Potential issue when (phase difference) fieldmap PE orientation does not match EPI

[effigies]

Thread on Neurostars: https://neurostars.org/t/phaseencodingdirection-i-and-j-in-json-file/24347

I have one fMRI (json: “PhaseEncodingDirection”: “i”, “ImageOrientationText”: “Tra>Cor(-21.6)>Sag(1.6)”) and three fmap - magnitude1 (json: “PhaseEncodingDirection”: “j-”), magnitude2 (json: “PhaseEncodingDirection”: “j-”), phasediff (json: “PhaseEncodingDirection”: “j-”).

[effigies]

I'm starting to think that this probably is related, to #340, and it is likely a result of an overly restrictive magnitude mask. This appears to present in two ways:

1. Empty masks lead to Tracking LAS phasediff issue #340.
2. Very small masks generate wild spline fields, mostly on the edges as magnitude images tend to have much higher intensity closer to the coils.

[effigies]

Should be resolved by #396.