Hello,
I want to use fMRIPrep to apply the susceptibility distortion correction (SDC) to my fMRI scan. In our project they acquired “reverse fMRI” (rfMRI), which would be an extra fMRI scan with opposite phase-encoding direction (PE) than that of the fMRI scan. Both images are GR scans, but the fMRI has PE=AP, and the rfMRI has PE=PA.
In this thread, a clear distinction is made on how to use GR and SE scans for SDC. However, as I do not have any phasediff nor magnitude.nii files, I tried to use the SE Field Maps approach with two GR scans (only first time volume) with opposite PE. My questions are: 1. What are the risks of doing this? 2. Is there a way to derive the phasediff and magnitude.nii files from the rfMRI?
The results I got seem good from the side of the geometrical deformation i.e. the fMRI scan with SDC seems to be more aligned with the T1w, but I want to be sure that the process is not affecting the intensities of the fMRI. Can someone with more experience give me advice on this?
In theory SE-EPI images with opposite PE would be better than two GR-EPI images with opposite PE for SDC with fmriprep. Indeed fmriprep is calling SDCFlows which is itself calling FSL topup for susceptibility distorsion correction, and topup is designed for pair of SE-EPI images. GR-EPI have similar distorsions as SE-EPI images, but they also get dropout in addition which makes the correction less accurate in theory. In practice I now several MRI center using pairs of GR-EPI images for SDC and they seem happy with this approach when using fmriprep.
Hello @jsein, thanks a lot for your reply! So I understand that the choice is up to us and it would be good to do quality checks on the intensities of the SDC-corrected and -uncorrected image, right?
And do you have any insights on my second question?
I am wondering whether these files can be extracted somehow from the rfMRI NIFTI/DICOM or if they should be provided from the scanner. I understand that this should not be possible if the only acquisition parameter that differs in the two scans (fMRI and rfMRI) is the PE. There would be a phase difference if the two scans were acquired at different echo times, right?
No, these files (phasediff and magnitude) come from the measurent made by a different sequence, a gradient-echo sequence, with two echo times, not a gradient-echo EPI image as it is the case for bold.
In the case where you have only GRE-EPI images with reversed PE, you can still try the PEPOLAR method , the correction should work.
Otherwise you can still do SDC with Syn-BOLD DisCo, for wich you don’t need any fieldmap:
Thanks a lot, @jsein !
I have tried both methods (I include details which could be useful to other users):
PEPOLAR method with GRE-EPI images with reversed PE;
estimation of the synthetic field map with –use-syn-sdc: first I ran fMRIPrep with the –use-syn-sdc flag and a T1w resampled to the BOLD dimensions, and then reused the fmap folder from the latter run in a new fMRIPrep run with the full-resolution T1w. This was done by adding an extra flag to the fmriprep-docker command: “–derivatives fmaps=derivatives/fmriprep/fmap”.
The SDC results are actually very similar across the two methods. This time, we will use the PEPOLAR method results, as for this method we only need to run fMRIPrep once.