I am facing issues with the fmriprep registration. I have a longitudinal data but I pass through an averaged T1 template across the time-points. I also included the distortion correction parameters. But it seems that the registration is poor between the EPI and T1 images. I also checked using the check reg option and it seems that the registration is not proper.
Please have a look to the attached report and let me know your suggestions.
Hi, what version of fMRIPrep are you using, and can you provide your full command-line? Are you running FreeSurfer through fMRIPrep or passing a pre-computed directory?
I am using the fmriprep version 20.0.6.
Th full command-line is
singularity run --cleanenv fmriprep-20.0.6.simg /home/upadhyayn/Projects/fmriprep_work/raw /home/upadhyayn/Projects/fmriprep_work/derivatives participant --participant-label 03ERIM05 04BEWO26 04ENNG22 04ERNS06 04KEEN22 04LIUS27 04RTAX08 05CKUT03 05ELUS17 05ERNT03 06CHIM07 06CHIM28 06ELCH20 06ERRD28 07ELAN15 07NGJE28 07RSEL02 07UDNS05 07WAAL30 08ELIT13 08ENCH10 08TZAV17 09ELUS12 09EZSE11 09LTMA30 09NNER24 09NNLF21 09RNOR18 --nthreads 12 --omp-nthreads 8 --low-mem --ignore slicetiming --dummy-scans 5 --use-aroma -w /home/upadhyayn/Projects/fmriprep_work/work
Yes, i am also performing the freesurferprocessing through fmriprep.
Thanks. Looking at this, it seems likely that your problem is too low of gray/white contrast, but the boundary-based registration refinement is not so bad that it raises flags. You can try the
--force-no-bbr option to see if the affine registration works better.
Just for experimentation, I changed the phase encoding direction and the direction of the registration error inverted as per attached images. So i think the issue is related with the distortion correction parameters which is not included in the last stage of co-registration.
Please have a look to the attached images with opposite phase encoding direction “j”.
What type of fieldmaps do you have? And could you provide orientation information (as well as
PhaseEncodingDirection metadata) for all of your BOLD and fieldmap input files?
To get orientation info:
import nibabel as nb
img = nb.load(fname)
Thanks for your suggestions. I did run the -force-no-bbr option and it provide volume driven registration and surface drive registration. The volume driven registration is quite promising and while checked using the check register in spm show good registration between EPI and T1. However, the surface driven registration got more bad.
attached images of both registration.
Please have a look. the information regarding the metadata provided and orientation information of EPI images are also provided in attached images.