Lars Kasper, Institute for Biomedical Engineering, University of Zurich & ETH Zurich

Model-based Physiological Noise Correction for fMRI

Physiological noise poses a major confound mechanism for fMRI [1], contaminating BOLD time series by fluctuations of cardiac and respiratory origin. This is a particular issue for analyses focused on subtle, trial-wise effects, as in computational fMRI. Several model-based correction methods exist, relying on peripheral recordings of physiology [2], such as ECG and pneumatic belts. However, evidence of group level benefits of such correction has been scarce, despite multiple reports on its efficacy at the single-subject level [3,4].
In this talk, I will address this discrepancy and show that physiological noise correction can indeed greatly improve group level results in computational fMRI. We found that robust preprocessing of the peripheral recordings is crucial for this finding, and therefore developed the PhysIO Toolbox at our lab for this purpose [5]. Furthermore, I will provide evidence that the group-level effect of noise correction is driven by reducing between-subject variability in parameter estimates for contrasts of interest. This addresses a common concern that "summary statistics" analyses based on the general linear model (GLM) do not benefit from physiological noise removal because they only consider parameter estimates – but not their variance – at the group level [6].
I will close with a discussion of physiological noise correction in studies of (voluntary) physiological manipulations, e.g. interoception.

[1] G. Krüger, G.H. Glover, Magnetic Resonance in Medicine, 2001, 46, 631.
[2] G.H. Glover, T.Q. Li, D. Ress, Magnetic Resonance in Medicine, 2000,44, 162.
[3] C. Hutton /et al./, NeuroImage, 2011, 57, 101
[4] T.E. Lund et al., NeuroImage, 2006, 29, 54.
[5] L. Kasper et al., Journal of Neuroscience Methods,2016, http://dx.doi.org/10.1016/j.jneumeth.2016.10.019
[6] A. Holmes, K. Friston, NeuroImage, 1998, 7, S754.

Additional Information

Part of the colloquium series of the GRK 1589 "Sensory Computation in Neural Systems"

Organized by

Alexander Genauck / Andreas Heinz