Functional MRI signatures of autonomic physiology in aging

Hemodynamic changes that underlie the blood-oxygenation-level-dependent (BOLD) contrast in functional Magnetic Resonance Imaging (fMRI) are not only sensitive to regional neuronal activity but also influenced by systemic physiological processes regulated by the autonomic nervous system, including slow variations in respiratory depth and rate, heart rate, and blood pressure^{1,2,3,4,5,6,7,8,9,10}. Such BOLD fluctuations driven by various physiological processes are often considered as “noise” and removed from the fMRI time series during preprocessing.

While BOLD physiological dynamics may act as “noise” that confound the extraction of neuronal information from fMRI data, they also provide valuable insights into cerebral blood flow regulation, vascular anatomy, and associated biomechanical properties, presenting an excellent opportunity to infer brain physiology in both health and disease. For instance, recent studies have proposed using respiratory “noise” in resting-state fMRI signals to characterize cerebrovascular reactivity (CVR)—the ability of cerebral blood vessels to dilate or constrict in response to vasoactive factors—and thereby inferring brain-wide vascular reserve^11,12. Complementing conventional studies that employ hypercapnic CO₂ inhalation^13,14,15, natural fluctuations in arterial CO₂ driven by spontaneous variations in respiration can also be used to map CVR^{11,12,16,17,18}. Respiratory variation—reflected in changes in the breathing rate and depth, and typically measured using a respiratory bellows placed around the chest or abdomen—has been shown to strongly correlate with end-tidal CO₂ (a surrogate for arterial CO₂ levels^19,20) and accounts for common aspects of the fMRI signal²¹. Accordingly, CVR can also be achieved by quantifying fMRI signal changes relative to these respiratory fluctuations. Beyond vascular reserve, the timing of respiratory fMRI responses—determined by the transit speed of CO₂ from the lungs to brain tissues and by regional vasogenic responses, i.e., the arteriolar dilation triggered by elevated partial pressure of CO₂ in arterial blood—also provides valuable information about arterial transit time and, ideally, regional delays in hemodynamic responses²².

In addition to respiratory “noise”, there is also rich information regarding brain-heart interactions and the vascular mechanical properties embedded in the cardiac “noise” in the form of widespread fMRI fluctuations following a change in heart rate^3,5. Distinct from active vasogenic responses to CO₂ associated with respiratory variation, these cardiac fMRI responses are linked to beat-to-beat variations in the blood “piped” from the heart through the cerebrovasculature. They largely reflect autonomic regulation and systemic, passive vasodilation due to arterial compliance caused by blood pressure waves. As such, changes in blood pressure can modulate the amplitude of these cardiac fMRI responses, and vascular elasticity influences their timing. Accordingly, we can gain insights into the biomechanical properties of the arterial system.

The goal of this study is to leverage the rich, multi-parametric information embedded in physiological fMRI “noise” to gain insights into age-dependent changes in the autonomic regulation and vascular physiology. Previous research has shown that aging could lead to changes in the tortuosity of penetrating arterioles^23,24,25 as well as a reduction in cerebral capillary and arteriole density in older brains^{26,27,28,29,30,31,32}, which could cause slower blood transit times from large arteries to capillaries³³ and reduced vascular reactivity to systemic physiological stimuli such as changes in CO₂ levels and other vasodilators^34,35,36,37. Additionally, age-related mechanism could alter biomechanical properties of the blood vessels, resulting in increased arterial stiffness³⁸ and consequently faster pulse wave velocity^39,40. This stiffening is attributed to the accumulation of collagen and degradation of elastin fibers within the vessels⁴¹, which in turn reduces cerebral blood flow and cerebrovascular reserve, impairs deep white matter perfusion, and compromise vascular integrity^42,43,44,45. The interaction between arterial stiffness and blood pressure is bi-directional⁴⁶, with both hypertension and vascular stiffening being common features of aging. Systolic blood pressure tends to rise with age while diastolic pressure stabilizes or declines, leading to a widened pulse pressure that is thought to increase the mechanical load on cerebral vessels^47,48. These age-related changes in vascular health and autonomic physiology can differentially affect the timing and amplitudes of systemic respiratory and cardiac fMRI “noise”. Thus, dynamic fMRI data may offer a means to infer distinct aspects of vascular or autonomic function. Surprisingly, existing fMRI-based aging studies often neglect the altered physiological components and predominantly focus on the neurogenic functional activity reflected in the fMRI time-series data^49,50.

Here, by analyzing a large, public cohort of resting-state fMRI datasets from the Lifespan Human Connectome Project Aging (HCP-A) study^51,52, we conduct a comprehensive evaluation of the impact of aging on fMRI signatures of autonomic physiology. By linking brain-wide fMRI dynamics with low-frequency peripheral respiratory and pulse measures, we show that aging induces significant changes in various physiological fMRI metrics, including globally slower fMRI responses driven by respiratory changes, faster fMRI responses following heart-rate variations, and an increased coupling between brain fMRI and cardiac signals. We further demonstrate that age-related alterations in these physiological fMRI metrics do not change linearly across the lifespan, with a distinct inflection in age dependence observed after age sixty. These physiological changes are interpreted in terms of age-related shifts in tissue perfusion and arterial compliance, and their connections to accompanying structural and arousal-state alterations are also examined and discussed. Together, our findings demonstrate and elucidate detailed changes in fMRI signatures of autonomic physiology in aging, highlighting altered vascular properties and autonomic function as significant features of late-life stages. Methodologically, our study also shows that resting-state physiological fMRI “noise” can be used to extract multi-parametric physiological information, providing a valuable complement to existing fMRI research that focuses primarily on brain neuronal function in clinical and neuroscientific contexts. A preliminary account of this study has been presented in abstract form⁵³.

Subjects and data

Resting-state fMRI data from a cohort of 400 adults from the HCP-A Lifespan 2.0 Release^51,52 were analyzed in this study to characterize changes in fMRI signatures of autonomic physiology in aging. This public dataset includes individuals who exhibit typical health for their age: common age-related health conditions such as hypertension and other vascular risk factors are included; in contrast, individuals diagnosed with less prevent conditions that could confound data interpretation—such as Alzheimer’s disease and symptomatic stroke—are excluded. This subset of subjects was chosen by visually inspecting the peripheral sensory recordings (pulse oximetry and respiratory bellows) to ensure the quality of extracted physiological metrics (as detailed in Supplementary Table S1). Subjects were further divided into two age groups: Aging I (36–60 yrs, N = 227, 125 females) and Aging II (60–90 yrs, N = 173, 91 females). Age 60 was chosen as the cut-off threshold because it has been previously marked as a pivotal point in the altered rate of decline in brain structural and functional changes^{54,55,56,57,58}.

Age effects on respiratory variation (RV)-coupled fMRI dynamics

We first examined how naturalistic respiratory responses—spatiotemporal dynamics of fMRI signals following an increase in respiratory variation—change with aging. RV was derived by computing time-windowed standard deviations of respiratory waveforms recorded by the bellows, tracking dynamic changes in respiratory volumes and depths³. A cross-correlation analysis was performed to characterize the temporal lags and couplings between fMRI signals and RV at both global and regional levels.

Cross-correlations between the cerebral cortical fMRI signal, i.e., the mean across gray-matter voxels, and RV exhibited a bimodal pattern across all age groups—an increase in RV led to an initial overshoot, followed by a pronounced undershoot in the global fMRI signal (Fig. 1a). Note that here positive lag values correspond to the fMRI responses trailing the RV, and vice versa for the negative lag. A significant age effect was identified in the temporal lags associated with the trough of the cross-correlation results (i.e., corresponding to the undershoot in the respiratory fMRI response), with the elder group exhibiting more delayed respiratory responses (Aging I: 13.4 ± 3.9 s; Aging II: 16.3 ± 4.1 s, two-sample t-test, p = 4.6 × 10^-12). No significant between-group differences were observed in the correlation values associated with the trough (Aging I: −0.39 ± 0.17; Aging II: –0.39 ± 0.15, two-sample t-test, p = 0.69), and no differences were observed in the peak of the cross-correlation plots including both the timing (Aging I: –0.06 ± 10.0 s; Aging II: –0.54 ± 8.39 s, two-sample t-test, p = 0.61) and correlation value (Aging I: 0.24 ± 0.10; Aging II: 0.26 ± 0.11, two-sample t-test, p = 0.074) (Fig. 1a).

a The cross correlation between the global cortical fMRI signal and RV (positive lag values suggest that fMRI signals lag RV) for different age groups, with shade denoting the standard errors across subjects. b Intra-cortical distributions of region-specific fMRI signal lag relative to RV (based on the Schaefer 300-parcel atlas). Regions exhibiting statistically significant between-group temporal lag differences were displayed at the bottom (“Aging II > I”, FDR < 0.05). c Tissue-type specific temporal lags relative to RV for each age group. Error bars indicate the standard errors of RV-fMRI temporal lags across subjects, and the shaded gray area highlights ROIs that exhibited statistically significant between-group differences (FDR < 0.05). ROI labels are shown at the bottom.

