Evaluating the Influence of Spatial Resampling for Motion Correction in Resting-State Functional MRI.

Head motion is one of major concerns in current resting-state functional MRI studies. Image realignment including motion estimation and spatial resampling is often applied to achieve rigid-body motion correction. While the accurate estimation of motion parameters has been addressed in most studies, spatial resampling could also produce spurious variance, and lead to unexpected errors on the amplitude of BOLD signal. In this study, two simulation experiments were designed to characterize these variance related with spatial resampling. The fluctuation amplitude of spurious variance was first investigated using a set of simulated images with estimated motion parameters from a real dataset, and regions more likely to be affected by spatial resampling were found around the peripheral regions of the cortex. The other simulation was designed with three typical types of motion parameters to represent different extents of motion. It was found that areas with significant correlation between spurious variance and head motion scattered all over the brain and varied greatly from one motion type to another. In the last part of this study, four popular motion regression approaches were applied respectively and their performance in reducing spurious variance was compared. Among them, Friston 24 and Voxel-specific 12 model (Friston et al., 1996), were found to have the best outcomes. By separating related effects during fMRI analysis, this study provides a better understanding of the characteristics of spatial resampling and the interpretation of motion-BOLD relationship.