Unsupervised modelling of physiological noise artifacts in fMRI data&[phi]
Madsen K, Lund T
Technical University of Denmark, Copenhagen University Hospital
This study suggests a way to obtain nuisance regressors to correct for artefacts such as cardiac and respiratory noise in fMRI time series without measures of cardiac and respiratory cycles. We show that including these regressors in the design matrix reduces correlation and non-normality in the residual errors. In addition, the obtained regressors are in fact similar to the ones obtained using measures of cardiac and respiratory cycles. Furthermore, we show that these regresssors surpass other methods in terms of reducing non-normality and correlations in the residual errors.