Learning a preconditioner to accelerate compressed sensing reconstructions in MRI

A Fast CS-Based Reconstruction Model with Total Variation Constraint for MRI Enhancement in K-Space Domain

Due to the fact that Magnetic Resonance Imaging (MRI) is still a relatively slow imaging modality, its application for dynamic imaging is restricted. The total variation is introduced into the CS-based MRI reconstruction model, and three regularization conditions are adopted to ensure that a high-quality reconstructed image is produced. In this paper, a simple yet fast CS-based optimization model for noisy MRI Enhancement is proposed. The alternative direction multiplier method is chosen to optimize the model, and the k-terms power series is applied in order to derive the LogDet function into the augmented Lagrange form. Following this, an approximation of the feature vector is achieved through the iterative process. The quality of the reconstructed image was much better than that of the CS-based MRI image reconstruction algorithm, as shown by experimental results under different noise conditions. The peak signal-to-noise ratio of the reconstructed image was able to be improved anywhere from 5 to 20 percent.