Full size image

Consistent with our previous report on the HCP young-adult data (18–36 yrs)⁵⁹, the lifespan dataset also revealed spatially heterogenous temporal lags in the respiratory responses, quantified according to the trough of the cross-correlation results. Within the cortex, association cortex generally exhibited slower responses than the sensory cortex (Fig. 1b, Aging I and Aging II); and across the brain, the most delayed fMRI responses were identified in the cerebellum and white matter (Fig. 1c). As subjects aged, a brain-wide increase in the RV-fMRI temporal lags was observed, with the largest between-group lags reaching up to approximately 3 seconds (Fig. 1b, Aging II > I, & 1c).

Age effects on heart rate-coupled fMRI dynamics

We next investigated how cardiac responses—spatiotemporal dynamics of fMRI signals following an increase in heart rate—change with aging under resting state conditions. During the experiments, the aging group exhibited slightly slower heart rates and higher heart-rate variability after age 60 (Fig. 2a and Supplementary Fig. S1), consistent with previous reports^60,61. A time-varying waveform of the heart rate variation was derived by computing time-windowed averages of inter-heart-beat intervals (HBIs, i.e., the inverse of heart rate) recorded by pulse oximetry. Akin to the RV-fMRI analysis, cross-correlations were calculated to characterize the temporal lags and couplings between fMRI signals and HBI at both global and regional levels.

a Mean heart rate (left) and heart rate variability (right) of different age groups. b The cross correlation between the global cortical fMRI signal and HBI (positive lag values suggest that fMRI signals lag HBI) for different age groups, with shade denoting the standard errors across subjects. c Intra-cortical distributions of region-specific fMRI signal lags (left) relative to and couplings (right) with HBI (based on the Schaefer 300-parcel atlas). Lag and peak correlation values were estimated based on the trough of cross-correlations (pointed by red arrow in (b)). Regions exhibiting statistically significant between-group temporal lag or correlation value differences were displayed at the bottom (“Aging II > I”, FDR < 0.05). d Tissue-type specific temporal lags (top) and cross-correlation peak correlation values (bottom) for each age group. Error bars indicate the standard errors of HBI-fMRI temporal lags and peak correlation values across subjects, and the shaded gray areas highlight ROIs that exhibited statistically significant between-group differences (FDR < 0.05). The labels and orders of ROIs are identical to Fig. 1c.

Full size image

Age effects were also prominent in the cardiac responses. In the older cohort, we observed a faster cardiac response and increased coupling (i.e., stronger absolute correlation values) between global fMRI signals and the HBI when focusing on the trough of the cross-correlation (Fig. 2b, temporal lag: Aging I: 1.3 ± 1.4 s; Aging II: 0.9 ± 1.5 s, two-sample t-test: p = 0.01; correlation value: Aging I: –0.22 ± 0.13; Aging II: –0.29 ± 0.15, two-sample t-test: p = 2.0 × 10⁻⁶). No statistically significant between-group differences were identified for the later positive peak in the cross-correlation plots (temporal lag: Aging I: 8.9 ± 2.5 s; Aging II: 9.4 ± 3.0 s, two-sample t-test, p = 0.041; peak correlation value: Aging I: 0.19 ± 0.14; Aging II: 0.20 ± 0.13, two-sample t-test, p = 0.19).

Consistent with our previous report⁵⁹, cardiac responses in the lifespan cohort were also inhomogeneous across the brain (Fig. 2c, Aging I & II, & 2d). The intra-cortical lag patterns of cardiac responses, quantified by the trough of the cross-correlation results, correlated positively with those of the respiratory responses (Spearman’s correlation r = 0.83, p < 10⁻²⁰). For both age groups, the somatosensory and visual network consistently exhibited stronger HBI-fMRI coupling, while the limbic network (temporal pole and orbitofrontal cortex) displayed relatively low correlations with the heart rate signal.

Statistically significant reductions in the temporal lags and increases in the HBI-fMRI coupling with age were observed in most brain regions (Fig. 2c, d). However, the specific cortical patterns associated with the between-group difference were modestly correlated (Spearman’s correlation r = 0.20, p = 5.06 × 10⁻⁴), suggesting distinct mechanisms underlying age-related changes in the cardiac response timing and magnitudes.

Linking spatial distributions of age-related changes in physiological fMRI metrics to anatomical structure, tissue perfusion, and arousal states

Having identified significant age-related effects in both respiratory and cardiac fMRI responses, we next tested their associations with accompanying changes in brain structure, tissue perfusion, and arousal state dynamics during aging.

We first assessed whether the observed age-related effects in physiological fMRI metrics might partially stem from potential anatomical biases. For example, reduced percentage of gray matter within each fMRI voxel due to cortical atrophy^62,63, which may diminish the relative contributions of the neuronal component to the total variance of voxel-wise fMRI signals and thereby increases HBI-fMRI correlations. To test this hypothesis, we quantified the parcel-level gray-matter partial volumes using each subject’s high-resolution structural MRI data and compared the resulting spatial distributions with those of different physiological metrics. As expected, the regions showing the most significant reductions in gray matter percentage (Fig. 3a, “MGMF”) aligned with those reported to exhibit the most severe cortical thinning, including the frontoparietal cortex, limbic cortex, dorsal attention, and somatomotor network^64,65,66. However, the spatial pattern of reduced gray-matter partial volumes in aging was only modestly correlated with those of physiological metrics (Fig. 3b), suggesting that the partial volume effect had a limited influence on the observed age-related physiological patterns.

a Group-level mean gray-matter fraction (MGMF), ATT, CBF, and eye camera-based arousal effects. “Aging I/II”: mean results within each age group; “Aging II > I”: spatial pattern of age effects. b Spearman’s spatial correlation of age effects between physiological fMRI metrics (Figs. 1b, 2c) and structure, blood perfusion and arousal effects (a). Pairwise correlation values were shown, with significant correlations indicated by an asterisk (*) following Bonferroni correction ((alpha) = 0.01).

Full size image

We next examined the vascular physiological origins of the characterized respiratory and cardiac fMRI responses by linking their timing and amplitudes to two baseline physiological metrics—arterial transit time (ATT) and cerebral blood flow (CBF), estimated from arterial spin labeling (ASL) MRI data from the same subject cohort of the HCP-A project³³. Since the timing of the BOLD responses to respiration is partially influenced by the speed of CO₂ delivery, a positive coupling between RV-fMRI temporal lags and ATT was expected. Regarding cardiac BOLD responses, since previous studies have consistently reported reduced baseline CBF with increased arterial stiffness and blood pressure^42,67,68,69—key determinants of arterial compliance to the cardiac pressure wave—a positive association between CBF and cardiac response metrics is expected. Consistent with our predictions, the frontal-parietal regions showing the most significant between-group differences in respiratory responses also exhibited the most delayed ATT in aging (Fig. 3); and cortical regions with the most CBF reductions also exhibited faster cardiac responses and stronger HBI-fMRI coupling (Fig. 3). These findings corroborate the physiological basis of age-related changes in cardiac and respiratory fMRI responses identified in this study.

Finally, we tested the potential impact of in-scan vigilance levels on the coupling between systemic physiology and fMRI signals. This analysis was motivated by emerging evidence that arousal can intensify the apparent correlation between fMRI signals and physiological responses⁷⁰. Thus, we assessed whether older subjects were drowsier during the experiments. Subjects’ vigilance levels during scans were inferred from eye camera data collected simultaneously during resting-state scans. For a subset of subjects (N = 88 in total, identical number of subjects from each age group) with high-quality eye camera data, we computed the instantaneous in-scan eye-closure frequency and correlated it with brain-wide fMRI signals to identify the fMRI signatures of vigilance. We considered this subset of subjects representative because the RV/HBI-fMRI coupling results derived from this subset were consistent with those from the full N = 400 group (see Supplementary Fig. S2a). In both age groups, we observed modest effects of fluctuating arousal (see Supplementary Fig. S2b); and contrary to our predictions, the older subjects exhibited a less pronounced arousal effect, evidenced by reduced correlations between fMRI signals and the eye-camera data (Fig. 3, “Arousal”). Additionally, the spatial alignment between age-related changes in arousal effects and HBI-fMRI correlation patterns was not statistically significant. Thus, differences in in-scan arousal levels are unlikely the primary drivers of the observed physiological fMRI changes in aging.

