A Linear Model with a Spatiotemporal Covariance Structure for the Analysis of Longitudinal Imaging Data

• Activity Number: 655
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Brandon George*+ and Inmaculada Aban
• Companies: University of Alabama at Birmingham and University of Alabama at Birmingham
• Keywords: spatiotemporal ; longitudinal ; imaging ; covariance ; linear model ; spatial

Abstract

Longitudinal imaging studies allow great insight into how the structure and function of a subject's internal anatomy changes over time. Unfortunately, the analysis of longitudinal imaging data is complicated by inherent spatial and temporal correlation from the repeated measures and the outcomes of interest being observed at multiple points in a patient's body. The issue is compounded by a small sample size characteristic of imaging studies, particularly ones employing MRI, which prevents the use of unstructured covariance functions. Inspired by a longitudinal cardiac imaging study looking at the remodeling in the left ventricle of mitral regurgitation patients, we propose a linear model with a separable parametric spatiotemporal covariance structure,. Simulation studies were run to examine the effects of covariance structure misspecification on statistical inference as well as the use of information criteria to choose a correlation structure. An example analysis is also considered as a demonstration of how the model can be applied, how the results can be interpreted, and what computational challenges (such as missing data) may arise in fitting the model.