Motion and distortion correction of Diffusion Weighted images
Motion and distortion correction of the data is very important in diffusion-weighted MR imaging. When scanning non-cooperative subjects or non-sedated newborns, resulting images are significantly distorted by motion. When scanning at b-values higher than 1500 s/mm², distortion from eddy currents becomes severe if no compensatory scheme is applied. In our research, we use and evaluate different state of the art motion and distortion correction approaches. Various techniques are tested on diffusion-weighted data acquired at b-values in the range [0,3000] s/mm², with and without suppression of eddy currents, for subjects which lay still and for patients which move a lot. Used correction methods are based on affine and non-linear transformation models. To evaluate results of correction, different indices and methods are used: mutual information, principal component analysis, chi-squared residuals with respect to the diffusion model prediction, violations of the diffusion model constraints etc. Our results show that appropriate distortion correction methods depend on the scanning settings and properties of the original data.