Undersampled MRI reconstruction using edge-weighted ι1 norm minimization

MRI, a widely used analytical tool for medical diagnosis, is burdened by slow data acquisition. An effective way to speed up MRI is to undersample k-space. However, undersampling often violates the Nyquist-Shannon sampling theorem, resulting in aliasing artifacts in reconstructed MR images. Some artifacts blur edges, which usually contain significant information for pathological diagnosis, such as vascular disease and tumor border detection. To better recover the edge information, we propose an edge-weighted iterative thresholding (EWIT) algorithm for undersampled MRI reconstruction. EWIT is implemented on the base of the iterative thresholding (IT) algorithm [1,2], a fast algorithm solving the 1 l norm minimization for compressed sensing (CS) MRI [3,4]. Reconstruction results indicate that EWIT yields more precise reconstructions than IT, especially for the edge recovery.