A hierarchical spatial Bayesian model for multisubjectfunctional MRI data

There are a rich collection of tools available for making inference for fMRI data, but most are based on the 'Mass Univariate' approach where univariate models are individually fit at each voxel. A critical shortcoming of these methods is that they cannot explicitly model the spatial structure of fMRI signals. For multi-subject fMRI analyses, I argue this is particularly crucial, since even after atlas warping there is considerable spatial variability in activation location over subjects. I will present a hierarchical spatial model for multi-subject fMRI analyses, where latent population- and individual centres fit the anticipated focal signals. The model uses priors for identifiability and full posterior sampling to provide inference on a variety of measures of interest unavailable in a mass-univariate framework, including population prevalence of activation and inter-subject spread of activation about population centres. I show evaluations of the model with simulations and demonstrate it with real data. Time permitting, I will also show how this framework generalizes to other settings, including 'spatial meta-analyses' and lesion modelling in Multiple Sclerosis.