Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 471 - Contemporary Statistical Methods for Imaging Data Analysis
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Statistics in Imaging
Abstract #312698
Title: Unveil the Intrinsic Connectivity Network with Bayesian Dynamic Latent Factor Model
Author(s): Meini Tang* and Chee-Ming Ting and Hernando Ombao
Companies: King Abdullah University of Science and Technology and King Abdullah Univ. of Sci. and Tech (KAUST) and King Abdullah Univ. of Science and Technology (KAUST)
Keywords: Intrinsic connectivity network; Dynamic functional connectivity; Bayesian factor models; Stochastic volatility
Abstract:

Intrinsic connectivity network is a specific type of functional brain networks that is present during resting state or elicited by stimuli. Indeed, some studies demonstrated that while some stimuli do not have an impact on intrinsic connectivity, others actually either suppress, activate, or moderate it. However, previous studies about the intrinsic connectivity networks either only used the task-independent data, or ignored the time-varying nature of brain networks. We propose a Bayesian dynamic latent factor model to unveil the intrinsic connectivity using both resting-state fMRI data and task-related fMRI data. The low-rank structure of the intrinsic network is described by a time-invariant factor loading matrix, and thus each factor captures a latent network. A factor variance can be interpreted as the strength of the corresponding latent network, which is time-varying and can be affected by some stimuli. The dynamics of the logarithm of a factor variance can be interpreted as a series of continuous-valued hidden brain states and modeled with an autoregressive process. With the proposed model, we can unveil and compare state-related intrinsic connectivity subnetworks.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program