Visual Report: CompCor

Component-based noise correction addresses issues of physiological noise from cardiac pulsation or respiration.

Principal Component Analysis (PCA) is used to compactly characterize the time series of both anatomically and temporally determined ROIs. These components, along with other confounds, are made available in text files and can be used with caution to further denoise data during analysis.

aCompCor uses anatomically-determined ROIs: Signals from WM, and CSF are unlikely to be related to neural activity and are primarily a function of physiological noise.

tCompCor uses temporally-determined ROIS: Time courses with a relatively high SD are dominated by physiological noise (edge voxels, ventricles, areas near large vessels).

See the Official fMRIPrep Confound documentation.

Legend for ROI Colors

● Red identifies the edge of the brain.

● Magenta (aCompCor) is placed in WM and CSF: Signal in these voxels is unlikely to be related to neural activity and is primarily a function of physiological noise.

● Blue (tCompCor) identifies the most variable voxels in the brain (edge voxels, ventricles, areas near large vessels). These are used for the tCompCor components.

● Green defines the `edge` or `crown` (area right outside the brain). Including voxels located in a closed band (or crown) around the brain improves motion regression because voxels outside the brain should not have any signal. If there is signal in the crown, it probably arises from large head movements or signal spillover (Provins 2021).

Watch this ~6-minute video clip

fMRIPrep 3 (10)