Interventional MRI with sparse sampling: an application of compressed sensing

INTRODUCTION In interventional MRI (I-MRI), a sequence of MR images is reconstructed in order to guide a diagnostic or therapeutic procedure where an invasive device is inserted in the body. The need for near-real-time image updates places two distinct constraints on I-MRI reconstruction: 1) high frame rate (several frames per second, depending on the application [2,3]); 2) causal reconstruction of the image sequence (as opposed to other dynamic MRI applications, where the complete image sequence can be recovered after all the data are collected [1]). Several methods have been proposed, which take advantage of the temporal correlations in I-MRI to reduce k-space coverage, thus allowing a higher frame rate [4]. We present a compressed sensing (CS) method, which exploits the redundancy present in many I-MRI acquisitions for reduced-encoding image reconstruction.