Dear Matthias,
there are two things that can possibly have gone wrong here.
1. The use of gradient echo fieldmaps is unfortunately still not always straightforward. In order to be used with eddy the resulting displacement fields need to be invertible, which is not always the case for measured fieldmaps pre-processed with fsl_prepare_fieldmap.
2. You should not denoise data before running eddy. The artificial nature of denoised data means that the gaussian process that is at the heart of eddy sometimes/often fail to find reasonable hyperparameters for data that has been denoised. Sometimes this results in suboptimal parameters that means that the motion/distortion correction is suboptimal. Sometimes it results in an infinite loop in the search for that hyperparameters, which I would guess is what has happened here.
I don’t know the details about the denoising you use, but I would also not be surprised if the denoising works suboptimally when there are motion and dropouts in your data. Which would be another reason to do it after eddy, if at all.
Jesper
> On 19 Jan 2019, at 03:20, Matthias Heyne <[log in to unmask]> wrote:
>
> Dear list,
>
> I have previously managed to run eddy_cuda on various DWI data sets with and without opposite PE blips or gradient echo field maps and various numbers of b0s interspersed or all collected at the beginning of the DWI scan. However, eddy_cuda fails to get through the 3rd iteration on a different data set with all b0s collected at the beginning of the scan. The scans were acquired on a 3T Siemens Tim Trio scanner along with gradient echo field map scans (magnitude + phase volumes) that I processed using fsl_prepare_fieldmap. The DWI volume was denoised and de-Gibbs-ed (in that order) using the commands from MRTrix and includes 60 diffusion directions at b=700 with 10 b0s collected at the beginning. The data were acquired with GRAPPA=2 but no SMS/cmrr and do not contain excessive motion. My GPU jobs fail at 48h while I've been able to process data with much more motion from a different data set within 4.5h max.
>
> I used the following parameters with a maximum time allowance of 48h and the tool didn't make it through the 3rd iteration before dying via signal kill: eddy_cuda --imain=[preprocessed_DWI_vol] --mask=DWI_mask.nii.gz --acqp=[…] --index=[…] --bvecs=[…] --bvals=[…] --field=field_map --niter=3 --fwhm=2,1,0 --repol --mporder=2 --cnr_maps --residuals --slspc=[…] --out=eddy_cuda_pre-processed_DWI_using_gre_field_map_w_denoising_and_degibbsing --very_verbose
> I have also tried excluding the --mporder flag (not running s2v correction), using --mporder=6 and/or using a larger fwhm for the first iterations (e.g. --fwhm=10,2,1,0,0) with the same result. I can share the --very-verbose piped output if that might be helpful.
> What other flags/settings might I be able to try?
>
> Thanks,
> Matthias
>
> ########################################################################
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
|