A new AI foundation model named Prima, developed by researchers at the University of Michigan, has demonstrated the ability to read and diagnose brain MRI scans within seconds, according to a 2026 health system-wide study. This model outperformed existing AI technologies in diagnosing major neurological disorders, with a diagnostic accuracy of up to 97.5%. Prima was trained on over 220,000 MRI studies and was tested on nearly 29,500 MRI studies, achieving a mean diagnostic area under the curve of 92% across 52 different conditions. Designed to assist radiologists by providing differential diagnoses and clinical recommendations, Prima integrates patient history to enhance its effectiveness. Future research aims to refine the model further and explore its potential application for various imaging types beyond MRIs. The development addresses the increasing demand for MRI studies, which has posed challenges for healthcare systems, particularly in underserved areas. However, researchers note that the model is still in its early evaluation phase.
Mon, 16 Feb 2026 13:20:20 GMT | European Medical Journal