AI Algorithm Reduces MRI Scan Time and Preserves Image Quality 

Researchers in Argentina studied the effects of implementing AI-based reconstruction into five MRI scanners across a multisite MRI network. Patricia M. Carrascosa, MD, PhD, assistant professor of imaging, medical director and head of the Research Department at Buenos Aires University, presented the findings in a Monday afternoon session.    

“Although our institution operates multiple MRI sites with high utilization and almost full schedules, maintaining immediate access for patients increasingly required adding new scanners—an expensive and slow solution,” Dr. Carrascosa said. “By implementing AI algorithms that shorten scan times, we were able to expand daily patient volume and improve access without additional hardware investment.”

The research team integrated AI algorithms into five MRI scanners across five branches of the institution from 2021 to 2025. They developed custom protocols to reduce acquisition time and improve image quality. 

Knee, shoulder, ankle, brain and spine, representing most clinical indications in the institution’s workflow, became the primary focus for protocol development and image quality improvements. They used a moderate level of noise reduction to improve image quality.

The team assessed annual MRI volume per site, comparing volume before and after implementing AI algorithms. They also analyzed ramp-up speed and growth patterns according to the time each site had been using the AI tools.

The analysis revealed a marked difference in ramp-up speed between early and later AI implementations. While the first sites showed steady year-over-year MRI volume growth, branches that adopted AI more recently reached comparable increases within only a few months. According to the investigators, this accelerated ramp-up reflects both the maturation of standardized workflows across the network and growing acceptance among patients and referring physicians, enabling newer sites to achieve operational benefits much faster than the early adopters.

In addition to these growth patterns, the researchers also quantified the impact of AI on scan duration across key MRI protocols.

Impact of AI on MRI Scan Efficiency

Implementation of AI-based reconstruction in MRI protocols resulted in substantial time savings across all major exam regions. For example, knee scans were reduced from 8 minutes 27 seconds to just 4 minutes 39 seconds—a 46.9% decrease. 

Spine imaging times dropped by 42.6% while shoulder and ankle protocols saw the most dramatic improvements, with scan times cut by 64% and 63.9% respectively. Even brain MRI experienced a 21.9% reduction, going from 10 minutes 31 seconds to 8 minutes 5 seconds.

Across all regions, the mean reduction in scan time was 47.9%, Dr. Carrascosa noted. After AI implementation, volume grew each year across all sites. “By implementing AI algorithms that shorten scan times, we were able to expand daily patient volume and improve access without additional hardware investment,” she said.

AI-driven time reduction also enhances the patient experience by minimizing time spent inside the scanner. “This is particularly valuable for symptomatic patients who cannot tolerate long examinations, as well as for claustrophobic or pediatric patients who, in many cases, no longer require sedation or anesthesia thanks to shorter AI-accelerated protocols,” Dr. Carrascosa said. “Institutional performance indicators, including our Net Promoter Score of 82, confirmed high patient satisfaction and positive perception of AI-enhanced services.”

Dr. Carrascosa and her colleagues observed that AI not only reduced exam duration but also enhanced image quality. “This enabled more accurate diagnoses that helped referring physicians make better therapeutic decisions for their patients, in alignment with my institution’s strategic goals of efficiency, quality and excellence in patient care,” she noted.

Dr. Carrascosa emphasized that she would recommend other institutions use a strategic, holistic approach to AI adoption in MRI.

“Begin with a pilot on high-volume protocols and establish baseline metrics—scan duration, completed exams per day, repeat rates and patient wait times—to objectively measure impact,” she said. “Combine AI with protocol standardization and continuous quality monitoring to achieve reproducible results across scanners and sites.”

Access the Learning Center Theater presentation, “Real-World Implementation of AI in MRI: Five Years of Productivity Gains Across a Multisite Imaging Network,” (M5-STCE1-2) at RSNA.org/MeetingCentral