Robert L. Barry1,2, John A. Sexton1,3, John C. Gore1,2
1Vanderbilt University Institute of Imaging Science, Nashville, TN, USA; 2Department of Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, USA; 3Department of Biomedical Engineering, Vanderbilt University, Nashville, TN, USA
Spatial smoothing (SS) is a crucial processing step preceding statistical analyses of blood-oxygenation-level-dependent (BOLD) fMRI data. Although previous works have shown that the optimal SS kernel size varies both between subjects and active regions of interest (ROIs) within subjects, the use of a single kernel size for one or many subjects is virtually ubiquitous. A simple and computationally efficient automated algorithm based on matched filter theory is presented that determines the SS kernel size to maximize t-statistics for a given ROI. The results also emphasize the benefits of locally adaptive SS techniques to optimize BOLD contrast-to-noise in single-subject analyses.