A frontier machine intelligence for human neuroimaging
Neuro-VL outperformed general-purpose and specialized medical models on neuroimaging tasks in the authors’ evaluations.
Trained on nearly one million studies from routine clinical care, the model interprets complete examinations and answers questions about their findings. Tests across internal and international datasets covered diagnosing, locating and describing abnormalities, alongside generating radiology reports. It also compared serial scans to assess disease progression. The authors report that it resisted misleading clinical assumptions more reliably than leading OpenAI and Google models.
HF Daily Papers' note
Trained on nearly one million studies from routine clinical care, the model interprets complete examinations and answers questions about their findings. Tests across internal and international datasets covered diagnosing, locating and describing abnormalities, alongside generating radiology reports. It also compared serial scans to assess disease progression. The authors report that it resisted misleading clinical assumptions more reliably than leading OpenAI and Google models.
HF Daily Papers' note
score 5