gastropy.neuro.fmri.regress_confounds

gastropy.neuro.fmri.regress_confounds#

gastropy.neuro.fmri.regress_confounds(bold_2d, confounds_df, confound_cols=None)[source]#

Remove confound signals from BOLD data via GLM regression.

Performs ordinary least squares regression of confound regressors on each voxel’s BOLD time series and returns the z-scored residuals. This follows the approach used in standard fMRI denoising pipelines.

Parameters:

  • bold_2d (array_like, shape (n_voxels, n_timepoints)) – BOLD time series for each voxel (e.g., from a masked NIfTI).

  • confounds_df (pandas.DataFrame) – Confounds table (e.g., from fMRIPrep *_confounds_timeseries.tsv). Must have at least n_timepoints rows.

  • confound_cols (list of str, optional) – Column names to use as regressors. Default uses 6 motion parameters + 6 aCompCor components (12 regressors total).

Returns:

residuals – Z-scored residuals after confound regression.

Return type:

np.ndarray, shape (n_voxels, n_timepoints)

Notes

NaN values in the first row of derivative columns (common in fMRIPrep confounds) are replaced with 0 before regression.

Contents