Machine Learning Improves Atypical Alzheimer's Diagnosis Accuracy

Researchers (Michael Evans et al.) propose a machine-learning approach to distinguish atypical Alzheimer's disease (atAD) from non-AD cognitive impairment using standard clinical testing and MRI, posting an arXiv preprint on Jan 21, 2026. They analyze 1,410 subjects across private, NACC, and ADNI datasets and show adding broad MRI features outperforms hippocampal volume alone. The model raises atAD recall from 52% to 69% (NACC) and 34% to 77% (ADNI), improving diagnostic utility.