Figure 3.

The removal of background low-frequency phase (Φ) variation increases the sensitivity to BOLD signal changes in phase images and enables the visualization of phase activity maps highly co-localized to magnitude (M/M₀) activity maps. Magnitude and phase activity maps during visual task (P < 0.05 Bonferroni corrected) and at rest (P < 0.005 Bonferroni corrected, for display purposes only) for two subjects (two slices shown, red/green = positive/negative activity) obtained: (A) without and (B) and (C) with physiological noise correction (temporal drifts were removed in (A–C). The same noise regressors were employed in (B) and (C), with the following exception: the effects related to the phase of respiratory cycle were modeled by (B) four RETROICOR respiratory regressors, and (C) a single Φnoise-regressor (derived from spatial polynomial fitting of phase images, polynomial order = 4). In (D), we show the percentage (mean ± standard error across subjects) of overlapping voxels between magnitude and phase activity maps relative to the number of voxels in magnitude activity maps (P < 0.05 Bonferroni corrected, for both conditions) with and without physiological noise correction (see legend). In 1, three to six data from eight subjects were pooled, in 2 only six subjects were included because of missing physiological recordings for two subjects. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]