Correction of low-frequency physiological noise from the resting state BOLD fMRI--Effect on ICA default mode analysis at 1.5 T

Abstract

Confounding low-frequency fluctuation (LFF) physiological noise is a concern for functional connectivity analyses in blood oxygen level-dependent (BOLD) functional magnetic resonance imaging (fMRI). Using estimates of LFF physiological noise derived from measured cardiac and respiration signals, noise can be filtered from the time series thus improving the results of functional connectivity analysis. The ability of spatial independent component analysis (ICA) to separate LFF physiological noise from the default mode network (DMN), which overlap each other spatially and occur at similar frequencies, has remained an open question. We aimed to define the net effect of physiological correction for spatial ICA DMN detection at 1.5 T by statistically testing obtained ICASSO centrotype DMN maps before and after physiological correction. Comparisons with 21 subjects were performed for ICA model orders 20, 30 and 40 and no statistically significant spatial difference was found after physiological correction, although slight DMN reduction in precuneus or sagittal sinus was detected in all dimensionalities. A confounding factor in the analysis is the susceptibility of the ICA decomposition for data changes yielding different DMN splitting between and after physiological correction conditions without comparable true change in the data. This issue is mitigated at higher ICA model orders. The results suggest that subject-level DMN can for some subjects be optimized by physiological correction, but on the group-level this contribution is minor.