Feasibility Study of a Machine Learning Approach to Predict Dementia Progression

We conducted a feasibility study of machine-learning to predict progression of cognitive impairment to Alzheimer´s disease (AD) among individuals enrolled in the Alzheimer´s Disease Neuroimaging Initiative (ADNI). Our approach uses diverse participant information including genetic, imaging, biomarker, and neuropsychological data to predict transition to dementia in three clinical scenarios: short-term prediction (half or one year) based on a single assessment (simulating a "new patient" visit), short-term prediction based on information from two time points (simulating a "follow up" visit), and long-term (multiple years) prediction (simulating ongoing follow-up with repeated opportunities for assessment).