Motion-robust diffusion tensor acquisition at routine 3T magnetic resonance imaging

We compared different acquisition and reconstruction methods in phantom and human studies in the clinical setting to validate our hypothesis that optimizing the k-space acquisition and reconstruction method could decrease motion artifacts.

Diffusion tensor images of a water phantom were obtained with three table displacement magnitudes: 1 mm, 2 mm, and 3 mm. Images were reconstructed using homodyne and zero-fill reconstruction. Overscanning in 8- and 16-k(y) lines was tested. We performed visual assessment of the artifacts using reconstructed coronal images and analyzed them with Wilcoxon signed-ranks test both for phantom and human studies. Also, fractional anisotropy (FA) changes between acquisition methods were compared.

Artifacts due to smaller displacement (1 and 2 mm) were significantly reduced in 16-k(y) overscan with zero filling. The Wilcoxon signed-ranks test showed significant differences (P < 0.031 for reconstruction methods and P < 0.016 for overscanning methods). FA changes were statistically significant (P < 0.037; Student's t-test). The Wilcoxon signed-ranks test showed significant reductions (P < 0.005) in the human study.

Motion-induced artifacts can be reduced by optimizing acquisition and reconstruction methods. The techniques described in this study offer an effective method for robust estimation of diffusion tensor in the presence of motion-related artifactual data points.