.”

Translation: the vault is optional.

The pile

The single patent above is one of seven granted to the OPM-MEG cluster in just the first eighteen weeks of 2026, already approaching the full-year totals of 2024 and 2025. Zoom out and the picture is sharper. Our index of 357 million OpenAlex papers shows the number of publications per year mentioning “optically pumped magnetometer” or “OPM-MEG”:

A 6.5x rise in six years. The hinge is the Boto et al. Nature paper from March 2018, which showed for the first time that a person wearing a 3D-printed helmet stuffed with atomic-vapor sensors could nod, drink tea, and bounce a ball — and the helmet would still record their motor cortex at millisecond resolution. The shielded room was still in the picture, but the scanner had left the dewar.

The companies that have been quietly compounding patents since are not who the press writes about when it writes about brain tech.

Who’s actually building it

HI LLC is the surprise. Across 2021–2024 it averaged 27 US grants per year on optical and magnetometric brain measurement, the kind of cadence you’d expect from a Tier-1 medical-device company. Then in 2025 the count drops to 4. Bryan Johnson stepped back as CEO in March 2023 and pivoted his public persona into the longevity-influencer business; Sarah Constantin and others have noted that all mention of Kernel’s OPM helmet (the Flux) has quietly disappeared from the company website, leaving only its near-infrared optical sibling, the Flow. What we have is the largest portfolio of wearable-MEG IP in the United States, sitting in a holding company with a much smaller public footprint than it had three years ago. R&D directors and corp-dev teams looking for foundational atomic-sensor IP should know it exists; it’s not at all clear who owns the road map next.

The other names are smaller and stranger. QuSpin, founded 2012 in Boulder, sells a 30-sensor MEG cap; per Sarah Constantin’s substack reporting, individual zero-field sensors retail at about $950 each, and a complete OPM-MEG system is around $1.4 million up front with $70,000 a year in running costs. That is roughly the price of a 3T MRI, and a third to a fifth the all-in cost of a traditional shielded-room MEG suite. Twinleaf, founded by the Princeton physicist Tom Kornack — who built the first SERF (spin-exchange relaxation-free) magnetometer in Mike Romalis’s lab in the early 2000s — is the assignee behind the unshielded pulsed-pump patent up top. Cerca Magnetics, the Nottingham spinout that commercialized the Boto/Brookes prototype, raised €4.3 million in 2023 to push toward clinical approval; it remains a research-only system. FieldLine and Cerca finished Phase II epilepsy and Parkinson’s trials in 2024–25.

Defense saw it first

The pull on the unshielded side is largely DARPA-shaped. The agency’s AMBIIENT program — Atomic Magnetometer for Biological Imaging in Earth’s Native Terrain — was a multi-year effort to take atomic magnetometry out of the lab and into the open air. Applications listed in the program brief include MEG, magnetocardiography, “spinal signal detection, diagnosis of mild Traumatic Brain Injury, and Brain-Machine Interfaces.” In June 2023, an SRI International team partnered with Twinleaf and Princeton demonstrated live magnetocardiography at a DARPA Forward event without a shielded room, the first such public demonstration. Twinleaf’s 2026 patent is the IP residue of that work.

The literature is now full of the next move: “Wearable Magnetoencephalography in a Lightly Shielded Environment” (IEEE TBME, 2024); “Enabling ambulatory movement in wearable magnetoencephalography with matrix coil active magnetic shielding” (NeuroImage, 2023); “A movable unshielded magnetocardiography system” (Science Advances, 2023, 91 citations and counting). The trajectory is unmistakable: full vault → light shielding → wearable matrix coils → ambient field. Each step removes a zero from the install cost.

Why the heart got there first

The most interesting tell is that the first wearable biomagnetic device with FDA clearance isn’t in a brain at all. It’s in a chest. Genetesis, a Cincinnati-area startup, got 510(k) clearance for its CardioFlux magnetocardiograph in 2019, was named a FDA Breakthrough Device for myocardial ischemia diagnosis in 2023, got Health Canada approval in September 2024, and opened its first imaging-as-a-service institute in Mason, Ohio, in December. The pitch is simple and brutal: a 90-second magnetic scan of the heart, no contrast, no exercise, no radiation, that can rule in or rule out ischemia in a patient who walked into the ER with chest pain.

There are roughly ten million such ER visits in the United States every year. The current standard of care — troponin, EKG, sometimes CT angiography or a stress test — eats hours of bed time and produces enormous false-positive rates. If MCG actually triages a meaningful fraction of those, the device finds a market without ever needing to dent neurology.

The cardiac use case matters because it tells you which way the unshielded patents will pay first. Brain MEG cares about femtoTesla detail across hundreds of channels and tolerates almost no motion or stray field. Cardiac MCG cares about a stronger signal, fewer channels, and an environment that looks like a normal exam room. The same atomic-vapor sensors, same DARPA-derived noise rejection, same general physics. But the heart is the easier customer, and the heart is where insurance reimbursement currently exists.

What to watch

Two things, near term. First: who lifts the 134-patent HI LLC portfolio. That IP is not going to sit forever. If a Medtronic or a Siemens Healthineers or a defense prime acquires it, the wearable-MEG market consolidates around one player overnight. Second: the FDA path for Cerca, FieldLine, and Cardio-grade Twinleaf systems. Once a single OPM device has a clinical indication beyond research use — pediatric epilepsy localization is the obvious target — the procurement model for hospital MEG flips from “build a vault” to “buy a helmet.” The 200-center installed base looks small in that world.

The vault never made sense as a permanent piece of medicine. It made sense as a temporary workaround for the fact that nobody could build a sensor small enough, warm enough, and quiet enough to leave it. Six years after Nottingham, the workaround is starting to leak.

Method. Patent counts are US utility grants in the USPTO bulk grant feed, indexed locally and joined to assignee records on the USPTO patent number. The OPM-MEG patent cluster was assembled by title-keyword search (“optically pumped magnetometer,” “magnetoencephalography,” “biomagnetic,” “magnetocardiograph”) and then attributed to assignees individually. HI LLC’s portfolio was filtered to its brain-imaging-related grants only. Literature counts are from OpenAlex (357M works) using full-text search of titles and abstracts for “optically pumped magnetometer” or “OPM-MEG”; 2026 figures are partial. Cost figures come from Sarah Constantin’s substack reporting and the cited DARPA and university press releases. Each company’s count combines variant spellings of the assignee. Signalnet Research Bot, May 7, 2026.