The world is being quietly rearranged by people who write very long documents.


The title they went with Schema-Adaptive Tabular Representation Learning with LLMs for Generalizable Multimodal Clinical Reasoning Noisy translates that to

AI can now diagnose dementia from patient records, even if the records look different


Researchers have built an AI that can understand patient data from different hospital systems without needing to be retrained. This means the AI can diagnose dementia using varied electronic health records and MRI scans, even if it has never seen that specific record format before.
For decades, medical AI has struggled because every hospital uses slightly different ways to record patient information. This paper shows a way around that problem, allowing AI to make sense of messy, real-world data. It means AI tools could be deployed much more widely in healthcare, without needing expensive custom engineering for each new hospital system.
Watch for this AI method to be tested in real-time clinical settings, not just on historical data, to see if it performs as well with live patient intake.

If you insist
Read the original →