Dwiextract error during dwifslpreproc (dwiextract: [ERROR] No dwi volumes present)

1

Hello,

I’m getting an unexpected error during the dwifslpreproc script. My DWI data have 3 shells (B=0,1000,2000) acquired in the AP phase-encoding direction, including one B=0 volume. I have two additional B=0 volume acquired with AP and PA phase-encoding directions (but with a shorter TR than my main DWI sequence).
When I run dwifslpreproc with option -rpe_pair and -se_epi, I get the following dwiextract error after topup has completed:

eddy correction / susceptibility distortion correction / motion compensation (requires FSL 6.0)
dwifslpreproc DWI_degibbs.mif DWI_corrected.mif -rpe_pair -pe_dir ap -readout_time 0.04438 -se_epi B0_degibbs.mif -eddy_options “–slm=linear --repol --data_is_shelled” -info
dwifslpreproc
…
Command: mrconvert /media/user/SharedDisk/data/PTD/PTD03/DWI/DWI_degibbs.mif /media/user/SharedDisk/data/PTD/PTD03/DWI/dwifslpreproc-tmp-SCL5ON/dwi.mif -json_export /media/user/SharedDisk/data/PTD/PTD03/DWI/dwifslpreproc-tmp-SCL5ON/dwi.json
mrconvert: [ 0%] copying from “/media/use…D/PTD03/DWI/DWI_degibbs.mif” to "/media/use… mrconvert: [100%] copying from “/media/use…D/PTD03/DWI/DWI_degibbs.mif” to “/media/use…lpreproc-tmp-SCL5ON/dwi.mif”
dwifslpreproc: Loading header for image file ‘/media/user/SharedDisk/data/PTD/PTD03/DWI/B0_degibbs.mif’
Command: mrconvert /media/user/SharedDisk/data/PTD/PTD03/DWI/B0_degibbs.mif /media/user/SharedDisk/data/PTD/PTD03/DWI/dwifslpreproc-tmp-SCL5ON/se_epi.mif
…
dwifslpreproc: Changing to scratch directory (/media/user/SharedDisk/data/PTD/PTD03/DWI/dwifslpreproc-tmp-SCL5ON/)
dwifslpreproc: Loading header for image file ‘dwi.mif’
dwifslpreproc: Loading header for image file ‘se_epi.mif’
dwifslpreproc: Command: ‘/home/user/matlab/mrtrix3/bin/mrinfo dwi.mif -shell_indices’ (piping data to local storage)
dwifslpreproc: Result: …
dwifslpreproc: Command: ‘/home/user/matlab/mrtrix3/bin/mrinfo dwi.mif -shell_bvalues’ (piping data to local storage)
dwifslpreproc: Result: 0 1000.02 2000.04
dwifslpreproc: First b=0 volume in input DWIs is volume index 64; this will be permuted to be the first volume (index 0) when eddy is run
Command: mrinfo dwi.mif -export_grad_mrtrix grad.b
Command: mrconvert se_epi.mif topup_in.nii -strides -1,+2,+3,+4 -export_pe_table topup_datain.txt
mrconvert: [100%] copying from “se_epi.mif” to “topup_in.nii”
Command: topup --imain=topup_in.nii --datain=topup_datain.txt --out=field --fout=field_map.nii --config=/usr/local/fsl/etc/flirtsch/b02b0.cnf --verbose
dwifslpreproc: Output of topup command:
…
Command: mrinfo dwi.mif -export_pe_eddy applytopup_config.txt applytopup_indices.txt
Command: dwiextract dwi.mif -pe 0.0,1.0,0.0,0.0443999991 - | mrconvert - dwi_pe_1.nii -strides -1,+2,+3,+4 -json_export dwi_pe_1.json
dwiextract: [ERROR] No dwi volumes present
mrconvert: [ERROR] no filename supplied to standard input (broken pipe?)
mrconvert: [ERROR] error opening image “-”

The script is apparently trying to extract volumes acquired in the PA phase-encoding direction from the main DWI dataset, but it only contains AP volumes.
There is no warning issued before this error. The same happens if I try with the -rpe_header (and -se_epi) options (I checked that all the the PE tables are as expected in the input files).

Cheers
Julien

2

Update:
if I add a volume with reversed phase-encoding direction ( PA ) at the end of the main DWI dataset, I still get the error, at last when using option -rpe_pair. But If I use option -rpe_header with the extra B=0 volume, then the script completes without error (and correction looks reasonable).
When using option -rpe_pair, the error occurs because the script is trying to extract a PA volume from the main DWI dataset, while at the same time manually setting all phase-encoding values to AP using option -import_pe_table, which is contradictory and results in an error:

dwifslpreproc DWI_degibbs_test.mif DWI_corrected.mif -rpe_pair -pe_dir ap -readout_time 0.04438 -se_epi B0_degibbs.mif -eddy_options “–slm=linear --repol --data_is_shelled” -info
dwifslpreproc:
…
Command: dwiextract dwi.mif -import_pe_table dwi_manual_pe_scheme.txt -pe 0.0,1.0,0.0,0.0444 - | mrconvert - dwi_pe_1.nii -strides -1,+2,+3,+4 -json_export dwi_pe_1.json
dwiextract: [ERROR] No dwi volumes present

In this case, dwi_manual_pe_scheme.txt only contains lines with AP phase-encoding direction.
This suggests that this is due to a bug in the script?

From what I can see of the script output when it completes, the extracted B=0 volumes (both AP and PA) are used to make the eddy mask. But when using the -se_epi option, shouldn’t these be extracted from the provided B=0 dataset rather than the main DWI dataset?
Alternatively, maybe dwiextract should exit without error when no volume matching the requested phase-encoded direction is found and the mask would be based only on the B=0 volumes present in the main DWI data (i.e. only in the AP direction in my case).

Cheers
Julien

OS: Ubuntu 16.04
MRtrix3* version: 3.02-193-gd63cc20

3

Hi Julien,

I have two additional B=0 volume acquired with AP and PA phase-encoding directions (but with a shorter TR than my main DWI sequence).

Please try to avoid such acquisitions. Using one set of volumes to estimate the inhomogeneity field for correction of a different set of volumes necessitates a rigid-body transformation between them; but the combined operation of topup and eddy do not permit such. And there’s no individual volume that you can use as the reference image for both field estimation and correction, as the change in TR between that volume pair and your DWI b=0 means that the image contrast is different and therefore the latter shouldn’t be included in the input to topup as that difference in image contrast could drive erroneous inhomogeneity field estimates. See also relevant blog post.

The script is apparently trying to extract volumes acquired in the PA phase-encoding direction from the main DWI dataset, but it only contains AP volumes.

I’m fairly confident that this would be fixed by upgrading to version 3.0.3, due in fact not to fixes to phase encoding handling but to specifrically this commit that completely changes how data are prepared for applytopup within dwifslpreproc.

Rob