Hi,
We ran QSIprep v1.0.1 on a sample of data acquired in 30 directions with two opposing phase-encoding directions (LR and RL). The BIDS directory structure is that both dMRI scans are in the /dwi/ folder. QSIprep runs without error on the majority of the subjects, except for a few with the following error:
Field of view of image #1 is different from reference FOV
Investigating further, we noticed the transform is slightly different in the raw images (by running mrinfo):
Transform: 0.9999 7.318e-11 0.01251 -121.1
0.003324 0.964 -0.2657 -83.14
-0.01206 0.2658 0.964 -132.4
And
Transform: 0.9999 -0.001745 0.01251 -120.9
0.005006 0.964 -0.2657 -83.35
-0.01159 0.2658 0.964 -132.4
The solution we are thinking of so far is to extract the transform of one of the images (using awk on mrinfo output) and align the other using this transform (with mrtransform), adjusting bval/bvec file as well. According to mrtransform:
If a DW scheme is contained in the header (or specified separately), and
the number of directions matches the number of volumes in the images, any
transformation applied using the -linear option will also be applied to
the directions.
So, that being said and considering img1 and img2 are our RL and LR sequences, respectively:
mrtransform -linear transform_img1 -fslgrad img2_bvec img2_bval -export_grad_fsl img2_bvec_transformed img2_bval_transformed img2 img2_to_img1
I have two questions:
- What do you think about this approach?
- Do you think this transformation may hinder the susceptibility distortion correction?
Best,
Amir