On the Midjourney whole-body scanner, and why doctors keep getting innovation backwards

Midjourney, the text-to-image company, announced a health division this week and showed off a prototype whole-body ultrasound scanner. Some of the marketing calls it "Ultrasonic CT," which is nonsense — no CT, no radiation, nothing to do with either. It runs on licensed ultrasound-on-chip hardware through a deal with Butterfly Network reportedly worth tens of millions. The pitch: a full internal scan in under a minute, for a few dollars, no radiation, no magnetic field. No FDA clearance for diagnosis. They're shipping it as a "general wellness" product — body-composition maps, not diagnoses — while they chase clearance later.

Radiologists were dunking on it within hours. And my first reaction, I'll be honest, wasn't to the scanner. It was to my own people.

I'm embarrassed by the doctors in this debate. Not a polite, professional embarrassment. The real kind. More innovation is better. Obviously. It will create demand for the gentler intermediate steps we don't have yet. Obviously. So how is it that so many physicians can't picture a world with something between a calcium score and an angiogram? Something between a scan that finds a growth and a PET scan that tells you if the growth wants to kill you? That's not some wild futurist fantasy. That's just... how medicine works. How it has always worked. Cheap and harmless first, sharp and invasive only for what the cheap stuff flags. That's the whole game. We invented it.

And look — I believe the doctors raising the alarm mean well. I do. But I don't think they have any idea how anti-health they sound. How cold. People are already exhausted by us. Every visit is an hour in a waiting room because the system is jammed, and patients gave up expecting better years ago, while we wave our hands at "admin" like the dysfunction is weather instead of something we built and could unbuild. So people stop coming. They stop coming until somebody's mom gets the stage four diagnosis that a scan could've caught at stage one, if scanning at scale had ever been cheap and easy enough to fit a normal human life. And here's the part that gets me: it takes almost no imagination to see how "let's restrict access because some patients might get anxious about an ambiguous result" turns into exactly that outcome. A profession protecting itself from inconvenience and calling it caution.

Here's the analogy I keep coming back to. A teacher who won't let anyone in the room read above a third-grade level, because one kid can only read at third-grade level. That's the logic.[^1] Lock down everyone's access to information because a few people might react to it in a way that's annoying for the institution to handle.

More data is useful. Full stop.[^2] Not because data is holy — it isn't. Some of the ugliest surveillance states and weapons programs in history were exquisitely data-driven. But every guideline, every evidence-based anything this field has ever produced came from more data, never less. "What do we do with this new information" is a completely normal question. It has never once been a good reason not to go get the information. It's the reason to go get it — so we can figure out what it means and when to act. That is the literal job description of evidence-based medicine. Saying "we don't know what to do with this yet, so let's not collect it" doesn't protect the discipline. It turns the discipline inside out.

Every real shift in medicine got this same reception, in real time, from the people with the most to lose if they were wrong. I'm not going to recite the whole list — Galen, Avicenna, Semmelweis, Pasteur, Koch, Fleming — you already see it. Semmelweis told Vienna's doctors to wash their hands between the corpses and the mothers, and they had him committed, while women kept dying from the infection his colleagues were ferrying around on their own fingers. Germ theory got laughed at by men who'd built whole careers on being wrong. Fleming's mold sat on a shelf for a decade. DNA, RNA, protein folding — same objection every time, just wearing new clothes: this threatens what we know, and what we know is who we are. Darwin got it. Watson, Crick, and Franklin got it. Same fear underneath, every time — that the data left too little room for the story people had organized their lives around. And I'm watching it happen again right now, except this round the fear wears a lab coat and calls itself "concern about incidentalomas."

I want to be in this field at the exact moment it hurts the most — not after some braver generation has already paid the bill. I know what I'm signing up for. And, setting aside medicine's weird conviction that being older is the same thing as being right: most of the technology we use every day was built by people decades younger than the physicians who keep citing their age and experience as the reason to slow down. Sit with that one.

So think about the actual mechanism here. No new data, no real demand for the softer steps between an incidentaloma and a cancer diagnosis — where, exactly, does the technology that fixes this come from? Who funds it? Nobody is saying go calcium-score-to-angiogram for the hell of it. The point is the opposite: the more these tools spread — calcium scoring, whole-body MRI, whole-body ultrasound — the better we get at managing the climb from "huh, what's that" to "here's what it is." More concern makes more data. More data is demand. Demand is the only thing that has ever conjured better tools into existence. And as that machine spins up, fewer moms get blindsided at stage four. In the meantime, yes, the flood of findings gets handled the hard, unglamorous way — actual counseling, actual coordination, doctors talking to each other to stop needless escalation. That work is not optional and it is not beneath us. The alternative to doing it isn't less disease slipping through. It's more.

The objection I find almost impossible to take seriously is "we need rigorous study before we embrace any of this."[^3] You cannot have the definitive study yet — the thing you'd study doesn't exist in mature form yet. And the second it does start to exist, the second a technology is good enough to generate demand for the next refinement, this exact crowd shows up to block it — strangling the process that would've produced the evidence they say they're waiting for. It's a closed loop. So let me ask the uncomfortable question: how many people died of cancers we could have cut out, in the years between an imaging tool being good enough and medicine deigning to use it?[^4]

Now — there is one good argument on the medical side, and I'm not going to pretend there isn't. Medicine can't just swallow every shiny new thing, because being wrong is expensive in human lives, most new toys don't pan out, and you do need solid data before you tell people something improves outcomes.

