Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jun 9.
Published in final edited form as: Neuroimage. 2026 Apr 30;334:121971. doi: 10.1016/j.neuroimage.2026.121971

Pupil Size is Linked to Opposite Signal Fluctuations Between Cortical BOLD and Ventricular CSF

Sukru Baris Demiral 1, Helene Benveniste 2, Nora D Volkow 1
PMCID: PMC13245481  NIHMSID: NIHMS2172894  PMID: 42066921

Abstract

Pupillometry has been used as a biomarker of activity in the Locus Coeruleus (LC), a noradrenergic nucleus crucial for arousal. As the LC has been implicated in modulating cerebrospinal fluid (CSF) flow, here we investigated the associations of pupil size with CSF T2-dependent signal in the ventricles, BOLD signal in cortical and subcortical brain regions (including LC), and peripheral physiology (heart rate and respiration) as a function of arousal.

We hypothesized that arousal changes indexed by pupil size would be associated with changes in CSF signal intensity and volume in opposition to brain BOLD signal. Analyses of fMRI data from healthy controls (HCP 7T) showed that changes in pupil size were linked with changes in LC, cortical and subcortical BOLD in opposition to signal changes in the lateral ventricles. CSF signal in the ventricles increased as the activity in LC, thalamus and cortical regions declined concomitant with decreases in pupil size. The coupling between pupil size and LC emerged earlier than for other ascending arousal nuclei and Granger causality analysis corroborated strong LC-pupil coupling. While physiological measures had stronger correlations with each other and with pupil size in the drowsy state, pupil-BOLD and CSF associations were stronger during the vigilant state. The weakened pupil-CSF association in the drowsy state suggests a segregation of central and peripheral signal modulations of CSF during low arousal states. Mental states that promote arousal might be beneficial in neuropsychiatric disorders with glymphatic system impairment and pupillometry could serve as a biomarker to monitor glymphatic function.

Keywords: CSF, BOLD, Pupil, Arousal, Respiration

Classification: Neuroscience

1. Introduction

Pupil size is used as a biomarker of arousal (McGinley, David, & McCormick, 2015; Podvalny, King, & He, 2021) such that pupil size increases correspond to increases in arousal and vice versa (Grujic, Polania, & Burdakov, 2024). The relationship between pupil size and arousal is mediated by the sympathetic and parasympathetic systems (McDougal & Gamlin, 2015; Podvalny et al., 2021) and by the ascending arousal network (AAN) including the locus coeruleus (LC) the main noradrenergic nucleus (van den Brink, Pfeffer, & Donner, 2019). Specifically, phasic and tonic pupil dilations index neural activity in norepinephrine (NE) and cholinergic (ACh) systems respectively; the temporal derivative (phasic increase or decrease) of pupil size precisely correlates with activity within these systems (Joshi, Li, Kalwani, & Gold, 2016; Reimer et al., 2016). Using fMRI we previously reported that the activity in AAN was temporarily associated with eye-blinks (Demiral, Kure Liu, Benveniste, Tomasi, & Volkow, 2023). Here we explored the associations between AAN activity (measured with BOLD) with pupil size and CSF signals to investigate their dynamic interactions as a function of arousal state. We also assessed their interaction with heart rate (HR) and respiratory variability (RV) as physiological markers of the peripheral autonomic system that have been shown to modulate CSF and BOLD signals (Bolt et al., 2025; Kandimalla et al., 2025).

Several studies have reported that the CSF signal is in opposition to cortical BOLD signals such that increased cortical signals are generally accompanied by reduced CSF signals in the ventricles (Bright, Bianciardi, de Zwart, Murphy, & Duyn, 2014) and cortical BOLD decreases when the CSF inflow signal increases--due to the spin-history-dependent signal intensity increase observable in the first few slices-- in the fourth ventricle (Fultz et al., 2019; Yang et al., 2022). Consistently, in studies of visual stimulation (on-off-on pattern) the CSF signal increased during the ‘off’ periods (Williams et al., 2023). The association between BOLD signal and CSF flow have mostly relied in inflow measures from the fourth ventricles (Gonzalez-Castillo, Fernandez, Handwerker, & Bandettini, 2022; Yang et al., 2022). Recently by using a deep learning method (‘SynthSeg’) (Billot, Greve, et al., 2023; Billot, Magdamo, et al., 2023) a novel study showed that temporal changes in the lateral and the fourth ventricular sizes were coupled to CSF slow oscillations (Cheng & Kennedy, 2025) providing evidence that CSF signal and the related ventricular volumetric estimations can be used to evaluate CSF related T2-weighted (T2w) changes. Moreover, CSF-global BOLD coupling has also been proposed as a potential biomarker of amyloid beta accumulation (Tanaka et al., 2025) and is disrupted in Alzheimer disease (Cho et al., 2025; Han et al., 2021).

Most of the clinical studies that have investigated the associations between BOLD and pupil size modeled it with the hemodynamic-response function (HRF). (DiNuzzo et al., 2019; Lloyd, de Voogd, Maki-Marttunen, & Nieuwenhuis, 2023; Murphy, O’Connell, O’Sullivan, Robertson, & Balsters, 2014; Schneider et al., 2016). However the pupil size shows fast (phasic) and slow (tonic) changes and a recent fMRI study (Lloyd et al., 2023) showed that pupil size and its temporal derivative had positive correlations with subcortical and cortical BOLD signals only when the HRF peak time was very short (~1s) or when HRF was not used. Weaker effects in brainstem nuclei were also observed, likely reflecting their smaller signal-to-noise (SNR) than the cortex (Brooks, Faull, Pattinson, & Jenkinson, 2013; Harvey et al., 2008). The temporal proximity between pupil size and BOLD changes suggest contributions from peripheral autonomic fluctuations that directly affect vascular networks (Ozbay et al., 2019; Picchioni et al., 2022), without the time-delays assumed by the HRF.

Peripheral autonomic signals contribute to the global BOLD signal (Bolt et al., 2025; Zhang & Northoff, 2022) through subcortical-cortical infra-slow phase-based mechanisms, which may reflect coordinated patterns of functional brain network activity (Zhang & Northoff, 2022) that are modulated by vigilance (Gu, Han, & Liu, 2019; Han, Gu, & Liu, 2019). Additionally, brain-stem arousal nuclei modulate vascular tone through sympathetic innervation of arteries on the brain’s pial surface (Ozbay et al., 2018).

