Accelerated first pass cardiac perfusion MRI using improved k #x2212; t SLR

摘要

Routinely trade-offs between the spatio-temporal resolution, volume coverage and SNR are done in first pass cardiac perfusion MRI due to the restricted imaging acquisition window (usually of the order of 300 to 400 msec per heart beat). In this paper, we demonstrate the use a low rank and sparse reconstruction scheme (k − t SLR) in obtaining highly accelerated first pass perfusion MR images and hence aim to reduce the above mentioned trade-offs. We introduce non-convex spectral norms and use a spatio-temporal total variation norm in recovering the dynamic signal matrix. We introduce an augmented Lagrangian optimization scheme in the context of matrix recovery to speed up the convergence of the algorithm. Extensive validations on in-vivo data are done to demonstrate the performance improvement of the proposed frame work.