Longitudinal Spatio-Spectral Analysis of Resting-State fMRI

• Activity Number: 590
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Hakmook Kang*+
• Companies: Vanderbilt University
• Keywords: resting-state fMRI ; spatio-temporal model ; longitudinal analysis

Abstract

In this talk, we propose a spatio-spectral mixed-effects model to explore patterns of functional brain network estimated by resting-state function MRI (fMRI). Typical resting-state fMRI analysis ignores spatial correlations within each region of interest, which can lead to inference error by underestimating the standard error of a functional connectivity measure. However, our spatio-spectral model can properly take the intrinsic spatial and temporal correlations in resting-state fMRI data. Moreover, employing the two-stage approach commonly used in longitudinal analysis allows us to explore changes in patterns of functional network over time. We compare our model to the typical approach to assess the advantage of the spatio-spectral mixed-effects model via simulated data. Also, it is applied to a resting-state fMRI data set to assess the effect of insulin on longitudinal changes in the limbic circuit of obese subjects with Type 2 diabetes.