Thanks for explanation, Mark! On Wednesday 17 August 2005 14:49, Mark Jenkinson wrote: > Hi Martin, > > There are indeed two methods for unwarping. I certainly > recommend the pixel-shift for unwarping, but the > current distributed version uses Fourier for forward warping > (of the magnitude fieldmap image in order to create a > distorted target image for registration purposes). The reason > for this is that it avoids the necessity to invert the warp > field, as taking a warped image and unwarping it is opposite > to warping an unwarped image, and what space the warp I am getting warped myself. :) So, if my field map and image are both in distorted/undistorted space and I want to unwarp/warp the image to undistorted/distorted space, the Fourier shift is my friend. But if the field map and image are in opposite spaces, I can use pixel shift. Do I have it right? > is in matters, so I either needed an inverse of the warp field > (in general difficult to impossible) or a different technique > for forward warping. I chose the latter and used a Fourier method. > In summary, both methods are used in fugue, but just got with the > defaults as they automatically choose the best method. > > The second question is all about units. OK. Now I understand. > My fieldmap is in radians per second, obtained by > taking phase diff in radians and dividing by asym_time > in seconds. My pixel_shift is then in voxels (dimensionless) > which is obtained by taking the fieldmap in rad/s and > multiplying by dwell time in seconds, and dividing by 2*pi > (radians). > In contrast, the field in Jezzard et al is in Telsa, so he takes phase_diff > (radians) and divides this by 2*pi (radians) and gyr_ratio > (Hz/T) and asym_time (seconds). Since Hz = 1/seconds > then the result is a quantity in units of Tesla. By taking my > fieldmaps in rad/s I just avoid the need for using the > gyromagnetic ratio, as I never need to convert back to Tesla. > However, you can certainly do it with mine in which case you > need to insert the gyromagnetic ratio again. > > All the best, > Mark All the best to you as well. Martin