Neuro-VL vision-language model trained on one million studies

Researchers introduce Neuro-VL, a vision-language model trained on nearly one million neuroimaging studies designed to interpret complete clinical examinations and engage in natural language conversations about its findings. In evaluations across internal and international datasets, the model outperformed frontier general-purpose and specialized medical models in diagnosing, localizing, and describing abnormalities, as well as generating comprehensive radiology reports. Additionally, Neuro-VL successfully assessed disease progression through serial examinations and proved more reliable at resisting misleading clinical assumptions than leading systems from OpenAI and Google. Open article →