The B-matrix must be rotated when correcting for subject motion in DTI data

To estimate diffusion tensor MRI (DTI) measures, such as fractional anisotropy and fiber orientation, reliably, a large number of diffusion-encoded images is needed, preferably cardiac gated to reduce pulsation artifacts. However, the concomitant longer acquisition times increase the chances of subject motion adversely affecting the estimation of these measures. While correcting for motion artifacts improves the accuracy of DTI, an often overlooked step in realigning the images is to reorient the B-matrix so that orientational information is correctly preserved. To the best of our knowledge, most research groups and software packages currently omit this reorientation step. Given the recent explosion of DTI applications including, for example, neurosurgical planning (in which errors can have drastic consequences), it is important to investigate the impact of neglecting to perform the B-matrix reorientation. In this work, a systematic study to investigate the effect of neglecting to reorient the B-matrix on DTI data during motion correction is presented. The consequences for diffusion fiber tractography are also discussed.