A Multi-modal approach to predicting Alzheimer's disease conversion and progression via machine and deep learning

Alzheimer's Disease (AD) or Dementia of Alzheimer's type (DAT) is a complex neurodegenerative disorder that accounts for 60% to 80% of all dementia cases. Because there is currently no cure available for AD, there is a substantial interest in developing new methods that can accurately detect those at risk at an early stage of the disease before the symptomatic onset. Multiple factors contribute to the development and progression of DAT, but the magnitude of each factor's impact on the disease is unknown. In this thesis, we use powerful machine and deep learning tools to extract and analyze information from multiple data modalities such as MRI, genetic, Cerebrospinal Fluid (CSF), cognitive tests, and demographic information to 1) develop novel biomarkers to predict future conversion to DAT (predict if the subjects will develop DAT); and 2) predict the time to conversion to DAT (predict when the subjects develop DAT) in the future.