Processing steps

[tsalo]

1. Collect data from preprocessing derivatives and raw dataset.
   • Raw phase data from raw dataset.
   • Preprocessed magnitude data from preprocessing derivatives.
   • Transforms from preprocessing derivatives.
2. Rescale from arbitrary units to radians.
3. Remove non-steady-state volumes?
4. Unwrap phase data with ROMEO/MEDIC.
   • Combine unwrapped phase data across echoes with MEDIC?
5. Calculate phase jump and phase jolt time series from rescaled, but not unwrapped, phase with LayNii's LN2_PHASE_JOLT.
6. Extract confounds
   • HighCor
   • Phase quality map calculation (Rodriguez et al., 2010)
7. Denoise the phase data.
   • RETROICOR
8. Apply transforms from fMRIPrep to get data into the appropriate spaces.
   • Probably need to ignore STC.
   • In idevicente/Phase-based_denoising_pipeline, they combine mag + phase into complex, then split into real + imag, then apply the transforms, then combine into complex, then split into mag + phase. Not sure if that's necessary.

Related to nipreps/fmriprep#3336.

[tsalo]

Useful links:

• https://neurostars.org/t/processing-for-complex-valued-fmri-data/5093
• https://github.com/tsalo/complex-flow (a past attempt at something similar)