On Medical Device Development

As a community-driven research lab we want to do things a little differently than usual. We want to post regular updates showing the community what we’re up to, even if that means publishing technical deep-dives about our product development processes and regulatory classifications. We also want our community to be maximally informed and provide as many opportunities for feedback (and criticism) as possible. The best way to do that is completely out in the open.

We hope that blog posts like this will offer an interesting look behind the scenes.

What can we do on day one?

The long-term potential for ultrasound CT technology is incredible, but we’re starting from humble beginnings. The biggest needle to thread is balancing user needs against time-to-market without compromising on safety.

After a great deal of thought and consideration, we landed on a simple strategy last year: launch a high-end body composition analyzer intended for general wellness use that lets people see an amazingly detailed 3D map of their body composition over time. By not performing any diagnosis or treatment with the first device, we can get a safe product out to the community much faster. Not only will this enable us to get feedback from hundreds of thousands of people, it also gives us a lot more time and data to help with what comes next.

With this functionality, we believe that we can already help a lot of people better understand what’s happening inside their bodies. Lifestyle changes that would only slowly and visibly affect our bodies over time can suddenly become monitored (and encouraged) at a much more granular level. We think this can immediately help us live happier, healthier lives.

Here’s what to expect with our first gen scanner:

• You change into swimwear, shower, and step into a private scanner that runs fully automatically.
• Once you step on the scanner platform, you will be gently lowered into water, which is warm, like a hot tub.
• A ring of ultrasound sensors images your body composition as you pass through. Today a scan takes 20+ minutes; Gen 2 will be much faster, and our final goal is 60 seconds.
• Afterward, you’ll get a detailed map of your body composition uploaded to your account. It’ll be visible in the Midjourney Medical app and on the website. No needles, no ionizing radiation and nothing invasive.

Our regulatory pathway

We’re already getting a lot of questions about what the Midjourney Scanner is from a regulatory standpoint. Here in the United States, medical devices fall into one of three classes: Class I (e.g., bandages), Class II (e.g., most imaging devices), and Class III (e.g., surgical implants). The class your device fits into determines what set of processes you need to follow in order to get a device to market.

Back in February we engaged directly with the FDA. The mechanism we used for this is called a “513(g) Request for Information”. You basically send the FDA a detailed document describing what your product is and does and the agency responds with its assessment of the product’s classification and how it should be regulated. FDA responded to our 513(g) on June 4th, 2026.

This is exactly what we hoped for. The Midjourney Scanner is a Class II medical device, but given the nature of the technology and limited general wellness uses, we are exempt from FDA’s premarket review process for the device.

Though we are exempt from premarket review, many aspects of FDA’s medical device regulation still apply to us, including:

• Establishment registration & device listing (21 CFR Part 807)
• Medical Device Reporting (we report adverse events (Part 803))
• Labeling requirements (Part 801)
• Additional processes (QMSR 820.180 & 820.198, tied to ISO 13485 Clause 8.2.2)
• Unique Device Identification (21 CFR Part 830)
• Corrections and Removals (21 CFR Part 806)

This response to our 513(g) submission isn’t a stamp of approval, but it does give us comfort that the FDA will not require premarket review for the Midjourney Scanner when used for general wellness purposes.

How this works regarding the regulatory ladder

Every new capability that we look to add beyond general wellness, including any diagnosis or treatment, will require moving up this regulatory ladder with more rigorous FDA premarket review processes. The burden of evidence may also increase depending on the feature(s).

How we do research and development

Medical devices aren’t built like a typical consumer product where you ship and iterate on any features you want in production. They are built with strict design controls in place. Design controls are basically a highly disciplined product development and maintenance process in which every user need, requirement, risk and test is traced, tracked and reviewed.

This involves developing a design and development file (DDF) including many hundreds of pages of documentation that gets created when you kickoff the project and rides along with the product during all phases of development, test and release. There are similar procedures for making sure that every part we manufacture and ship is tightly controlled and tested after launch. If you want to change a single feature on the Midjourney Scanner, you need to thoroughly test and track the change.

We break this process of releasing things into the market into seven distinct phases.

There is a detailed and formal design review with the cross-disciplinary team required after every phase of the project. We don’t move onto the next phase until that gate is passed and documented.

The shape these phases trace is what’s known as the V-model. You travel down the left side, breaking it down from broad human needs into detailed specifications and then build that into the thing itself. Then you climb the right side of the V, testing your way back up. Each level on the right proves out the corresponding level on the left. All of the solid and dashed lines provide traceability.

Down the left side: human needs become specifications, which in turn become the device. Up the right side: each test covers the level across from it. Hover or tap any stage to see what it means.

The V-model vs. the spiral development method

There is a tension that exists whenever teams build something like this. The V-model is linear and well-defined. It’s how you build a safe and effective product if everything is known ahead of time. The science of turning terabytes of messy sound waves into the most accurate possible images is genuine research that doesn’t follow such a process. For that, we need an iterative process that looks something like a spiral.

