AI tools boost dementia diagnosis accuracy and flag medical data bias

University of Florida researchers have created an artificial‑intelligence system called Automated Imaging Differentiation for Dementia (AIDD) that analyzes specialized MRI scans to distinguish Alzheimer’s disease from dementia with Lewy bodies. Tested on 519 brain scans, the tool achieved near‑perfect accuracy, addressing a high rate of misdiagnosis that can lead to inappropriate treatment.

Separately, scientists at Johns Hopkins University together with the U.S. Food and Drug Administration have developed G‑AUDIT, a method for scanning large medical‑AI training datasets for hidden biases. The system can reveal unintended cues—such as reliance on non‑clinical features like mascara—that may skew model predictions, helping to improve the reliability and safety of AI‑driven clinical tools.