 |
 |
Dear all, Briefly, inhomogenities of the static magnetic field result in distorted coding of the k-space, which transforms into geometric displacements in image space. Based on the fieldmap a voxel displacement image is generated, which is then used to "unwarp" the images. Assuming a default acquisition scheme and looking from the side voxels seem indeed to be moved "forward" as illustrated in Jim's picture (it's by no means a simple linear shift though). There might have been some modifications since SPM2, but with later SPM versions the outcome should be very similar, as the algorithm tries to account for the same physical effect. Realign & unwarp includes a reslice step. When the voxel displacement image is applied to the data to correct for the geometric distortions, undistorted = "unwarped" data is interpolated. If you go with that preprocessing step for some subjects, but not for others, this additional interpolation step would be another confound. In principle you could still go with Realign & unwarp without a fieldmap specified, the necessary parameters are estimated from the EPI series then. This is not identical to a aquired fieldmap though. Of course you could interpret this a between-subject confound and add a variable to your second-level models. But I agree with Donald and Jim, preprocessing should be identical. Best, Helmut