Golden-Angle Radial Sparse Parallel MR Image Reconstruction Using SC-GROG Followed by Iterative Soft Thresholding

Abstract

Golden-angle radial sparse parallel (GRASP) magnetic resonance imaging (MRI) is a recent MR image reconstruction technique which integrates parallel imaging, compressed sensing and golden-angle radial scheme to reconstruct the dynamic contrast-enhanced MRI (DCE-MRI) data. Conventionally, GRASP exploits non-uniform fast Fourier transform to grid and de-grid the golden-angle radial data and employs nonlinear conjugate gradient method to recover the unaliased images. GRASP performs gridding and de-gridding operations of golden-angle radial data in every iteration which increases the computational complexity of the conventional GRASP and takes a long image reconstruction time. In this paper, self-calibrated GRAPPA operator gridding (SC-GROG) followed by iterative soft thresholding (IST) is proposed for faster GRASP reconstruction of the golden-angle radial DCE-MRI data. In the proposed method, firstly SC-GROG maps the undersampled golden-angle radial data to a Cartesian grid and then reconstructs the solution image using the IST technique. The proposed method does not require gridding and de-gridding in each iteration; therefore, it is computationally less expensive as compared to the conventional GRASP reconstruction approach. The proposed method is tested for undersampled DCE golden-angle radial liver perfusion data (at acceleration factors 11.8, 19.1 and 30.9). The reconstruction results are assessed visually as well as using mean square error, line profiles and reconstruction time. The reconstruction results are compared with the conventional GRASP reconstruction. The results show that the proposed method provides better quality reconstruction results in terms of reconstruction time and spatio-temporal resolution than the conventional GRASP approach.