AI System Reads and Diagnoses Brain MRIs in Seconds with 97.5% Accuracy

Researchers at the University of Michigan have developed Prima, an artificial intelligence system that can read and diagnose brain MRI scans in just seconds, achieving accuracy as high as 97.5 percent. The system, published in Nature Biomedical Engineering, represents a significant leap forward in medical imaging technology and could transform how neurological emergencies are handled in hospitals worldwide.

Prima is classified as a vision language model (VLM), a type of artificial intelligence that can process images, video, and text together in real time. The team trained it on every available MRI collected since radiology records were digitized at University of Michigan Health — more than 200,000 MRI studies encompassing 5.6 million individual imaging sequences, along with their corresponding patient histories and radiologist reports.

“5 percent.”

Across more than 50 different radiologic diagnoses involving major neurological disorders, Prima delivered stronger diagnostic performance than other advanced AI models. The system can identify conditions ranging from brain tumors and strokes to multiple sclerosis and neurodegenerative diseases, providing detailed reports that match the quality of experienced radiologists.

One of Prima's most valuable capabilities is its ability to triage cases by urgency. Certain neurological conditions, including strokes and brain hemorrhages, demand immediate medical attention. The system was designed to automatically notify the most appropriate subspecialist — such as a stroke neurologist or neurosurgeon — ensuring that critical cases receive rapid attention even during overnight hours or weekends.

The implications for healthcare delivery are profound. In many hospitals, MRI interpretation backlogs can mean patients wait hours or even days for results. Prima could dramatically reduce this bottleneck, providing preliminary readings almost instantly while allowing radiologists to focus their expertise on the most complex and ambiguous cases.

The researchers emphasize that Prima is designed to augment rather than replace human radiologists. By handling routine interpretations and flagging emergencies, the system frees physicians to spend more time on challenging diagnoses and direct patient care, ultimately improving outcomes for patients with neurological conditions.