*Example of spatial "bias correction" ("intensity homogeneity correction"). More relevant for anatomical data and tissue segmentation.
*Examples of "fieldmap-based undistortion". Toggling is your friend 😁. Side by side is terrible.
Effect of gradient nonlinearity correction. It's overall a very small minorish difference. (At least for this MR scanner.)
*Example of "pushing" the EPI distortion around depending on what you choose for the phase-encode direction.
Push the flow artifact around by just changing the phase-encoding direction. (The subject is not moving significantly.)
Phase component of some data. And big phase wraparounds (as expected). If you use a circular colormap, this weird visual thing will be dampened.
Add a caption...
Sampling below, in the middle, and above gray matter, this is T2* measurements. Above the GM we see clear effects of vessels and CSF and it's complex.
Reconstruction of the vessels, visualized in 3D
Add a caption...
Add a caption...
Add a caption...
*this binary mask has a weird outlier bumps at the bottom. this could easily cause problems downstream.
Example visualization of all volumes in a sample fMRI session (after pre-processing). Notice the instability in the lateral parts of the coronal slices; this was caused by an interaction between the run-by-run ANTS warp determination and a poor setting of parameters controlling the bias correction of EPI inhomogeneiy. The lesson is: carefully inspect results and don't blindly assume that things will be fine.
In one very rare instance, the fieldmap unwrapping algorithm catastrophically failed (upper right). This led to complete failure in the undistortion (lower right) and complete garbage thereafter. The lesson here is: 99% success rate is not good enough -- you have to do a comprehensive check of every case.
This shows a raw fMRI volume; magnitude component of fieldmap; and pseudocolor rendering of the fieldmap. Notice the substantial artifact in the fMRI volume related to the rapidly changing magnetic field near the frontal lobe.