Diffusion imaging techniques such as DTI and HARDI are difficult to implement in infants because of their sensitivity to subject motion. A short acquisition time is generally preferred, at the expense of spatial resolution and signal-to-noise ratio. Before estimating the local diffusion model, most pre-processing techniques only register diffusion-weighted volumes, without correcting for intra-slice artifacts due to motion or technical problems. Here, we propose a fully automated strategy, which takes advantage of a high orientation number and is based on spherical-harmonics decomposition of the diffusion signal.

The correction strategy is based on two successive steps: 1) automated detection and resampling of corrupted slices; 2) correction for eddy current distortions and realignment of misregistered volumes. It was tested on DTI data from adults and non-sedated healthy infants.

The methodology was validated through simulated motions applied to an uncorrupted dataset and through comparisons with an unmoved reference. Second, we showed that the correction applied to an infant group enabled to improve DTI maps and to increase the reliability of DTI quantification in the immature cortico-spinal tract.

This automated strategy performed reliably on DTI datasets and can be applied to spherical single- and multiple-shell diffusion imaging.