Back to the stories

Neuro-VL uses nearly one million clinical neuroimaging studies to support broad diagnostic and longitudinal interpretation

Score 9.0,

AI applications in healthcare

A new AI for brain scans can look beyond one exam and compare scans over time.

That is the idea behind Neuro-VL, a model described in a research preprint. It can interpret complete imaging examinations and answer questions about what it finds.

A scan is a snapshot. Comparing it with earlier scans can help show whether a condition is changing. That is a more clinically useful ambition than simply asking AI to name what appears in one image.

The researchers trained Neuro-VL on nearly one million neuroimaging studies from routine clinical care. Think of comparing two photographs of the same place: one shows what is there, while the pair can reveal what has changed.

In the authors’ tests, Neuro-VL performed better than general-purpose and medical AI systems at identifying abnormalities and producing radiology reports. For now, that is evidence from a research preprint, not proof that the model improves care in hospitals.

The next test is whether it can help clinicians follow changes reliably in real-world practice. Can AI become useful not just at reading a scan, but at helping make sense of a patient’s story over time?

AI Above 9 score: 9.0