MAGNETIC RESONANCE IN MEDICINE 8, zyxwvu DCBA I zyxwvutsr 10-1 zyxwv EDCBA 15 ( 1 988) Snapshot Head Imaging at 0.5 T Using the Echo Planar Technique R. J. ORDIDGE, R. COXON, A. HOWSEMAN, B. CHAPMAN, R. TURNER, M. STEHLING, AND P. MANSFIELD zyx LKJ Departmenf of zyxwvut GFEDC Physics, University ofNottingham, Nottingham NG7 ZRD, United Kingdom Received April 22, 1988; revised June 17, 1988 The echo planar imaging (EPI) method and related variants of this technique can produce complete two-dimensional images from the data collected in a single expe lasting a fraction of a second. EPI methods are used at 0.5 T to produce snapshot images of the human head with a suatial resolution of less than 2 mm. o 1988 Academic Press. Lnc. There have been several variations of the echo planar imaging (EPI) techn published in recent years zyxwv ONML (1-5). The difference between these methods can be best visualized by using two-dimensional k space (6, 7) to represent the sampling of time data required to produce a complete image. BEST (blipped echo planar singl technique) (2,3) covers k space in regularly spaced trajectories which traverse k space along the zyxwvu JIHGFED x direction. This is illustrated in Fig. 1 a. Reconstruction by a two-dimen- sional Fourier transform provides an undistorted image. The equivalent coverage of k space for reconstruction by a one-dimensional Fou- rier transform is shown in Fig. l b. In both cases alternate lines of data mus reversed in order to preserve a common direction of signal evolution in appa equal sign gradients throughout the experiment. The principal difference between these two experiments is that the sample evolution occurs in a single large gradient, k,, in Fig. la, and two orthogonal gradi- ents, k, and a smaller ky, in Fig. lb. For nonrectangular gradient waveforms, the principles of nonlinear sampling ( 8 ) may be applied to prevent image distortion. However, in the experiment of Fig. 1 b, the gradient waveforms along the x and y axes must be matched in shape for proper nonlinear sampling in both gradien BEST experiment of Fig. la is therefore preferred. However, it has the disadvantag that a two-dimensional Fourier transform is needed rather than a one-dimen Fourier transform, which can be performed more quickly. Fortunately the di between one- and two-dimensional Fourier transformations is small, and the visible effect of applying a one-dimensional Fourier transform to the data co in the BEST experiment is a spatial rotation of the image by an angle tan-' ( A x / N a y ) , where Ax and Ay are the spatial resolutions along the x and y axes and N is the size of the image matrix along the y axis. A further variation of the EPI experiment produces unphased data using a tude Fourier transform. This method is called MBEST ( 9 ) and is illustrated in Fig. 2. The MBEST technique allows symmetric coverage of two-dimensional k space by 0740-3 194/88 $3.00 I10 Copyright 0 1988 by Academic Press, Inc. All rights of reproduction in any form reserved.