Figure 1.

Importance of phase (Φ) preprocessing to remove noise instability. Shown for each of the preprocessing steps are: left panel) the phase image acquired at a specific single time‐point (eighth TR, in this example) during visual stimulation for an example data‐set; right panel) an example time‐course (black curve) extracted from the voxel indicated by the red arrow in the top left panel. Preprocessing steps: (A) raw phase data obtained after image reconstruction; (B) phase data after subtraction of the first phase image of the time‐series, voxel‐by‐voxel unwrapping, and removal of linear drift over time; (C) spatial polynomials (sixth model order) fitted on a slice‐by‐slice basis to the phase image shown in (B): this fit accounts for background spatial low‐frequency phase variation mostly due to respiration as shown in (C), right panel (black: resulting phase time‐course from spatial polynomial fitting = Φnoise‐regressor; blue: respiratory trace sampled at the slice acquisition timing); in (C) right panel, we also show a zoomed view (60 s only) of the Φnoise‐regressor and of the respiratory trace; (D) phase image obtained after subtraction of spatially fitted polynomials shown in (C) from phase image shown in (B). In (D), right panel, black: phase time‐course; magenta: magnitude (M/M ₀) time‐course in the same voxel; red: stimulus regressor (arbitrary units and offset). [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]