In this paper we hypothesized that pupil size fluctuations, as an index of arousal would be temporarily associated with an opposing pattern between cortical, thalamic and AAN BOLD signals as well as CSF signal and ventricle volume. Specifically, BOLD increases would lead to CSF signal drops in the ventricles when pupil size increased. We also hypothesized that CSF-BOLD coupling with pupil size would be decreased and delayed in drowsy versus alert state due to reduced vascular reactivity. For this purpose we analyzed brain fMRI images and pupillometry data from the HCP 7T resting state data set and used a deep-learning approach (https://github.com/rgbayrak/deep-physio-recon) to extract heart rate (HR) and respiratory variability (RV) measures (Bayrak et al., 2021; Salas, Bayrak, Huo, & Chang, 2021), which were obtained to monitor spontaneous systemic autonomic fluctuations (Bianciardi, Fukunaga, van Gelderen, de Zwart, & Duyn, 2011; Bianciardi et al., 2009; Liu et al., 2018; Ozbay et al., 2019). We also estimated lateral and fourth ventricle volume changes (LLV and FVV respectively) in time in T2w signal, we used SynthSeg (Billot, Greve, et al., 2023; Billot, Magdamo, et al., 2023) utilizing series of 3D T2w images to calculate brain segmentation--as used in Cheng and Kennedy (2025). Scans were classified as vigilant when eye closures were <10% and compared with those classified as drowsy (eye closures between 10–40%). Specifically, we examined the following dynamic associations: (1) CSF signal changes in the ventricles and BOLD changes in cortex, thalamus, and AAN including LC; (2) pupil size changes with the CSF and BOLD signals; and (3) BOLD-CSF associations during vigilant and drowsy states and their association with HR and RV; (4) LVV and FV changes and their relationship with each other and LC in different states.

2. Materials and methods

2.1. 7T HCP dataset

The HCP 7T dataset contains data from 184 participants who underwent up to four resting state fMRI runs done over two separate days -two runs in each day interspersed with task scans (Glasser et al., 2013; Van Essen et al., 2013). Participants were positioned supine in the MR scanner and during the resting state scan stared at a cross on the center of the screen. In total, we found that 177 subjects had four runs of resting state fMRI scans, two had three runs, four had two and one participant had only one run in the database. Among these runs some lacked eye tracking data leading to 131 participants with four runs of resting state and eye tracking data, 14 participants with 3 runs, one with two runs and two participants with one run (leading to 570 runs across 148 participants). Of these, 10 runs had to be excluded due to problems in the eye-tracking file either not being recorded properly, missing data points, or showing problems in opening/importing files into the computer software. Final set comprised 560 runs across 148 participants.

2.2. Eye-tracking

Eye-tracking was conducted with the Eye Link 1000 system with 1 kHz sampling rate (some runs were with 500 Hz) from the right eye (www.humanconnectome.org/storage/app/media/documentation/s1200/HCP_S1 200_Release_Reference_Manual.pdf). We conducted the following preprocessing steps: (i) used R package ‘eyelinker’ (https://cran.r-project.org/web/packages/eyelinker/vignettes/basics.html) to extract all runs with usable continuous eye-tracking data, (ii) followed by MATLAB routines, runs with 500 Hz sampling rate were up-sampled to 1 kHz, (iii) time period between 40–400ms of pupil loss were marked as eye-blinks, (iv) those missing data points due to eye-blinks were linearly interpolated with the neighboring points, (v) any other missing values were replaced by zero due to filtering requirements, finally (vi) pupil size time series were down-sampled to 1Hz to match with the fMRI TR resolution. We used the temporal derivative of the pupil time series.

2.3. Final selection of runs and participants

Each resting scan/run was classified into one of three states: (a) vigilant (when less than 10% time of the run the pupil was not detected); (b) drowsy (10–40% pupil loss in a run); (c) discarded (>40% pupil loss in a run). Similar analysis was reported as in (Demiral & Volkow, 2024).

Of the total 560 runs from the 7T HCP dataset, 291 were classified as vigilant (52% of the total runs across 111 subjects) and 111 as drowsy (19.8% of the total runs across 75 subjects). 58 subjects had both vigilant and drowsy runs available. Thus, a total of 402 runs across 128 participants were used in the final analysis, split into 291 vigilant and 111 drowsy states.

2.4. fMRI preprocessing

The HCP fMRI data has been further processed by i) applying weak high-pass temporal filtering (> 2000s FWHM) achieving slow drift removal, ii) running Independent Component Analysis (ICA) based correction method (MELODIC ICA, where artefact components were identified the procedure labelled as ICA-FIX), and iii) regressing out artefact and motion-related time courses (Smith et al., 2013). Thus, we used the final cleaned datasets labelled as “rfMRI_RESTx_7T_PA_hp2000_clean.nii.gz”. Here the ‘RESTx’ specifies the order of the resting state scan of the participant (i.e., REST1, REST2 etc.), and ‘AP’ specifies either AP or PA phase direction, and ‘hp2000’ is the high-pass filtering value (T = 2000s, f = 1/2000 Hz). Framewise displacement (FD) motion estimates are also reported. We additionally applied spatial smoothing (FWHM = 5mm) for the whole brain analysis (not for AAN ROI analysis, see below).

2.5. Extraction of heart rate and respiratory variability from fMRI

HCP 7T minimally processed data (before the application of ICA-FIX) (Glasser et al., 2013) , which includes distortion correction, motion correction, bias field correction and spatial normalization to the MNI template space was used for the deep-physio-recon application (Bayrak et al., 2021) (https://github.com/rgbayrak/deep-physio-recon). This approach uses a bidirectional Long Short-Term Memory (BiLSTM) model that processes sequential data in both directions in time. In the original training, the HCP 3T data was used, and a set of brain region of interests were selected from a set of human brain atlases (i.e., cortical and subcortical segmentations, white matter diffusion maps, brainstem arousal nuclei), and their BOLD signal time series were fed into the model, which contained multiple deep layers. Expected/known physiological responses were used as the outcomes to train the model. The model learns the associations and connection strengths in the layers and can be used to infer the unknown physiological signals (Respiratory variability; RV, and heart rate; HR) from the fMRI data in the HCP 7T dataset. The model needs minimally processed data without motion or other nuisance regressors as such signals can bear physiological information driven by factors, such as respiration.

2.6. Cross-correlation analysis of pupil, HR and RV

Extracted HR, RV and pupil size and derivative values are then combined to calculate correlations between each other in lags from −5 to 5 TRs for each individual scan in MATLAB. Correlations (r) were Fisher’s z transformed as 0.5 * log((1 + r) / (1 − r)). Group level plots and statistics for each pair-wise correlation were reported such that any ANOVA result comparing vigilant and drowsy states was only reported if the p-value was less than 0.0045 (to minimize false positives due to multiple comparisons across lags; alpha=.05/11 lags).

2.7. Statistical analysis of whole brain pupil activity analysis

SPM statistical software was used to generate t-tests separately for vigilant and drowsy scans for the pupil time series as follows, i) for pupil size analysis; in the 1st level SPM analysis, we used pupil size, RV time series and 12 motion estimates as regressors, and fMRI time series was used as the outcome variable, ii) for pupil derivative analysis; in the 1st level SPM analysis we used pupil time series derivative, RV time series derivative and motion estimates as regressors and the derivative of the fMRI time series as the outcome variable. SPM does not do HRF convolution for regressors, so our analysis aims for direct changes in T2w signal. In the pupil derivative analysis, we wanted to explore phasic and instantaneous pupillary effects on the BOLD signal, while in the pupil size analysis our goal was to explore slower changes.

