Improving fMRI activation detection sensitivity using intervoxel coherence mapping

Spatiotemporal data coherence widely exists in fMRI data and can be potentially used to increase brain activation detection. To assess this possibility, a two-stage fMRI data analysis method was presented in this work. Standard voxelwise general linear model (GLM) was used to first exclude voxels with low correlation to the functional design, and an intervoxel coherence map (ICM) was then calculated voxel by voxel to locate brain regions, which were most consistently activated during the functional experiment. Population inference about the detected effects was provided through random effect analysis or permutation testing. Evaluations using synthetic data and scene-encoding memory task fMRI data both showed enhanced activation detection performance for the proposed method when compared with standard GLM. © 2012 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 22, 33–36, 2012