ATREMIA is a motion-gating cradle for portable ultra-low-field MRI. Two independent sensors must agree the patient is still before a single line of image data is kept. When they disagree — or when either one falters — the scanner holds.
Ultra-low-field MRI brings imaging to the NICU, the bedside, the emergency bay — places fixed magnets can't reach. Its patients are often the ones least able to stay motionless: neonates, infants, the agitated and the critically ill. And because low field means low signal, the scans run long. Motion is the binding constraint on every image.
Nearly every motion tracker in use watches through a single channel — one camera, one navigator, one sensor.
A blocked line of sight, a slipped marker, a lost lock — and a single-channel tracker can keep reporting as if nothing is wrong.
It fails silently, passing motion-corrupted data downstream dressed as clean. The corrupted scan that looks fine is more dangerous than the scan that visibly fails, because nothing flags it for a second look.
ATREMIA doesn't try to be a more accurate tracker. It changes what the scanner does when it can't be sure.
Two sensing channels built on two different physical principles, chosen so that nothing capable of blinding one can blind the other. Redundancy only counts when the failure modes don't overlap.
Image data is kept only while both channels agree, within tolerance, that the patient is still. Agreement is the entry condition for every window of acquisition — not an after-the-fact correction.
Lose a channel, lose agreement, lose confidence — and the cradle holds. Acquisition is never the resting state. Concordance is the only thing that earns it, every window, or nothing is recorded.
This is the gate, live. Pull the channels apart, drop one entirely, or tighten the tolerance — and watch the resting state. Only concordance opens it.
The gate decides every window. Your inputs can only break agreement — never force acquisition.
Each channel produces its own motion reading of the same patient, in a shared reference frame fixed by the cradle. Neither relies on the other to function.
An arbitration unit maps both readings to a common measure of motion and asks one question: do they agree, closely enough, right now — and is each confident?
On agreement, the current window is acquired. On disagreement, that window is rejected or flagged. The decision is made before the data is trusted, not after.
Whenever agreement isn't affirmatively established — at power-up, on a fault, on a dropped channel, on low confidence — the cradle defaults to not acquiring. Silence is the resting state; every recorded window is one the system had a positive reason to trust.
A self-test confirms both channels before a session begins; a state machine degrades gracefully toward hold; the gated signal is all the scanner ever sees.
Redundancy is only real if the two channels can't fail for the same reason. Pairing an optical channel with a strain channel makes their blind spots disjoint — each one sees what the other can't.
| Disruption | Channel A · optical | Channel B · strain | Both lost? |
|---|---|---|---|
| Line of sight occluded | degraded | unaffected | no |
| Marker slips or detaches | degraded | unaffected | no |
| Illumination change | degraded | unaffected | no |
| Strain element debonds / fatigues | unaffected | degraded | no |
| Sensor interrogation lost | unaffected | degraded | no |
When a disruption degrades one channel, the channels stop agreeing — and disagreement is exactly what makes the gate hold. The blind spot becomes a hold, not a silent error.
What matters is failure-mode independence, not any one sensor. Optical-plus-strain is the reference pairing; the gate logic is the same for any independent pair, or for three or more channels voting.
Camera tracking paired with a fiber-optic strain element. Disjoint blind spots; the baseline embodiment.
Tracking paired with an air-bladder or bellows pressure channel for contact-based motion.
Tracking paired with a capacitive proximity or displacement channel.
Two contact principles with independent transduction and mounting.
Tracking paired with ultrasonic ranging for a non-optical distance reference.
Three or more channels; acquire on a required majority, flag the outlier.
The cradle doesn't choose between working and failing. It moves through defined states, and the unsafe direction is never the default one.
Both channels healthy and in agreement. Acquisition is enabled, window by window, for as long as concordance holds.
One channel impaired. Acquisition is held while the system attempts recovery and notifies the operator. It does not guess.
No trustworthy basis to acquire. The gate stays shut. Return to nominal requires re-established health and agreement — nothing less.
Low field isn't a limitation here — it's the enabler. The same permissive environment that frees a scanner from the shielded suite is what makes two unlike, MR-compatible sensors practical to place side by side.
At ULF, sensors and materials that would perturb or be unsafe near a high-field magnet can co-locate in the bore — which is exactly what heterogeneous redundancy needs.
Point-of-care siting puts the scanner next to the patients least able to hold still. The motion problem and the portability premise arrive together.
Low signal means long acquisitions. Every rejected-but-flagged window saved from silent corruption is time and certainty recovered.
The original motivation: bedside imaging of the smallest patients, who can't be asked to hold still.
Holding on motion rather than discarding whole scans reduces the pressure to sedate young children.
ICU patients who can't be moved to fixed MRI — and often can't stay still — imaged where they are.
Uncontrolled settings where a silent motion error is hardest to catch and most costly to miss.
Conventional motion correction optimizes how well it fixes motion while the sensor works. ATREMIA optimizes what happens when it can't be trusted.
A cradle carrying two channels of independent failure mode and an arbitration unit that enables acquisition only on agreement within tolerance, and defaults to a non-acquiring state otherwise.
A method of enabling acquisition windows only while the channels agree, rejecting or flagging them on disagreement, and defaulting to inhibition — a control process, determining no clinical condition.
A reusable cradle paired with a single-use liner that carries a disposable sensing element of one channel — renewing the patient-contacting sensor with each subject.
Documented in a provisional patent application now in preparation. Claim descriptions above are plain-language summaries for orientation, not the claim language itself, and describe an application in preparation rather than an issued or granted right.
ATREMIA is an early-stage medical-imaging venture. The motion-gating architecture is documented in a provisional patent application now in preparation, built around heterogeneous-redundant sensing and fail-safe acquisition gating for portable ULF MRI.
For technical, clinical-collaboration, or licensing inquiries, get in touch.