Undistort EPI in seconds
- SuCor.
SuCor turns an opposite-PE b0 pair into one distortion-free image and a B0 fieldmap that's sharp at air–tissue edges and smooth in tissue — using closed-form optimal transport instead of slow iterative registration.
Drag to undistort a real brain.
Wipe between the distorted input and the SuCor-corrected output, scrub slices, or switch to the estimated field.
23.1× faster than topup.
Wall-clock to estimate the field on identical input.
One field, estimated end to end.
SuCor estimates a single B0 displacement field from the opposite-PE pair and applies it to the whole series. A few design principles, without the recipe:
Direct start
A fast, non-iterative first estimate — no slow registration to converge.
Dropout-robust
Stays coherent where signal has piled up or dropped out.
Physics-like field
Smooth within tissue, sharp at air–tissue boundaries.
Refined
A polish pass tightens the fit without roughening the field.
Signal-conserving
Applied across the full 4-D series while preserving total signal.
It recovers the true brain.
On a physics simulation where the undistorted brain is known exactly, SuCor turns the distorted input back into the ground truth.
The two encodings snap together.
AP and PA spin-echo EPIs distort in opposite directions. A correct field maps both back to the same true geometry — so red (AP) and green (PA) collapse to gray. PE-axis mismatch drops 77%. This is the same consistency FSL topup is scored on, with no external reference or co-registration in the loop.
And against an undistorted T2 spin-echo acquired in the same space — the inferior slab where susceptibility distortion is worst:
Piled-up brainstem and cerebellum decompress toward the spin-echo's true shape; orbitofrontal tissue is pulled back forward.
On par with FSL topup — faster.
Want to try SuCor on your data?
SuCor is in active development at Indiana University. For early access or to collaborate, reach out.
srchig@iu.edu