Ultra-low-field MRI · In development

The scanner is already under the patient.

SomniScan is building a whole-body MRI into the hospital bed. An open-face permanent magnet array sits flat beneath the mattress, so imaging happens where the patient already is, as often as the care team needs it.

Field
Roughly 64 mT from a passive permanent magnet array. No cryogens, no quench, no chiller.
Geometry
Open face rather than a bore. Nothing surrounds the patient, and nothing has to be disconnected.
Siting
Designed to run in an ordinary patient room, without a shielded enclosure or a dedicated suite.
Coverage
Whole body along the length of the bed, with the patient supine and stationary.

The problem

Imaging is a trip, so it only happens once.

Moving a critically ill patient to the MRI suite takes a nurse, a respiratory therapist, a transport team, an elevator, and the better part of an hour. Lines get clamped. Drips get paused. The monitoring gap is real, and the adverse event rate is high enough that clinicians ration the trip.

So the patients whose brains and bodies are changing fastest are the ones who get imaged least. What comes back is a snapshot, a single frame of a process that is still moving, and the team spends the next two days inferring the trend from labs, exams, and a bedside monitor.

Portable scanners solved part of this by bringing the machine to the room. We think the next step is to stop moving anything at all.


How it works

Four subsystems, one of which is genuinely hard.

Magnet

A flat, passive array under the mattress

Permanent magnets in a Halbach configuration project a working field above the array instead of inside a bore. The assembly is flat, always on, and draws no power. It lives in the bed platform, under the mattress, out of the way of every line, drain, and ventilator circuit already attached to the patient.

Signal

Built to work in a noisy room

A hospital room is full of electromagnetic noise, and at ultra-low field there is very little signal to hide it behind. Our approach combines external reference sensing with active cancellation so the system can operate without a shielded enclosure. This is the hardest unsolved part of the problem, and it is where most of our engineering effort goes.

Reconstruction

Learned reconstruction, tuned for low field

Deep learning reconstruction and denoising pull usable contrast out of low-SNR acquisitions. The goal is not to match a 3 T diagnostic exam. It is to make each scan good enough to answer the question the team is asking today, and consistent enough to compare against yesterday.

Workflow

A trend line, not an event

When a scan costs nothing but time, imaging can repeat on a schedule instead of on a referral. The output a clinician opens is not one study in isolation. It is the same anatomy, in the same position, across the length of the stay.

Admission
6h
12h
24h
48h
Discharge

Same bed, same position, same sequence. Serial imaging without the registration guesswork, because the patient never moved between studies.

Where we think it matters first

The patients who are hardest to move.

Neurocritical care

Hemorrhage, edema, hydrocephalus, and shift all evolve over hours. Serial imaging at the bedside would let teams watch the direction of travel instead of scheduling around it.

Post-operative and trauma

Patients who are sedated, ventilated, or externally fixated are exactly the ones for whom transport is most costly and most risky, and for whom a repeat study is most often deferred.

Long-stay and rehabilitation

Recovery is measured in weeks. Imaging built into the bed makes a weekly quantitative check-in a routine part of the stay rather than a separate procedure.

These are development targets that define our engineering requirements. They are not indications for use, and no clinical performance has been established.

Where we are

Early, and looking for the right people.

We are building the subsystem prototype, pursuing federal non-dilutive funding, and raising a pre-seed round. If you are an investor in hard medtech, a hospital willing to think about serial bedside imaging, or an engineer who has spent real time inside a low-field system, we would like to talk.