Parsimonious Spatiotemporal Modeling Through Matrix Decompositions

• Activity Number: 441
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Russell Shinohara*+ and Simon Vandekar and Ted Satterthwaite
• Companies: University of Pennsylvania and University of Pennsylvania and University of Pennsylvania
• Keywords: imaging ; biostatistics ; spatiotemporal

Abstract

To study healthy trajectories of brain cortical thickness, blood perfusion, and morphological development in adolescents, we consider spatially smooth and temporally smooth models of imaging outcomes across ages, genders, and other demographic variables. We study these patterns in novel frameworks for parsimonious joint spatiotemporal modeling based on principal component and canonical correlation analyses. We demonstrate the generalizability of these techniques through the multimodal analysis of hundreds of subjects studied as part of the Philadelphia Neurodevelopmental Cohort, a large longitudinal study of healthy adolescents and subjects who develop psychiatric disorders.