fmri_tempsmooth

Description

FMRI_TEMPSMOOTH Temporal Gaussian smoothing of fMRI time series. Convolves each voxel time series with a Gaussian kernel, suppressing high-frequency noise while preserving low-frequency BOLD signal. Useful as a preprocessing step before GLM analyses with slow HRFs.

Usage

newdata = fmri_tempsmooth(data)
newdata = fmri_tempsmooth(data, sigma)

Inputs

  • DATA: fMRI time series, vectorized (228453 x T) or volume (91 x 109 x 91 x T).

  • SIGMA: Standard deviation of the Gaussian kernel in scans (not seconds). Default: 5 scans. For TR-specific smoothing: sigma = FWHM/(2.355*TR).

Outputs

  • NEWDATA: Temporally smoothed data, same format as DATA.

Examples

data = fmri_tempsmooth(data);        % default sigma=5
data = fmri_tempsmooth(data, 3);     % lighter smoothing

Notes

  • For frequency-domain bandpass filtering use fmri_bandpassfilter instead.

  • The kernel is zero-padded at the edges (boundary effects possible for

  • first and last ~sigma scans).

See Also

  • fmri_bandpassfilter

  • fmri_detrend