Skip to main content
. 2021 May 6;15:625737. doi: 10.3389/fnins.2021.625737

FIGURE 1.

FIGURE 1

Example Voxel Inclusion Probability Calculation. Shown are spatial maps for three independent component analysis (ICA) components (ICs), each with a color bar thresholded at the 95th percentile and correlations with cerebrospinal fluid (CSF) and white matter (WM) tissue probability maps. From left to right, displayed are spatial maps for component number 36 from ICA model order 60 (ICA60,36), ICA270,81, and ICA2,1, respectively. ICA60,36 and ICA270,81 correlated strongly with CSF and WM, respectively, while ICA2,1 did not correlate substantially with either noise source (rCSF = 0.07, rWM = 0.03). ICA weights for a voxel marked by an asterisk are mapped onto each color bar. For the first IC, this voxel is less than the cutoff for the 95th percentile (IC weight = 0.09, | cutoff| = 0.26, range = ± 1.6). For the second IC, the voxel is within the 95th percentile (IC weight = 0.41, | cutoff| = 0.22, range = ± 1.1). For the third IC, the voxel is outside of the 95th percentile (IC weight = 0.02, | cutoff| = 1.22, range = ± 3.8). Therefore, for this voxel, event A is true for the second IC, while event B is true for the first and second ICs. The marked voxel then has a probability of ½ of being strongly weighted by a noise component, given the three ICs shown. Repeating this calculation for all voxels and all ICs in the analysis yields the Voxel Inclusion Probability Volumes displayed in Figures 4, 5 and Supplementary Figure 6.