Dynamic Gaussian Graphical Models to Study Time-Varying Clinical Symptom and Imaging Networks

• Activity Number: 155 - Section on Statistics in Imaging Student Paper Award Winners
• Type: Topic-Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Erin McDonnell* and Shanghong Xie and Karen Marder and Yuanjia Wang
• Companies: Columbia University and Columbia Unviersity and Columbia University and Columbia University
• Keywords: Gaussian graphical models; L0 penalty; Revival process; Adaptive lasso; Structural MRI; Huntington's disease

Abstract

Our goal is to examine how the complex interrelationships between clinical symptoms and brain imaging biomarkers change over time leading up to diagnosis of a disease in subjects with a known genetic risk of disease. We propose a time-dependent Gaussian graphical model that ensures smoothness across time-specific networks to examine the trajectories of interactions between markers aligned at the time of disease onset. Specifically, we anchor subjects relative to the time of diagnosis (anchoring time) as in a revival process, and we estimate networks at each time point of interest relative to the anchoring time. We apply kernel weights to borrow information across observations that are close to the time of interest. Adaptive lasso weights encourage temporal smoothness in edge strength, while a novel elastic fuse L0 penalty removes spurious edges and encourages temporal smoothness in network structure. Our approach can handle practical complications such as unbalanced visit times. We conduct simulation studies to compare our approach with existing methods. We then apply our method to identify symptom and imaging network changes that precede diagnosis of manifest Huntington’s disease.