Image Quality Assessment and the Conditions of Image Reconstruction in the Application of Compressed Sensing to MR Multi-slice Imaging

Abstract

MR multi-slice imaging is a technique which can reduce the total signal acquisition time by acquiring signals slice by slices sequentially. That imaging method is based on the two-dimensional (2D) imaging. However, the application of compressed sensing (CS) to three-dimensional (3D) objects is feasible by introducing an adequate signal acquisition trajectory and regarding a number of sliced images as three-dimensional objects. In this paper, comparison of image quality between multi-slice three-dimensional CS reconstruction and two-dimensional slice by slice CS reconstruction were performed using three MR model images by assuming that the same signal trajectories is used. Simulation experiments showed that the reconstruction error in multi-slice three-dimensional CS reconstruction reduces as the number of slice images increases, or the slice spacing decreases, which results in remarkable improvements in image quality compared to two-dimensional slice by slice CS reconstruction. In addition, 3D-FREBAS transform showed better performances than 3D-wavelet transform when they were used as a sparsifying transform function in CS reconstruction.