How can we model the brain and its PET images

Outline:

Positron Emission Tomography is a powerful brain imaging modality which produces enormous data sets. A single scan from a single subject can produce a quarter of a million time series. In order to analyse this data it is typical to either treat the small volume elements (voxels) within these images as being completely independent of one another (a poor modelling assumption) or to aggregate large numbers of voxels within regions of interest -- both losing spatial resolution and imposing an assumption of homogeneity over large parts of the brain volume. This project will investigate modelling approaches which balance the competition between computational efficiency and realistic modelling.