Evidence grade: SPECULATION
computed from the encoded evidence by the Atlas grading rule.
Generalist medical AI and pathology/retina foundation models show strong transfer; promise for label-efficient neuroimaging
GPT-4V and medical VLMs show inconsistent/poor accuracy on real image-based radiology questions; hallucination and calibration unquantified in neuro-oncology
Risk of premature clinical trust; opportunity if rigorously measured
This is an educational research resource from Resonant Labs, not clinical advice. Evidence grades and controversies are summarized from the literature and may change as the field evolves.