Rapid Magnetic Resonance Imaging Using Undersampled Projection-Onto-Convex-Sets Reconstruction

Scan time reduction is important in clinical magnetic resonance imaging (MRI). Partial Fourier data acquisitions rely on the conjugate symmetry of Hermitian data, allowing for shorter scan times due to fewer phase-encoding steps needed during the MRI data acquisition. Estimation of the missing k-space MR data is then usually accomplished by direct conjugate synthesis (e.g., homodyne reconstruction) or by iterative constraint-based algorithms (e.g., projection-onto-convex sets, POCS). Faster image acquisition has been achieved by combining partial Fourier acquisitions with parallel MRI. In this study, an iterative partial Fourier reconstruction method based on POCS technique is presented for sharply decreasing MRI scan times. It is achieved by asymmetrically undersampling k-space MR data acquired using a multielement MRI receive coil. The results obtained from experimental models (phantoms) and humans shows that the proposed method can rapidly produce high-quality MR images with high acceleration rates.