A standard MRI machine is a huge magnet that you can't move, plus a powerful computer that turns signals from the body into a picture. In recent years, portable scanners with weaker fields have appeared: they are light, cheap, and can work in a field hospital or an ambulance. But they have a problem with the computing part — the built-in processor can't handle modern algorithms that make the image sharp.

Engineers found a way out: instead of carrying a computer along, send the data over the network to the cloud. The scanner transmits raw signals to a remote server, which quickly processes them and returns the finished image. For testing, they used three tasks: noise reduction with a neural network, correcting distortions from the weak magnetic field, and reconstructing an image with a non-standard readout trajectory.

This approach removes limitations and allows using the most advanced algorithms even on cheap equipment. A doctor in a remote village gets an image processed just as well as in a capital city clinic. And the scanner itself stays light and inexpensive — all the heavy work moves to the data center.