Anisotropic spatial 3D Matern Priors for fMRI
Title: Anisotropic spatial 3D Matern Priors for fMRI
SNIC Project: LiU-compute-2020-29
Project Type: LiU Compute
Principal Investigator: Per Sidén <>
Affiliation: Linköpings universitet
Duration: 2020-08-27 – 2021-09-01
Classification: 10106


Functional magnetic resonance imaging (fMRI) data are brain imaging data that are correlated both in space and time. Most existing analyses of fMRI data analyse each voxel separately without taking the spatial dependency between voxels into account. Bayesian methods are popular for fMRI, but are extremely time-consuming for spatial data. In this work we develop Bayesian methods based on anisotropic extensions of the Matern covariance function for analyzing large 3D datasets that take advantage of recent advances in sparse numerical algebra, including multigrid methods. This extends our previous work (Sidén et al., 2017), which uses a simpler spatial model. The current project evaluates these methods and compares to previous approaches on a number of fMRI datasets. References: Sidén, P., Eklund, A., Bolin, D., and Villani, M. (2017). Fast Bayesian whole-brain fMRI analysis with spatial 3D priors. NeuroImage, 146:211–225.