Combined interactions of respiratory and cardiac signals measured by high-temporal resolution fMRI

Resultsrespiratory artifacts, Fourier regressors up to order 2 were associated with statistically significant clusters in all subjects. In some subjects, even higher-order regressors up to order 5 were also significant. For cardiac artifacts, Fourier regressors up to order 7 were significant in all subjects, with higher-order regressors also significant in some subjects up to order 10. The effect of the high-order terms was widespread; the percentage volume with a significant effect was 23% for respiratory regressors of order greater than 2, and 99% for cardiac regressors of order greater than 2. After removing these common fluctuations across the respiratory and cardiac cycles, a principal component analysis revealed that 69% of the studied brain voxels showed a correlation coefficient greater than 0.25 between the respiratory signal and the amplitude of the first principal component of the cardiac residuals (see Figure 1). This indicates that cardiac fluctuations were modulated by the respiration. In a GLM analysis, cardiac regressors of Fourier orders up to 7 and modulated by the respiratory signal were associated with statistically significant fMRI clusters accounting for an average of 81% of the studied brain volume. The addition of these modulation regressors resulted in small, but statistically significant (p<0.05) increases in activation volumes (average 1.5%) and t-values (average 5.4%) related to the visual paradigm. Discussion The fMRI signal is confounded by significant cardiac and respiratory artifacts. These artifacts are aliased in standard fMRI acquisitions, but the current study shows that high-order harmonics account for a significant amount of signal variance and require a high temporal sampling rate for their adequate characterization and removal. Additionally, fMRI cardiac fluctuations were found to be significantly correlated with the respiratory amplitude in widespread areas. The modeling of this modulation effect only resulted in a modest increase in sensitivity for the detection of the BOLD response to the slow visual paradigm, but this could potentially have a higher impact for faster event-related paradigms near the respiratory frequency.