A New Combination Of GRAPPA And Ompressed Sensing For Accelerating Magnetic Resonance Imaging

It is important to accelerate scanning speed in Magnetic Resonance Imaging(MRI)and the scanning time is related to the field of view(FOV)in traditional MRI.However,if we shorten FOV to reduce the scanning time,the final medical image is overlapped and cannot be used in the diagnosis directly.Parallel Imaging(PI)utilizes the spatial information of phased array coils and the specific reconstruction algorithm to get the de-overlapped medical image.SENSE(Sensitivity Encoding for Fast MRI)and GRAPPA(Generalized Auto-calibrating Partially Parallel Acquisition)have been mostly used approaches in PI.Compressed sensing(CS)is another approach to shorten scanning time.Based on the sparsity of MRI,CS under-samples the k-space randomly and uses the optimized algorithm to reconstruct the final image.As CS and PI are based different theories and use different strategy to under-sample k-space,some researchers have combined CS and PI to speed up the scanning time further and reconstructed medical images.In this paper,a new approach was proposed to combine CS and PI to reconstruct the under-sampled k-space.I designed a creative sampling strategy to under-sample the k-space,reconstructed partial k-space by GRAPPA,filled all k-space by CS for each coil,and got the final reconstruction.For sampling strategy,a local sampling pattern with fixed interval along the phase-encoding direction was designed to sample local k-space and this was used to under-sample the whole k-space randomly,which was helpful for GRAPPA to reconstruct as many as un-sampled k-space data.I also updated cost function based the reconstruction k-space data by GRAPPA in order to improve CS performance.In-vivo MRI data with different module were used to prove that the proposed approach can get better medical images compared to other state-of-art approaches at same amount of k-space data.