Compressed sensing in parallel magnetic resonance imaging

Abstract

Magnetic resonance imaging (MRI) is a non-invasive method that allows the study of interior structures of matter. MRI is based on magnetizing a sample, manipulating the magnetization, and detecting the magnetic field that the sample produces. Today, MRI is widely used in medical imaging. Reductions in the MRI measurement time are important because they correspond to, e.g. higher throughput of MRI scanners or enhanced spatial resolution of images.

It is well-known that natural images, such as MR images, are compressible. Compressed sensing (CS) is a method that exploits the compressibility of signals in order to measure and reconstruct them efficiently. Moreover, in parallel MRI (pMRI), the magnetic fields produced by the magnetized sample are measured with multiple coils simultaneously. The differing sensitivity profiles of the coils allow a faster rate of information flow compared to single-coil receiving.

In this work, CS and pMRI are combined to achieve shortened measurement times. Theory of CS, pMRI, and their combination are reviewed. Simulations were conducted to demonstrate the performance of the method in idealistic conditions. In addition, experimental measurements were performed to validate the technique in practice. Simulation results indicate that the combination of CS with pMRI reduces the measurement time by 30-50 %. Experimentally, reductions of 10-20 % were obtained.