FAST REGULARIZED RECONSTRUCTION OF NON-UNIFORMLY SUBSAMPLED PARTIAL-FOURIER PARALLEL MRI DATA

We present an adaptation of our previous fast, regularized parallel MRI reconstruction approach (LSQR-hybrid) to encompass the reconstruction of partial-Fourier data. Reconstructions of partial-Fourier data require a constraint on the signal phase variation in the reconstructed image. Here, we employ a two-parameter Tikhonov regularization formulation to constrain the real and imaginary components separately. A variation of the LSQR-hybrid algorithm is used to rapidly reconstruct a set of images across the two-dimensional regularization parameter search space. The "best" image available is chosen as the one with the least system error using the most regularization. Results are demonstrated with subsampled parallel MR data acquired on a 1.5 T MRI system.