Bayesian Multi-Object Regression with Applications in Multi-Modal Imaging Data

• Activity Number: 379 - Single and Multi-Object Regression and Clustering with Applications in Neuro-Imaging Data
• Type: Topic Contributed
• Date/Time: Wednesday, August 10, 2022 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Rajarshi Guhaniyogi and Aaron Scheffler and Rene Gutierrez*
• Companies: Texas A & M University and UCSF and Texas A & M
• Keywords: Bayesian; Object Oriented Regression; Multi-Object; Brain Imaging; sMRI; Connectivity Network

Abstract

Clinical researchers collect multiple images from separate modalities to investigate questions of human health. Viewing the collection of images as objects, the integration of multi-object data can produce a sum of information greater than the individual parts. This article focuses on a multi-modal imaging application where structural/anatomical information from grey matter (GM) and brain connectivity information in the form of a brain connectome network from structural magnetic resonance imaging are available for multiple subjects. We develop a regression model to predict a language score, used as a measure to assess the degree of primary progressive aphasia (PPA), from these image predictors and to identify brain regions (ROIs) significantly related to language ability. We build a Bayesian regression framework exploiting network information of the brain connectome while leveraging linkages among connectome network and anatomical information from GM to draw inference on significant ROIs and offer predictive inference on the language score. This Bayesian framework allows precise characterization of the uncertainty n prediction and identification of significant brain ROIs.