Characterizing the “aging rate” of physiological fMRI metrics

As a final investigation, we examined whether age-related changes in various physiological fMRI metrics progress linearly as a function of age. This analysis was motivated by previous evidence that brain shrinkage, molecular profiling, and vascular aging change more quickly after age 60^{54,55,56,57,58}.

Specifically, we tested whether the “aging rate”, defined as the linear dependence on age, of different physiological fMRI metrics altered after age 60 (i.e., Aging I vs. Aging II) using a linear regression model that incorporated distinct slopes for the two age segments, and mapped brain regions exhibiting the most prominent between-group differences (see “Methods”). Apparent inflection in the linear dependence on age emerged for all physiological fMRI metrics after age 60 (Fig. 4). For the RV-fMRI temporal lag, increased aging rates were identified in midline structures (posterior cingulate cortex (PCC), anterior cingulate cortex (ACC), and precuneus), and at the temporal-frontal boundary (secondary somatosensory cortex (S2) and lateral sulcus). For the HBI-fMRI temporal lag, reduced aging rates were identified in the mid cingulate cortex (MCC), prefrontal cortex, and temporal lobe. Compared to the more focal patterns associated with the temporal lags, widespread cortical regions, except for a few visual and prefrontal patches, exhibited increased aging rates of HBI-fMRI correlation values.

a Scatterplots of RV-fMRI temporal lags, HBI-fMRI temporal lags, and HBI-fMRI peak correlation values as a function of age. Results of the global cortical gray matter are shown, and the age dependence was fitted with non-linear loess regression. Shaded pink regions indicate standard errors of fitting. Linear dashed lines (fitted using the Aging I group data) were included in the plots to help visualize apparent inflections in the aging rate after age sixty. b Regions exhibiting statistically significant changes in linear aging rates after age 60 (FDR < 0.05).

Full size image

At the global level, a shift in aging rate around age 60 is also apparent when employing a more flexible local polynomial regression model. The overall age trajectories of different physiological fMRI metrics are well approximated by two linear segments intersecting near age 60, in line with prior studies reporting altered rate of brain decline around this age^{54,55,56,57,58}. Notably, RV- and HBI-coupled fMRI metrics show more pronounced inflection in aging rates after 60 than most anatomical and perfusion features (Fig. 5). This observation implies a prominent role of declined physiological functions in the later stages of brain aging and highlights the importance of considering brain-body interactions in understanding the aging process and its impact on brain health.

Different metrics were normalized and rescaled to 0-100 for display. Results of absolute global cortical HBI-fMRI correlation values (“|Correlation with HBI | ”, the absolute value was computed to emphasize the strength of correlations), RV-fMRI temporal lags (“RV temporal lags”), and HBI-fMRI temporal lags (“HBI temporal lags”) are shown. Structural metrics include the total brain volume (“Brain volume”), ventricular volume (“Ventricular volume”), cerebral white-matter volume (“Cerebral WMV”), mean cortical thickness (“Mean cortical thickness”), and cortical gray-matter volume (“Cortical GM”). Brain perfusion metrics include cortical mean CBF (“Mean cortical CBF”) and ATT (“Mean cortical ATT”). For each metric, the trajectory of age dependence was fitted using loess, with the shaded area indicating standard errors of fitting. Individual metric fitting results are shown on the right, ranked by the statistical significance of altered linear aging rates after age 60, from the highest to the lowest (uncorrected p values are shown; * indicates statistical significance after Bonferroni correction, (alpha) = 0.05). Linear dashed lines (fitted using the Aging I group data) were included in the plots to help visualize apparent inflections in the aging rate after age sixty.

Full size image

In this study, we utilized the HCP-A resting-state dataset to characterize age-related changes in fMRI signatures of autonomic physiology. Our findings revealed significant age effects on the dynamics of physiological fMRI signals, identifying slower respiratory responses, faster cardiac responses, and stronger brain-cardiac signal coupling in aging. These physiological fMRI patterns mirror known age-related changes in vessel biomechanical properties, tissue perfusion, and autonomic function. We further demonstrated that the impact of aging on respiratory and cardiac fMRI dynamics does not progress uniformly across the lifespan, with a notable turning point around the age of 60, highlighting physiological declines in the cerebrovasculature as a key feature of aging. These results deepen our understanding of age-related changes in brain physiology and present new biomarker opportunities for investigating cardiovascular and cerebrovascular health in aging populations in future research.

Overall, the slower respiratory responses observed in our study are consistent with both prolonged arterial transit time and delayed local hemodynamic effects in aging. Aging leads to increased arteriolar tortuosity^23,24,25 and reduced density of the cerebral capillary and arterioles^{26,27,28,29,30}, causing both longer blood transit and arrival time, thus longer time for CO₂ delivery, from lungs to the brain and from the large arteries to capillaries. The spatial consistency of age effects in ATT and RV-fMRI temporal lag patterns in gray matter (Fig. 3b), as well as similarly delayed responses of both measures in the white matter^33,71, all supported a prominent effect of ATT in impacting the respiratory response speed. However, it appears that ATT alone (with <1-s lags across the cortex) cannot fully account for the differential RV-fMRI lags observed between two age groups (Fig. 3 vs. Fig. 1, “Aging I vs. II”). Thus, respiratory response delays may also stem partly from slower arteriolar dilation and blood delivery caused by elevations in CO₂ level, i.e., slower vascular reactivity generating slower hemodynamic responses to CO₂. This possibility aligns with previous evidence showing that older subjects exhibit more sluggish stimulus-driven hemodynamic responses, with a time-to-peak delay of ~1 s compared to healthy young controls^72,73, although it should be noted that vasogenic and neurogenic responses involve distinct mechanisms of active vasodilation⁷⁴. Collectively, the widespread delayed RV respiratory responses in the elder cohort may reflect a combined effect of both delayed ATTs and local vasogenic hemodynamic responses in aging. Notably, although the group-level RV-fMRI correlation values were smaller in the Aging II compared to the Aging I cohort, this between-cohort difference did not reach statistical significance (see Supplementary Fig. S3), which was unexpected given the reduced vascular reserve shown by previous hypercapnic studies in aging^13,34. Since correlation values depend on the fractional contribution of the respiratory response to total fMRI variance rather than its absolute amplitude, our results suggested that total fMRI signal variance also decreases to a comparable extent with aging. This possibility was further supported by our observations that both the amplitudes of fMRI fluctuations and respiratory signals decreased in the older cohort of the HCP-A data (see Supplementary Fig. S4).

The accelerated global cardiac response in aging is consistent with vascular stiffening and reduced arterial compliance to the incoming cardiac pressure waves. Using phase contrast MRI^75,76 and sonography⁷⁷, previous studies have demonstrated faster pulse-wave velocity in several major vessels in old adults^39,40. Our findings suggest that, in addition to the fast pulse wave associated with each heartbeat (~1 Hz), slower pressure waves (<0.1 Hz) driven by heart-rate variations also propagate more rapidly with aging. These slow blood pressure waves, when transmitted to the cerebrovasculature, induce local vasodilation that displaces tissue and cerebrospinal fluid signals. Because these compartments have distinct T₂* values, such displacements result in an apparent change in nominally BOLD-weighted fMRI signals. Beyond these dynamic partial volume effects, there is evidence that lower CBF could lead to faster hemodynamic responses^78,79. Thus faster passive vasodilation may also be linked to the reduced baseline CBF seen in aging. This effect of lower baseline CBF may play a bigger role in the passive vasodilation caused by increased blood flow than in neurogenic or CO₂-triggered vasogenic changes, which involve active stimulation of arteriole smooth muscle dilation that also undergoes significant alterations in aging.

