Improving image quality in low snr parallel acquisition using a weighted least squares GRAPPA reconstruction

We analyze the performance of a Weighted Least Squares (WLS) GRAPPA calibration for improving the g-factor of GRAPPA reconstruction from low Signal-to-Noise-Ratio (SNR) acquisition. Our method is particularly useful when the SNR degradation is influenced by noise correlation across channels. When applied with regularization, this form of GRAPPA reconstruction is found to be superior compared to other recent variants of GRAPPA such as coefficient penalized GRAPPA. We demonstrate this feature quantitatively using GRAPPA derived g-factors.