2.8. Statistical thresholding and multiple comparison correction of whole brain and ROI data

We generated brain maps of the significant clusters using AFNI 3dClustSim with the global mask image only containing the brain voxels (total of 517125 voxels in the mask each with size of 1.60mm iso; NN=2 using both corner and side neighborhood, 2-tailed, FWHM=5). We report a total of 11 lags [from −5 to +5; as previous studies, (Bolt et al., 2025; Yang et al., 2022) showed CSF-BOLD correlations close to 2–3s] for two states (vigilant and drowsy) thus we chose voxel p-threshold value of p=.002, under cluster level alpha value of .05. This analysis led us to use a final selection of clusters k>60, voxel-t > 3.1. For the ROI analysis and plots we used t-value of 3 as a threshold; df =110 for drowsy runs and df=290 for vigilant states; under p=.0045 (corrected for 11 lags), t =~3)

2.9. Ascending arousal nuclei and selected ROIs

In addition to the whole-brain analysis, we used various ROIs to depict time-dependent changes in CSF and blood related fMRI signal. Four 4mm-spherical ROIS (MNI; x, y, z in mm); two of them the right anterior calcarine sulcus (x=10, y=−53, z=12) (Demiral et al., 2023) and the right dorsal medial thalamus (x=10, y=−12, z=8), which have been linked to arousal (Najdenovska et al., 2018; Setzer et al., 2022) and two the fourth ventricle (x=0, y=−42, z=−28), and the right lateral anterior ventricle (x=6, y=14, z=14), which correspond to CSF comprising region (Supp. Fig 1). We also selected fours ROIs from catecholaminergic and cholinergic brainstem nuclei in the AAN (Bianciardi et al., 2015) (https://www.nmr.mgh.harvard.edu/resources/aan-atlas) corresponding to the serotonergic dorsal Raphe (DR), the noradrenergic locus coeruleus LC, the dopaminergic VTA, and the cholinergic periaqueductal gray (PAG) (Supplementary material for anatomical locations). Values were extracted from the non-smoothed brain images.

We compared differences in the temporal associations between pupil size and activity in AAI nuclei for the correlations within 0s to 2s lags with ANOVA and t-tests for subsequent paired comparisons

Granger causality analysis was done between the pupil measures and the AAN nuclei of DR, LC, PAG, VTA, and the fourth ventricle ROI. We conducted this analysis for the pupil size and the pupil derivative separately. For this purpose, we used “bruceR” package in R, with model order of 3. χ2 values for each pair of causality response per scan was calculated and the mean values reported. We compared vigilant and drowsy states using t-tests in the mean χ2 values calculated for all the 10 pairs of causality measures (i.e., LC->Pupil or Pupil->LC) so that the threshold alpha value after multiple-comparisons was set as .005.

2.10. SynthSeg temporal ventricle volumetric analysis (LLV/FVV) of the fMRI data

Lastly, to estimate lateral and fourth ventricle volume changes (LLV and FVV respectively) in time in T2w signal, we used SynthSeg (Billot, Greve, et al., 2023; Billot, Magdamo, et al., 2023) utilizing series of 3D T2w images to calculate brain segmentation--as used in Cheng and Kennedy (2025). First, HCP 7T minimally-processed data was inverse warped back to the individual T1w space with fsl function “applywarp”. Then, each volume was saved as a separate image with AFNI command “3dTsplit4D”. We applied the function “mri_synthseg” function which is the version of SynthSeg available as a build-in function in freesurfer and bandpass filtered the volume time series (.01Hz-.1Hz). Finally, we examined cross-correlation between the SynthSeg FVV and LVV with the T2w signals in LC and corresponding ROIs (lateral ventricle and fourth ventricle ROIs extracted above) to detect arousal-state dependent temporal changes between these measures (In xcorr function, SynthSeg values were always the second argument; positive lag values indicated SynthSeg leading the other signal.)

3. Results

3.1. Cross-correlation analysis of pupil, HR and RV

Fig 1. shows cross-correlation results between HR and RV and their derivatives. We observed the maximal correlation for both time series and their derivatives at lag +1 (HR followed RV 1s). Correlations in the drowsy state were higher than in the vigilant state reaching maximal values in lag +1: mean(se): vigilant; r=.408(.009), drowsy; r=.498(.0142), and for derivative, vigilant; r=.30(.008), drowsy; r=.409(.017). The ANOVA comparing vigilant and drowsy states (after correcting for multiple comparisons; alpha=.05/11 lags) showed significant differences between states, which are identified in Fig. 1 with a red star for significant lags.

Figure 1. Correlation between HR and RV.

Figure 1.

Open in a new tab

Time-series on the left and their derivative time-series on the right for lags between −5s to +5s for two vigilance states. Error bars represent standard errors. Positive values indicate HR follows. Red stars indicate vigilant-drowsy significant difference corrected for multiple comparisons; .05/11 lags; alpha=.0045). Correlations were Fisher’s z transformed.

The cross-correlation results between pupil size and RV and their derivatives are shown in Fig. 2. We observed the maximal correlation at around 0s lag. These correlations were higher in the drowsy than in the vigilant state for peak values in lag 0s: mean(se): vigilant; r=.163(.011), drowsy; r=.326(.0186), and for their derivatives at lag +1s, vigilant; r=.071(.004), drowsy; r=.118(.011). ANOVA comparing vigilant and drowsy states (corrected for multiple comparisons; alpha=.05/11 lags) showed significant differences for the lag shown in Fig. 2 with red stars.

Figure 2. Correlation between Pupil and RV.

Figure 2.

Open in a new tab

Time-series on the left and derivatives of time-series on the right for lags between −5s to +5s for two vigilance states. Error bars represent standard errors. Positive values indicate pupil follows. Positive values indicate HR follows. Red stars indicate vigilant-drowsy significant difference corrected for multiple comparisons; .05/11 lags; alpha=.0045). Correlations were Fisher’s z transformed.

3.2. Pupil BOLD associations

Temporal associations between changes in pupil size and BOLD signals in cortical, subcortical and ventricular regions and their derivatives for the vigilant and drowsy state for the SPM t-test results are shown in Figs. 3–5 (also see Supplementary Figs 2-5 for whole brain sagittal and axial images). Whole brain activation maps for the 0 to 2s lag showed earlier negative CSF correlation with the pupil followed by positive cortical correlations mainly in the default mode network (DMN), prefrontal cortex, calcarine sulcus, and thalamus, and negative correlations with sensory-motor cortices. Occipital lobe and early visual areas showed opposing patterns in two different arousal states; post-pupil negative correlations in the vigilant state and positive correlations in the drowsy state (Fig. 3).

Figure 3. fMRI signal correlations with pupil measures.

