fMRI denoising

Collection of Matlab functions for denoising fMRI data, such as resting- or steady-state data.

• fmri_cleaning performs data denoising - including frequency filtering, removal of signals of no interest (e.g., motion or physiological estimates) and censoring motion-contaminated volumes – via a single linear regression model. This is the main function; the other functions allow you to create appropriate input variables for this function.

• fmri_compcor extracts signals from specified regions of interest. You can extract the average signal, the median signal or the principal components (PCs; aCompCor or tCompcor). Before extracting PCs, you can orthogonalize the data with respect to other variables that will compose the final regression model, so that the extracted PCs will be maximally predictive.

• fmri_rp_metrics computes various framewise displacement metrics.

• fmri_censoring_mask creates temporal masks for volume censoring (e.g., using the framewise displacement).

The code was developed for the following publication:

Evaluation of denoising strategies for task‐based functional connectivity: Equalizing residual motion artifacts between rest and cognitively demanding tasks. Mascali, D., Moraschi, M., DiNuzzo, M., Tommasin, S., Fratini, M., Gili, T., Wise, R.G., Mangia, S., Macaluso, E. and Giove, F. Human Brain Mapping (2021).