Compressed Sensing and Beyond

Magnetic Resonance Imaging (MRI) and Spectroscopy (MRS) have become well-established tools in basic cardiovascular research to non-invasively provide multi-parametric characterization of hearts in small animal models of human disease. Technical challenges and long scan times frequently preclude a comprehensive, multi-parametric investigation in the same animal, increasing noise, and ultimately the number of animals required to conduct a pre-clinical study. Dedicated acceleration techniques that do not compromise diagnostic accuracy are urgently needed to fully realize the potential of these powerful techniques. While hardware-based solutions (i.e., parallel imaging (PI) techniques) typically provide only two to threefold accelerations in the data acquisition speed, Compressed Sensing (CS) is an alternative approach, which potentially allows for higher scan time reductions. Importantly, it does not rely on specific hardware requirements for the data acquisition, and can in principle be applied on any (pre-) clinical MR optimisation system. CS utilises the fact that the majority of medical images are compressible (i.e., sparse) in some transform domain. The reconstruction uses a non-linear optimisation algorithm, exploiting both the undersampled, incoherent data acquisition and the sparsity model. This chapter explains the basics of CS in the context of MRI, reviews the role of CS in pre-clinical/clinical cardiac MR, and provides examples for the successful applications in cardiac phenotyping of mouse and rat hearts.