RSPIRIT: Robust self-consistent parallel imaging reconstruction based on generalized Lasso

In this paper, we propose a novel approach called robust iterative self-consistent parallel imaging reconstruction (RSPIRiT) in parallel magnetic resonance imaging (pMRI). Different from the smooth Tikhonov regularization used in SPIRiT, our model utilizes generalized Lasso to fix calibration errors in the reconstruction process. It results in a non-smooth optimization problem, which we introduce a new primal-dual pMRI algorithm to solve. We conduct extensive experiments to demonstrate the effectiveness of our approach, compared to state-of-the-art methods.