Ghost artifact removal using a parallel imaging approach.

Abstract

Parallel imaging techniques, which use several receive coils simultaneously, have been shown to enable a significant scan time reduction by subsampling k-space. Nevertheless, the data acquired with multiple coils in parallel exhibit some redundancy if the number of receive coils exceeds the subsampling factor. This redundancy leads to an overdetermination of the reconstruction problem, which is generally used to optimize the signal-to-noise ratio (SNR). However, it can yield further information about the quality of the reconstructed image, and can thus be used to identify and correct image artifacts. While some known approaches try to solve this problem in k-space, this study addresses it in the spatial domain and uses a modified SENSE reconstruction to reduce or completely remove ghost-type artifacts arising from processes such as motion or flow during data acquisition. Phantom and in vivo studies show significant improvements in image quality after correction, and serve as a basis for the discussion of the performance and limitations of this new approach.