Mingwu Jin1, Rajesh Nandy2, Dietmar Cordes1
1Radiology, University of Colorado Denver, Aurora, CO, USA; 2Biostatistics and Psychology, UCLA, Los Angeles, CA, USA
Gaussian spatial smoothing using a fixed FWHM is a common preprocessing step in fMRI data analysis. In this work, we investigate the effects of spatially correlated noise and the size of activation patterns to detect activations with and without spatial smoothing. The detection power is measured using ROC curves on simulated data and activation maps of real 3T fMRI data. Results indicate that spatial smoothing with any fixed FWHM is far less effective when the noise is spatially correlated as in 3T resting-state data and more advanced locally adaptive smoothing kernels should be applied.