Totally fair. The Midjourney scanner might be a dead end. It might also be the first link in a chain of demand that ends somewhere genuinely great. Nobody knows. I don't know. That's the point. So why is it so hard to just say: this is exciting, let's see where it goes?[^5] And while everyone's busy planting flags against giving it the chance — my grandmother died because her cancer was found far too late. So did a lot of people's grandmothers. You can dislike hearing this, I don't care: when you've buried someone to a late catch, and you watch a physician stand in the path of the tool that might've caught it sooner, you do not feel grateful for their prudence. You feel like the person you trusted with your life is the same person blocking the thing that might have saved it.

This is what happens when doctors get so sealed off from how technology and markets actually get built that they lose the plot entirely on what building anything requires. It's a big reason we get called pharma shills and sales reps. It's a big reason trust in healthcare is in the gutter. People are not stupid. They notice when the profession that's supposed to guard their health flinches away from the thing most likely to serve it.

The next few years are going to be ugly while medicine drags itself through this. I'm walking in anyway, eyes open. If the cost of catching deadly disease earlier than we do today is a stretch of real institutional discomfort while we build the right thresholds and guardrails — that's a price worth paying, and I think refusing to pay it is the less defensible choice, not the responsible one. And if you're certain I'm the one out of line here, fine. I'd just ask you to sit with the reality that someone’s loved one died because of the absence of early detection, because of all the structural issues within healthcare that we brush aside as “out of our control” whilst we roadblock tech-first solutions because they will increase the burden on our already duct-tapped system, all the while more preventable deaths keep happening.

[^1]: The analogy doesn't fully hold, and I'd rather say so than let it slide. In the classroom, letting everyone read at their own level costs basically nothing — a book on a shelf hurts no one. A false positive on a body scan is not free. It's a biopsy, or a surgery, each with a real complication rate, landing on a real person who never had the disease you were hunting. So the honest objection from the medical side isn't "patients get anxious." It's that acting on a positive finding carries physical risk, and there's no version of that in a reading-level rule. A better analogy would carry that asymmetry. Mine doesn't, and I'm leaving it in anyway because the policy logic — restrict everyone over a manageable few — is still the part I'm attacking.

[^2]: South Korea is the cleanest real-world test of this, and you have to sit with both halves or you're cherry-picking. Starting in the '90s, Korea made cheap thyroid ultrasound widely available at routine checkups. By 2011, thyroid cancer diagnoses were fifteen times the 1993 rate — and the death rate didn't budge. The scans were finding small, lazy disease that autopsy studies show sits quietly in a huge chunk of the population and never does a thing, and for years the only response on offer was surgery: tens of thousands of thyroidectomies, real complications and all, on people who were never in danger. That's not "the data was bad." That's "data with no normalization layer can do real harm." And the fix wasn't to stop scanning — it was to build the exact threshold this essay says demand should produce: risk-stratification like K-TIRADS, which scores a nodule before anyone reaches for a needle, plus active surveillance for the small, low-risk stuff instead of reflexive surgery. That took decades and a lot of needless operations to get right. It's still not clean: after screening pulled back hard, mortality among the people diagnosed after 2015 crept back up — meaning the screening wave was catching some genuinely dangerous disease alongside the overdiagnosis, and overcorrecting away from data has its own body count. So the honest read isn't "screening won" or "screening lost." It's that building the right threshold is real, ongoing, expensive work — not a freebie that materializes the moment the data shows up.

[^3]: This one overstates itself. You can study an immature screening technology — if you point it at a population whose risk justifies it and track outcomes prospectively. Inova Health System is doing exactly that right now, using whole-body MRI and ultrasound to screen firefighters for cancer, a group chosen for elevated occupational risk, with downstream outcomes built into the design. That's the model: deploy where prevalence earns it, watch what happens, let evidence accumulate. The real failure mode isn't "we can't study this yet." It's blasting it out to a general, unstratified population before that study exists. Which is a choice. Not fate.

[^4]: This fear isn't hypothetical — there's a case that maps onto it almost perfectly, just from the opposite side: too much caution, not too much access. In 2012 the USPSTF recommended against routine PSA screening for prostate cancer at any age, mostly over overdiagnosis and overtreatment — the same logic aimed at whole-body scanning today. Screening cratered. In the years after, multiple independent analyses found metastatic and high-grade prostate cancer rising at diagnosis, and prostate-cancer mortality — which had been falling for two decades — flattening or turning back up. The USPSTF walked it partway back in 2018. The evidence is mostly observational, not a randomized smoking gun, so I won't oversell it. But the direction and timing line up tightly enough to take seriously the idea that excess caution kills people too — not just excess access.

[^5]: The generosity is right in spirit, but be precise about what Midjourney is actually pulling, because it changes what "give it a chance" means. No FDA clearance for diagnostic use, and a deliberate "general wellness" label — a category that conveniently dodges the clearance pathway that would've forced them to generate the exact prospective evidence this essay keeps demanding, all before scaling toward a stated goal of ~50,000 units worldwide. That's not a scrappy innovator getting throttled by stuffy doctors. That's a company skipping the validation route that already exists so it can hit consumers first — which is close enough to Theranos to deserve hard scrutiny on its own, separate from whether whole-body screening as a category deserves a shot. So the real version of "let's see where this leads" is "let's see where it leads once it's tested the way Inova is testing it" — not "once it's been sold to fifty thousand people with zero follow-up."