Space-dependent effects of motion on the standard deviation of fMRI signals: a simulation study

摘要

In fMRI, any fluctuation of signal intensity, not recognized as a result of a specific task, is treated as noise. One source for "noise" is subject motion. Normally, motion effects are reduced by applying realignment. We investigate how apt a realignment procedure is in removing motion-related effects by comparing the distribution of the normalized standard deviation of each voxel, NSD (=standard deviation/mean), before and after realignment. We use data acquired from a simulation program developed previously. This program covers the effects of spin history, slice profiles, and motion in (f)MRI. We simulated inter-volume motion perpendicular to the slices of a digital, artificial head phantom, with and without spin history. In all cases, fluctuations in signal intensity were reduced after standard realignment. This effect was strongest when no spin history was present. Hence, spin history has a marked effect on the "noise". The spatial distribution of NSD showed similarities to the structures of the brain. This indicates that (residual) motion effects were largest at the transitions between tissues. Possibly, the spatial distribution of NSD can provide a (independent) tool to investigate brain structure. Furthermore, the method presented can be used to qualitatively compare different "noise" reduction steps in fMRI data analysis.