Adaptive Sampling and Reconstruction for Sparse Magnetic Resonance Imaging

An adaptive acquisition sequence for Sparse 2D Magnetic Resonance Imaging (MRI) is presented. The method combines random sampling of Cartesian trajectories with an adaptive 2D acquisition of radial projections. It is based on the evaluation of the information content of a small percentage of the k-space data collected randomly to identify radial blades of k-space coefficients having maximum information content. The information content of each direction is evaluated by calculating an entropy function defined on the power spectrum of the projections. The images are obtained by using a non linear reconstruction strategy, based on the homotopic L0-norm, on the sparse data. The method is tested on MRI images and it is also compared to the weighted Compressed Sensing. Some results are reported and discussed.