Current standard sensitivity-encoded parallel imaging (SENSE) utilizes a fully sampled low-resolution reference scan to estimate the coil sensitivities. This reference scan adds scan time and may introduce misregistration artifacts. The purpose of this study was to investigate the feasibility of estimating the coil sensitivities for spiral SENSE directly from an undersampled k-space center. The limited spatial frequencies of the coil sensitivities, and the undersampling beyond the Nyquist radius cause image artifacts. A point spread function (PSF) analysis and experiments on both phantoms and humans identified an optimal radius for the k-space center by minimizing these image artifacts. The preliminary data indicate that self-calibrated SENSE is as accurate as standard SENSE, which uses a fully sampled reference scan.
