5 Bayesian Treatments of Neuroimaging Data

In this chapter we discuss the application of Bayesian methods to neuroimag-ing data. This includes data from positron emission tomography (PET), functional magnetic resonance imaging (fMRI), electroencephalography (EEG), and magnetoencephalography (MEG). We concentrate on fMRI but the concepts, methodologies, and modeling approaches apply to all modalities. A general issue in the analysis of brain imaging data is the relationship between the neurobiological hypothesis one posits and the statistical models adopted to test that hypothesis. One key distinction is between functional specialization and integration. Briefly, fMRI was originally used to provide functional maps showing which regions are specialized for specific functions, a classic example being the study by Zeki et al. [27] who identified V4 and V5 as being specialized for the processing of color and motion, respectively. More recently, these analyses have been augmented by functional integration studies, which describe how functionally specialized areas interact and how these interactions depend on changes of context. A recent example is the study by Buchel et al. [1] who found that the success with which a subject learned an object-location association task was correlated with the coupling between regions in the dorsal and ventral visual streams. Functionally specialized brain responses are typically characterized using the general linear model (GLM). An fMRI data set comprises a time series of volumetric data. In "mass-univariate" approaches GLMs are fitted to fMRI time series at each voxel (volume element), resulting in a set of voxel-specific parameters. These parameters are then used to form posterior probability maps (PPMs) that characterize regionally specific responses to experimental manipulation. Figure 5.5, for example, shows a PPM highlighting regions that are sensitive to visual motion stimuli. Analyses of functional integration are implemented using multivariate approaches that examine the changes in multiple brain areas induced by experimental manipulation. Although there a number of methods for doing this we