In this study, we first demonstrate using resting state fMRI (rsfMRI) “null” datasets, that serial correlation in fMRI time-series arises from non-stochastic signals (e.g., coordinated activity within brain function networks unrelated to the fMRI paradigm of interest). Using this principle, we then advance a method to obtain whitened GLM first-level analysis regression residuals in task fMRI studies, by accounting for non-stochastic brain signals through principal components analysis. Importantly, the proposed methods is insensitive to the temporal resolution of fMRI time-series, unlike conventional stochastic models of serial correlation, whose parameters have to be modified depending on fMRI scan-TR.
This abstract and the presentation materials are available to members only; a login is required.