next up previous
Next: Introduction

Cluster Analysis Revisited

FMRIB Technical Report TR00DF1

David E. Flitney and Mark Jenkinson

Oxford Centre for Functional Magnetic Resonance Imaging of the Brain (FMRIB),
Department of Clinical Neurology, University of Oxford, John Radcliffe Hospital,
Headley Way, Headington, Oxford, UK


To complete the analysis of any FMRI experiment it is necessary to compensate for the multiple comparisons problem inherent in making a large number of simultaneous statistical tests. One method for accomplishing this is to employ cluster analysis to bolster one's confidence in any particular statistical result. The following review combines the well known works in this field to form the basis of an implementation of the Gaussian random field theory applied to the analysis of FMRI statistic images.

keywords: FMRI, Gaussian Random Fields, smoothness estimation, multiple comparisons, statistical significance.

David Flitney 2001-11-29