An Improved MRI Reconstruction Algorithm Combining Partial Fourier and Parallel Imaging

The relatively long duration of magnetic resonance imaging (MRI) limited its application in many important clinical and research scenarios, and many methods have been proposed to reduce the scan time of MRI, such as partial Fourier imaging(PF), parallel imaging and compressed sensing(CS). In order to obtain higher acceleration factors, several reconstruction methods combining half Fourier and parallel imaging have been proposed in recent years. In these methods, PF_POCSENSE is a very effective method, which apply the phase constraining to reconstruct partial Fourier and parallel sampling data. Compared with partial Fourier imaging and parallel imaging, the SNR of reconstructed image is obviously improved. However, the traditional PF_POCSENSE method converges slowly and spends longer reconstruction time, because the initial image has less information. In this paper, an iterative partial Fourier parallel imaging (PFPI) method is proposed. Firstly, the amplitude of low resolution image is used as the initial guess image. Secondly, both phase constraint and region of interest (ROI) constraint are added into the iterative process to improve image quality. Moreover, the data convex sets projection operators are optimized. More priori information is included in reconstruction process and the data projection operations also are improved, so the number of iterations is greatly reduced, and the reconstruction speed is obviously improved. The reconstruction results of phantom and in vivo data showed that the proposed algorithm can reconstruct the image with higher SNR and faster reconstruction speed than the traditional algorithm under the same acceleration factor.