Rotation Space: Detecting Functional Activation by Searching Over Rotated and Scaled Filters

Introduction The aim of this paper is to extend the concept of scale space and multi-resolution analysis to rotation space, that is, rotating as well as scaling a Gaussian-shaped smoothing filter. Using random field theory, we have derived an accurate P -value for the rotation space maximum that allows us to detect non-isotropic signals of arbitrary scale or rotation in functional PET and fMRI images. The advantage of our method is increased sensitivity at detecting elliptically shaped regions of activation that might be missed by a circular shaped filter. These results are applied to a simple fMRI experiment.