FIGURE 9.

Illustration of different 2D k‐space undersampling schemas for two‐fold acceleration (R = 2) and four‐fold acceleration (R = 4). GRAPPA/SENSE can reconstruct coherently sampled k‐space data (e.g., entire rows or columns are not acquired), while CAIPIRINHA can reconstruct even coherently undersampled data with any other patter and benefits from controlled aliasing. All of them use information about sensitivity profiles of the individual receive channels to remove spatial aliasing either in the image (e.g., SENSE) or k‐space domain (e.g ., GRAPPA, CAIPIRINHA). This translates into better reconstruction with lower g‐factors (e.g ., less lipid aliasing and higher SNR). In contrast, CS can reconstruct incoherently (e.g., random‐like) undersampled k‐space data without knowledge about the coil receive profiles of the individual receive channels. Courtesy of Lukas Hingerl