The widespread increase in brain-cardiac signal coupling cannot be fully attributed to altered autonomic nervous system activity in aging, in part because the associated peak timing fell outside the expected neurovascular coupling range, also because brainstem regions and cortices whose neuronal activity drives autonomic signaling (such as the insula and anterior cingulate cortex) did not exhibit the most pronounced increases in HBI-fMRI coupling (Fig. 2c, d). One possibility lies in hypertension common in aging⁸⁰. Pulse pressure—the difference between systolic and diastolic pressure—showed a marked increase after age 60 in the HCP-A data (see Fig. 6). This increase, together with less effective cushioning of the arterial system in aging due to vascular stiffening, may increase the transmission of cardiac pressure wave energy to the cerebrovasculature⁷. This possibility is supported by a significant correlation between pulse pressure and HBI-fMRI coupling strength (see Fig. 6), as well as the role of pulse pressure as a significant mediator of age-related changes in HBI-fMRI coupling (p = 2.4 × 10⁻³, tested using the mediation package in R). A second and intertwined possibility relates to age-related changes in heart rate variability. Consistent with previous reports^60,61, we observed a U-shaped relationship between heart rate variability and age, with a notable increase in heart rate variability after age 60 (see Fig. 2a RMSSD measures and Supplementary Fig. 1), which has been previously attributed to heightened irregularity and fragmentation of heartbeats in aging⁸¹ and can contribute to overall amplified brain-cardiac signal coupling. This possibility may help explain the discrepancy between our findings and those of a separate study, which reported that older adults showed both reduced heart rate variability and weaker fMRI-heart rate correlations⁸². A third possibility is that neuronal and synaptic loss, along with diminished neurovascular coupling in aging, reduces the proportion of neurogenic hemodynamic changes contributing to the total variance in fMRI signals. This reduction reflects both decreased neural activity that directly consumes oxygen⁸³, and impaired neurovascular coupling that actively controls arteriole smooth muscle dilation^84,85. In contrast, passive vascular dilation in response to systemic changes in cerebral blood supply is less susceptible to deficits in neurovascular coupling. Thus, this shift in neurogenic BOLD fractional contributions may elevate the correlation between fMRI and heart rate signals.

a Mean and standard deviations of pulse pressure within each age group. b Pulse pressure vs. global HBI-fMRI coupling (Spearman’s correlation = –0.22, p = 2.4 × 10⁻³). Each dot represents the results of a single subject.

Full size image

By introducing the metric of “aging rate”, our results revealed that fMRI-based physiological metrics, particularly brain-cardiac signal coupling, do not progress linearly with age. Notably, the inflection in aging rates was more pronounced in physiological fMRI metrics compared to structural and perfusion-based imaging measures. This observation highlights altered autonomic function and vascular physiology as key markers of advanced brain aging. Furthermore, we observed spatially varying aging rates across all physiological fMRI metrics (contrasting age-related changes in Figs. 2–4). For instance, the insula and ACC were among those exhibiting the most significant rate changes in the temporal lags of physiological responses, possibly due to their role in autonomic regulation⁸⁶ and location within cerebral arterial territories that are particularly vulnerable to late-stage aging⁸⁷.

From a methodological perspective, our study shows that resting-state fMRI signals can provide rich, multi-parametric physiological information, extending beyond the extensively studied functional connectivity estimates. These physiological fMRI metrics leverage the high spatiotemporal resolution and sensitivity of BOLD fMRI, offering new insights into certain vessel mechanical properties that are difficult to assess with conventional techniques (e.g., arterial compliance inferred from fMRI signal timing). In particular, they may serve as potential proxies when complex experimental setups or maneuvers, such as gas inhalation in CVR mapping, are impractical or unsuitable for certain patient populations. As such, physiological fMRI signals collected in the resting state hold significant clinical promise, or at least, can serve as a valuable complement to existing imaging modalities and invasive procedures for assessing brain physiology and cardiovascular health.

Finally, our study also opens up several directions that can be explored in future studies. First, while the HCP-A dataset represents a typical, healthy aging cohort, excluding individuals diagnosed with common forms of atypical cognitive impairment and symptomatic stroke (see Supplementary Fig. S5 for the cognitive test results of subjects examined in this study), it does not exclude those with common age-related vascular risks^51,52. Therefore, the age-related changes in physiological fMRI metrics observed here may, in part, reflect the influence of prevalent vascular conditions in late-stage aging, such as atherosclerosis in addition to hypertension examined above. Studies with finer age range divisions (encompassing different stages of aging) and mediation analyses incorporating detailed clinical cardiovascular metrics, when available, will provide deeper insights into the complex interplay between age and vascular disease. Second, we primarily characterized physiological and biomechanical properties of the vessels by focusing on the temporal features of physiological fMRI metrics outside the neurovascular coupling range. Future investigations employing rapid fMRI sampling and multi-modal acquisitions with ground-truth neuronal information will enable further opportunities to uncover a comprehensive picture of autonomic physiology by including the neurovascular coupling components associated with focal autonomic nervous activity, such as sympathetic activity involved in regulating low-frequency heart rate variations and blood pressure waves^88,89,90. Third, in this study, we examined the influence of age on respiratory and cardiac responses by separately linking RV and HBI to global and local fMRI signals, without accounting for interactions between the two processes. However, it is known that these two processes can be coupled due to mechanisms such as respiratory sinus arrhythmia or autonomic arousal^91,92,93, which may account for the resembling intra-cortical temporal lag patterns of RV-fMRI and HBI-fMRI correlations observed within each age group. A preliminary analysis revealed insignificant age effects in RV-HBI cross-correlations (see Fig. 7a, p = 0.40 at peak correlation). Moreover, as detailed in Fig. 7b, any potential interactions between RV and HBI are more likely to affect the early phase of respiratory fMRI responses (corresponding to the positive peak in the RV-fMRI cross-correlations in Fig. 1a) and the later phase of cardiac fMRI responses (corresponding to the positive peak in the HBI-fMRI cross-correlations in Fig. 2b), where no significant between-group differences were observed in the current dataset. Thus, we expect only a modest influence of RV-HBI interactions on our primary findings and interpretations. This is further supported by a linear partial correlation analysis that revealed statistically significant age effects on different physiological fMRI metrics, even after controlling for dependencies between RV and HBI (see Supplementary Fig. S6). Future studies might consider developing more sophisticated models to better account for the intricate interactions among various physiological processes and teasing apart associated age effects, respectively. Finally, sex is known to influence vascular properties through the effects of sex hormones such as estrogen and testosterone^94,95. Significant sex differences were observed in the physiological fMRI metrics across both age groups (see Supplementary Fig. S7). However, at this sample size, the interaction between age and sex was not statistically significant. Future studies with larger cohorts may provide sufficient power to detect meaningful sex-specific modulation at different stages of aging, such as the impact of menopause on vessel biomechanical properties in females^96,97.

a Fisher’s z-transformed cross-correlation between RV and HBI at varying time lags (seconds) for two distinct age groups. The shaded areas indicate the standard errors across subjects within each age group. An inset highlights the central region (−5 to 5 s) to better visualize the trough correlation values of two age groups (between-group t test of correlations at 0-s lag: p = 0.40). b Hypothesized influence of RV-HBI interactions on measured respiratory and cardiac fMRI responses, illustrated using cortical physiological fMRI responses from Chen et al., NeuroImage, 2020. Arrows indicate possible interactions: fMRI responses driven by increased arterial pressure waves (orange) may contribute to the early peaks in measured respiratory fMRI responses; while the trough of CO₂ responses (purple) may influence the later phase of the measured cardiac fMRI responses.

Full size image

Data and acquisition

Functional imaging data from a subset of 400 adults from the HCP-A Lifespan 2.0 Release were analyzed in this study^51,52. All participants provided informed, written consent for this Institutional Review Board-approved study. The selection criteria and list of subjects are detailed in Supplementary Table S1. Distributions of cardiovascular measures and cognitive test scores across all subjects are summarized in Supplementary Figs. S1 and S5. Each participant underwent two eyes-open resting-state sessions (REST1 and REST2), and each session contained two ~6 min fMRI scans with opposite phase-encoding directions. These resting-state fMRI scans were acquired using multiband accelerated 2D echo-planar imaging (EPI): 2 mm isotropic voxels, matrix size = 104 × 104, 72 slices, no in-plane acceleration, TR = 800 ms, TE = 37 ms, flip angle = 52°, multi-band factor = 8, echo spacing = 0.58 ms. Resting-state fMRI scans from the first session (REST1_AP and REST1_PA) were analyzed in this study. For each subject, physiological fMRI metrics from the two resting-state scans were averaged to yield a single sample for assessing age-related effects at the group level (see below). During these fMRI scans, concurrent cardiac signals were monitored using a fingertip pulse oximeter and respiratory signals were recorded through chest bellows, both at a sampling rate of 400 Hz.