Figure 3.

Open in a new tab

Summary plot of the time-dependent correlations between pupil and the whole brain for the 0 to 2s lag and its derivative for the vigilant and drowsy states.

Figure 5. Correlations of the T2*w signal in AAN with pupil measures.

Figure 5.

Open in a new tab

Pupil size (top) and its derivative (bottom) and the BOLD signal in AAN ROIs for vigilant (left) and drowsy (right) states. Positive values indicate T2*w follows. T-threshold was set to t=3 (i.e., minimal is df=110 -drowsy scans- and alpha value set to .0045 under multiple comparison of 11 lags).

Average t-values calculated across voxels in the preselected ROIs were plotted per lag. We found positive correlations in the cortical and AAN ROIs around 2s-3s following pupil size changes while negative CSF correlations were observed 1s-2s before the maximal cortical peak correlations (Figs 4 and 5).

Figure 4. Correlations of the pupil measures with the T2*w signals in gray matter and ventricle’s ROIs.

Figure 4.

Open in a new tab

With pupil size (top) and its derivative (bottom), and t-test results for the different lags for the vigilant (left) and drowsy (right) states. Positive values indicate T2w* signal follows. T-threshold was set to t=3 (i.e., minimal is df=110 - drowsy scans- and alpha value set to .0045 under multiple comparison of 11 lags).

We then explored differences between the AAN-nuclei pupil correlations for the vigilant and drowsy states (Figs. 6 and 7). One-way ANOVA showed that the pupil size correlations with the AAN nuclei were significant both in the vigilant [F(3,1160)=3.08, p=.026], and the drowsy states [F(3,440)=4.58, p=.0036]. Similarly, for the pupil derivative the correlations with AAN nuclei were significant in the vigilant [F(3,1160)=5.79, p<.0001], and drowsy states [F(3,440)=3.56, p=.014]. The strength of the correlations for the vigilant state in the LC (paired t-tests) were stronger than for DR (p<.0001), PAG (p<.005), and VTA (p<.001) and for the drowsy state the LC correlations were stronger than for DR (p<.001), PAG (p<.0001), and VTA (p<.0001) (Fig. 6). For the pupil derivative the strength of the correlations was also stronger for the LC during the vigilant state in DR (p<.005), PAG (p<.002) and VTA (p<.002) and for the drowsy state in VTA (p<.005), and marginally PAG (p=.02) (Fig. 7).

Figure 6. Pupil size and LC activity.

Figure 6.

Open in a new tab

Pupil size – AAN BOLD correlations within 0s to +2s lags (blue boxes on the right side) show that in vigilant state (top) and drowsy (bottom) LC-Pupil correlations emerged earlier compared to the other AAN nuclei. Multiple correction applied paired t-test results are shown comparing LC with other AAN nuclei. Shaded regions in the plots on the right show 95% confidence intervals (CI).

Figure 7. Pupil derivative and LC activity.

Figure 7.

Open in a new tab

Pupil derivative – BOLD correlations within 0s to +2s lags (blue boxes on the right side) show that in vigilant state (top) and drowsy (bottom) LC-Pupil derivative correlations emerge earlier compared to the other AAN nuclei. Multiple correction applied paired t-test results are shown comparing LC with other AAN nuclei. Shaded regions in the plots on the right show 95% confidence intervals (CI).

3.3. Causality analysis

We found strong causality between pupil measures with LC (mean χ2 >9 for pupil size and >10 for pupil derivative) that did not differ significantly between the vigilant and drowsy states (Supplementary Tables 1 and 2). In contrast stronger causal relations emerged in the vigilant compared with the drowsy state between pupil measures and DR (p<.001), PAG (p<.005) and marginally in VTA (p<.01; multiple-comparisons under 10 causal relations of interest; threshold-alpha=.005). There was also a non-significant trend in causality between the fourth ventricle and pupil size derivative in the drowsy state (p=.021).

3.4. SynthSeg temporal LVV and FVV change results

SynthSeg cross-correlation between the temporal dynamics of LLV and the FVV showed that in the drowsy state, correlations were larger than in the vigilant state and that the FVV change happened 1s before the LLV change. In the vigilant state the correlations were smaller, and volume changes happened simultaneously as maximum correlation emerged at 0 lag. SynthSeg volume cross-correlation analysis with LC showed that there was a strong negative correlation (larger for the lateral ventricles and in the drowsy state). The FVV correlation was minimal at 1s lag while LVV and LC showed minimal correlation at 0 lag (Supplementary Figure 6). This indicated that the LVV (and its correlation with LC) followed the FVV. SynthSeg volume cross-correlation with corresponding ventricle ROIs showed positive correlation (stronger for the lateral ventricles and drowsy state). The FVV change was at around 1s lag (preceding) while LVV and ROI at the right lateral ventricle showed minimal correlation at 0 lag (Supplementary Figure 7). Again, this indicated that lateral volumes followed the fourth ventricle in volumetric changes (and its correlation with the T2w/T2* signal observed in these ROIs).

4. Discussion

This study supports a link between changes in pupillary size and CSF-BOLD co-fluctuations as examined in two arousal states (vigilant and drowsy) and provides preliminary evidence that physiological systemic fluctuations might play a role in this coupling. Specifically, we showed that the RV and HR correlations were stronger in the drowsy than in the vigilant state with RV followed by HR and the maximal correlation between them at lag +1s. Moreover, we showed that pupil size and its derivative were correlated with RV, such that RV preceded pupil size changes and drowsiness increased this correlation; RV synchronized with (in size modulation) or slightly preceded (1s; in derivative analysis) the pupil size in an arousal state-dependent manner consistent with the sympathetic modulation of RV.

In the fMRI signal analysis, we showed that pupil size increases were associated with CSF signal decreases. This was shown in the ventricles within 1s followed by a signal increase in major cortical and thalamic regions and in subcortical arousal nuclei within 2s. We did not see this in primary visual and sensory motor cortices, which showed the opposite associations. Remarkably, the strength of these associations was weaker for the drowsy compared to the vigilant state. We also showed that the correlation in the AAN was strongest for LC. Overall, our study demonstrate that pupil measures provide information about CSF and BOLD fluctuations which occurred in opposition of one another and were stronger during vigilant than drowsy states.

Our analysis shows supporting evidence that the pupil size-BOLD associations in LC exerts a stronger role than the other AAN nuclei evaluated. Optogenetic and chemo-genetic studies of the LC have confirmed a strong linkage between its activation with pupillary diameter and its necessary role in changing arousal levels (Poe et al., 2020). The LC has also been hypothesized to play a gating role in the activation of dopaminergic VTA under novel and salient environmental cues (Fois et al., 2022).

