Generation of optimized, cost-sensitive feature sets to enable quick screening of mild dementia

Michael J. Kleiman, Elan Barenholtz, James E. Galvin, for the Alzheimer’s Disease Neuroimaging Initiative

Published in the Journal of Alzheimer’s Disease

Link to PDF
Link to preprint: DOI
Link to supplemental code: DOI

Two-thirds of dementia cases are undetected globally. A contributing factor is that our current methods for detecting dementia in clinical settings target moderate-to-severe impairment, resulting in mild impairment often being missed or mistaken for normal aging. To address this, we utilized feature selection techniques guided by random forests to identify the fewest number of cognitive measurements with the highest sensitivity ratings.