Extracting physiological signals from sensor recordings

An outlier-replacement filter was first applied to eliminate noise in the physiological traces due to subject motion or improper attachment of the respiratory belt or pulse oximeter. The pulse oximeter traces were then bandpass filtered (0.5–10 Hz) using a 4th-order Butterworth filter, and the cardiac-peak detection was performed using the MPSTD (Multi-Scale Peak and Trough Detection) detector from the open-source PPG-beats package⁹⁸.

The computation of subject-specific respiratory variation (RV) and heartbeat interval (HBI) time courses followed previous studies^3,59. RV reflects time-dependent changes in respiratory depths and volumes, and HBI reflects the mean heartbeat intervals, i.e., the inverse of heart rates. Both metrics were computed across 6 s sliding windows centered at each measured fMRI time point, to be consistent with previous studies^3,59.

Mean heart rate (beats per minute, BPM) and heart rate variability (measured as the root-mean-square of successive differences of normal-to-normal intervals, RMSSD⁹⁹) were also calculated and compared between age groups.

Preprocessing of resting-state fMRI data

All fMRI datasets underwent standard preprocessing steps, including slice-time correction and rigid-body motion realignment using AFNI¹⁰⁰. Magnetic susceptibility field-induced geometric distortion was corrected using the TOPUP toolbox in FSL¹⁰¹. Quasi-periodical physiological fluctuations time-locked to cardiac and respiratory cycles were modeled using RETROICOR¹⁰² (up to second order, eight regressors in total) and linearly projected out of each voxel’s fMRI time series along with six rigid-body motion parameters and low-frequency scan drifts. The initial five and the final frames of each scan dataset were removed prior to analyses. No temporal filtering was applied to the time-series data.

Brain parcellations

All cortical and subcortical regions of interest were first delineated based on each subject’s high-resolution T₁-weighted anatomical reference data. A multi-resolution functional network atlas was used to parcellate the cortical gray matter¹⁰³. Both the 100- and 300-parcel atlases were used for fMRI data analysis. Masks of the subcortical structures, cerebellum, brainstem, and white matter were extracted from the “aseg” automatic segmentation generated by FreeSurfer¹⁰⁴. To mitigate the potential contamination from gray-matter signals, the white-matter mask was further eroded by one voxel. All brain regions of interest (ROIs) defined in the high-resolution anatomical space were then registered to each subject’s native functional space based on the cross-modal transformation estimated by FreeSurfer’s bbregister.

Cross-correlations between fMRI signals and autonomic physiology

Cross-correlations were utilized to quantify the strengths and temporal lags of fMRI responses elicited by different physiological processes. All ROI-wise fMRI signals and physiological recordings (RV and HBI) were up-sampled to a temporal resolution of 200 ms for the cross-correlation analyses, and the correlation coefficients at each time shift were Fisher-z transformed prior to group averaging. The peak/trough detection for cross-correlations was constrained to 6-s windows centered at the group mean, and a retrospective check confirmed that the vast majority of peaks/troughs fell within the set temporal boundaries. The temporal window for peak/trough detection was adjusted for white-matter and subcortical regions.

The cross-correlation analyses were performed at both global and ROI levels, and two-sample t-tests were performed to test for significant age effects.

Quantifying the “aging rates” of physiological metrics

To examine whether the “aging rate”, defined as the linear dependence of different physiological fMRI metrics on age, significantly differs between the two age groups, we performed the following linear regression tests. For each physiological fMRI metric, we fit it with a general linear model consisting of a constant baseline term, years of education, sex, a binary age group indicator (“Aging I” or “Aging II” group), and an age covariate (to model the linear dependence on age). We then tested whether the inclusion of an additional conditional age covariate (i.e., derived by setting the ages of all “Aging I” participants zero) significantly improved the linear model fitting using an F test. The test results informed whether the aging rate of any specific physiological metric changed significantly after age 60.

Relating physiological fMRI metrics to age-dependent changes in anatomical structure, perfusion, arousal

Cortical thickness and gray-matter partial volume

Parcel-wise cortical thickness was quantified automatically based on the high-resolution T₁-weighted anatomical reference data using FreeSurfer¹⁰⁴. For each cortical parcel, the level of gray-matter volume fraction was computed by averaging across all voxels within the parcel using mri_compute_volume_fractions in FreeSurfer.

Arterial transit time and cerebral blood flow

Multi-inversion-time ASL data of the participants (N = 400) from HCP-A were employed to compute regional ATT and CBF as detailed previously³³. Briefly, ATT was calculated using a two-stage cross-correlation method, comparing the multi-delay ASL signal with a simulated signal based on the flow-modified Bloch equation to identify the shift that yielded the maximum correlation for each voxel. CBF was derived using a two-compartment model that factored in physiological constants and relaxation time, with results averaged over appropriate post-labeling delays.

Arousal effects

To track the subjects’ instantaneous arousal levels during the scans, the eye camera data from a subset of subjects (N = 88) were also analyzed. Subjects were selected based on the visibility of the eye (minimal head coil blockage), synchronization with fMRI acquisition, and minimal interference from eyelashes or artifacts. Eye closure was detected using a customized eye detection algorithm¹⁰⁵, and the percentage of eye closure was calculated across 0.8-s windows (TR length) to indicate the subject’s instantaneous arousal level. The eye-closure arousal metrics were temporally shifted (Aging I: 6.4 s and Aging II: 5.6 s for all cortical parcels) and then correlated with fMRI signals to characterize arousal effects. The relative temporal lag for each group was optimized by maximizing the peak cross correlation values between the global fMRI signal and the arousal metric (see Supplementary Fig. S2b).

Spatial correlations

The spatial similarity between age-dependent changes in physiological metrics and structural, perfusion, and arousal measures was estimated using the Spearman’s rank correlation computed over space.

Trajectories of age dependence

For each physiological fMRI, structural, and perfusion metric, its trajectory across the lifespan was mapped by fitting the age dependence of each metric using local polynomial regression, implemented via the loess function in R.

Influence of blood pressure on the brain-heart signal coupling

Individual pulse pressure was calculated as the difference between systolic and diastolic pressure, and examined in relation to global HBI-fMRI coupling using Spearman’s rank correlation. To further assess the influence of pulse pressure on age-related changes in HBI-fMRI coupling, a mediation analysis was conducted using the mediation package in R.

Statistics and reproducibility

In the main text, the statistical significance of age-related effects on various physiological fMRI metrics is assessed using two-sample t-tests based on metrics estimated from two distinct age groups (Aging I and Aging II). This group-wise comparison facilitates direct visualization of brain-wide and quantitative differences in physiological parameters across age. As a complementary approach, age-related effects were also evaluated using age as a continuous variable, by testing the significance of nonparametric Spearman’s correlations between age and ROI-wise physiological fMRI metrics. The results of this analysis are presented in supplementary Fig. S6 (top row). Notably, this correlation-based approach yielded spatial patterns of age effects that closely resemble those derived from the group comparisons in Figs. 1, 2: the Spearman’s spatial correlation is 0.80 (p < 10⁻¹⁵) for RV-fMRI temporal lags, 0.39 (p = 3.5 × 10⁻¹²) for HBI-fMRI temporal lags, and 0.85 (p < 10⁻¹⁵) for HBI-fMRI peak correlation values. These observations suggest that the between-group analysis sufficiently captures the overall age-related trends across the examined age range.

In the main text, statistical significance of spatial similarity between age-dependent changes in physiological fMRI metrics and structural, perfusion, and arousal measures was assessed using Spearman’s rank correlation computed across brain regions. As a further confirmatory analysis, we tested the partial correlation between age and various physiological fMRI metrics while controlling for structural and perfusion factors, using the partialcorr function in MATLAB. Arousal was excluded from this analysis due to the small sample size. The results are summarized in supplementary Fig. S6 (middle row). Statistically significant age effects remained evident in most brain regions for all RV-/HBI-related physiological parameters, albeit with slight reductions. These results support the interpretation that, although age-related changes in physiological fMRI metrics are associated with structural and perfusion alterations (particularly perfusion), the examined factors are not the primary drivers of the RV-/HBI-related age effects characterized in this study.

Statistical testing for other analyses, including aging rates and the influence of blood pressure, is detailed in the corresponding method sections above.

