Super-pixel algorithm and group sparsity regularization method for compressed sensing MR image reconstruction

Exploiting the sparsity of MR signals, Compressed Sensing MR imaging (CS-MRI) is one of the most promising approaches to good quality MR image reconstruction from highly under-sampled k-space data. The group sparse method, which exploits additional sparse representations of the spatial group structure, can increase the overall degrees of sparsity, thereby leading to better reconstruction performance. In this work, an efficient superpixel/group assignment method, simple linear iterative clustering (SLIC), is incorporated to CS-MRI studies. A variable splitting strategy and classic alternating direct method are employed to solve the group sparse problem. This approach, termed Group Sparse reconstructions using Super-Pixel or SP-GS algorithm, was tested on three different types of MR images with different undersampling rates to validate its performance in reconstruction accuracy and computational efficiency. The results indicate that the proposed SP-GS method is capable of achieving significant improvements in reconstruction accuracy and computation efficiency when compared with the state-of-the-art reconstruction methods.