Many strokes are "cryptogenic," the fancy medical term for we don't know why. A new study using A.I. of brain MRI images points to the ability to accurately determine if atrial fibrillation was causal. https://www.thelancet.com/journals/eclinm/article/PIIS2589-5370(25)00050-1/fulltext
Comments
what if AI diagnostics turns out to have similar results, like, for illustration:
* 51% have major issues
* 19% with factual errors/faults
* 13% indications don’t exist in patient?
I‘m not saying humans do better: they often don’t.
Blind trust won’t help.
https://www.bbc.com/mediacentre/2025/bbc-research-shows-issues-with-answers-from-artificial-intelligence-assistants