A closed loop inside the magnet: what United Imaging's MRI brain interface can and cannot do

The word doing the work here is "evaluation". A read-only interface can only ask whether the decoder guessed the intent correctly. A loop that includes measurement can ask whether the intervention changed the tissue, which is the claim the platform is built around: neural plasticity made measurable and controllable, with the scanner acting as the perception and feedback hub rather than a diagnostic camera ^[11][6].

That claim is checkable in principle and unchecked in practice. The stack is described as four parts, one of which is an "MRI-guided neuromodulation method" with no modality named ^[11]. The announcement carries no field strength, no patient numbers, no regulatory status and no peer-reviewed result ^[3]. The two figures on offer, millisecond capture and sub-millimetre structural precision, attach to acquiring activity and measuring anatomy respectively ^[8]. Neither is a round-trip number for the loop ^[4], and round-trip is the only latency that matters when stimulation is supposed to respond to what the imaging just saw.

The engineering detail that reads as genuine rather than promotional is the interference work. Implanted hardware distorts the field and degrades the image, and the platform claims to compensate for that, with a magnetically compatible toolbox meant to cut mutual interference between the interface and the scanner ^[9][10]. This is the unglamorous part of making an implant and a magnet share a room, and it is the part a competitor cannot skip.

What leaves that room is data, not therapy. An electrode array validated in the bore can then run in a clinic or a home; the loop itself cannot follow the patient anywhere. So the honest description of the product is an instrument for the people building implants, and the buyers are the programmes that currently infer stimulation effects from behavioural scores.

Which is why the application list sits awkwardly. The near-term clinical areas cited are motor rehabilitation, neuro-critical care, psychiatry, ophthalmology and audiology ^[13], all plausible in a hospital that already owns a scanner. The later list adds education, sports, gaming, sleep improvement and industrial safety ^[14], none of which happen inside a superconducting magnet. Those belong to wearable products whose evidence base a machine like this might one day supply. Putting both lists in one launch makes the platform sound wider than its footprint allows.

Provenance deserves the same scepticism. The account is the developers' own, relayed through Global Times, including the forecast that the BCI industry sees rapid development in 2026, which comes from Ming Dong of the Haihe Laboratory, whose institution co-released the system ^[12][2]. The accelerants cited by unnamed industry insiders, ultra-high-field MRI, faster sequences, real-time processing and AI decoding ^[15], are real trends, but they are trends anyone with a strong scanner can cite. The description "world's first full-stack" is the developers' framing ^[7], and full-stack is not a specification anyone can fail.