Sensitivity and performance time in MRI dephasing artifact reduction methods

Although shimming can improve static field inhomogeneity, local field imperfections induced by tissue susceptibility differences cannot be completely corrected and can cause substantial signal loss in gradient echo images through intravoxel dephasing. Dephasing increases with voxel size so that one simple method of reducing the effect is to use thin slices. Signal-to-noise ratio (SNR) can then be increased by averaging over the subslices to form the final, thick slice. We call this method subslice averaging or SSAVE. Alternatively, a range of different amplitude slice select rephase gradients can be used to compensate for different susceptibility induced gradient offsets. The final image can then be formed by combining individual images in a variety of ways: summation, summation of the squares of the images, forming the maximum intensity projection of the image set, and Fourier transformation followed by summation. We show here that, contrary to previous claims, the theoretical sensitivity (i.e., SNR divided by the square root of the imaging time) of all these alternative methods is very similar. However, performance time (i.e., minimum-imaging time) of the simplest method, SSAVE, is much shorter than that of alternatives. This is confirmed experimentally on phantoms and anesthetized mice. Magn Reson Med 45:470-476, 2001.