AI reduces MRI's carbon footprint

CHICAGO - Using AI with MRI imaging -- across a variety of exam types -- reduces exam time and therefore energy consumption and carbon dioxide (CO2) emissions, according to research presented December 3 at the RSNA meeting.

"[Our study showed] CO2 emissions savings equivalent to 19.4 million kilometers of car traveling," said presenter Tiziano Polidori, MD, of Sapienza University of Rome-Sant'Andrea University Hospital in Italy.

Healthcare makes up 4% to 10% of overall CO2 emissions, Polidori noted, but medical imaging makes up 10% of all healthcare CO2 emissions. MRI consumes the most energy, so decreasing scanning time -- without compromising diagnostic accuracy -- is a key strategy for reducing a radiology department's carbon footprint. Deep learning can help do this, he said.

Polidori and colleagues performed a study that included 148 patients. Of these, 45 underwent upper abdomen MRI, 53 underwent cardiac MRI, and 50 underwent lumbar spine MRI. Each patient had both conventional and AI-assisted MRI acquisitions. The team used Air Recon DL (GE HealthCare, or GEHC) with T2 and DWI sequences for upper abdomen MRI and T1, T2, and STIR sequences for lumbar spine MRI. It also used Sonic DL (also GEHC) with steady-state free precession (SSFP) sequences for cardiac MRI. The investigators compared time saved per patient, energy reduction (kW/h), and CO2-equivalent emissions between conventional MRI sequences and AI-assisted sequences.

The group reported that the AI-assisted MRI sequences improved performance measures across all three types of exams:

The research results underscore AI's potential for "enhancing sustainability in medical imaging practices and lowering the healthcare sector's carbon footprint," Polidori concluded.

To view full coverage of RSNA 2025, visit our RADCast.