Bayesian Neuroimaging Feature Selection for Characterizing Alzheimer's Disease

• Activity Number: 360 - Contributed Poster Presentations: Section on Bayesian Statistical Science
• Type: Contributed
• Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Daniel Baer* and Andrew Lawson
• Companies: Medical University of South Carolina and Medical University of South Carolina
• Keywords: High dimensional feature data; Longitudinal outcome data ; Bayesian feature selection model; neuroimaging; Alzheimer's disease

Abstract

There is an ongoing need to characterize the progression of Alzheimer’s disease (AD) in order to prevent its onset in patients. We can accomplish this by selecting regions of the brain via neuroimaging which are most associated with measures of patient cognition over time. However, such neuroimaging data are high-dimensional, and longitudinal measures of patient cognition are correlated over time. As such, we propose and evaluate novel Bayesian multivariate group lasso models using multivariate spike-and-slab prior distributions. In particular, we consider a model which can select groups of features via group-specific probability parameters, and a model which allows for heterogeneity in measurement occasions amongst patients. Preliminary results suggest this class of Bayesian models are promising in terms of gleaning the neurodegenerative process underlying AD and thus providing a means to characterize AD risk.