Highly Accelerated Multi-shot EPI based Diffusion MRI Using SMS and Joint k-q Under-sampling Enabled Using Deep Learned Manifold Priors

We propose a new acceleration and reconstruction method for highly accelerated multi-shot dMRI. The acceleration makes use of multi-band excitation in combination with joint k-q undersampling. We develop new iterative reconstruction with deep learned q-space manifold priors to enable the recovery for the DWIs from 12-fold under-sampled data. The reconstruction error is shown to be less than 3%. The proposed method enables utilization of multi-shot EPI trajectories for diffusion microstructure and connectivity studies requiring high q-space coverage, without prolonging scan time.