Researchers at the University of Michigan in the United States have developed an AI model, dubbed Prima, which is capable of swiftly reading and analyzing brain magnetic resonance imaging (MRI) scans within seconds, to rapidly evaluate the urgency of a patient's required treatment. This model boasts a remarkable detection accuracy rate of up to 97.5% for neurological diseases, a feat documented in the esteemed journal Nature Biomedical Engineering. Prima adeptly integrates imaging data, textual information, and the patient's medical history in real-time, facilitating a holistic judgment akin to that of an experienced radiologist.

To train this advanced model, the research team leveraged an extensive dataset comprising over 200,000 MRI examinations and 5.6 million image sequences amassed by the University of Michigan Health System. This robust training regimen has significantly enhanced the model's versatility and adaptability across a spectrum of tasks. In diagnostic tasks involving images related to more than 50 major neurological diseases, Prima outshone several state-of-the-art AI models currently in use. Moreover, the system is equipped to automatically assess the priority of cases, issuing timely alerts to medical personnel in critical situations such as brain hemorrhage and stroke, thereby ensuring prompt intervention.