Scientists at the University of Michigan have introduced the AI model Prima, which can analyze brain MRI scans in seconds and diagnose a wide range of neurological conditions. Over the course of a year, the system was tested on more than 30,000 MRI exams, and in some scenarios its accuracy reached 97.5%. The model recognizes more than 50 conditions and outperformed other modern AI systems in some tasks.
Prima’s key feature is not only detecting abnormalities, but also sorting cases by urgency. For example, if the scans show signs of stroke or bleeding, the system can automatically alert doctors that immediate intervention is needed. It also suggests which specialist is best to bring in for the patient — a neurologist or a neurosurgeon.
The technology could significantly reduce the wait for results, lower the number of urgent cases that go unnoticed, and make high-quality diagnosis more accessible where there is a shortage of specialist doctors. The researchers plan to expand the system using electronic medical record data and, in the future, adapt a similar approach for mammography, X-ray, and ultrasound.
