FMRIPREP distortion correction issues with higher resolution MB-EPI

Hi NeuroStars,

I’ve been noticing some issues with SDC in OFC with 2 mm isotropic multiband T2*-EPI which might be related to the move from TOPUP to 3dQwarp correction in sdcflow. You’d expect a roughly hemispheric area of severe dropout in posterior/caudal OFC with this sequence, partly due to intravoxel dephasing and partly to a k-space center shift out of the sampled k-space FOV from local field gradients in this area. Running the SDC with TOPUP on the SBRef volume outside of FMRIPREP produces the expected dropout pattern and distortion correction of anterior OFC (below left, corrected and uncorrected sbref)

If we run the associated multiband BOLD series through FMRIPREP (these images are fully preprocessed, bias corrected etc) we see anterior OFC signal pushed back into the dropout region in posterior OFC (below right, corrected).

What’s wierd is that SDC in the rest of the brain seems to be fine (frontal and temporal poles, brain stem and cerebellum). We originally thought this might be something simple like a flipped PE orientation in the JSON sidecar, but inverting the PE direction results in globally worse distortion.

Any thoughts? I feel I might be missing something fundamental about the 3dQwarp approach or the polarity inverted SE-EPI pair. Happy to provide raw data if it’ll help with debugging.

  • Mike
    Caltech Brain Imaging Center

Moved to and previously reported in