Abstract
Alzheimer’s disease (AD) is a neurodegenerative disease and the leading cause of dementia (50-70% of cases). Despite worldwide efforts, there is no progress in developing a cure for AD and dementia. Machine learning, hand in hand with magnetic resonance imaging (MRI), come to the aid of disease diagnostics. In the scope of AD, many efforts have been dedicated to the automated detection of mild-cognitive impairment and dementia. In our research, instead we focus on the prediction of AD in its preclinical stage using machine learning classification. Another key innovation is that we will work with a longitudinal pipeline. In addition to classification, the project focuses on detecting the most relevant imaging voxels for classification, that is, to help us locate where AD-specific structural brain changes occur. We have improved classification performance i n comparison with results obtained with cross-sectional datasets in previous studies and we have identified possible regions of interest based on feature scores obtained from feature selection.
