 |
 |
Hi Jonathan, I wish I could offer some help with this, but we abandoned that pipeline within a few weeks of my post. We simply modified our protocol to collect 1 additional B0 image with the opposite phase encoding as the rest of our sequence (which is 10 B0 images and 60 DWI). This adds something like 30 seconds at most on a 3T MAGNETOM. In exchange for that 30 seconds, you gain the ability to use TOPUP. In my experience with both healthy subjects, tumor patients, and AVM patients, estimating the susceptibility field using TOPUP and feeding the results to Eddy works far, far, far better than using the Siemens fieldmaps. That said, if you are working with already collected data and/or cannot change your sequence going forward, I would defer to Jesper and the other experts for your problem. I never used the field_mat option because the fieldmap I was giving Eddy was already in the space of the diffusion data (from having previously run FLIRT). So I assumed using field_mat would be redundant. Further, because the problem was unsolved back when I emailed, we settled for using eddy_correct in place of eddy. Shortly after that we changed our protocol as I described above. Good luck! -Ben Ben Chernoff Ph.D. Student, Concepts, Actions, and Objects Lab Rochester Center for Brain Imaging, Room 2B231 University of Rochester On Aug 31, 2017, at 2:13 PM, Jonathan Repple <[log in to unmask]<mailto:[log in to unmask]>> wrote: Dear Ben, Dear Jesper, we are running into the same kind of problem with our DTI data (same issue: only Siemens fieldmap, no reverse phase encoding). We get the same error message after a while running eddy (segmentation error). A) Ben proposed using the fieldmap separately after running eddy without any fieldmap correction (see below). Would you recommend this strategy? Ben's proposal: 1. Smooth the (prepared with fsl_prepare_fieldmap) fieldmap with fugue, using -s 4. Call this fieldmap_fugue 2. Use fugue to warp the magnitude image (brain extracted and eroded) based on the smoothed field map (--loadfmap=fieldmap_fugue) I use 0.0075 for dwell time because our protocol says “Echo spacing: 0.75ms.” Call this warped magnitude image FMM_bet_fugue 3. Use flirt (-dof 6) to generate a conversion matrix from fieldmap space to diffusion space. I flirt “in" FMM_bet_fugue and “ref” the brain extracted B0 image 4. Use flirt to apply this conversion matrix to fieldmap_fugue (from step 1). Call this fieldmap_fugue_dif 5. Use fugue to unwarp the eddy-corrected diffusion brain using --loadfmap=fieldmap_fugue_dif. B) Also, I noticed that in Ben's original code, he did not report to use the field_mat option in eddy. When we try to incorporate the .mat-file created in Ben's first FLIRT-step (flirt -in <FMM_BET_ero> -ref <NoDiffBet> -omat <fieldmap2dif.mat>) in our eddy calculation, we get an error message that the file does not contain a rigid transformation. The matfile itsself looks okay: 1.017922091 0.008996027587 -0.006016609064 -3.750376717 -0.007675637977 1.045219252 -0.01051661839 -4.681491907 0.001407249022 0.01309824726 1.028790877 -4.956612221 0 0 0 1 Thanks for your help. A little more detailed information on what to do exactly with the output of the fsl_create_fieldmap command to incorporate that file in eddy on the eddy-userguide page, would be really helpful! If you like me to upload the files, I am happy to do so! Best wishes, Jonathan