Digital Tuning to Suppress Ghost Artifacts in EPI by Minimization of Total Variation

Introduction Echo-planar imaging (EPI) is widely used in functional MRI studies, most often in single-shot form. Interleaved multi-segment EPI becomes desirable when higher spatial resolution and less geometric distortion are crucial. However, interleaved EPI is more susceptible to ghost artifacts. The first major cause of ghost artifacts is even-odd echo data discrepancy, mainly due to gradient eddy current. It corresponds to the familiar N/2 ghost seen in the single-shot EPI. Several post processing methods exist for the N/2 ghost correction, by performing phase correction based on information that is derived either from reference scans [1-3] or from the ghosted image itself [4-6]. The second major cause of ghost artifacts is inter-segment data discrepancy due to hardware instability, subject motion, and field fluctuation due to physiological motion. The ghosts are similar to but spatially shifted from the even-odd echo ghosts. Navigators can be added in EPI sequences to measure the inter-segment discrepancy. The information is then used in image reconstruction for ghost suppression [7]. We propose a digital tuning approach to suppress both types of ghost artifacts altogether. The method uses minimization of total variation (MiniTV) which has been recently shown capable of reconstructing a compressible digital image from a small number of random linear measurements [8].