Visualization and Exploration of Shape Variance for the Analysis of Cohort Study Data

Abstract

In epidemiological studies, a group of people with common characteristics or experiences (a cohort) is studied including an analysis of socio-demographic as well as biological factors and correlations indicating per subject the absolute risk of getting a disease. Longitudinal studies are carried out over years or even decades comprising up to thousands of individuals. More recently, such studies include the acquisition of image data such as MRI to answer crucial epidemiological questions. For instance, how is the shape of an anatomical structure related to behavioral or clinical factors, e.g., liver shape related to drinking habits and obesity? We propose a pipeline for shape variance analysis in cohort study data which comprises the definition of groups of individuals and control groups based on socio-demographic and biological factors or attributes derived from the image data as well as the visualization of intra-group shape variance and inter-group shape difference. We employ different shape variance models and investigate the applicability of the pipeline for liver and spine related epidemiological research.