Isolation and minimization of effects of motion on fMRI using multiple reference images

Cost functions implemented in most fMRI motion correction algorithms often fail to account explicitly for the signal variability between images not due to motion, but coming from brain activation. In this paper, we explore the performance of a newly proposed motion correction method, named motion-corrected independent component analysis (MCICA), which allows for brain activation to be explicitly modeled using multiple reference images, and compare MCICA with the conventional square of difference-based measures like LS-SPM and LS-AIR. We demonstrate that in simulations, MCICA is more robust to the addition of simulated activation. With actual data from a motor fMRI experiment, the time course of the derived task-related ICA component become significantly more correlated with the underlying behavioral task, and the associated activation map is more clustered in the primary motor and supplementary motor cortices without spurious activation at the brain edge.