Research on A Multi-slice MRI Reconstruction Algorithm Based on Deep Learning

Because the high information-dominated correlation between two adjacent slices can compensate for under-sampled, multi-slice MRI reconstruction are more likely than single-slice MRI reconstruction to reconstruct more accurate images by interpolating the missing K-space data into the adjacent slices and estimating the missing sample of the middle slice based on the information from its adjacent slices. Traditional multi-slice MRI reconstruction relies on coil sensitivity map and structured K-space sampling. In this paper, the DISN model engages in network training through deep learning and carries out multi-slice reconstruction under different conditions through controlling the variables. The experimental results of multi-slice and single-slice reconstruction by DISN comparatively show that the effect of multi-slice network reconstruction is always better than that of single-slice network reconstruction.