 |
 |
Dear Rodrigo, > > # --> > Information about my DWIs and eddy parameters > > I am working with conventional DWIs data (5 b0s, 30 at 700 b/mm^2 -correct me if this bvecs units aren't correct - ). All the volumes images were acquired in the A>>P phase encoding direction, echo spacing is 0.8 ms and EPI factor is 128. > > Using fsl 5.0.7, I executed the following eddy command: > > eddy --imain=<SUBJ>.nii.gz --mask=<SUBJ_BINARY_MASK> --index=index.txt --acqp=acqp.txt --bvecs=<SUBJ>.bvecs --bvals=<SUBJ>.bvals --repol --out=<SUBJ>_eddied.nii.gz > > My index.txt is: > <terminal>$ cat ../params/index.txt > 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 > > My acqp.txt is: > <terminal>$ cat ../params/acqp.txt > 0.0 -1.0 0.0 0.102 > > > The output looks like this: > > <SUBJ>_eddied.eddy_outlier_free_data.nii.gz > <SUBJ>_eddied.eddy_outlier_report > <SUBJ>_eddied.eddy_parameters > <SUBJ>_eddied.nii.gz > > > #--> > My questions: > > 1) Based on a single acquisition parameter and as explained in this website (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/topup ), I believe the topup step is not necessary. Am I correct? I also tried it but the image seem more distorted than the original one. That is true. If all your images are acquired in the same way you cannot run topup. However, this also means that your data will not be corrected for susceptibility induced distortions. > > 2) After using the --repol argument, should I use the output "<SUBJ>_eddied.eddy_outlier_free_data.nii.gz" for further processing steps instead of the <SUBJ>_eddied.nii.gz? You should use <SUBJ>_eddied.nii.gz. The "<SUBJ>_eddied.eddy_outlier_free_data.nii.gz” is only for the case where someone might want to use eddy to correct for outliers, but some other software for correcting for distortions and movement. > > 3) Many of these images have an empty <SUBJ>_eddied.eddy_outlier_report, which I believe it means that no outliers exist, am I correct? That is correct. > > 4) If outliers exist, then the <SUBJ>_eddied.eddy_outlier_report looks like this: > Slice 47 in scan 14 is an outlier 0.00235832 standard deviations off > Slice 51 in scan 14 is an outlier -0.0346646 standard deviations off > Slice 53 in scan 14 is an outlier 7.89147e-05 standard deviations off > > 4a)*I believe "scan" 14, refers to the specific DWI volume. In fslview indexing starts at 0. So is it the same for the outlier_report? (I am trying to understand what volume had the correction whether vol 13 in fslview (if indexed in the outlier report starts at 1) or 14 (if indexing starts at 0) This looks very strange. It should only reject slices as outliers if they are <-4 std off. > > 4b) What axis does "Slice" refer to? My guess will be the z-axis but I am not very sure about it :/ It refers to the z-axis. > > 5) Should I try running the same command in fsl 5.0.9, would I see any significant changes in the output? Yes, you definitely should. We never intended to release outlier rejection for 5.0.7. It was work in progress, and the functionality for using it was hidden such that you would not see the options if you typed the command. People started using it anyway, so we disabled it for 5.0.8. You should only use it with 5.0.9. or later. > > 6) Also, after correction for slice outliers, I would like to also implement RESTORE when reconstructing the tensor model. It is not very clear whether the --repol options will have a similar effect than the RESTORE outlier correction algorithm. If so, this step could be avoided. I tried Camino RESTORE and Dipy RESTORE but estimating the sigma (in Camino) and fitting the tensor (in Dipy) seems to add more complications. Any advice will be greatly appreciate it. The outlier correction in eddy is _instead of_ things like RESTORE. You should use one or the other, not both. Jesper > > > > > Thanks in advance, > > Rodrigo