Compressed sensing techniques for accelerated magnetic resonance imaging

Abstract

Magnetic resonance imaging has seen a growing interest in the recent years due to its non-invasive and non-ionizing nature. However, imaging speed remains a major concern. Recently, compressed sensing theory has opened new doors for accelerated imaging applications. This dissertation studies compressed sensing based reconstruction strategies for accelerated magnetic resonance imaging, speci cally for angiography and multiple-acquisition methods. For magnetic resonance angiography, we propose a novel approach that improves scan time e ciency while suppressing background signals. In this study, we attain high-contrast angiograms from undersampled data by utilizing a two-stage reconstruction strategy. Simulations and in vivo experiments demonstrate that the developed strategy is able to relax trade-o s between image contrast and scan e ciency without compromising vessel depiction. For multiple-acquisition balanced steady state free precession imaging, we develop a framework that jointly reconstructs undersampled phasecycled images. This approach is able to improve banding artifact suppression while maintaining scan e ciency. Results show that the proposed method is able to attain high-quality reconstructions even at high acceleration factors. Overall, the ndings presented in this thesis indicate that compressed sensing reconstructions represent a promising future for rapid magnetic resonance imaging. Consequently, compressed sensing reconstruction techniques hold a great potential to change the time-consuming clinical imaging practices.