AI Reads Brain MRIs in Seconds with 97.5% Accuracy

Getting a brain MRI often means waiting days or even weeks for results, but a groundbreaking AI system is about to change that reality for patients nationwide.

Researchers at the University of Michigan have developed Prima, an artificial intelligence model that reads brain MRIs and delivers diagnoses in seconds. In testing across more than 30,000 MRI studies over a year, Prima achieved up to 97.5% accuracy across 50 different neurological conditions.

The speed matters most for life-threatening situations. When Prima detects brain hemorrhages or strokes requiring immediate attention, it automatically alerts the right specialist, whether that's a stroke neurologist or neurosurgeon. This instant notification happens the moment a patient completes their scan.

"As the global demand for MRI rises and places significant strain on our physicians and health systems, our AI model has the potential to reduce burden by improving diagnosis and treatment with fast, accurate information," said Dr. Todd Hollon, a neurosurgeon who led the project.

Prima works differently than previous AI attempts. Earlier systems analyzed only specific slices of MRI data to perform narrow tasks like detecting lesions. Hollon's team trained Prima on every single MRI taken at University of Michigan Health since digitization began, more than 200,000 studies containing 5.6 million image sequences.

The system also reads patients' medical histories and understands why doctors ordered the imaging. This comprehensive approach mirrors how experienced radiologists actually work, integrating multiple information sources to reach accurate conclusions.

The Ripple Effect

The timing couldn't be better. Millions of MRI studies happen globally each year, but the number of qualified neuroradiologists hasn't kept pace. This shortage creates dangerous delays and increases the risk of diagnostic errors.

Rural hospitals face the steepest challenges, often lacking neuroimaging specialists entirely. Patients in these areas can wait longest for results, even when facing urgent conditions. Prima offers these underserved communities access to expert-level diagnostic capabilities instantly.

Dr. Vikas Gulani, chair of radiology at University of Michigan Health, sees this as a solution with "tremendous, scalable potential." Whether patients receive scans at overwhelmed major hospitals or resource-limited rural facilities, Prima could democratize access to fast, accurate diagnoses.

The research team plans to integrate even more patient data from electronic medical records, which should improve accuracy further. They're also exploring adapting Prima for other imaging types including mammograms, chest X-rays, and ultrasounds.

Hollon describes Prima as "ChatGPT for medical imaging," a broad AI assistant rather than a narrow tool for single tasks. It's designed to work as a co-pilot for doctors, supporting rather than replacing human expertise.

The technology still requires further evaluation before widespread adoption, but early results published in Nature Biomedical Engineering show remarkable promise for transforming how quickly and accurately patients receive life-changing diagnoses.