TESTING FOR SPACE-TIME SEPARABILITY IN FUNCTIONAL MRI

In functional magnetic resonance imaging (fMRI), a main goal of data analysis is to determine which regions of brain are activated by presented temporal stimuli, equivalently, to build up an activation map from an observed spatiotemporal data. In most studies in fMRI, space and time separability has been implicitly assumed and accepted without a proper justification. Space-time separability allows a substantial amount of simplifications in modeling and analysis for activation map. For the first time in fMRI, we propose a method to build up a detection statistic for testing the separability of space and time with full consideration of spatial and temporal correlations. The developed detection statistic is tested by a simulation and applied to a human dataset.