In resting-state fMRI nuisance regression, a General Linear Model (GLM) is employed to fit and remove the variance associated with a noise model. Without "ground-truth" knowledge, the noise models must be tested and improved to obtain accurately cleaned datasets without "throwing the baby out with the bath-water." Valid statistical inference on a GLM fit requires normally-distributed residuals, which is not the case when intrinsic brain fluctuations are present. We demonstrate that existing pre-whitening tools can be appropriately applied to account for serial autocorrelation in resting-state fluctuations during nuisance regression, allowing statistical differentiation of true and simulated noise models.
This abstract and the presentation materials are available to members only; a login is required.