Fig. 5.

The iRESTORE algorithm provides partial remedy for pulsation artifacts.

(A) Color coded maps of the orientation of the principal eigenvector are displayed for the triggered (top middle) and non-triggered acquisitions obtained using OLLS tensor fitting (bottom left), as well as for the non-triggered acquisition after employing the iRESTORE algorithm (bottom right). The iRESTORE algorithm corrects some of the artifacts that are clearly present in the non-triggered data set without employing the algorithm. Still the triggered data set provides the most reliable results, Paired fractional anisotropy difference images are also presented among the 3 methods. The bottom middle figure depicts the improvement due to the iRESTORE algorithm. The white arrows indicate the acquisition methods compared. A coregistered slice of the Cincinnati Children's Hospital Medical Center infant brain template (https://irc.cchmc.org/software/infant.php) is shown in inset as reference for anatomical localization. For comparison of the principal eigenvectors' orientation, see Supplementary Fig. 1. (B) Mean (and range) of absolute residuals of OLLS tensor fit errors for the non-triggered, triggered and non-triggered with iRESTORE experiments. The y axis scale is arbitrary but identical for all 3 plots while the x axis represents the 15 diffusion encoding directions. Note that mean residual is largest for the 3rd diffusion direction, which is along z gradient axis. The bars are color coded according to the diffusion encoding direction using the standard DTI color scheme (see inset). See Results for quantitative analysis of the error introduced in FA values and the principal tensor directions within the regions of interest outlined on the color coded maps.