New AI Predicts MRI Energy Use — Could Cut Hospital Bills and Carbon

MRI machines are some of the hungriest electricity users in any hospital. They run giant magnets, keep them chilled around the clock, and scan patients all day. A team of German researchers asked a simple question: can we predict how much energy a scan will use before it even happens? They trained two AI models on the settings technicians punch into the scanner — things like how long the scan runs, image resolution, and how the machine pulses its magnetic fields.

The answer was yes. The energy model predicted actual consumption almost perfectly, matching real measurements about 96% of the time; the power model was less precise but still useful. More interesting was what the AI revealed: total energy use is driven mainly by timing — how long a scan runs — while peak power depends on messy, tangled interactions between magnetic field changes and radio pulses.

That matters because it tells hospitals which few dials are actually worth turning. Instead of rebuilding scanners or buying new machines, radiology departments could tweak their protocols, schedule electricity-heavy scans during off-peak hours, and shave costs and carbon emissions without touching image quality. For patients, that could eventually mean shorter waits and less pressure on hospital budgets.

The catch: this is a prediction tool, not a fix. It doesn't save a single kilowatt on its own. The findings come from a limited set of scanners and data, and any real change to scanning routines has to pass strict safety and image-quality checks first. Hospitals also rarely see an electricity bill per scan, so pure cost savings may not be enough to motivate change.