Sympathetic activity contributes to the fMRI signal

Figures

• 

  Fig. 1. Covariation of fMRI, EEG, and autonomic signals during NREM2 sleep (subject S2). a Relationship between raw PPG and EEG (channel Fp1). K-complexes, indicated by asterisks, are typically followed by drops in the envelope of the PPG signal. b Relationship between fMRI and the various other signals. Episodic low-frequency (LF) EEG increases are apparently followed by drops in PPG-AMP and fMRI gray-matter (GM) signal. Relationship with variations in heart rate (HR, zero mean) and respiratory volume (RV) is less apparent. LF-EEG was calculated as the power in the 0.5–2 Hz spectral band for each 3 s (1 fMRI TR) interval, and clipped at two standard deviations for scaling purposes

• 

  Fig. 2. Relationship between fMRI and either PPG-AMP (a), or LF-EEG (b) during NREM2 sleep (n = 7). Single slice fMRI correlations are shown at various lags. Strong and spatially similar (anti-)correlation patterns are seen for both PPG and EEG, with the latter occurring at later lags. These observations are consistent with the notion that LF-EEG and PPG-AMP changes are strongly related, with the former preceding the latter. Note that that the color scale was inverted between the two comparisons to emphasize similarities. Correlations with LF-EEG are predominantly negative, consistent with a vasoconstrictive effect on cerebral arteries associated with sympathetic activity

• 

  Fig. 3. Lag-dependent cross-correlations between various signals, resulting from group level analysis (n = 7) of NREM2 sleep data. Cross-correlations which are above or below the dashed lines indicate statistical significance (p < 0.0125, corrected for multiple comparisons) against null hypothesis (see Methods). Strongest correlations of LF-EEG with all other signals occur at positive lags, indicating LF-EEG changes precede all other signals. Black lines are the mean temporal correlations of LF-EEG, PPG-AMP, and fMRIGM with all other signals. Colored areas indicate standard deviations across subjects. Strongest correlations (mean Fisher’s z) with corresponding lags in seconds, within parentheses, are given on the upper left corner of each subfigure. As example, for LF-EEG vs. PPG-AMP, a negative z-score (z = −0.49) at a positive lag (lag = +6) indicates an increase in LF-EEG followed by a PPG-AMP drop

• 

  Fig. 4. Confirmation of temporal relationship between K-complexes and fMRI with K-complex-aligned averaging of fMRI signals during NREM2 sleep. a Increased power around t = 0 s in the K-complex aligned average LF-EEG signal supports the use of LF-EEG as a surrogate for K-complexes. Black line and blue colored area give means and standard deviations (n = 7). b K-complex triggered average of fMRI signal shows pattern resembling that observed LF-EEG-fMRI correlation (Fig. 2b), suggesting that in much of gray matter K-complexes precede fMRI changes by about 6–12 s

• 

  Fig. 5. Schematic of arousal system showing sympathetic connections from superior cervical ganglion (SCG) to heart, peripheral, and CNS vessels, and to SCG from brainstem and input to nucleus tractus solitarius (NTS) from cortex and periphery. “Top-down” cortical input well as “bottom-up” receptor input to the brainstem can result in sympathetic output that simultaneously constricts vasculature in CNS and finger skin, resulting in joint changes in PPG-AMP and fMRI GS. In addition, pathways emanating from the brainstem exist that may effectuate coincident changes in heart rate and respiration. Freely adapted from ref. 34