|
Abstract:
|
Accurate prediction of conversion to Alzheimer's Disease (AD) in subjects with mild cognitive impairment (MCI), who do not yet match criteria for AD, but are at high risk of progression to AD, has received increasing attention as early detection of the disease can dramatically increase the effectiveness of future treatments. To this end, we propose a selection method that effectively identifies subregion of the brain highly associated with the conversion using baseline Alzheimer's Disease Neuroimaging Initiative (ADNI) data. Here, we take functional view point, taking each individual's brain as a 3D functional object. Different biomarkers for AD, including neuroimaging, demographic, genetic and cognitive measures, are known to complement each other for diagnosis and prognosis of the disease, so we integrate these scalar covariates as potential predictors, in addition to the images, into the selection procedure. To determine the best subset of the predictors, a subsampling technique is used. Our results show that a combination of relevant predictors achieves the best prediction with an AUC score larger than 0.88.
|
