Performance analysis of CS-MRI reconstruction using particle swarm optimization for different sampling patterns

Abstract

This paper provides an analysis of reconstruction of Magnetic Resonance Imaging (MRI) from Compressive Sensed k-space data using Particle Swarm Optimization (PSO) for different sampling patterns. Compressive Sensing (CS) gives a framework for the recovery of signal or image from far fewer measurements than that are required according to Nyquist sampling theorem. k-space data of MRI are the individual Fourier coefficients. Compressive Sensing of k-space data is done by undersampling using radial and Cartesian sampling patterns to get fewer samples much less than its actual dimension. The basic use of PSO is to resolve the optimization problems and it is designed to simulate processes in natural systems that are required for evolution. PSO is based on the social manners of the collections of population in nature such as animal herds or bird flocking or schooling of fish. Compressive sensing can make accurate reconstructions by using PSO from a small subset of k-space, rather than an entire k-space grid. Experimental outcomes showed that dense sampling must be done at the centre of k-space to get better reconstruction of CS-MRI.