An Efficacious MRI Sparse Recovery Method Based on Differential Under-Sampling and k-Space Interpolation

A proposed method that reduces the lengthy scan time characterizing Magnetic Resonance Imaging (MRI) is presented in this paper. The method also improves the robustness to noise and reconstruction artifacts associated with conventional MRI. It employs a differentially structured k-space under-sampling technique to reduce the imaging time. The under-sampled k-space is then interpolated using a bicubic method in order to obtain an estimate of the part of k-space that is not acquired in the under-sampling step. The interpolation improves the resilience of the proposed procedure to noise and reconstruction artifacts. The method exploits the sparsity of MR images in the wavelet transform domain to reconstruct the images from the interpolated k-space using a greedy Compressive Sampling (CS) method. The Peak Signal to Noise Ratio (PSNR) and the Structural SIMilarity (SSIM) objective quality metrics are used to evaluate the performance of the proposed method in comparison to other reported CS based MRI methods. Computer simulation experimental results reveal that the proposed method exhibits better robustness to noise by a 1.6 dB PSNR improvement and an imaging acceleration of at least 9% while maintaining image quality.