Describir: Exploiting sparsity and low-rank structure for the recovery of multi-slice breast MRIs with reduced sampling error.