 |
 |
Thanks for your detailed response Jesper - Much appreciated! Best, Danny Kim ________________________________ From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Jesper Andersson <[log in to unmask]> Sent: Tuesday, February 6, 2018 3:41 AM To: [log in to unmask] Subject: Re: [FSL] EDDY --data_is_shelled Dear Danny, I had a question about the --data_is_shelled flag in eddy from 5.0.10. My DTI set consists of 10 b=300, 40 b=700 and 90 b=2000 and after initiating eddy, I've got the "data is not shelled" error. I was thinking of using the data_is_shelled flag to bypass the heuristic shell check and process onward. What are some things that I should be checking for once eddy is completed? Is it mainly for good alignment / EC correction? What are some things that could fail if shell check was not done on a data? I think that in your case you will be fine. The "story" behind the "data is not shelled" error is that people started using eddy for DSI data, which it will currently not work for. It didn't seem to make any difference that the documentation said that it was only working for shelled data. So I came up with a heuristic inside eddy to try and work out if the data that was passed in was shelled or not. It turned out that heuristic was a little too strict for protocols that include a small number of directions with low b-values (something that we rarely do). So I then added the--data_is_shelled flag, such that when someone explicitly sets that they are "guaranteeing" that data is shelled. Your protocol looks very reasonable. 10 directions is sufficient for eddy to make good predictions about a shell with a b-value as low as 300 (which has very little angular contrast). In general though, what I usually do is to run a movie of my data after correction. I would run it at two different speeds as you tend to see slightly different things when the frames go past fast compared to when they go past quickly. Since you have a wide range of b-values you will need to manipulate the scaling so that you see the different shells, alternatively you can use "select_dwi_vols" to divide up your data on the different shells and run the movies separately. Look for anything that looks like movement and/or distortions that differ between the different frames. The latter would be best seen along edges that are perpendicular to the PE-direction. Eddy will optionally also correct for outliers (slices with signal dropout) and intra-volume movement. If your data has those problems and you attempt to correct them with eddy you should also look for slices with dropout (easiest to see as dark stripes in a sagittal/coronal view) or telltale zig-zag patterns (also easiest seen in a sagittal/coronal view). Jesper Many thanks for your input! Danny Kim