Stability effects on results of diffusion tensor imaging analysis by reduction of the number of gradient directions due to motion artifacts: an application to presymptomatic Huntington’s disease

In diffusion tensor imaging (DTI), an improvement in the signal-to-noise ratio (SNR) of the fractional anisotropy (FA) maps can be obtained when the number of recorded gradient directions (GD) is increased. Vice versa, elimination of motion-corrupted or noisy GD leads to a more accurate characterization of the diffusion tensor. We previously suggest a slice-wise method for artifact detection in FA maps. This current study applies this approach to a cohort of 18 premanifest Huntington’s disease (pHD) subjects and 23 controls. By 2-D voxelwise statistical comparison of original FA-maps and FA-maps with a reduced number of GD, the effect of eliminating GD that were affected by motion was demonstrated. We present an evaluation metric that allows to test if the computed FA-maps (with a reduced number of GD) still reflect a “true” FA-map, as defined by simulations in the control sample. Furthermore, we investigated if omitting data volumes affected by motion in the pHD cohort could lead to an increased SNR in the resulting FA-maps. A high agreement between original FA maps (with all GD) and corrected FA maps (i.e. without GD corrupted by motion) were observed even for numbers of eliminated GD up to 13. Even in one data set in which 46 GD had to be eliminated, the results showed a moderate agreement.