Particle swarm optimization based multilevel MRI compression using compressive sensing

A multilevel compression method, for magnetic resonance imaging (MRI) images, is presented in this paper. First, the image is segmented into frames of equal size. Then, the sparsity of each frame is computed. Based on the sparsity index value, each frame is compressive sensing (CS) compressed/reconstructed at one level of four. Particle swarm optimization (PSO) is used to optimize the amount of information to be used in the CS reconstruction process, and to optimize the sparsity thresholds, that separate the different compression levels. Two-dimensional sigmoid function is suggested as a fitness function for the PSO. Six MRI images are used to evaluate the performance of the proposed method. The results show considerable gain in both peak signal to noise ratio (PSNR) and compression level (CL), when compared to single level compression, which is commonly considered in the literature.