Assessing dynamic effects on a Bayesian matrix-variate dynamic linear model: An application to task-based fMRI data analysis

Abstract

A modeling procedure for task-based functional magnetic resonance imaging (fMRI) data analysis using a Bayesian matrix-variate dynamic linear model (MVDLM) is presented. With this type of model, less complex than the more traditional temporal-spatial models, it is possible to take into account the temporal and, at least locally, the spatial structures that are usually present in this type of data. Thus, every voxel in the brain image is jointly modeled with its nearest neighbors, as defined by a Euclidean metric. MVDLM's have been widely used in applications where the interest lies in performing predictions and/or analysis of covariance structures among time series. However, in this context, the interest is rather to assess the dynamic effects related to voxel activation. In order to do so, two algorithms are developed and an already-existing one is adapted to simulate on-line trajectories related to the state parameter. With those curves or simulated trajectories, a Monte Carlo evidence for voxel activation is computed. Through two practical examples of auditory- and motor-cortex activations and two different types of assessments using resting-state and simulated fMRI data, it is shown that the proposed method can be viewed by practitioners as a reliable tool for task-based fMRI data analysis. Despite the assessments and applications being illustrated just for a single-subject analysis, a description is given of how general group analysis can be implemented, exemplified with a group analysis for 21 subjects.