Heart rate and respiratory variability are controlled by sympathetic and parasympathetic signals (Borovkova et al., 2022; Charkoudian & Rabbitts, 2009; de Bernardo et al., 2025). Falling asleep and awakening have been associated with differing degrees of synchronizations of respiration and heart rate with brain BOLD signals (Soon et al., 2021). The coupling of the autonomic and central nervous systems is influenced by arousal and affected by sleep deprivation (Borovkova et al., 2022; Bright, Whittaker, Driver, & Murphy, 2020; Gold et al., 2024; Shokri-Kojori, Tomasi, & Volkow, 2018). Thus, our findings of increased HR-RV correlations during the drowsy state support the view that as people lose vigilance and become drowsy, the autonomic system increases its influence on these physiological systems. This is consistent with prior studies that reported that the influence of the autonomic system differed in various conditions (i.e., aging, sleep stage) (Ozbay et al., 2018; Peng et al., 2008) and its influence in the brain was dependent on the vascular architecture of the different cortical regions (Tong et al., 2016).

Using Magnetoencephalography (MEG), a study that explored the link between pupil size and central arousal systems in humans found that peak timing for spontaneous pupil dilations were associated with decreases in low-frequency (2–8 Hz) activity in temporal and lateral frontal cortex followed by increased high-frequency (>64 Hz) activity in mid-frontal regions (Pfeffer et al., 2022). An optical imaging study in mice found that variations in pupil dilation during awake stillness were preceded by activation of NE terminals whereas the activation of ACh axons was associated with persistent pupil dilation during locomotion (Reimer et al., 2016). In an fMRI study, peaks of pupil dilation coincided with opposite responses between the DMN and sensorimotor cortex (Yellin, Berkovich-Ohana, & Malach, 2015).

Covariation between activity in AAN and in subcortical and cortical regions and the autonomic nervous systems have been previously reported by various studies in humans (Demiral et al., 2023; Demiral & Volkow, 2024; DiNuzzo et al., 2019; Lloyd et al., 2023; Murphy et al., 2014; Schneider et al., 2016). Recently we showed that the pupil size changes that follow eye-blinks were heritable, which we ascribed to a genetic modulation of the autonomic nervous system (Demiral & Volkow, 2024).

We found that LC BOLD-pupil coupling was stronger for the LC than for the other AAN nuclei. This might reflect a crucial role of this particular adrenergic nuclei in initiating arousal (Fois et al., 2022; Poe et al., 2020) and its strong association with pupil activity support the use of pupillometry as an arousal measure. In parallel, we observed that Granger causality analysis with model order of three (3 seconds lag) showed a missing pupil-LC causal relationship at this lag for pupil size. This indicates that the immediate pupil-LC correlation should emerge in shorter time scales (i.e., order <2), whereas significant PAG, VTA and DR causality with pupil size emerged after, perhaps reflecting a more delayed tuning of the activity of these other nuclei. This interpretation would be consistent with previous findings in mice that in the cortex norepinephrine (NE) preceded cholinergic (ACh) activity linked to pupil size changes (Reimer et al., 2016).

Cortical BOLD and the ventricles’ CSF signal have previously been shown to have inverse patterns (Fultz et al., 2019; Yang et al., 2022). These signals change either due to the T2* signal fluctuations (i.e., susceptibility changes in the deoxy- and oxy-hemoglobin ratio in the blood; neurovascular coupling) or due to an inflow-related effect driven by cardiac driven CSF pulsations or due to water outflow from the Choroid Plexus into and through the ventricles. However the T2w signal intensity changes in CSF can also be explained by an increased number of CSF water molecules as partially described by the Monro–Kellie doctrine where a change in one compartment (i.e., blood) will cause changes in the other compartments (i.e., CSF) (Mokri, 2001).

When we examined the relationship between the slow volumetric changes found in SynthSeg and their relationship with the LC and selected ROIs we found that the fourth ventricle volume changes preceded lateral ventricles and that the drowsy state had the strongest correlations. These findings indicated that ventricle T2w signal bear information related to CSF and ventricular volumetric changes. Interestingly volumetric changes in the fourth ventricle could happen slightly earlier than the T2w signal increases in the close regions. Leading LC signal was maximally correlated with pupil size as well as anti-correlated with FVV/LVV indicated the existence of arousal driven effects on the autonomic system and ventricle size. In addition, in their study Cheng & Kennedy (2025) reported that maximal positive correlation (r=.73) emerged when <mi> preceded fourth ventricle CSF inflow signal by ~1.4s. In our study, we found that FVV and LVV volume changes emerged in proximity in an arousal state dependent manner (i.e., correlation between FVV and LVV reached maximum levels when LVV followed FVV in the drowsy state). Combining our findings with Cheng and Kennedy’s findings, we think that both studies are complementing each other; LC and arousal system initiates the enlargement of ventricles (and potentially vasoconstrict the blood vessels) forming the necessary and sufficient condition for the upcoming pulsatile CSF inflow to emerge. CSF inflow is a phasic pulsatile measure mainly driven by cardiac cycle, and its observable strength may partially depend on the state and size of the ventricular spaces.

Given the pupil and cardio-pulmonary relationships modulated through the autonomic nervous system and the link between CSF and global signal anti-correlations, our study demonstrated that pupil size changes directly or indirectly relate to CSF fluctuations. Our findings provide evidence that association between pupil size and associated changes in BOLD activity might serve as indicator of CSF dynamics and glymphatic flow in the brain. This might be a useful clinical biomarker as coupling between the global fMRI signal and CSF influx was found to correlate with AD-related pathology (Han et al., 2021) and amyloid beta accumulation (Tanaka et al., 2025). CSF flow, production, and ventricular volumetric CSF changes can influence their synchronicity with the cortical (and related vascular and neural) BOLD signals due to reduced efficacy of the waste-clearance mechanisms in pathological cases.

One potential limitation of our study relates to BOLD measures in very small regions such as AAN nuclei which make them susceptible to partial volume effects (Ludwig et al., 2024; Yi et al., 2023). There are MR sequences like neuromelanin imaging (Clewett et al., 2016) which can provide more accurate information about the size and location of brainstem nuclei. While HCP 7T spatial resolution is relatively good, additional MNI normalizations and potential low SNR effects may lead to biased estimation of the actual activity. Thus, our results should be interpreted with caution. In addition, using multi-echo images in the future can segregate T2w proton density effects and T2* susceptibility effects in and near the ventricular regions and cortical areas near the large arteries and veins.

Future studies that evaluate the association between pupil size and CSF flux and its changes in neurodegenerative diseases are needed to determine its potential as a clinical biomarker for assessing disruption in arousal in neurodegenerative diseases and their consequences to glymphatic flow.

Supplementary Material

1

Highlights.

  • BOLD and ventricular CSF T2w signals are counter-associated with changes in pupil size.

  • Vigilance changes pupil-linked arousal effects on resting BOLD and CSF signal.

  • LC-pupil coupling precedes and leads other arousal nuclei activation.

  • Pupil-BOLD associations suggest pupillometry could serve as a biomarker of glymphatic function.

Acknowledgments

We would like to thank Human Connectome Project team for providing the data. This work utilized the computational resources of the NIH HPC Biowulf cluster (https://hpc.nih.gov). We also thank NIH Biowulf cluster and Biowulf team for their support.

Data Source and Ethical Approval

Data used in this study were provided by the WU-Minn Human Connectome Project (HCP), funded by the NIH Institutes and Centers that support the NIH Blueprint for Neuroscience Research, and by the McDonnell Center for Systems Neuroscience at Washington University. The HCP 7T data were collected under the approval of the Washington University in St. Louis Institutional Review Board (IRB #201201036). All participants provided informed consent, including consent for sharing de-identified, restricted, and imaging data with the scientific community.

Data Privacy and Protection

The authors accessed the HCP 7T data through the ConnectomeDB database, adhering to the HCP Open Access and/or Restricted Data Use Terms. While the HCP removes explicit identifiers, the investigators acknowledge that high-resolution 7T imaging and detailed behavioral data are not considered fully anonymized and may allow for the identification of individuals. Therefore, all data were handled in accordance with the strict data use terms, ensuring secure storage, and maintaining confidentiality throughout the analysis.

Compliance

This study complies with all relevant ethical regulations for human subject research. The authors have ensured that data access and usage align with the IRB guidelines of NIH.

Footnotes

Competing Interest Statement: No competing interests.

Ethics Statement

Declaration of interest

We do not have any conflicting interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Data Availability

Human dataset used in this study is publicly available from the HCP1200 database where data can be downloaded via the Connectome DB software (https://www.humanconnectome.org/study/hcp-youngadult/article/reprocessed-7t-fmri-data-released-other-updates). Rest of the data processing scripts and summary data produced by the authors will be available at https://github.com/demiralsb/Blink-Arousal upon the receipt of this manuscript.

References

  1. Bayrak RG, Hansen CB, Salas JA, Ahmed N, Lyu I, Huo YK, & Chang CT (2021). From Brain to Body: Learning Low-Frequency Respiration and Cardiac Signals from fMRI Dynamics. Medical Image Computing and Computer Assisted Intervention - Miccai 2021, Pt Vii, 12907, 553–563. doi: 10.1007/978-3-030-87234-2_52 [DOI] [Google Scholar]
  2. Bianciardi M, Fukunaga M, van Gelderen P, de Zwart JA, & Duyn JH (2011). Negative BOLD-fMRI signals in large cerebral veins. J Cereb Blood Flow Metab, 31(2), 401–412. doi: 10.1038/jcbfm.2010.164 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bianciardi M, Fukunaga M, van Gelderen P, Horovitz SG, de Zwart JA, & Duyn JH (2009). Modulation of spontaneous fMRI activity in human visual cortex by behavioral state. Neuroimage, 45(1), 160–168. doi: 10.1016/j.neuroimage.2008.10.034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bianciardi M, Toschi N, Edlow BL, Eichner C, Setsompop K, Polimeni JR, . . . Wald LL. (2015). Toward an In Vivo Neuroimaging Template of Human Brainstem Nuclei of the Ascending Arousal, Autonomic, and Motor Systems. Brain Connect, 5(10), 597–607. doi: 10.1089/brain.2015.0347 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Billot B, Greve DN, Puonti O, Thielscher A, Van Leemput K, Fischl B, . . . Adni. (2023). SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining. Medical Image Analysis, 86, 102789. doi: 10.1016/j.media.2023.102789 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Billot B, Magdamo C, Cheng Y, Arnold SE, Das S, & Iglesias JE (2023). Robust machine learning segmentation for large-scale analysis of heterogeneous clinical brain MRI datasets. Proc Natl Acad Sci U S A, 120(9), e2216399120. doi: 10.1073/pnas.2216399120 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bolt T, Wang S, Nomi JS, Setton R, Gold BP, de BFB, . . . Chang C. (2025). Autonomic physiological coupling of the global fMRI signal. Nat Neurosci, 28(6), 1327–1335. doi: 10.1038/s41593-025-01945-y [DOI] [PubMed] [Google Scholar]
  8. Borovkova EI, Prokhorov MD, Kiselev AR, Hramkov AN, Mironov SA, Agaltsov MV, . . . Penzel T. (2022). Directional couplings between the respiration and parasympathetic control of the heart rate during sleep and wakefulness in healthy subjects at different ages. Front Netw Physiol, 2, 942700. doi: 10.3389/fnetp.2022.942700 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Bright MG, Bianciardi M, de Zwart JA, Murphy K, & Duyn JH (2014). Early anti-correlated BOLD signal changes of physiologic origin. Neuroimage, 87, 287–296. doi: 10.1016/j.neuroimage.2013.10.055 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Bright MG, Whittaker JR, Driver ID, & Murphy K (2020). Vascular physiology drives functional brain networks. Neuroimage, 217, 116907. doi: 10.1016/j.neuroimage.2020.116907 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Brooks JC, Faull OK, Pattinson KT, & Jenkinson M (2013). Physiological noise in brainstem FMRI. Front Hum Neurosci, 7, 623. doi: 10.3389/fnhum.2013.00623 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Charkoudian N, & Rabbitts JA (2009). Sympathetic neural mechanisms in human cardiovascular health and disease. Mayo Clin Proc, 84(9), 822–830. doi: 10.4065/84.9.822 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cheng H, & Kennedy DP (2025). Temporal fluctuation in lateral ventricle volume and its coupling with CSF inflow and global BOLD signal. Scientific Reports, 15(1). doi:ARTN 30537 10.1038/s41598-025-15842-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Cho G, O’Dell RS, Liu X, Benveniste H, Kay D, van Dyck CH, . . . Mecca AP. (2025). Coupling between global gray matter BOLD activity and CSF movement is associated with Aβ and tau burden in preclinical Alzheimer’s disease. Alzheimer’s & Dementia, 21(S1), e106025. doi: 10.1002/alz70855_106025 [DOI] [Google Scholar]
  15. Clewett DV, Lee TH, Greening S, Ponzio A, Margalit E, & Mather M (2016). Neuromelanin marks the spot: identifying a locus coeruleus biomarker of cognitive reserve in healthy aging. Neurobiol Aging, 37, 117–126. doi: 10.1016/j.neurobiolaging.2015.09.019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. de Bernardo N, de la Rubia Orti JE, Villaron-Casales C, Privado J, Maset-Roig R, Canabate M, . . . Alarcon-Jimenez J. (2025). Autonomic nervous system and mediating role of respiratory function in patients with ALS. Sci Rep, 15(1), 10513. doi: 10.1038/s41598-025-94844-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Demiral SB, Kure Liu C, Benveniste H, Tomasi D, & Volkow ND (2023). Activation of brain arousal networks coincident with eye blinks during resting state. Cereb Cortex, 33(11), 6792–6802. doi: 10.1093/cercor/bhad001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Demiral SB, & Volkow ND (2024). Blink-induced changes in pupil dynamics are consistent and heritable. Sci Rep, 14(1), 28421. doi: 10.1038/s41598-024-79527-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. DiNuzzo M, Mascali D, Moraschi M, Bussu G, Maugeri L, Mangini F, . . . Giove F. (2019). Brain Networks Underlying Eye’s Pupil Dynamics. Front Neurosci, 13, 965. doi: 10.3389/fnins.2019.00965 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Fois GR, Bosque-Cordero KY, Vazquez-Torres R, Miliano C, Nogues X, Jimenez-Rivera CA, . . . Georges F. (2022). Locus coeruleus activation during environmental novelty gates cocaine-induced long-term hyperactivity of dopamine neurons. iScience, 25(4), 104154. doi: 10.1016/j.isci.2022.104154 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Fultz NE, Bonmassar G, Setsompop K, Stickgold RA, Rosen BR, Polimeni JR, & Lewis LD (2019). Coupled electrophysiological, hemodynamic, and cerebrospinal fluid oscillations in human sleep. Science, 366(6465), 628–631. doi: 10.1126/science.aax5440 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Glasser MF, Sotiropoulos SN, Wilson JA, Coalson TS, Fischl B, Andersson JL, . . . Consortium, W.-M. H. (2013). The minimal preprocessing pipelines for the Human Connectome Project. Neuroimage, 80, 105–124. doi: 10.1016/j.neuroimage.2013.04.127 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Gold BP, Goodale SE, Zhao C, Pourmotabbed H, de Zwart JA, Özbay PS, . . . Chang C. (2024). Functional MRI signals exhibit stronger covariation with peripheral autonomic measures as vigilance decreases. Imaging Neuroscience, 2, 1–25. doi: 10.1162/imag_a_00287 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Gonzalez-Castillo J, Fernandez IS, Handwerker DA, & Bandettini PA (2022). Ultra-slow fMRI fluctuations in the fourth ventricle as a marker of drowsiness. Neuroimage, 259, 119424. doi: 10.1016/j.neuroimage.2022.119424 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Grujic N, Polania R, & Burdakov D (2024). Neurobehavioral meaning of pupil size. Neuron, 112(20), 3381–3395. doi: 10.1016/j.neuron.2024.05.029 [DOI] [PubMed] [Google Scholar]
  26. Gu Y, Han F, & Liu X (2019). Arousal Contributions to Resting-State fMRI Connectivity and Dynamics. Front Neurosci, 13, 1190. doi: 10.3389/fnins.2019.01190 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Han F, Chen J, Belkin-Rosen A, Gu Y, Luo L, Buxton OM, . . . Alzheimer’s Disease Neuroimaging, I. (2021). Reduced coupling between cerebrospinal fluid flow and global brain activity is linked to Alzheimer disease-related pathology. PLoS Biol, 19(6), e3001233. doi: 10.1371/journal.pbio.3001233 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Han F, Gu Y, & Liu X (2019). A Neurophysiological Event of Arousal Modulation May Underlie fMRI-EEG Correlations. Front Neurosci, 13, 823. doi: 10.3389/fnins.2019.00823 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Harvey AK, Pattinson KT, Brooks JC, Mayhew SD, Jenkinson M, & Wise RG (2008). Brainstem functional magnetic resonance imaging: disentangling signal from physiological noise. J Magn Reson Imaging, 28(6), 1337–1344. doi: 10.1002/jmri.21623 [DOI] [PubMed] [Google Scholar]
  30. Joshi S, Li Y, Kalwani RM, & Gold JI (2016). Relationships between Pupil Diameter and Neuronal Activity in the Locus Coeruleus, Colliculi, and Cingulate Cortex. Neuron, 89(1), 221–234. doi: 10.1016/j.neuron.2015.11.028 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kandimalla M, Lim S, Thakkar J, Dewan S, Kang D, In MH, . . . Min HK. (2025). Cardiorespiratory dynamics in the brain: Review on the significance of cardiovascular and respiratory correlates in functional MRI signal. Neuroimage, 306, 121000. doi: 10.1016/j.neuroimage.2024.121000 [DOI] [PubMed] [Google Scholar]