To ensure reproducibility of our findings, source data and code used to generate the main results are publicly available (see Code and Data Availability statements). The raw HCP lifespan dataset is available through the NIH NIMH Data Archive.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

All imaging and behavioral data are available through the Human Connectome Project Lifespan studies. Subject IDs are provided in Supplementary Tables S1 and S2. Source data of this study are available at: https://github.com/CANDYlaboratory/Physiology/tree/main/Functional-MRI-Signatures-of-Autonomic-Physiology-in-Aging.

The following publicly available tools were utilized for data preprocessing and analysis: fMRI: FreeSurfer (https://github.com/freesurfer, version 2016-12-29); AFNI (https://afni.nimh.nih.gov, version 17.2.05); FSL (https://fsl.fmrib.ox.ac.uk/fsl, version 6.0.4); ANTs (http://stnava.github.io/ANTs, version 2.3.5); Rstudio (http://www.rstudio.com/, version 2023.06.0 + 421); PPG-beats (https://ppg-beats.readthedocs.io/). Custom code for this project is available at: https://github.com/CANDYlaboratory/Physiology/tree/main/Functional-MRI-Signatures-of-Autonomic-Physiology-in-Aging.

1. Lee, Q. N., Chen, J. E., Wheeler, G. J. & Fan, A. P. Characterizing systemic physiological effects on the blood oxygen level dependent signal of resting-state fMRI in time-frequency space using wavelets. Hum. Brain Mapp. 44, 6537–6551 (2023).

   PubMed  PubMed Central  Google Scholar 

2. Birn, R. M., Diamond, J. B., Smith, M. A. & Bandettini, P. A. Separating respiratory-variation-related fluctuations from neuronal-activity-related fluctuations in fMRI. Neuroimage 31, 1536–1548 (2006).

   PubMed  Google Scholar 

3. Chang, C., Cunningham, J. P. & Glover, G. H. Influence of heart rate on the BOLD signal: The cardiac response function. Neuroimage 44, 857–869 (2009).

   PubMed  Google Scholar 

4. Liu, T. T. Noise contributions to the fMRI signal: an overview. Neuroimage 143, 141–151 (2016).

   PubMed  Google Scholar 

5. Shmueli, K. et al. Low-frequency fluctuations in the cardiac rate as a source of variance in the resting-state fMRI BOLD signal. Neuroimage 38, 306–320 (2007).

   PubMed  Google Scholar 

6. Wise, R. G., Ide, K., Poulin, M. J. & Tracey, I. Resting fluctuations in arterial carbon dioxide induce significant low frequency variations in BOLD signal. Neuroimage 21, 1652–1664 (2004).

   PubMed  Google Scholar 

7. Whittaker, J. R., Driver, I. D., Venzi, M., Bright, M. G. & Murphy, K. Cerebral autoregulation evidenced by synchronized low frequency oscillations in blood pressure and resting-state fMRI. Front. Neurosci. 13, 434073 (2019).

   Google Scholar 

8. Tong, Y., Hocke, L. M. & Frederick, B. B. Low frequency systemic hemodynamic “noise” in resting state BOLD fMRI: Characteristics, causes, implications, mitigation strategies, and applications. Front. Neurosci. 13, 437402 (2019).

   Google Scholar 

9. Power, J. D. et al. Characteristics of respiratory measures in young adults scanned at rest, including systematic changes and “missed” deep breaths. Neuroimage 204, 116234 (2020).

   PubMed  Google Scholar 

10. Das A., Murphy K., Drew P. J. Rude mechanicals in brain haemodynamics: non-neural actors that influence blood flow. Philos. Transact. R. Soc. B. 376, 20190635 (2021).

11. Liu, P. et al. Cerebrovascular reactivity mapping using intermittent breath modulation. Neuroimage 215, 116787 (2020).

    PubMed  Google Scholar 

12. Liu, P. et al. Cerebrovascular reactivity mapping without gas challenges. Neuroimage 146, 320–326 (2017).

    PubMed  Google Scholar 

13. Raut, R. V., Nair, V. A., Sattin, J. A. & Prabhakaran, V. Hypercapnic evaluation of vascular reactivity in healthy aging and acute stroke via functional MRI. Neuroimage Clin. 12, 173–179 (2016).

    PubMed  PubMed Central  Google Scholar 

14. Thomas, B. P. et al. Life-long aerobic exercise preserved baseline cerebral blood flow but reduced vascular reactivity to CO2. J. Magn. Reson. Imaging 38, 1177–1183 (2013).

    PubMed  Google Scholar 

15. Yezhuvath, U. S., Lewis-Amezcua, K., Varghese, R., Xiao, G. & Lu, H. On the assessment of cerebrovascular reactivity using hypercapnia BOLD MRI. NMR Biomed. 22, 779–786 (2009).

    PubMed  PubMed Central  Google Scholar 

16. Liu, P. et al. Cerebrovascular reactivity mapping using resting-state BOLD functional MRI in healthy adults and patients with moyamoya disease. Radiology 299, 419–425 (2021).

    PubMed  Google Scholar 

17. Golestani, A. M., Wei, L. L. & Chen, J. J. Quantitative mapping of cerebrovascular reactivity using resting-state BOLD fMRI: validation in healthy adults. Neuroimage 138, 147–163 (2016).

    PubMed  Google Scholar 

18. Zvolanek, K. M. et al. Comparing end-tidal CO2, respiration volume per time (RVT), and average gray matter signal for mapping cerebrovascular reactivity amplitude and delay with breath-hold task BOLD fMRI. Neuroimage 272, 120038 (2023).

    PubMed  Google Scholar 

19. Peebles, K. et al. Human cerebrovascular and ventilatory CO2 reactivity to end-tidal, arterial and internal jugular vein PCO2. J. Physiol. 584, 347–357 (2007).

    CAS  PubMed  PubMed Central  Google Scholar 

20. Mcswain, D. et al. End-tidal and arterial carbon dioxide measurements correlate across all levels of physiologic dead space. Respir. Care. 55, 288–293 (2010).

    PubMed  Google Scholar 

21. Chang, C. & Glover, G. H. Relationship between respiration, end-tidal CO2, and BOLD signals in resting-state fMRI. Neuroimage 47, 1381–1393 (2009).

    PubMed  Google Scholar 

22. Donahue, M. J. et al. Time delay processing of hypercapnic fMRI allows quantitative parameterization of cerebrovascular reactivity and blood flow delays. J. Cereb. Blood Flow. Metab. 36, 1767 (2015).

    PubMed  PubMed Central  Google Scholar 

23. Bell, M. A. & Ball, M. J. Neuritic plaques and vessels of visual cortex in aging and Alzheimer’s dementia. Neurobiol. Aging 11, 359–370 (1990).

    CAS  PubMed  Google Scholar 

24. Bullitt, E. et al. The effects of healthy aging on intracerebral blood vessels visualized by magnetic resonance angiography. Neurobiol. Aging 31, 290–300 (2010).

    PubMed  PubMed Central  Google Scholar 

25. Thore, C. R. et al. Morphometric analysis of arteriolar tortuosity in human cerebral white matter of preterm, young, and aged subjects. J. Neuropathol. Exp. Neurol. 66, 337–345 (2007).

    PubMed  Google Scholar 

26. Abernethy, W. B., Bell, M. A., Morris, M. & Moody, D. M. Microvascular density of the human paraventricular nucleus decreases with aging but not hypertension. Exp. Neurol. 121, 270–274 (1993).

    CAS  PubMed  Google Scholar 

27. Bell, M. A. & Ball, M. J. Morphometric comparison of hippocampal microvasculature in ageing and demented people: diameters and densities. Acta Neuropathol. 53, 299–318 (1981).

    CAS  PubMed  Google Scholar 

28. Brown, W. R., Moody, D. M., Thore, C. R., Challa, V. R. & Anstrom, J. A. Vascular dementia in leukoaraiosis may be a consequence of capillary loss not only in the lesions, but in normal-appearing white matter and cortex as well. J. Neurol. Sci. 257, 62–66 (2007).

    PubMed  PubMed Central  Google Scholar 

29. Buée, L. et al. Pathological alterations of the cerebral microvasculature in Alzheimer’s disease and related dementing disorders. Acta Neuropathol. 87, 469–480 (1994).

    PubMed  Google Scholar 

30. Mann, D. M. A., Eaves, N. R., Marcyniuk, B. & Yates, P. O. Quantitative changes in cerebral cortical microvasculature in ageing and dementia. Neurobiol. Aging 7, 321–330 (1986).

