Distortion correction of DKI data: affine approach

Publication Type:

Conference Abstract


10th Belgian Day on Biomedical Engineering – joint meeting with IEEE EMBS Benelux Chapter (2011)


medical imaging


In this work, the performance of FMRIB’s Linear Registration Tool (FLIRT) is evaluated with respect to motion and distortion correction of diffusion kurtosis imaging (DKI) data. FLIRT is tested on two DKI datasets heavily distorted by motion and eddy currents artefacts. Six different criteria are used for evaluation. Our results show that distortion correction using FLIRT substantially improves the original data. However, being an affine registration tool, FLIRT cannot cope with the amount of non-linear distortion present in DKI datasets. Hence, we conclude that, for correction of DKI data, non-linear registration techniques are more suitable.