AI startup is hoping to replace MRI's with its full-body scanner — uses a shallow pool of water to map the human body. But some doctors aren't so sure

Patients are lowered into the pool and a ring of microscopic squares generates and records ultrasonic waves. That data is then sent to supercomputer clusters and processed into body images, MarketWatch reports.

Founder David Holz’s ambitions are enormous: a fleet of 50,000 scanners running one billion scans per month by 2031, beginning with a spa location in San Francisco by the end of 2027.

Why MRI access is a real problem

The pitch lands against a backdrop of ongoing frustration. MRIs are among the most diagnostically useful tools in medicine, but they are expensive, slow and in short supply.

Without insurance, an MRI can cost anywhere from $400 to $12,000 depending on things like the body part, scan type and location, with hospital-based scans averaging $2,250 compared to roughly $650 at outpatient centers, according to New Choice Health. Even insured patients routinely face hundreds of dollars in out-of-pocket costs after deductibles and coinsurance.

Access is the other half of the problem. According to Tesla MR Institute analysis, approximately 17.4% of MRI technologist positions across the U.S. are currently unfilled — roughly 7,200 open positions — with vacancy rates at or near all-time highs.

Studies in the Journal of the American College of Radiology project the radiologist shortage will persist through 2055 without intervention.

And rural communities are especially exposed. According to the American Hospital Association (AHA), a large portion of patients served by rural critical access hospitals experience significant transportation barriers when traveling for medical imaging — a challenge it noted at a 25-bed rural Idaho hospital that had been relying on a mobile rented MRI trailer precisely because patients couldn’t easily reach imaging facilities elsewhere.

What the technology actually does

The underlying hardware is supplied by medical-device company Butterfly Network, which confirmed that each Midjourney scanner prototype embeds 40 of its proprietary Ultrasound-on-Chip imaging modules, MarketWatch notes.

The software layer relies on full-waveform inversion — a technique widely developed by the petroleum industry to locate underground gas and oil reserves — which converts raw acoustic wave data into three-dimensional images.

A 2020 peer-reviewed study demonstrated that full-waveform inversion can produce detailed brain images from ultrasound data, validating the core scientific premise. Until recently, the technique was considered commercially impractical because of the enormous computing power it required — a barrier now falling rapidly as AI supercomputer infrastructure expands.

What experts are saying and why the caveats matter

The response from medical professionals has been measured.

Eric Topol, a cardiologist and director with Scripps Research Translational Institute, told MarketWatch that extending ultrasound technology in this direction is “a good idea” — but said he remains “somewhat skeptical.”

His key reservation: ultrasound cannot penetrate dense bone structures to image brain tissue the way an MRI can, which means describing Midjourney’s scanner as a “full-body” alternative to MRI is misleading. Topol believes Midjourney’s scanner is more accurately compared to a DEXA scan — a tool used to assess body composition, bone density and muscle mass.

Nisha Mehta, a radiologist and founder of the Physician Side Gigs community, was cautiously optimistic in a LinkedIn post, but drew a clear line: “I’m not an advocate for deploying party tricks at a spa and using it to guide medical advice or care.”