Fluctuations in Arousal Correlate with Neural Activity in the Human Thalamus

Abstract

The neural basis of consciousness has been explored in humans and animals; however, the exact nature of consciousness remains elusive. In this study, we aimed to elucidate which brain regions are relevant to arousal in humans. Simultaneous recordings of brain activity and eye-tracking were conducted in 20 healthy human participants. Brain activity was measured by resting-state functional magnetic resonance imaging with a multiband acquisition protocol. The subjective levels of arousal were investigated based on the degree of eyelid closure that was recorded using a near-infrared eye camera within the scanner. The results showed that the participants were in an aroused state for 79% of the scan time, and the bilateral thalami were significantly associated with the arousal condition. Among the major thalamic subnuclei, the mediodorsal nucleus (MD) showed greater involvement in arousal when compared with other subnuclei. A receiver operating characteristic analysis with leave-one-out crossvalidation conducted using template-based brain activity and arousal-level data from eye-tracking showed that, in most participants, thalamic activity significantly predicted the subjective levels of arousal. These results indicate a significant role of the thalamus, and in particular, the MD, which has rich connectivity with the prefrontal cortices and the limbic system in human consciousness.

Keywords: awake; blink; consciousness; fMRI; rest.

Figure 1

A 3D atlas of the human thalamus is superimposed on the MNI brain template of MRIcron. Some small subnuclei are merged into a large VOI for the MD (left, red; right, green), VAL (left, yellow; right, cyan), and PUL (left, blue; right, magenta) in the left and right hemispheres. Coronal (y = −18), sagittal (x = −1), and axial (z = 7) images and 3D rendering of the VOIs are shown in the figure.

Figure 2

A large matrix with 1920 (scan number) rows and 1718 (voxels in the thalamus template) columns represents the data of a single participant with four scans. In an individual analysis, the time-series data were extracted from voxels in the thalamus template, z-transformed, and concatenated across the scans. A separate row in the bottom indicates z-transformed BOLD activity of the thalamus template (1718 voxels) averaged across the 17 participants in a group analysis (thalamus group mean). Pearson’s correlation coefficients between the thalamus group mean data and the individual thalamic activity data were computed from the first to the 1920th scan. The correlation coefficient was Fisher-transformed to create a single vector that represented the similarity between the template and an individual’s brain activity (a separate column on the right-hand side). The similarity vector [r(z)] was used in the ROC analysis.

Figure 3

A box plot of the mean blink rate (time per minute, A) and the proportion of drowsiness during the scan time (B) for 20 participants. The box indicates the first and the third quartiles. The horizontal bar and the cross represent the median and mean, respectively. The whiskers indicate plus/minus 1.5× interquartile range. Dots indicate outliers. (C) The histogram of the EC period across 20 participants is shown. The horizontal and vertical axes indicate the time bin (0.01 s, from 0.08 to 9.98 s) and counts, respectively. The black arrow indicates the 2-s threshold that separates blinks and drowsiness. Total number, mean, SD, and range of the EC period are shown in the figure. Approximately, 7% of the EC periods are labeled as drowsiness. (D) The proportion of arousal periods during the scan averaged across 78 runs from 20 participants. The horizontal axis indicates the scan number, and the vertical axis indicates the proportion of aroused participants. A black curved line indicates a fitted second-order polynomial function (Y = 1E − 06X² − 0.0011X + 0.9626, R² = 0.87, P < 0.01).

Figure 4

Left: A SPM glass-brain view of significant positive correlation with the arousal condition with GSR is shown (threshold at P = 0.001, uncorrected, for the peak-level, and P = 0.05, FWE-corrected, for the cluster level). The neuromorphometrics atlas of the left and right thalami proper in SPM12 is applied in the analysis. Right: The mean (column) and SD (bar) of the activation in each of the VOIs in the thalamic subnuclei are shown. L/R, left/right hemispheres.

Figure 5

Significant positive correlation with the arousal condition in the thalamus is superimposed on the canonical brain of SPM12. The results are with GSR. The statistical threshold was set at P = 0.001, uncorrected, for the peak-level, and P = 0.05, FWE-corrected, for the cluster level. The left side on the figure shows the left side of the brain. Axial images are shown from z = −4 to z = 18 mm in 2-mm increments.

Figure 6

A regional overlap between the thalamus mask (significant positive correlation with the arousal condition at P = 0.001, uncorrected, for the peak-level, and P = 0.05, FWE-corrected, for the cluster level) and each of the VOIs (MD, VAL, and PUL) is shown. The thalamus mask is colored green, the VOIs red, and the overlap yellow. The MD shows the maximal overlap with the mask, while the PUL shows the least overlap, and the overlap of the VAL falls between those of the MD and PUL. Numbers in the bottom indicate levels at the z-axis (mm). The results are with GSR.

Figure 7

Left: The results of the ROC analysis with GSR for each participant (n = 18) are shown. The horizontal and vertical axes indicate the false-positive rate (FPR) and true-positive rate (TPR), respectively. Each colored and curved line represents the result for each participant. In 16 of the 18 participants, the AUC was significantly (P < 0.05) greater than the chance level (0.5). A diagonal black line indicates a chance level of prediction. Right: A box plot of the AUC value across the 18 participants is shown. The mean AUC was significantly (P < 0.001) greater than the chance level (0.5).