    CAS  PubMed  Google Scholar 

31. Thorin-Trescases, N. et al. Impact of pulse pressure on cerebrovascular events leading to age-related cognitive decline. Am. J. Physiol. Heart Circ. Physiol. 314, H1214–H1224 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

32. Bennett, H. C. et al. Aging drives cerebrovascular network remodeling and functional changes in the mouse brain. Nat. Commun. 15, 1–20 (2024). 2024 15:1.

    Google Scholar 

33. Juttukonda, M. R. et al. Characterizing cerebral hemodynamics across the adult lifespan with arterial spin labeling MRI data from the Human Connectome Project-Aging. Neuroimage 230, 117807 (2021).

    PubMed  Google Scholar 

34. Bhogal, A. A. et al. The BOLD cerebrovascular reactivity response to progressive hypercapnia in young and elderly. Neuroimage 139, 94–102 (2016).

    PubMed  Google Scholar 

35. De Vis, J. B. et al. Age-related changes in brain hemodynamics; a calibrated MRI study. Hum. Brain Mapp. 36, 3973–3987 (2015).

    PubMed  PubMed Central  Google Scholar 

36. Gauthier, C. J. et al. Age dependence of hemodynamic response characteristics in human functional magnetic resonance imaging. Neurobiol. Aging 34, 1469–1485 (2013).

    PubMed  Google Scholar 

37. Lu, H. et al. Alterations in cerebral metabolic rate and blood supply across the adult lifespan. Cereb. Cortex. 21, 1426–1434 (2011).

    PubMed  Google Scholar 

38. Lee, H. Y. & Oh, B. H. Aging and arterial stiffness. Circ. J. 74, 2257–2262 (2010).

    PubMed  Google Scholar 

39. Kornemann N. et al. Cerebral microcirculatory pulse wave propagation and pulse wave amplitude mapping in retrospectively gated MRI. Sci. Rep. 13, 1–10(2023).

40. Mitchell, G. F. Arterial stiffness in aging: does it have a place in clinical practice? Recent advances in hypertension. Hypertension 77, 768–780 (2021).

    CAS  PubMed  Google Scholar 

41. O’Rourke, M. F. & Hashimoto, J. Mechanical factors in arterial aging: a clinical perspective. J. Am. Coll. Cardiol. 50, 1–13 (2007).

    PubMed  Google Scholar 

42. Jefferson, A. L. et al. Higher aortic stiffness is related to lower cerebral blood flow and preserved cerebrovascular reactivity in older adults. Circulation 138, 1951–1962 (2018).

    PubMed  PubMed Central  Google Scholar 

43. DuBose, L. E. et al. Higher aortic stiffness is associated with lower global cerebrovascular reserve among older humans. Hypertension 72, 476–482 (2018).

    CAS  PubMed  Google Scholar 

44. O’Rourke, M. F. & Safar, M. E. Relationship between aortic stiffening and microvascular disease in brain and kidney: cause and logic of therapy. Hypertension 46, 200–204 (2005).

    PubMed  Google Scholar 

45. Ito, S. Cardiorenal syndrome: an evolutionary point of view. Hypertension 60, 589–595 (2012).

    CAS  PubMed  Google Scholar 

46. Kim, H. L. Arterial stiffness and hypertension. Clin. Hypertens. 29, 31 (2023).

    PubMed  PubMed Central  Google Scholar 

47. Pearson J. D., Morrell C. H., Brant L. J., Landis P. K., Fleg J. L. Age-associated changes in blood pressure in a longitudinal study of healthy men and women. J. Gerontol. Series A. 52A, M177–M183 (1997).

48. Franklin, S. S. et al. Hemodynamic patterns of age-related changes in blood pressure: The Framingham Heart Study. Circulation 96, 308–315 (1997).

    CAS  PubMed  Google Scholar 

49. Ferreira, L. K. & Busatto, G. F. Resting-state functional connectivity in normal brain aging. Neurosci. Biobehav. Rev. 37, 384–400 (2013).

    PubMed  Google Scholar 

50. Dennis, E. L. & Thompson, P. M. Functional brain connectivity using fMRI in aging and Alzheimer’s disease. Neuropsychol. Rev. 24, 49–62 (2014).

    PubMed  PubMed Central  Google Scholar 

51. Bookheimer, S. Y. et al. The Lifespan Human Connectome Project in aging: an overview. Neuroimage 185, 335–348 (2019).

    PubMed  Google Scholar 

52. Harms, M. P. et al. Extending the Human Connectome Project across ages: imaging protocols for the Lifespan Development and Aging projects. Neuroimage 183, 972–984 (2018).

    PubMed  Google Scholar 

53. Fan J. et al. Characterizing fMRI signatures of systemic physiology in healthy aging. In Proc. of Annual Conference of Organization for Human Brain Mapping. p. 2305 (2023).

54. Fjell, A. M. et al. Development and aging of cortical thickness correspond to genetic organization patterns. Proc. Natl. Acad. Sci. USA 112, 15462–15467 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

55. Roe, J. M. et al. A symmetric thinning of the cerebral cortex across the adult lifespan is accelerated in Alzheimer’s disease. Nat. Commun. 12, 1–11 (2021). 2021 12:1.

    Google Scholar 

56. Bae, J. H. & Jeon, Y. J. Pulse sharpness as a quantitative index of vascular aging. Sci. Rep. 11, 1–10 (2021). 2021 11:1.

    Google Scholar 

57. Blinkouskaya, Y., Caçoilo, A., Gollamudi, T., Jalalian, S. & Weickenmeier, J. Brain aging mechanisms with mechanical manifestations. Mech. Ageing Dev. 200, 111575 (2021).

    CAS  PubMed  PubMed Central  Google Scholar 

58. Shen, X. et al. Nonlinear dynamics of multi-omics profiles during human aging. Nat. Aging 4, 1619–1634 (2024). 2024 4:11.

    PubMed  PubMed Central  Google Scholar 

59. Chen, J. E. et al. Resting-state “physiological networks. Neuroimage 213, 116707 (2020).

    PubMed  Google Scholar 

60. Almeida-Santos, M. A. et al. heart rate variability and patterns of autonomic regulation of the heart. Arch. Gerontol. Geriatr. 63, 1–8 (2016).

    PubMed  Google Scholar 

61. Hayano J., Kisohara M., Ueda N., Yuda E. Impact of heart rate fragmentation on the assessment of heart rate variability. Appl. Sci. 10, 3314 (2020).

62. Thambisetty, M. et al. Longitudinal changes in cortical thickness associated with normal aging. Neuroimage 52, 1215–1223 (2010).

    PubMed  Google Scholar 

63. McGinnis, S. M., Brickhouse, M., Pascual, B. & Dickerson, B. C. Age-related changes in the thickness of cortical zones in humans. Brain Topogr. 24, 279–291 (2011).

    PubMed  PubMed Central  Google Scholar 

64. Hogstrom, L. J., Westlye, L. T., Walhovd, K. B. & Fjell, A. M. The structure of the cerebral cortex across adult life: age-related patterns of surface area, thickness, and gyrification. Cereb. Cortex. 23, 2521–2530 (2013).

    PubMed  Google Scholar 

65. Salat, D. H. et al. Thinning of the cerebral cortex in aging. Cereb. Cortex. 14, 721–730 (2004).

    PubMed  Google Scholar 

66. Shaw, M. E., Sachdev, P. S., Anstey, K. J. & Cherbuin, N. Age-related cortical thinning in cognitively healthy individuals in their 60 s: the PATH through life study. Neurobiol. Aging 39, 202–209 (2016).

    PubMed  Google Scholar 

67. Kielstein, J. T. et al. ADMA increases arterial stiffness and decreases cerebral blood flow in humans. Stroke 37, 2024–2029 (2006).

    PubMed  Google Scholar 

68. Badji A. et al. Arterial stiffness and brain integrity: a review of MRI findings. Ageing Res. Rev. 53, 100907 (2019).

69. Pires, P. W., Dams Ramos, C. M., Matin, N. & Dorrance, A. M. The effects of hypertension on the cerebral circulation. Am. J. Physiol. Heart Circ. Physiol. 304, H1598 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

70. Gold, B. P. et al. Functional MRI signals exhibit stronger covariation with peripheral autonomic measures as vigilance decreases. Imaging. Neuroscience 2, 1–25 (2024).

    Google Scholar 

71. Özbay, P. S. et al. Contribution of systemic vascular effects to fMRI activity in white matter. Neuroimage 176, 541–549 (2018).

    PubMed  Google Scholar 

72. Handwerker, D. A., Gazzaley, A., Inglis, B. A. & D’Esposito, M. Reducing vascular variability of fMRI data across aging populations using a breathholding task. Hum. Brain Mapp. 28, 846–859 (2007).

