Abstract
The aim of compressed sensing (CS) is to reconstruct signals and images from significantly fewer measurements than usual based on the sparsity in measured signals or images. In general, magnetic resonance images themselves are not sparse; however, they can be transformed to sparse signals by applying an appropriate function. We propose an improved CS method using a FREBAS transform as the sparsifying function rather than a wavelet transform or finite difference transform. Compressed sensing is an iterative reconstruction technique, so it is more computationally intensive than traditional inverse Fourier reconstruction. One barrier to the routine adoption of CS MRI is the delay between data acquisition and the reconstruction of acceptable images. In the present study, we attempted to accelerate CS reconstruction by employing a graphics processing unit (GPU) for general-purpose computing. The results showed that the use of the GPU dramatically accelerated CS reconstruction. A single image could be reconstructed in 1.3 s, indicating that the speed was increased by a factor of 9.
