Non-Rigid Motion Estimation and Spatio-Temporal Realignment in FMRI

Existing approaches to the problem of subject motion artefacts in FMRI data have applied rigid-body registration techniques to what is a non-rigid problem. We propose a model which can account for the non-linear characteristics of movement effects, known to result from the acquisition methods used to form these images. The model also facilitates the proper application of temporal corrections which are needed to compensate for acquisition delays. Results of an implementation based on this model reveal that it is possible to correct for these effects, leading to accurate re-alignment and timing correction.