used a hybrid radial-Cartesian 3D EPI trajectory with a golden ratio based
angle update to perform retrospective motion correction in severely motion
corrupted fMRI data. Motion estimates were based on high temporal resolution image
timeseries and and k-space based estimates. The calculated rotations and
translations were corrected in k-space prior to the final
reconstruction, allowing the correction of both inter- and intra-volume motion
artifacts. This approach is self-navigated, requires no additional hardware and
is suitable for correction in fast fMRI acquisition.