📷 Midjourney’s Medical Imaging Bet

Editors’ Note: We’re excited to share our website with you! It features a healthspan-focused job board, an archive of past newsletters, a growing dictionary of terms we’ve highlighted in previous editions and more to come!

Visit us at: https://stacked-health.com/

Or copy and paste this link to others: {{rp_refer_url}}

On June 18, Midjourney, the company you last thought about when your feed filled with AI-generated portraits in 2022, announced that it is getting into medical imaging. A full-body scanner that uses sound instead of radiation or magnets, plus a chain of spas to put the machines inside. They have a stated goal of 50,000 scanners and a billion scans per month worldwide by 2031.

The announcement is written artistically. You step into "a shallow pool of golden light," descend through "a ring of underwater sensors, each acting like a dolphin," and 60 seconds later you have a 3D map of your insides "at nearly a hundred times the speed" of an MRI. The company argues that with enough early imaging, the world "could avoid 30% of all deaths and 50% of all healthcare costs."

Strip the mysticism off and there are three separate businesses bundled into one press release: a novel imaging device, a consumer real-estate and wellness operation, and a regulated diagnostics company. Each is hard. None of them is something Midjourney has done before.

The details that are buried

Midjourney's announcement reads as if the scanner sprang fully formed from its own research lab. It did not.

The core hardware comes from Butterfly Network, the publicly traded ultrasound-on-chip company (BFLY). Butterfly was apparently irritated enough by the omission that it issued its own press release "following Midjourney's public announcement" to take credit. Per Butterfly, the current prototype runs on 40 of its Ultrasound-on-Chip modules under a co-development agreement, and a 2025 SEC filing values that deal at roughly $74 million to Butterfly over five years.

Butterfly's filing is also blunt about the risk in its own forward-looking language: the payments depend on Midjourney hitting development and commercial milestones it has not hit yet, and on regulatory clearance it does not have.

Building on a partner's silicon is a perfectly good strategy, but it tells you what kind of company Midjourney wants to be. It is an interface, software, and brand wrapping around someone else's sensor. That is a very different bet from "we invented a new imaging modality," and it changes who actually captures value if this works.

Where the physics pushes back

Researchers have built whole-body and breast ultrasound CT systems for years, so the concept is not science fiction. The problem is that ultrasound has hard physical limits that no amount of compute removes.

Sound does not travel cleanly through bone, air, or deep soft tissue. Francis Deng, a neuroradiologist at Johns Hopkins, put it plainly after the announcement: the approach has a real technical basis, but ultrasound's inability to penetrate bone, air, and deep tissue renders many body parts inaccessible. A scan that cannot see clearly past the rib cage or into air-filled organs is not "in many ways superior to MRI," whatever the marketing says.

Then there is the gap between the demo and the dream. Reporting on the current prototype suggests the scan takes closer to 20 minutes than 60 seconds, that roughly a dozen people have been through it, and that there is no FDA clearance. Midjourney itself admits in the announcement that turning the raw wave data into images is still an unsolved "major computational task." None of that makes the project fake. It makes the timeline, SF spa by late 2027, 50,000 scanners by 2031, a statement of ambition rather than a schedule.

The spa is a regulatory instrument

The spa sounds like a branding flourish, but it’s doing legal work.

Anything that makes a diagnostic claim, "you have a tumor," "this is fatty liver," needs FDA clearance, and clearance is slow, expensive, and specific to each capability. Midjourney's move is to launch the scanner as a producer of "body composition maps," which describe what is physically there without telling you what it means. Body composition is not a diagnosis, so it sidesteps the diagnostic approval pathway at launch. The company says it will then submit results to the FDA to unlock "increased capabilities" over time.

The same framing quietly steps around HIPAA. Medical providers are covered entities with strict obligations. A spa selling a wellness experience that happens to generate "body composition" data may not be, which means the privacy rules you would assume apply might not. Privacy researchers who looked at the announcement flagged this: no published retention policy, no stated access controls, no clarity on whether your scans get used to train the models, and no consent form to read. The legal grey zone is deliberate. It is the wedge that lets a non-medical company start scanning bodies in 2027 without waiting on a decade of approvals.

Calling it a spa is the cleverest part of the plan.

The actual unsolved problem is interpretation, not imaging

Scanning a healthy person who has no symptoms reliably surfaces "incidentalomas," findings that look abnormal but are usually harmless. We have years of data on this from full-body MRI, the most direct comparison. Prenuvo, the incumbent, sells a whole-body MRI for about $2,499 and pitches it as routine preventive care.

In a large population study, roughly 30% of people got at least one incidental finding, and of those who went on to biopsy, about 62% showed no cancer. A 2025 meta-analysis of more than 9,000 asymptomatic people found whole-body MRI catches confirmed cancer in about 1.6% of those screened, around 1 in 63. Life-changing for that person. For the other 98.4%, the scan ranges from useless to a source of anxiety, follow-up procedures, and cost.

Now picture Midjourney's version of this, except cheaper, faster, marketed for weekly use, and pointed at a population with no symptoms and no doctor in the room. The scanning is the easy part. The hard part, the part nobody has solved, is telling a person what a finding means: whether it is signal or noise, whether it changed since last month, whether it warrants doing anything. A weekly scan does not reduce that problem, and it may multiply it by 52.

Why you should care

If you only take one thing from this: the scan is going to become a commodity, and the interpretation will not.

Midjourney, Butterfly, Prenuvo, and whoever comes next are racing to drive the cost of capturing data about your body toward zero. That race will probably succeed at least partway. What it does not solve is the layer sitting on top: what the data means, how it compares to a baseline, what actually changed, and what a non-expert should do about it. Every new scanner and every cheaper panel makes that interpretation layer more valuable, not less, because it dumps more raw data on people who cannot read it.

Notice that Midjourney's own pitch is, almost word for word, the thesis behind every serious longevity and metabolic-health product: get as much data about your body as cheaply as possible, watch how it changes over time, compare it to a baseline, and act earlier.

The durable position is likely at the opposite end. Own the trusted, longitudinal, plain-language layer that turns a flood of body data into a small number of decisions a normal person can act on.

Or copy and paste this link to others: {{rp_refer_url}}

Current Referral Count: {{rp_num_referrals}}