Dwifslpreproc : Single-Direction Phase Encoding Issue in DWI Without Opposite Direction

Dear MRtrix experts,

We acquired DWI data in the AP direction using the HCP sequence (1.5x1.5x1.5 mm³, 92 slices, 99 volumes, TE=89). For most of the subjects, we also collected b0 images in the opposite direction P->A. Some subjects are missing PA b0 images but have spin echo images collected in AP and PA directions during the same scanning session but with the different scanning parameters (2x2x2mm³, 72 slices, TE=66). We compared distortion correction using a pair of b0 images and a pair of SE images in the same subject and found that the corrected images looked differently (although such parameters as motion and the number of outliers were comparable).

I would be grateful for your guidance on how to handle image distortion correction using SE images when reverse phase b0 images are not available.

We thought that the differences could be due to the differences in the voxel sizes and the number of slices. Therefore, we resized the SE images to match the DWI images. Specifically, I resized SE_dir-AP_epi_den_degibbs.mif and SE_dir-PA_epi_den_degibbs.mif to match the voxel size resolution of AP_den_degibbs.mif after denoising and Gibbs deringing. The resized AP and PA spin echo images were combined into SE_AP_PA_epi_den_degibbs_b0_pair_resized.mif and used in dwifslpreproc with the options: -pe_dir AP -rpe_pair -se_epi (as shown below).

dwifslpreproc {subj}_AP_den_degibbs.mif {subj}_AP_den_degibb_epi_preproc.mif -nocleanup -pe_dir AP -eddyqc_all eddyqc_epi -rpe_pair -se_epi ${subj}_SE_AP_PA_epi_den_degibbs_b0_pair_resized.mif -readout_time $TotalReadoutTime -eddy_slspec my_slspec.txt -eddy_options " --slm=linear --mporder=4 --repol --fep --cnr_maps --ol_type=both --data_is_shelled --b0_peas " -align_seepi

For DWI data with both AP and PA images available, b0 images were merged into “AP-PA_dwi_den_degibbs_meanb0_pair.mif” and used in dwifslpreproc.

dwifslpreproc {subj}_AP_den_degibbs.mif {subj}_AP_den_degibb_preproc.mif -nocleanup -pe_dir AP -eddyqc_all eddyqc -rpe_pair -se_epi ${subj}_AP-PA_dwi_den_degibbs_meanb0_pair.mif -readout_time $TotalReadoutTime -eddy_slspec my_slspec.txt -eddy_options " --slm=linear --mporder=4 --repol --fep --cnr_maps --ol_type=both --data_is_shelled --b0_peas " -align_seepi

Finally, I compare the ’ AP_den_degibb_preproc.mif ’ images from the dataset with two PE directions (A->P and P->A) b0_pair to ’ AP_den_degibb_epi_preproc.mif ’ image from the dataset with only A->P PE direction and SE AP/PA epi b0_pair within the same subject. The SE epi b0 pair image exhibits inferior geometric distortion correction.

Is it correct to use the SE AP and PA epi b0 pair image in the option ‘-rpe_pair -se_epi’ of the ‘dwifslpreproc’ command? I attempt to use SE PA epi b0 (with the same resolution as the DWI images) and the first PE b0 from the DWI (A->P) to generate an AP-PA b0 pair file. However, the issue arises due to the discrepancy in echo times (TE) between the SE PA epi b0 image (TE=66) and the PE AP b0 image (TE=89).

Does 'dwifslpreproc ’ offer any options for addressing this issue, or do you have a better solution for it?

Any advice on the optimal approach to handling distortions in single PE direction (A->P) DWI data would be greatly appreciated!

Thank you for your assistance.

Best regards,



Dear Deyu,

Thank you for your suggestion. I’ll give ‘synb0-disco’ a try.

Best regards,