Improving image reconstructions for simultaneous multi-slice readout-segmented diffusion MRI data with phase errors

Readout-segmented echo planar imaging (RS-EPI) combined with controlled aliasing simultaneous multi-slice (SMS) acquisition improves spatial resolution of diffusion-weighted images (DWIs) with a scan time that is reduced by a factor proportional to the number of simultaneous slices. Split slice-GRAPPA (SSG) is a commonly used method to de-alias SMS DWIs using kernels trained from baseline b=0 images. When applying SSG to datasets acquired from a RS-EPI sequence, we found that SSG kernels trained from baselines do not de-alias DWIs effectively due to baseline phase errors. To overcome this issue, in this work we propose an iterative approach, termed iterative Split slice-GRAPPA (I-SSG), to train improved kernels using estimated DWIs rather than only the baseline images. Our results from two stroke patients show that the proposed I-SSG algorithm produces consistently better reconstructions in the presence of baseline phase errors. The proposed I-SSG algorithm yields over 50% improvement over the SSG method in Fractional anisotropy (FA) and Mean Diffusion (MD) estimations for slice reduction factors of up to R = 4.