A dual echo approach to removing motion artefacts in fMRI time series

In fMRI, subject motion can severely affect data quality. This is a particular problem when movement is correlated with the experimental paradigm as this potentially causes artefactual activation. A method is presented that uses linear regression, to utilise the time course of an image acquired at very short echo time (TE) as a voxel-wise regressor for a second image in the same echo train, that is acquired with high BOLD sensitivity. The value of this approach is demonstrated using task-locked motion combined with visual stimulation. Results obtained at both 1.5 and 3 T show improvements in functional activation maps for individual subjects. The method is straightforward to implement, does not require extra scan time and can easily be embedded in a multi-echo acquisition framework.