Behind the scenes with the Midjourney scanner [video]

### TL;DR A full-body imaging system using 40 modified ultrasonic transducers scanning simultaneously in a water tank can rapidly generate 3D body reconstructions. The goal is not to replace MRI but to make medical scanning cheaper and more accessible. ### Project Introduction: From Hot Tub to Full-Body Scanner Video creator Semkas demonstrates real hardware he helped develop at Midy — a full-body ultrasound scanner. The prototype looks crude (a modified hot tub with an elevator), but it packs 40 disassembled and reworked ultrasound transducers. The core idea is to expand a standard B‑scan from a single angle to 40 probes working simultaneously, then combine 2D slices into a complete 3D internal structure by vertical movement. If successful, it will be faster, cheaper, and more accessible than existing medical imaging technologies. ### Ultrasound Principles & System Assembly An ultrasound transducer works by emitting high‑frequency sound waves and listening for echoes reflected by tissue. In this system, the transducers do not fire all at once; instead, they take turns emitting a few beams while all other probes are in receive mode, allowing precise localization of where and when the wave appears. The build process included: purchasing medical‑grade ultrasound transducers, building a box with a water tank, modifying a waterproof elevator, constructing a crane to lower the elevator into the tank, assembling the 40 probes into a ring mounted on the tank, and connecting numerous computers to process the data. The first scans were poor — subsequent steps required math, a better‑designed probe ring, more computers, compression algorithms (because the system produces terabytes of data per second), and fixing water leaks. Eventually, clear images of organs and bones emerged. ### Why Water? — The "Wet Elephant" Traditional B‑mode ultrasound requires gel and direct skin contact, but experiments showed that warm, degassed water conducts ultrasound just as well. The subject simply soaks in warm water for a few minutes — a much more pleasant experience than applying gel. Regarding the obstacles of air and bone inside the body, the team explains that the 40 probes provide a 360° aperture and plane‑wave beamforming. Combined with dense integrated circuits on the ultrasound chip, they can observe from any angle, even penetrating through air pockets and bone. ### Team & Experiments The scanner is not a solo project — it’s a team of researchers and engineers, including Jin Hua, co‑first author of the original full‑body ultrasound tomography paper from Caltech, co‑first author David Garry, and software engineer Alex Kristofferson (involved from machine programming to cloud reconstruction). They demonstrated a small 8‑probe mini‑tank in real time: a hand placed in the water, and the screen immediately showed bone and muscle structure, without any radiation shielding. The team emphasizes that the current system is far more integrated than the Caltech version — which used hand‑soldered polymer PZO piezoelectric arrays — whereas collaboration with Butterfly Network brought chip‑based ultrasound technology, vastly improving integration. ### Founder’s Motivation & Future Vision Founder David states bluntly: "I’ve been building hardware my whole life. I kept trying to get someone else to do this, but nobody would, so I had to do it myself." The scanner’s goal is not to replace MRI but to offer a completely different type of image — some tissues are seen more clearly, others less so — but its value lies in simplicity and low cost. Imagine doctors scrolling through a sequence of images over time instead of guessing from a single static picture. None of the components in the current system are exotic: a pool pump, an ordinary water tank, off‑the‑shelf Butterfly probes, a stepper motor, a Raspberry Pi for triggering, and pre‑assembled computers. The core is not about achieving the best image, but about making scanning ordinary and feasible. If it becomes widespread, large‑scale longitudinal studies could collect unprecedented sample data. ### Current Status & Next Steps The prototype is far from finished — held together with clamps, with random sensors taped to its sides — but it’s already much better than three months ago. The team develops in a community‑driven way, regularly publishing blog posts, maintaining an open Discord channel, and listening to feedback. Regarding the FDA: they submitted a 513(g) form, hoping to be classified as a general wellness product (for body composition analysis only). The FDA replied that this seems reasonable, but full Class II medical device procedures (development, verification, validation, etc.) must still be followed. Short‑term plans: complete a large number of scans, demonstrate the spa function, then release a second‑generation scanner (much, much better). The third generation will use custom‑chips and silicon, and ultimately mass‑produce 10,000 units. The team even plans to film the assembly line. ### Conclusion This machine illustrates the hardware innovation process: constant failure, learning, and problem‑solving — even with millions of dollars invested. It may one day change the accessibility of medical imaging, but more importantly, it already proves that you can build something disruptive starting from a pool pump and a Raspberry Pi. **Source:** https://www.youtube.com/watch?v=4nzzpUKhj1M