It is necessary to perform correction of
eddy-current and motion artefacts before analysing DW-MR data, but none of the
commonly used correction techniques have been evaluated quantitatively using direct measures of correspondence. Here we
apply a recently proposed simulation framework to evaluate four correction techniques.
We found the three techniques that register to a b=0 image (Eddy_correct, ACID,
ExploreDTI) perform worse than a technique that registers to predicted DWIs
(eddy). Furthermore, we found that one
of the most commonly used methods for registration to b=0, eddy_correct,
performs significantly worse than the other methods considered.