Can an image-making AI company build a medical scanner?

The company recently shared a video tour of this hardware, which essentially looks like a specialized tank of water equipped with rows of ultrasound probes. An engineer at the company described the device as a collection of standard sensor parts connected to everyday desktop computers. The team plans to market the scanner as a wellness tool to measure things like body composition, rather than a diagnostic medical device. By framing it this way, they aim to avoid the years of clinical trials and government oversight that are normally required for machines used to diagnose diseases.

Ultrasound technology uses sound waves to create images of the inside of the body. You can imagine it like active sonar used by ships, where a sensor pulses sound into a medium—in this case, your body tissues—and listens for the echoes that bounce back. By measuring how long those echoes take to return, a computer can build a map of what lies beneath the skin. The challenge with ultrasound is that it is notoriously difficult to get clear, sharp images of deep structures because sound waves scatter and degrade as they travel through bone and dense tissue. To make this work at the quality Midjourney suggests, they must process gargantuan amounts of raw data, filtering out background noise and precisely stitching together signals from dozens of sensors simultaneously. This is very different from generating an artistic image, where the AI just needs to create something that looks plausible to the human eye; in medicine, the output must be an accurate representation of reality.

The company is attempting to skip the traditional, slow process of medical approval by labeling their device for wellness rather than diagnosis. If they succeed, it could provide people with unprecedented access to frequent scans. However, when health information is involved, the line between a wellness tool and a medical device can become dangerous. If people start using these scans to monitor their own health, they need to know if the data they are receiving is medically sound or simply a visually pleasing estimation. Without clinical testing, we are left looking at a high-tech tool that has yet to prove it can actually see beyond the surface level.