Brainstem functional magnetic resonance imaging: disentangling signal from physiological noise

Abstract

Purpose: To estimate the importance of respiratory and cardiac effects on signal variability found in functional magnetic resonance imaging data recorded from the brainstem.

Materials and methods: A modified version of the retrospective image correction (RETROICOR) method (Glover et al, [2000] Magn Reson Med 44:162-167) was implemented on resting brainstem echo-planar imaging (EPI) data in 12 subjects. Fourier series were fitted to image data based on cardiac and respiratory recordings (pulseoximetry and respiratory turbine), including multiplicative terms that accounted for interactions between cardiac and respiratory signals. F-tests were performed on residuals produced by regression analysis. Additionally, we evaluated whether modified RETROICOR improved detection of brainstem activation (in 11 subjects) during a finger opposition task.

Results: The optimal model, containing three cardiac (C) and four respiratory (R) harmonics, and one multiplicative (X) term, "3C4R1X," significantly reduced signal variability without overfitting to noise. The application of modified RETROICOR to activation data increased group Z-statistics and reduced putative false-positive activation.

Conclusion: In addition to cardiac and respiratory effects, their interaction was also a significant source of physiological noise. The modified RETROICOR model improved detection of brainstem activation and would be usefully applied to any study examining this brain region.