An evaluation of the use of magnetic field maps to undistort echo-planar images

When a head is placed in an MRI scanner, differences between the magnetic susceptibility of tissue, bone, and air distort the magnetic field. While some of the resulting inhomogeneity can be corrected by the shimming process, much of it cannot, and this causes distortion (sometimes referred to as geometric distortion) of echo-planar images (EPIs). One strategy for the correction of distortion is to acquire a map of the magnetic field achieved in each subject and then to use this to undistort their EPIs after reconstruction (Jezzard and Balaban, 1995). Here, we present five experiments to evaluate the application of such a strategy on data from a 3-T scanner. We show that after undistortion, the shape of EPIs is more similar to the true shape of the brain, and we investigate the effect of head movement on the efficacy of undistortion. If undistortion was applied first, it was found that less nonlinear warping was required to transform EPIs into a standard space, particularly in the phase-encode direction. We show that if SPM 99 normalization is used to perform a nonlinear warp to standard space, the prior application of undistortion increases the statistical power of group studies with motor and auditory tasks. We show that this increase in power is due to an increase in the overlap of activation of different subjects. Finally, we evaluate where in the brain undistorting EPIs might be expected to have the greatest effect, in terms both of mislocalization of activation and of a reduction in power. Overall, undistorting EPIs using field maps has proved extremely successful, improving the anatomical localization of activation and increasing statistical power. (C) 2002 Elsevier Science (USA).