Project by Guan Yu
Alzheimer's disease (AD) and its prodromal stage, i.e., mild cognitive impairment is very important for early treatment. By collaborating with the Biomedical Research Imaging Center in UNC-Chapel Hill, we develop a new graph-guided multi-task learning method using structural magnetic resonance imaging (MRI) to predict disease status and clinical scores jointly.
Our proposed method incorporates both the graphical structure information among MRI features and the intrinsic correlation information among different response variables. Besides predicting the disease status, our proposed method could identity the most discriminative brain regions for the diagnosis of AD as shown in the multi-slice view of brain MRI.