To determine the effect of cardiorespiratory activity on the distribution of frequency band power for the entire brain, HR and RVT signals were averaged at every TR and the resulting vectors were removed voxel-wise from the BOLD using Matlab linear regression. Of the 17 subjects, 12 subjects were included in the analysis, 1 subject had missing data, and 4 subjects had temporal artifacts in the autonomic signal that could not be corrected by interpolation.
Z-scored SDMs were generated from rapid TR fMRI scans in a similar manner as those obtained with slower TRs. Z-scores for each frequency band were generated by subtracting the mean voxel-wise power for the midline sagittal slice and dividing by the standard deviation. Subject z-score maps were averaged using a fixed-effects model to create group level z-statistic maps. BOLD signal significantly affected by physiological noise was determined by observing the distribution of power around frequency bands known to be associated with cardiorespiratory activity and showing relevant peaks for the group average midline slice BOLD frequency spectrum (0.2–0.4 Hz, and 0.75–1.5 Hz). Regions exceeding z>2.3 for these frequency bands were masked from further analysis of the midline slice. The group average z-score distribution for the slow fMRI data set was obtained using the same mask. BOLD SDMs of the right leg were made in an identical manner to the brain SDMs, without physiologic correction.
Z-scored SDMs were generated from rapid TR fMRI scans in a similar manner as those obtained with slower TRs. Z-scores for each frequency band were generated by subtracting the mean voxel-wise power for the midline sagittal slice and dividing by the standard deviation. Subject z-score maps were averaged using a fixed-effects model to create group level z-statistic maps. BOLD signal significantly affected by physiological noise was determined by observing the distribution of power around frequency bands known to be associated with cardiorespiratory activity and showing relevant peaks for the group average midline slice BOLD frequency spectrum (0.2–0.4 Hz, and 0.75–1.5 Hz). Regions exceeding z>2.3 for these frequency bands were masked from further analysis of the midline slice. The group average z-score distribution for the slow fMRI data set was obtained using the same mask. BOLD SDMs of the right leg were made in an identical manner to the brain SDMs, without physiologic correction.
