An open-source 3D-printed MRI machine costs under $70,000

A scanner you can build, not just buy

Full-sized MRI machines start at $1.1 million and climb past $3 million. The Open Source Imaging Initiative has been chipping away at that wall for years with the OSI2 ONE, an open scanner that teams around the world have already replicated. The latest version puts a 3D-printed core inside a portable shell and comes in under $70,000. That is less than 7% of the price of a commercial unit.

What you give up for the price

The portable builds run at around 50 millitesla. Hospital scanners sit between 1.5 and 3 tesla. Lower field strength means lower spatial resolution and a noisier signal. The images are not as crisp, and nobody is claiming this replaces a $2 million machine for a detailed brain study.

The trade is deliberate. A refurbished full-sized MRI still costs $100,000 before you build the shielded room to house it. For clinics that cannot clear that hurdle, a noisier picture beats no picture at all.

Where AI steps in

This is the part that changes the math. Tech analyst Brian Roemmele pointed out on X that low-field MRI lives exactly where modern AI does well. Deep networks trained on high-field data, or built on the actual physics of the machine, can denoise the raw scan, correct for field gaps, and sharpen resolution past what the hardware captured.

That is not a generic chatbot bolted onto a medical device. The models are trained specifically on MRI data, some on as many as 1.6 million brain scans, or on physics-informed reconstructions. Because the OSI2 ONE is open, researchers without access to patient datasets can generate synthetic data and build their own physics model instead.

Who it is actually for

Critics note that no garage build is clearing regulatory review in the US or Europe anytime soon. Roemmele's answer was blunt: no one can stop people from building in garages. The real target is regions with little access to imaging, or the money to maintain it. A $70,000 open scanner plus a trained model could put diagnostic imaging in clinics that today have nothing.

It will not end the reign of the giant superconducting magnet. It might, however, widen the map of who gets scanned, and that is the point.

The takeaway

3D printing did not just cut the cost of a plastic bracket here. It helped shrink a $1.1 million medical instrument into something a small clinic or a research lab can assemble. Pair that with open AI reconstruction, and the resolution gap starts to close. We will be watching whether the first deployed units prove the model holds up in real diagnoses.