  32. Liu X, de Zwart JA, Schölvinck ML, Chang CT, Ye FQ, Leopold DA, & Duyn JH (2018). Subcortical evidence for a contribution of arousal to fMRI studies of brain activity. Nature Communications, 9. doi:ARTN 395 10.1038/s41467-017-02815-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Lloyd B, de Voogd LD, Maki-Marttunen V, & Nieuwenhuis S (2023). Pupil size reflects activation of subcortical ascending arousal system nuclei during rest. eLife, 12. doi: 10.7554/eLife.84822 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Ludwig M, Yi YJ, Lusebrink F, Callaghan MF, Betts MJ, Yakupov R, . . . Hammerer D. (2024). Functional locus coeruleus imaging to investigate an ageing noradrenergic system. Commun Biol, 7(1), 777. doi: 10.1038/s42003-024-06446-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. McDougal DH, & Gamlin PD (2015). Autonomic control of the eye. Compr Physiol, 5(1), 439–473. doi: 10.1002/cphy.c140014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. McGinley MJ, David SV, & McCormick DA (2015). Cortical Membrane Potential Signature of Optimal States for Sensory Signal Detection. Neuron, 87(1), 179–192. doi: 10.1016/j.neuron.2015.05.038 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Mokri B (2001). The Monro-Kellie hypothesis: applications in CSF volume depletion. Neurology, 56(12), 1746–1748. doi: 10.1212/wnl.56.12.1746 [DOI] [PubMed] [Google Scholar]
  38. Murphy PR, O’Connell RG, O’Sullivan M, Robertson IH, & Balsters JH (2014). Pupil diameter covaries with BOLD activity in human locus coeruleus. Hum Brain Mapp, 35(8), 4140–4154. doi: 10.1002/hbm.22466 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Najdenovska E, Aleman-Gomez Y, Battistella G, Descoteaux M, Hagmann P, Jacquemont S, . . . Bach Cuadra M. (2018). In-vivo probabilistic atlas of human thalamic nuclei based on diffusion- weighted magnetic resonance imaging. Sci Data, 5, 180270. doi: 10.1038/sdata.2018.270 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Ozbay PS, Chang C, Picchioni D, Mandelkow H, Chappel-Farley MG, van Gelderen P, . . . Duyn J. (2019). Sympathetic activity contributes to the fMRI signal. Commun Biol, 2, 421. doi: 10.1038/s42003-019-0659-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Ozbay PS, Chang C, Picchioni D, Mandelkow H, Moehlman TM, Chappel-Farley MG, . . . Duyn JH. (2018). Contribution of systemic vascular effects to fMRI activity in white matter. Neuroimage, 176, 541–549. doi: 10.1016/j.neuroimage.2018.04.045 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Peng T, Ainslie PN, Cotter JD, Murrell C, Thomas K, Williams MJ, . . . Payne SJ. (2008). The effects of age on the spontaneous low-frequency oscillations in cerebral and systemic cardiovascular dynamics. Physiol Meas, 29(9), 1055–1069. doi: 10.1088/0967-3334/29/9/005 [DOI] [PubMed] [Google Scholar]
  43. Pfeffer T, Keitel C, Kluger DS, Keitel A, Russmann A, Thut G, . . . Gross J. (2022). Coupling of pupil- and neuronal population dynamics reveals diverse influences of arousal on cortical processing. eLife, 11. doi: 10.7554/eLife.71890 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Picchioni D, Ozbay PS, Mandelkow H, de Zwart JA, Wang Y, van Gelderen P, & Duyn JH (2022). Autonomic arousals contribute to brain fluid pulsations during sleep. Neuroimage, 249, 118888. doi: 10.1016/j.neuroimage.2022.118888 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Podvalny E, King LE, & He BJ (2021). Spectral signature and behavioral consequence of spontaneous shifts of pupil-linked arousal in human. eLife, 10. doi: 10.7554/eLife.68265 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Poe GR, Foote S, Eschenko O, Johansen JP, Bouret S, Aston-Jones G, . . . Sara SJ. (2020). Locus coeruleus: a new look at the blue spot. Nat Rev Neurosci, 21(11), 644–659. doi: 10.1038/s41583-020-0360-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Reimer J, McGinley MJ, Liu Y, Rodenkirch C, Wang Q, McCormick DA, & Tolias AS (2016). Pupil fluctuations track rapid changes in adrenergic and cholinergic activity in cortex. Nat Commun, 7, 13289. doi: 10.1038/ncomms13289 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Salas JA, Bayrak RG, Huo Y, & Chang C (2021). Reconstruction of respiratory variation signals from fMRI data. Neuroimage, 225, 117459. doi: 10.1016/j.neuroimage.2020.117459 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Schneider M, Hathway P, Leuchs L, Sämann PG, Czisch M, & Spoormaker VI (2016). Spontaneous pupil dilations during the resting state are associated with activation of the salience network. Neuroimage, 139, 189–201. doi: 10.1016/j.neuroimage.2016.06.011 [DOI] [PubMed] [Google Scholar]
  50. Setzer B, Fultz NE, Gomez DEP, Williams SD, Bonmassar G, Polimeni JR, & Lewis LD (2022). A temporal sequence of thalamic activity unfolds at transitions in behavioral arousal state. Nat Commun, 13(1), 5442. doi: 10.1038/s41467-022-33010-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Shokri-Kojori E, Tomasi D, & Volkow ND (2018). An Autonomic Network: Synchrony Between Slow Rhythms of Pulse and Brain Resting State Is Associated with Personality and Emotions. Cereb Cortex, 28(9), 3356–3371. doi: 10.1093/cercor/bhy144 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Smith SM, Beckmann CF, Andersson J, Auerbach EJ, Bijsterbosch J, Douaud G, . . . Consortium, W.-M. H. (2013). Resting-state fMRI in the Human Connectome Project. Neuroimage, 80, 144–168. doi: 10.1016/j.neuroimage.2013.05.039 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Soon CS, Vinogradova K, Ong JL, Calhoun VD, Liu T, Zhou JH, . . . Chee MWL. (2021). Respiratory, cardiac, EEG, BOLD signals and functional connectivity over multiple microsleep episodes. Neuroimage, 237, 118129. doi: 10.1016/j.neuroimage.2021.118129 [DOI] [PubMed] [Google Scholar]
  54. Tanaka Y, Kamagata K, Saito Y, Takabayashi K, Iseki R, Uchida W, . . . Alzheimer’s Disease Neuroimaging, I. (2025). The coupling of global brain activity and cerebrospinal fluid flow as a potential predictive marker of brain amyloid-beta accumulation. J Prev Alzheimers Dis, 12(8), 100228. doi: 10.1016/j.tjpad.2025.100228 [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Tong Y, Hocke LM, Lindsey KP, Erdogan SB, Vitaliano G, Caine CE, & Frederick B (2016). Systemic Low-Frequency Oscillations in BOLD Signal Vary with Tissue Type. Front Neurosci, 10, 313. doi: 10.3389/fnins.2016.00313 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. van den Brink RL, Pfeffer T, & Donner TH (2019). Brainstem Modulation of Large-Scale Intrinsic Cortical Activity Correlations. Front Hum Neurosci, 13, 340. doi: 10.3389/fnhum.2019.00340 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Van Essen DC, Smith SM, Barch DM, Behrens TE, Yacoub E, Ugurbil K, & Consortium WU-MH (2013). The WU-Minn Human Connectome Project: an overview. Neuroimage, 80, 62–79. doi: 10.1016/j.neuroimage.2013.05.041 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Williams SD, Setzer B, Fultz NE, Valdiviezo Z, Tacugue N, Diamandis Z, & Lewis LD (2023). Neural activity induced by sensory stimulation can drive large-scale cerebrospinal fluid flow during wakefulness in humans. PLoS Biol, 21(3), e3002035. doi: 10.1371/journal.pbio.3002035 [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Yang HC, Inglis B, Talavage TM, Nair VV, Yao JX, Fitzgerald B, . . . Tong YJ. (2022). Coupling between cerebrovascular oscillations and CSF flow fluctuations during wakefulness: An fMRI study. Journal of Cerebral Blood Flow and Metabolism. doi:Artn 0271678×221074639 10.1177/0271678×221074639 [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Yellin D, Berkovich-Ohana A, & Malach R (2015). Coupling between pupil fluctuations and resting-state fMRI uncovers a slow build-up of antagonistic responses in the human cortex. Neuroimage, 106, 414–427. doi: 10.1016/j.neuroimage.2014.11.034 [DOI] [PubMed] [Google Scholar]
  61. Yi YJ, Lusebrink F, Ludwig M, Maass A, Ziegler G, Yakupov R, . . . Hammerer D. (2023). It is the locus coeruleus! Or... is it?: a proposition for analyses and reporting standards for structural and functional magnetic resonance imaging of the noradrenergic locus coeruleus. Neurobiol Aging, 129, 137–148. doi: 10.1016/j.neurobiolaging.2023.04.007 [DOI] [PubMed] [Google Scholar]
  62. Zhang J, & Northoff G (2022). Beyond noise to function: reframing the global brain activity and its dynamic topography. Commun Biol, 5(1), 1350. doi: 10.1038/s42003-022-04297-6 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS2172894-supplement-1.docx (3.3MB, docx)

Data Availability Statement

Human dataset used in this study is publicly available from the HCP1200 database where data can be downloaded via the Connectome DB software (https://www.humanconnectome.org/study/hcp-youngadult/article/reprocessed-7t-fmri-data-released-other-updates). Rest of the data processing scripts and summary data produced by the authors will be available at https://github.com/demiralsb/Blink-Arousal upon the receipt of this manuscript.

RESOURCES