fsl_topup_epi_eddy Issue

[katerendall]

Hi there,

We are running fsl_topup_epi_eddy on our DWI data (108 directions, b-values from 5 - 3000 mm/s^2) with the following code:
fsl_topup_epi_eddy --imain $dwi_PA,$dwi_AP
--mask $dwi_PA_mask,$dwi_AP_mask
--bvals $dwi_PA_bvals,$dwi_AP_bvals
--bvecs $dwi_PA_bvecs,$dwi_AP_bvecs
--numb0 -1 --whichVol 1,2
--acqp $work_dir_dwi/acqparams.txt
--out $epi_out
--config $work_dir_dwi/eddy_config_eddy_cuda.txt
--eddy-cuda

The main problem that we see in our raw data is horizontal striping, likely due to motion, and when we run this command it smooths out some of these artifacts. In other cases, however, it does not fully remove these motion artifacts.

Please see attached 4 photos representing 2 of our cases before and after pre-processing:

Image 1: Case 1 - An example of our pre-processed data up-until fsl_topup_epi_eddy (note the horizontal lines visible in the coronal and sagittal sections)

Image 2: Case 1 - The same data after fsl_topup_epi_eddy (note that the horizontal lines seem to be improved slightly but not fully resolved)

Image 3: Case 2 - An example of our raw data

Image 4: Case 2- The same data after fsl_topup_epi_eddy

These are the parameters in our configuration file: $ eddy_openmp: --verbose --data_is_shelled --niter=8 --fwhm=10,8,4,2,0,0,0,0 --repol --mporder=6 --slspec=slspec.txt --s2v_niter=5 --s2v_lambda=1 --s2v_interp=trilinear

Is is possible that there is a particular tag that we could change to get this slice-to-volume motion correction to be stronger? Or, is there one that we could add? Might there be another way to correct these artifacts?

Thank you very much in advance for your input!

Best,

Kate

[tashrifbillah]

The main problem that we see in our raw data is horizontal striping, likely due to motion, and when we run this command it smooths out some of these artifacts. In other cases, however, it does not fully remove these motion artifacts.

Kate, your question is actually for FSL developers. Google FSL Jisc Mail and post there.

However, I know that the --repol flag does not work well below bshell 500. Is it the case for the screenshots?

In any case, I advise you to use our luigi-pipeline where we adopted official HCP pipeline's topup-eddy method. The latter is proven to work well in our office across many projects.

[katerendall]

Dear Tashrif,

Thank you for the advice, we have already posted on Jisc with no luck, but I will try again and will look into the Luigi pipeline.

However, I know that the --repol flag does not work well below bshell 500. Is it the case for the screenshots?

This is the case for many of our poor quality images, but not all. For example, Image 1 and 2 are from a bshell of 3005 and Image 3 and 4 are from a bshell of 5.

Thank you for the support and suggestions.

Best,

Kate