Thomas Maullin-Sapey

Spatial Confidence Regions and Uncertainty in fMRI

Functional Magnetic Resonance Imaging (fMRI) is a popular medical imaging technique used to capture the physiological processes (function) of the human brain. Central to fMRI analysis is the modelling of random `clusters', contigouous regions within the brain over which significant activity is observed in response to a task. Although cluster `inference' is a mainstay of fMRI analysis, conventional tools can only provide p-values and no notion of spatial uncertainty of cluster location. To combat this, in recent years, methods have been developed which create spatial Confidence Regions (CRs) for ‘excursion sets’ (collections of clusters), which provide, with a desired confidence, outer and inner bounds on regions of brain activation.

In this talk, I shall first provide a brief introduction to fMRI analysis covering the common mass-univariate analyses employed for voxel-wise inference. Following this, I shall introduce the notion of a confidence region, providing a short summary of the existing literature and explaining the practicalities of computing CRs via a wild-T bootstrap. Finally, I will discuss my own contributions to this area, including confidence regions for conjunction analysis, piecewise processes and individual components of an excursion set.