Prediction of Amyloidosis from Neuropsychological and MRI Data for Cost Effective Inclusion of Pre-symptomatic Subjects in Clinical Trials

We propose a method for selecting pre-symptomatic subjects likely to have amyloid plaques in the brain, based on the automatic analysis of neuropsychological and MRI data and using a cross-validated binary classifier. By avoiding systematic PET scan for selecting subjects, it reduces the cost of forming cohorts of subjects with amyloid plaques for clinical trials, by scanning fewer subjects but increasing the number of recruitments. We validate our method on three cohorts of subjects at different disease stages, and compare the performance of six classifiers, showing that the random forest yields good results more consistently, and that the method generalizes well when tested on an unseen data set.