Compressed sensing dynamic MRI reconstruction using multi-scale 3D convolutional sparse coding with elastic net regularization

In this paper, we introduce a fast alternating method to reconstruct highly undersampled dynamic MRI data by using multi-scale 3D convolutional sparse coding. The proposed method concurrently builds a multi-scale 3D dictionary as the MRI reconstruction proceeds by using a variant of the alternating direction method of multipliers algorithm. In addition, elastic net regularization is also applied to take the advantages of both lasso and ridge regularizations for promoting better sparse approximation to the measurement data. We demonstrate that the reconstruction quality of our method is higher than the state-of-the-art dictionary-based MRI reconstruction algorithms.