Gen 1 prototype → Gen 2 prototype → Released Product: Each loop confronts the next-biggest unknowns and ends more refined and capable.

So we make use of both types of development on purpose. The research, software and hardware mature on the spiral. And when a capability is ready to become part of a product that people can rely on, it enters the V-model, where it’s pinned to requirements, verified, validated, and documented. The spiral is how we do research and discover issues during development; the V is how we make sure that we have a safe and reliable final product. Knowing which mode we’re in and how to balance the two is a big part of the job.

Risk, taken seriously

Every credible medical-device program runs on formal risk management. For each thing that could go wrong, we score how bad it would be (severity) against how likely it is (probability), and the result tells us how hard we must work to control it. Green is good; Red is bad; nothing in the red zone is released without a control that brings it down to a lower-risk zone. Anything in yellow or orange gets a great deal of attention and many of those things are brought into the green.

Green = low risk · Yellow = medium risk · Orange = med/high risk · Red = high risk

For the Midjourney Scanner, the common risks include physical hazards such as slippery surfaces, skin irritation, and discomfort in the water. We engineer each of the identified risks down with the design of the product and how it is operated. For this product there is no ionizing radiation, nothing invasive, and we have designed our ultrasound output to remain consistent with FDA’s guidance on diagnostic ultrasound safety limits. Finally, the information presented in software needs to be safe with a minimum risk of inducing anxiety or misinforming the users.

Real people: our first IRB study

Here’s an important fact: the scan images presented in our announcement were not AI-generated. They came from real volunteers. Whenever you involve real people in research (even just to collect images for a product reveal) the ethical thing, and the right thing, is to put it under independent oversight. So we did that.

Our image-collection study was reviewed and approved by WCG IRB, an independent Institutional Review Board operating under FDA, HHS, and ICH good-clinical-practice standards. An IRB (institutional review board) exists to protect the people in a study, independent of the people running it.

Key details of our first IRB-reviewed study.

Here are some of the ways in which we protect the people who volunteer for studies:

• Everyone learns about what will happen, the risks, and their rights; and signs consent before anything begins.
• Anyone can decline or stop at any moment with no penalty. Employees’ jobs are never affected by their choice.
• We keep it safe. Low power ultrasound waves, head stays above water, warm and clean water, a trained operator up top and a second operator nearby, and the ability to exit at any time.
• The current study is for adults only, with documented exclusions for things like weight limit, open wounds, and active infections.
• Published scans don’t include any identifying information like names, and every volunteer signs an explicit image release with the ability to change their mind regarding their data.
• Any future studies generating data intended for regulatory submissions will be conducted under separate IRB protocols and, where applicable, will be assessed for Investigational Device Exemption (IDE) requirements under 21 CFR Part 812.

Let’s talk about your data

The IRB above governs research and the ethics surrounding collecting scan data from volunteers. IRB review protects the people taking part in the research, while HIPAA — and the idea of Protected Health Information, or PHI — governs how health data is stored, shared, and safeguarded by entities like hospitals, insurers, and the clinicians who bill them. As a direct-to-consumer product intended for general wellness, which doesn’t diagnose disease or bill insurance, the initial release of the Midjourney Scanner will generally sit outside of HIPAA’s strict legal scope, much like a smart scale or a fitness band.

As our capabilities climb the regulatory ladder toward clinical use, formal HIPAA and PHI obligations will begin, so we’re building to that standard now, while the product is small and early, rather than wedging it in later. The general ideas:

• Your scans, and the body-composition data we generate from them, are yours. You can get a copy and you can ask us to delete it at any time.
• Data lives in encrypted, access-controlled systems with administrative, technical, and physical safeguards, reachable only by authorized personnel.
• Images used in research, demos, or blog posts like this one are stripped of information that identifies you.

Privacy isn’t an afterthought. It’s something we’re putting our efforts into now; because the whole point is creating a future where billions of people can trust us with the most personal data there is.

Where we are in development

We’re building the scanner and the spa together as two projects that are closely linked. Below is our current target timeline for both the Midjourney Scanner and the spa, including where we are today. We’re taking bets on which will be ready to release first!

Why we’re telling you all of this stuff

Most companies keep this part of development hidden away. All of the classifications, matrices, review boards, and the careful language about a product are filed away in a back office like bureaucratic paperwork. The information is kept from the people it’s designed to protect and those people just need to trust the process.

We don’t operate like a typical company though. To us, the process is one of many things that we want to share with the community. A Midjourney scan you can trust is built on a thousand careful decisions, and you have every right to see how those decisions are made. Working openly is also how we keep ourselves accountable. We can take feedback from all of you even while development is still underway.

As a community-driven research lab, we answer to you, not to any investors. So we’ll keep publishing updates like this one alongside the beautiful and inspirational ones. We think knowing how carefully something is being built is its own kind of reassurance. Midjourney Medical might even become a template for how to do things transparently in a future that increasingly moves too fast to follow.

We’re building this the right way, in the open, together because we are all Midjourney.