    PubMed  Google Scholar 

73. West, K. L. et al. BOLD hemodynamic response function changes significantly with healthy aging. Neuroimage 188, 198–207 (2019).

    PubMed  Google Scholar 

74. Polimeni J. R. Neurovascular coupling and functional neuroimaging. Ref. Module Neurosc. Biobehav. Psychol. 306–324 (2025).

75. Rivera-Rivera, L. A. et al. Assessment of vascular stiffness in the internal carotid artery proximal to the carotid canal in Alzheimer’s disease using pulse wave velocity from low rank reconstructed 4D flow MRI. J. Cereb. Blood Flow. Metab. 41, 298–311 (2021).

    CAS  PubMed  Google Scholar 

76. Björnfot, C. et al. Cerebral arterial stiffness is linked to white matter hyperintensities and perivascular spaces in older adults—a 4D flow MRI study. J. Cereb. Blood Flow. Metab. 44, 1343–1351 (2024).

    PubMed  PubMed Central  Google Scholar 

77. Rasulo, F. A., De Peri, E. & Lavinio, A. Transcranial Doppler ultrasonography in intensive care. Eur. J. Anaesthesiol. 25, 167–173 (2008).

    Google Scholar 

78. Cohen, E. R., Ugurbil, K. & Kim, S. G. Effect of basal conditions on the magnitude and dynamics of the blood oxygenation level-dependent fMRI response. J. Cereb. Blood Flow. Metab. 22, 1042–1053 (2002).

    CAS  PubMed  Google Scholar 

79. Vazquez, A. L. et al. Vascular dynamics and BOLD fMRI: CBF level effects and analysis considerations. Neuroimage 32, 1642–1655 (2006).

    PubMed  Google Scholar 

80. Nyhan, D., Steppan, J., Barodka, V. & Berkowitz, D. E. V. ascularS. tiffness and Increased Pulse Pressure in the Aging Cardiovascular System. Cardiol. Res Pract. 2011, 263585 (2011).

    PubMed  PubMed Central  Google Scholar 

81. Stein, P. K., Domitrovich, P. P., Hui, N., Rautaharju, P. & Gottdiener, J. Sometimes higher heart rate variability is not better heart rate variability: results of graphical and nonlinear analyses. J. Cardiovasc Electrophysiol. 16, 954–959 (2005).

    PubMed  Google Scholar 

82. Song R. et al. The physiological component of the BOLD signal: impact of age and heart rate variability biofeedback training. bioRxiv. 2025-04 (2025).

83. Camandola, S. & Mattson, M. P. Brain metabolism in health, aging, and neurodegeneration. EMBO J. 36, 1474–1492 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

84. Iadecola, C. The neurovascular unit coming of age: a journey through neurovascular coupling in health and disease. Neuron 96, 17–42 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

85. Tarantini, S., Tran, C. H. T., Gordon, G. R., Ungvari, Z. & Csiszar, A. Impaired neurovascular coupling in aging and Alzheimer’s disease: Contribution of astrocyte dysfunction and endothelial impairment to cognitive decline. Exp. Gerontol. 94, 52–58 (2017).

    CAS  PubMed  Google Scholar 

86. Beissner, F., Meissner, K., Bär, K. J. & Napadow, V. The autonomic brain: an activation likelihood estimation meta-analysis for central processing of autonomic function. J. Neurosci. 33, 10503–10511 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

87. Bernier, M., Cunnane, S. C. & Whittingstall, K. The morphology of the human cerebrovascular system. Hum. Brain Mapp. 39, 4962–4975 (2018).

    PubMed  PubMed Central  Google Scholar 

88. Tsvetanov et al. Vascular and neuronal effects of age on fMRI BOLD signals. Philos. Transact. R. Soc. B. 376, (2021).

89. Reyes del Paso, G. A., Langewitz, W., Mulder, L. J. M., van Roon, A. & Duschek, S. The utility of low frequency heart rate variability as an index of sympathetic cardiac tone: A review with emphasis on a reanalysis of previous studies. Psychophysiology 50, 477–487 (2013).

    PubMed  Google Scholar 

90. Julien, C. The enigma of Mayer waves: Facts and models. Cardiovasc Res. 70, 12–21 (2006).

    CAS  PubMed  Google Scholar 

91. Bolt, T. et al. Widespread autonomic physiological coupling across the brain-body axis. bioRxiv. 2023.01.19.524818 (2024).

92. Gu, Y. et al. An orderly sequence of autonomic and neural events at transient arousal changes. Neuroimage 264, 119720 (2022).

    PubMed  Google Scholar 

93. Raut, R. V. et al. Global waves synchronize the brain’s functional systems with fluctuating arousal. Sci. Adv. 7, eabf2709 (2021).

94. Robison, L. S., Gannon, O. J., Salinero, A. E. & Zuloaga, K. L. Contributions of sex to cerebrovascular function and pathology. Brain Res. 1710, 43–60 (2019).

    CAS  PubMed  Google Scholar 

95. Orshal J. M., Khalil R. A. Gender, sex hormones, and vascular tone. Am. J. Physiol. Regul. Integr. Comp. Physiol. 286, R233-49 (2004).

96. Takahashi, K. et al. Impact of menopause on the augmentation of arterial stiffness with aging. Gynecol. Obstet. Investig. 60, 162–166 (2005).

    Google Scholar 

97. Moreau, K. L. & Hildreth, K. L. Vascular aging across the menopause transition in healthy women. Adv. Vasc. Med. 2014, 204390 (2014).

    PubMed  PubMed Central  Google Scholar 

98. Charlton, P. H. et al. Detecting beats in the photoplethysmogram: benchmarking open-source algorithms. Physiol. Meas. 43, 085007 (2022).

    PubMed Central  Google Scholar 

99. van Ravenswaaij-Arts, C. M. A., Kollee, L. A. A., Hopman, J. C. W., Stoelinga, G. B. A. & van Geijn, H. P. Heart rate variability. Ann. Intern Med. 118, 436–447 (1993).

    PubMed  Google Scholar 

100. Cox, R. W. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput. Biomed. Res. 29, 162–173 (1996).

     CAS  PubMed  Google Scholar 

101. Smith, S. M. et al. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 23, S208–S219 (2004).

     PubMed  Google Scholar 

102. Glover, G. H., Li, T.-Q. & Ress, D. Image-based method for retrospective correction of physiological motion effects in fMRI: RETROICOR. Magn. Reson Med. 44, 162–167 (2000).

     CAS  PubMed  Google Scholar 

103. Schaefer, A. et al. Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cereb. Cortex. 28, 3095–3114 (2018).

     PubMed  Google Scholar 

104. Fischl, B. FreeSurfer. Neuroimage 62, 774–781 (2012).

     PubMed  Google Scholar 

105. Chang, C. et al. Tracking brain arousal fluctuations with fMRI. Proc. Natl. Acad. Sci. USA 113, 4518–4523 (2016).

     CAS  PubMed  PubMed Central  Google Scholar 

Download references

Authors and Affiliations

1. Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, USA

   Jiawen Fan, Meher R. Juttukonda, Divya Varadarajan, Jonathan R. Polimeni, David H. Salat & Jingyuan E. Chen

2. Department of Radiology, Harvard Medical School, Boston, MA, USA

   Meher R. Juttukonda, Divya Varadarajan, Jonathan R. Polimeni, David H. Salat & Jingyuan E. Chen

3. Department of Biomedical Engineering, Vanderbilt University, Nashville, TN, USA

   Sarah E. Goodale, Shiyu Wang & Catie Chang

4. Centre for Sleep and Cognition, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore

   Csaba Orbán

5. Harvard-MIT Program in Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, USA

   Jonathan R. Polimeni

6. Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN, USA

   Catie Chang

Contributions

J.F., J.E.C., D.H.S., and C.C. conceived and designed the study. J.F. performed the formal analyses with support from M.R.J., S.E.G., S.W., and D.V.. J.R.P., C.C., C.O., M.R.J., and D.H.S. helped interpret the results. J.E.C. supervised and funded this study. J.F. and J.E.C. wrote the manuscript. All authors contributed to editing the manuscript.

Corresponding author

Correspondence to Jingyuan E. Chen.

Competing interests

The authors declare no competing interests.

Peer review information

Communications Biology thanks Donna Y Chen, Ali M. Golestani, and the other anonymous reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Sahar Ahmad and Jasmine Pan. A peer review file is available.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions