Multimodal Free-Water Imaging Links Cardiometabolic Risk to Periarterial Dysfunction and Amyloid Accumulation in Early Alzheimer’s

Multimodal Free-Water Imaging Links Cardiometabolic Risk to Periarterial Dysfunction and Amyloid Accumulation in Early Alzheimer’s | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Multimodal Free-Water Imaging Links Cardiometabolic Risk to Periarterial Dysfunction and Amyloid Accumulation in Early Alzheimer’s Yaqiong Chai, Hedong Zhang, Andrew S. Kim, Quanhao Sun, Shuting Chen, and 15 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9374985/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract The brain’s waste-clearance (glymphatic) system removes metabolic byproducts via periarterial influx, interstitial exchange, and perivenous efflux. Although dysfunction is implicated in Alzheimer’s disease (AD), current imaging markers emphasize perivenous changes and may overlook earlier periarterial impairment. We developed a diffusion MRI framework to quantify periarterial fluid mobility, white matter free water, and perivenous integrity, and applied it to 546 cognitively normal adults (HCP-Aging) and 173 participants across the AD spectrum (ADNI). Periarterial mobility was reduced with higher cardiometabolic risk and amyloid positivity, particularly in AD-vulnerable regions. Free water increased with aging and metabolic burden, whereas perivenous dysfunction was most pronounced in AD. Combined measures predicted amyloid positivity and cognitive impairment (AUC = 0.82). Mediation analyses showed that blood pressure influenced cognition through periarterial dysfunction and amyloid burden. These findings support a staged, compartment-specific trajectory of glymphatic dysfunction, with early periarterial impairment representing a potential biomarker and therapeutic target. Health sciences/Biomarkers Health sciences/Diseases Health sciences/Neurology Biological sciences/Neuroscience Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Alzheimer’s disease (AD) is defined by progressive impairment of memory and cognitive function and by the pathological accumulation of amyloid-β (Aβ) plaques and tau neurofibrillary tangles, hallmarks central to disease pathogenesis. Emerging evidence indicates that impaired clearance of Aβ via the glymphatic system significantly contributes to AD onset and progression 1 , 2 . Insufficient removal of tau aggregates may further accelerate cognitive decline 3 . The glymphatic system supports brain waste clearance through three principal phases: cerebrospinal fluid (CSF) influx into periarterial spaces (pAs), bulk solute exchange with interstitial fluid (ISF) in brain parenchyma, and efflux via perivenous spaces (pVs) toward lymphatic drainage. Yet, the relative vulnerability and distinct contributions of these three components to protein clearance impairments in AD, especially in humans, remain poorly understood. Current in vivo human imaging methods of the glymphatic system predominantly rely on diffusion tensor imaging along the perivascular space (DTI-ALPS) 4 , which quantifies water diffusivity within pVs near paraventricular white matter. While the ALPS index is significantly reduced in AD, its sensitivity to glymphatic dysfunction in early AD stages is limited 5 . Studies in individuals with mild cognitive impairment (MCI) or subjective cognitive decline have yielded inconsistent results, partly due to the ALPS index’s regional specificity. It primarily captures changes near the lateral ventricles and may miss early alterations in regions more vulnerable to initial amyloid pathology, such as the precuneus, or tau pathology, such as the entorhinal cortex 6 . Additionally, emerging hypotheses propose distinct underlying mechanisms of glymphatic dysfunction across disease stages. For instance, early-stage impairment may stem from reduced periarterial CSF influx due to arterial stiffening and diminished pulsatility, resulting in pressure reversal, impaired solute clearance, and cerebral amyloid angiopathy. In contrast, disruptions detected by ALPS in the perivenous system may reflect downstream consequences of extensive parenchymal Aβ accumulation in later disease stages. Moreover, the presence of other comorbidities may additively impair the glymphatic system (e.g., hypertensive arteriolosclerosis may decrease pulsatility in pAS, while obstructive sleep apnea or traumatic bleeding into the subarachnoid space may reduce CSF absorption in the pVs). For instance, pAs function, reliant on arterial compliance, may be more vulnerable to vascular risk factors such as hypertension, hyperlipidemia, diabetes and obesity 7 , 8 . Conversely, pVs dynamics are influenced by respiratory mechanics and may be adversely affected by sleep-disordered breathing, which commonly precedes cognitive impairment and may exacerbate glymphatic dysfunction 9 . While increased free water volume fraction (FWVF) in white matter (WM) has been proposed as a potential imaging correlate of glymphatic dysfunction 10 , the field still lacks a direct, validated in vivo imaging framework that can evaluate water diffusivity across all three compartments of the glymphatic pathway. This creates a critical knowledge gap regarding glymphatic dysfunction mechanisms across various AD stages, age groups, and comorbidities. To address this gap, we leveraged two complementary imaging cohorts: the Human Connectome Project in Aging (HCP-A) to examine glymphatic vulnerability to cardiometabolic risk factors and sleep in cognitively normal older adults 11 , and the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to characterize glymphatic alterations associated with Aβ and tau pathology, clinical disease severity, and cognitive impairment 12 . This dual-cohort approach allowed us to evaluate both early vulnerability and AD-specific dysfunction in a stage-sensitive manner. In this study, we set out to quantify glymphatic clearance across the three core compartments using multimodal MRI. Specifically, we aimed to: (1) identify glymphatic changes linked to aging, sleep, and cardiometabolic syndrome (CMS) in cognitively normal (CN) adults; (2) determine glymphatic components most strongly associated with Aβ and tau pathology; (3) characterize glymphatic disruption across CN, mild cognitive impairment (MCI), and AD cohorts; (4) evaluate the predictive value of glymphatic markers in classifying MCI or MCI/AD from CN individuals; and (5) map regional patterns of glymphatic dysfunction linked to aging and CMS, and AD progression. Results Study participants For the analysis of free-water diffusivity using multi-shell DWI and bi-tensor modeling, we analyzed 579 cognitively normal subjects (age range: 36–100 years; 58.6 \(\:\pm\:\) 14.1 years; 312 females) from the HCP-A Release 2.0 and 10 AD dementia patients, 68 MCI patients, and 95 CN controls from the ADNI-3 and ADNI-4 datasets (age range: 36–100 years; 76.5 \(\:\pm\:\) 7.3 years; 78 females). Initially, 687 subjects in HCP-A had MRI data available. However, 63 participants were excluded due to poor data quality following quality assurance protocols; 48 were excluded due to lack of the CMS and sleep data; 28 were excluded due to segmentation errors; and 2 were excluded due to anatomical abnormalities, resulting in 546 subjects from the HCP-A cohort in the present analysis. The demographic and clinical characteristics of the HCP-A and ADNI samples are summarized in Table 1 and Table 2 , respectively. Table 1 Characteristics of HCP-Aging participants Age (years) N = 546 58.61 (14.08) Women, n(%) 312 (57.14) BMI (kg/m 2 ) a 27.10 (4.84) Systolic (mmHg) b 129.05 (16.42) Diastolic (mmHg) b 79.29 (10.45) Mean arterial pressure (mmHg) b 95.00 (14.34) High density lipoprotein (mg/dL) 60.05 (16.47) Hemoglobin A1c (%) 5.36 (0.53) Total cholesterol (mg/dL) 198.69 (36.41) Triglycerides (mg/dL) 108.41 (68.34) Fasting-glucose (mg/dL) 101.91 (17.94) PSQI c 4.68 (2.69) MoCA 26.40 (2.55) Data are presented as mean (s.d.) or number (percentage). a BMI was missing for two subjects. b Blood pressure was missing for five subjects. c PSQI was missing for one subject. Table 2 Characteristics of ADNI participants Age, years All (n = 173) CN (n = 95) MCI (n = 68) AD (n = 10) 76.49 (7.33) 76.40 (7.48) 76.44 (7.00) 77.70 (8.60) Women (%) 78 (45.1%) 45 (47.4) 28 (41.2) 5 (50.0) APOE-ε4 carriers, n (%) a 54 (42.5) 24 (36.9) 24 (43.6) 6 (85.7) Education, years b 16.73(2.31) 16.99 (2.33) 16.35 (2.65) 16.90 (2.01) MMSE c 27.90 (2.68) 28.81 (1.32) 27.26 (2.89) 23.50 (4.88) MoCA d 24.45 (4.15) 26.06 (2.45) 22.75 (4.86) 17.86 (3.13) Aβ-PET measures n = 158 n = 92 n = 59 n = 7 Amyloid positive (n, %) 0.45 (0.50) 0.33 (0.47) 0.58 (0.50) 1.00 (0.00) Aβ SUVR (whole brain) 1.20 (0.27) 1.12 (0.23) 1.27 (0.28) 1.49 (0.34) Aβ SUVR (precuneus) 1.29 (0.32) 1.21 (0.28) 1.37 (0.32) 1.59 (0.37) Aβ SUVR (AD-Signature) 1.16 (0.27) 1.08 (0.23) 1.24 (0.27) 1.45 (0.36) Tau-PET measures n = 160 n = 90 n = 62 n = 8 Tau SUVR (whole brain) 1.19 (0.23) 1.13 (0.12) 1.23 (0.29) 1.48 (0.38) Tau SUVR (precuneus) 1.19 (0.23) 1.13 (0.12) 1.23 (0.29) 1.48 (0.38) Tau SUVR (AD-Signature) 1.32 (0.32) 1.21 (0.13) 1.40 (0.37) 1.96 (0.49) BMI e 28.23 (11.95) 30.02 (15.89) 26.22 (3.88) 26.69 (6.97) Systolic Blood Pressure f 136.91 (18.77) 135.97 (17.42) 138.03 (19.98) 138.00 (23.67) Diastolic Blood Pressure f 73.53 (9.60) 73.65 (9.00) 73.15 (10.25) 75.00 (11.21) Data are presented as mean (s.d.) or number (percentage). Aβ-PET positivity was determined as composite AV45 SUVR > 0.1.11 or FBB SUVR > 1.478. a APOE-ε4 was missing for 30 subjects in CN, 13 in MCI, and 3 in AD. b Education was missing for five subjects in CN and two in MCI. c MoCA was missing for two subjects in CN, seven subjects in MCI and three in AD. d MMSE was missing for two participants in MCI. e BMI was missing for two subjects in AD, 30 in CN and 17 in MCI. f Blood pressures were missing for two subjects in CN. Multimodal MRI metrics to understand glymphatic function across the three compartments We estimated the following complementary metrics: 1) pAs diffusivity, quantified using free-water axial diffusivity (FWAD) and additional diffusion parameters (radial diffusivity, mean diffusivity), along with pAs-FWVF; 2) WM-FWVF, reflecting parenchymal fluid accumulation and ISF–CSF exchange efficiency; and 3) DTI-ALPS index, reflecting perivenous efflux integrity. Motivated by prior evidence that MRI-visible perivascular spaces are predominantly periarteriolar in nature 13 , we quantified pAs diffusivity within perivascular spaces segmented from T1-weighted MRI. Multi-shell DWI data were analyzed using tissue tensor imaging (TTI), an advanced bi-tensor approach that separates tissue and free-water components while allowing anisotropic diffusion in the free-water compartment 14 – 16 . In contrast to conventional free-water–eliminated DTI, TTI enables directional characterization of fluid diffusivity along CSF-filled conduits such as the pAs ( Supplementary Methods ). Figure 1 provides a conceptual illustration of glymphatic alterations in healthy versus pathological states based on our findings in the following sections (Fig. 2 – 4 ). In healthy brains, robust CSF mobility along the pAs (high FWAD), low FWVF in pAs and WM, and high pVs-ALPS indices signify effective glymphatic circulation. In contrast, pathological states exhibit diminished pAs FWAD (impaired influx), elevated FWVF (fluid retention and greater flow impedance), and reduced ALPS (impaired pVs-mediated efflux), consistent with progressive glymphatic failure (Fig. 1 a–b). To minimize partial-volume contamination in pAs diffusivity measurements, it was essential that the selected pAs segments be sufficiently large relative to the spatial resolution of the DWI data. By selecting the four largest pAs per region of interest (ROI), we observed that the mean pAs diameters in both HCP-A and ADNI cohorts exceeded the corresponding DWI voxel sizes ( Tables S1–S2 ; HCP-A: pAs diameter = 1.7–2.2 mm vs. 1.5 mm isotropic DWI; ADNI: 2.7–3.0 mm vs. 2.0 mm isotropic DWI). We further validated that this choice provided optimal stability and sensitivity of associations with clinical variables compared with other choices of n largest pAs (see Methods and Supplementary Fig. S1 ), supporting the use of the four-largest-pAs threshold in subsequent analyses. Association of whole white matter glymphatic measures with aging, sleep, cognition, and CMS in HCP-A Whole WM glymphatic measures were examined in healthy aging cohorts after adjusting for effects of age and sex. In the periarterial compartment (Fig. 1 c; Supplementary Fig. S2 ), FWAD exhibited significant negative correlations with multiple CMS indicators, including body mass index (BMI; β = − 0.36, p = 2.8×10⁻ 17 ), systolic blood pressure (β = − 0.18, p = 1.0×10 − 4 ), diastolic blood pressure (β = − 0.17, p = 1.2×10 − 4 ), mean arterial pressure (MAP; β = − 0.19, p = 3.0×10⁻⁵), hemoglobin A1c (HbA1c; β = − 0.16, p = 5.1×10 − 4 ), and fasting glucose (β = − 0.15, p = 0.001), while showing a positive correlation with high-density lipoprotein (HDL; β = 0.17, p = 5.1×10 − 4 ). FWAD was not associated with age (p = 0.10). The effects of fasting glucose and blood pressure on pAs-FWAD were significant (p = 5.1×10⁻ 6 and 6.7×10⁻ 4 , respectively), but after adding antihypertensive and diabetic medication use as covariates, these correlations were no longer significant (p = 0.236 and 0.968). On the other hand, FWVF increased with age (β = 0.48, p = 1.2×10⁻ 24 ), and was positively associated with BMI (β = 0.19, p = 1.2×10 − 4 ) and triglycerides (β = 0.17, p = 2.0×10 − 5 ), but negatively correlated with HDL (β = − 0.15, p = 1.1×10 − 4 ). All p-values reported after false discovery rate correction. After adjusting for age (but not sex), cognitive performance (MoCA) was positively correlated with FWAD (β = 0.11, p = 0.02). This relationship was observed in males (β = 0.11, p = 0.02) but not in females. No significant correlations between global pAs metrics and sleep quality (PSQI) were found. Comparisons with other diffusivity metrics, including free-water radial diffusivity (FWRD), mean diffusivity (FWMD), and fractional anisotropy (FWFA) revealed similar directional trends but weaker associations ( Supplementary Fig. S3 ). Fewer CMS factors were significantly correlated with these metrics, and their absolute β-values were consistently smaller than those for FWAD. Thus, FWAD was selected as the representative marker of glymphatic diffusivity for subsequent analyses. In WM parenchyma, FWVF positively correlated with age (β = 0.55, p = 4.9x10 -47 ), BMI (β = 0.17, p = 8.7x10 -6 ), and triglyceride levels (β = 0.18, p = 8.7x10 -6 ), and negatively with HDL (β = − 0.11, p = 0.01). No correlation with PSQI was found. This pattern was similar to that of FWVF in pAs. Finally, the ALPS index, reflecting perivenous efflux, showed a robust negative association with age (β = − 0.37, p = 3.3x10 -20 ) but no correlation with PSQI, MoCA, or CMS factors. Regional association of glymphatic metrics with aging, sleep, cognition, and CMS in HCP-A. As we divided the entire WM into nine functional ROIs (see Methods ), region-specific analyses (Fig. 1 d) revealed differential spatial patterns of glymphatic vulnerability in pAs and WM parenchyma to CMS factors. Note that no regional pattern is available for pVs ALPS as this index is a single global measure. Across nine ROIs, age showed widespread positive associations with FWVF in both pAs and WM parenchyma, spanning the frontoparietal (FPN), dorsal attention, ventral language, sensorimotor, auditory, visual, salience, default mode (DMN), and limbic networks (p < 0.05). No regional FWAD associations with age were detected. Of the CMS variables, higher BMI was associated with lower FWAD and higher FWVF in pAs across all ROIs, paralleling the WM-FWVF pattern, though effects were most pronounced in the DMN, salience, sensorimotor, dorsal attention, and visual networks. Higher diastolic blood pressure was linked to widespread lower FWAD across networks and to higher FWVF restricted within the dorsal attention network (pAs) and FPN (WM). Higher triglyceride levels were associated with lower pAs-FWAD in the sensorimotor, FPN, visual, and DMN regions, and with higher FWVF across all WM networks except the limbic region. Furthermore, as in Supplementary Fig. S4 , lower HDL cholesterol correlated with lower FWAD and higher FWVF in both pAs and WM, particularly in frontal and temporal cortices. Elevated HbA1c and fasting glucose were also linked to widespread reductions in FWAD, primarily along periarterial pathways. Sleep quality (PSQI) showed a positive correlation with FWAD in the dorsal attention network and a negative correlation with periarterial FWVF in the visual network. Cognitive performance (MoCA) positively correlated with pAs-FWAD within the dorsal attention, salience, auditory, and default mode networks, whereas no significant relationships were observed for WM-FWVF or pVs-ALPS. Glymphatic alterations associated with Alzheimer's pathology and clinical stages in ADNI We examined group differences in glymphatic metrics across CN, MCI, and AD participants from the ADNI dataset (Fig. 2 ), with age, sex, years of education, and APOE genotype (allele ε4 vs. others) as covariates. pAs-FWAD was significantly lower in Aβ-positive compared to Aβ-negative individuals within both the precuneus (p = 0.038) and AD-signature regions (p = 0.003). pAs-FWVF in these same regions was significantly higher in the Aβ-positive group than the Aβ-negative (p = 0.008 in precuneus and 0.014 in AD-signature regions). Similarly, WM-FWVF was significantly higher in Aβ-positive participants within the precuneus (p = 0.041) and AD-signature regions (p = 0.009). However, the global pAs-FWAD, pAs-FWVF, and WM-FWVF showed no significant differences between Aβ-positive and Aβ-negative groups. The ALPS index showed no significant difference between Aβ groups, though a non-significant trend toward higher values in Aβ-positive individuals was observed. When stratified by diagnostic group, pAs and WM-FWVF in the whole brain, precuneus and AD-signature regions showed significant group differences between all pairs of CN, MCI and AD groups (whole brain pAs/FWVF: p = 0.002/0.011; precuneus: p = 0.002/0.011; AD-signature: p = 0.003/0.009). pAs-FWAD in the whole brain, precuneus, and AD-signature regions exhibited significant differences observed between CN and MCI (whole brain: p = 0.006; precuneus: p = 0.002; AD-signature: p = 0.003). We observed significant differences in ALPS between CN and AD (p = 0.009) and between MCI and AD (p = 0.014) but not between CN and MCI (p = 0.738). Classification of Aβ positivity and clinical diagnosis We next evaluated the ability of glymphatic metrics to classify Aβ-PET positivity using pairwise DeLong tests. The integrated glymphatic model combining all three components achieved the highest discriminative performance (AUC = 0.817 when including subjects with all stages of CN + MCI + AD, AUC = 0.823 for those with non-demented stages of CN + MCI only), significantly outperforming models using WM-only features (AUC = 0.633 for all and 0.626 for non-demented, ΔAUC = 0.184–0.197; p = 3.2×10 − 4 ) and ALPS (AUC = 0.606 for all and 0.612 for non-demented, ΔAUC = 0.211; p = 1.3×10⁻⁵). A combined pAs + WM model (AUC = 0.802 for all, and 0.803 for non-demented) also performed significantly better than either WM alone (ΔAUC = 0.169–0.177; p = 1.6×10 − 2 ) or ALPS (ΔAUC = 0.191–0.196; p = 2.7×10 − 4 ). Among single-compartment models, pAs (AUC = 0.74 in all and 0.727 in non-demented) outperformed both WM (ΔAUC = 0.101–0.107; p = 0.031) and ALPS (ΔAUC = 0.115–0.134; p = 0.001). Although the integrated model achieved the numerically highest AUC, its improvement over pAs + WM (ΔAUC = 0.015–0.02; p > 0.1) or pAs alone (ΔAUC = 0.077–0.096; p = 0.06) did not reach statistical significance. These findings indicate that pAs features provide the most robust discriminative signal, while WM integration enhances stability, and full glymphatic integration yields only marginal gains. For comparison with a PET biomarker, three combined features of the precuneus, AD-signature and whole brain tau-PET SUVR yielded AUCs of 0.812 (all participants) and 0.796 (non-demented), displaying significant difference in the discriminative performance between integrated glymphatic measures and established molecular imaging biomarkers. In differentiating clinical diagnostic groups (CN vs. MCI + AD and CN vs. MCI; Fig. 3 ), the integrated glymphatic features again achieved the highest overall accuracy (AUC = 0.812–0.821), significantly outperforming both pVs-only (ΔAUC = 0.200-0.208; p = 3.1×10 − 4 ) and WM-only (ΔAUC = 0.186–0.188; p = 8.0×10 − 5 ) models. Among individual components, pAs features (AUC = 0.795–0.802) exhibited markedly higher discriminative ability than pVs (ΔAUC = 0.181–0.191; p = 3.7×10 − 4 ) and WM (ΔAUC = 0.169; p = 0.04). Adding WM to pAs modestly improved performance (AUC = 0.787–0.789) but achieved a marginal significance (ΔAUC = 0.008–0.013; p = 0.05). For comparison with AD pathological biomarkers, combined Aβ- and tau-PET SUVR measures achieved lower AUCs (CN vs. MCI + AD: Aβ = 0.669, tau = 0.743; CN vs. MCI: Aβ = 0.648, tau = 0.698) than the integrated glymphatic features (CN vs. MCI + AD: ΔAUC = 0.076; p = 0.031; CN vs. MCI: ΔAUC = 0.124; p = 0.009). Combining tau with Aβ did not substantially improve performance compared to tau only. Mediation analysis: blood pressure, glymphatic function, amyloid burden, and cognition We conducted a serial mediation analysis to examine whether the relationship between elevated blood pressure, a key factor that dysregulates arterial pulsatility, and cognitive decline (MoCA) was mediated by glymphatic dysfunction and amyloid burden (Fig. 4 ). Diastolic blood pressure (BP) was positively associated with periarterial FWAD in the precuneus (a₁ = 0.00002, p = 0.03), which, in turn, was negatively associated with Aβ SUVR (a₂ = − 64.01, p < 0.05). Higher Aβ SUVR was associated with poorer cognitive performance (b = − 2.733, p = 0.02). The total effect of diastolic BP on MoCA was not significant (c = − 0.031, p = 0.39), and the direct effect after accounting for both mediators remained non-significant (c’ = − 0.039, p = 0.27). However, the indirect effect through both mediators was significant (0.0032, 95% CI, 0.0002, 0.008), indicating that elevated diastolic BP contributes to cognitive decline indirectly by reducing periarterial water mobility (lower FWAD) and subsequently promoting greater Aβ accumulation on PET. Discussion In this study, we used multi-compartment glymphatic metrics derived from a bi-tensor model using multi-shell diffusion MRI to characterize glymphatic alterations across pAs, WM parenchyma, and pVs in two complementary cohorts. Our principal findings are fourfolds: first, measures reflecting increased pAs free-water volume and reduced pAs water mobility were significantly linked to aging, CMS factors, and cognition in healthy aging adults and were also observed in Aβ-positive and cognitively impaired participants, with early changes observed in the precuneus and AD-signature regions. Second, pVs diffusivity decline was associated with aging and was most marked at later clinical AD stages, suggesting a downstream parenchymal and efflux failure following early periarterial compromise. Third, integrated glymphatic metrics combining periarterial, WM, and perivenous glymphatic features classified Aβ-PET positivity, with performance comparable to that of tau PET and clinical diagnosis, and outperformed combined tau and amyloid PET measures. Finally, a serial mediation analysis implicated a directional association linking higher vascular risk factors, pAs glymphatic dysfunction, Aβ tissue accumulation, and cognitive decline, with a statistically significant indirect effect. Together, using multi-compartment diffusion MRI in two large cohorts, our results support a stage-dependent compartmental model of glymphatic dysfunction in human aging and AD. Impact of Aging and Cardiovascular Risk Factors on Glymphatic Function Aging is a well-established contributor to glymphatic system decline, and our findings from pAs-FWVF, WM-FWVF, and pVs-ALPS measurements are consistent with this framework. These metrics revealed age-related deterioration in pAs, pVs, and WM microstructural integrity, supporting the notion that advancing age compromises all three compartments of glymphatic system. Beyond aging, specific CMS risk factors demonstrated distinct associations with pAs functionality, a marker closely linked to arterial pulsatility 17 , 18 . In particular, reduced pAs-FWAD was strongly associated with elevated diastolic blood pressure and fasting glucose levels, suggesting that vascular stiffening and metabolic dysregulation jointly compromise the pulsatile forces driving periarterial CSF transport. Hypertension alters the hemodynamic profile of arterial pulsation in ways detrimental to perivascular flow. In animal models, two-photon microscopy has shown that CSF transport along periarterial channels is driven by rhythmic arterial wall motion, which becomes markedly attenuated under hypertensive conditions due to reduced arterial compliance despite increased pulsation amplitude 17 . A 7T MRI human study similarly revealed elevated microvascular volumetric pulsatility in deep WM among hypertensive and older adults 19 , consistent with the increased pulsation amplitude observed in animal models. Collectively, these findings suggest that hypertension amplifies local vascular stress yet disrupts the coordinated wall motion required for efficient glymphatic transport. Chronic hyperglycemia, as seen in diabetes and poorly controlled glucose levels, may exacerbate glymphatic dysfunction through mechanisms such as microangiopathy, oxidative stress, and alterations in the perivascular basement membrane composition, all of which can impede interstitial solute clearance 20 – 22 . This interpretation aligns with animal studies demonstrating that experimental diabetes leads to perivascular AQP4 mislocalization and reduced glymphatic transport efficiency 22 . Notably, the associations between fasting glucose, blood pressure, and pAs-FWAD were no longer significant after adjusting for antihypertensive and antidiabetic medication use. This attenuation implies that pharmacological control of vascular and metabolic risk factors may help preserve periarterial fluid mobility. Effective management of blood pressure and glucose has been shown to improve cerebrovascular compliance and reduce arterial stiffness 23 , which may in turn stabilize the mechanical driving forces underlying glymphatic transport. Other CMS-related factors, including lower HDL cholesterol, elevated BMI, and higher triglyceride levels were associated with both lower pAs-FWAD and higher pAs- and WM-FWVF, but not with pVs-ALPS. These patterns suggest a combination of microvascular compromise and extracellular space expansion, possibly reflecting chronic low-grade inflammation and tissue remodeling. The convergence of these vascular–metabolic insults may not only hinder clearance but also promote amyloid accumulation and white matter damage over time 24 . Taken together, our findings support the concept that both aging and CMS components exert additive and potentially synergistic effects on glymphatic function, with periarterial pathways appearing especially vulnerable to hemodynamic and metabolic stressors. Role of Sleep Disturbances in Glymphatic Dysfunction Sleep disturbances critically affect glymphatic clearance and, consequently, AD-related neurodegeneration. In our study, poor sleep quality, indicated by high PSQI scores, was associated with pAs-FWAD in the frontoparietal network and FWVF in the visual network. Animal studies have robustly demonstrated that sleep deprivation impairs glymphatic clearance 25 , though human evidence remains limited. Recent work by Ma et. al. indicates that sleep quality directly influences glymphatic function and brain networks critical for visual memory in older adults 26 . Likewise, lighter sleep in middle-age and elderly individuals has been linked to an increased burden of enlarged perivascular spaces 27 . Aging exacerbates the adverse effects of poor sleep by impairing microglial and macrophage function, fostering a pro-inflammatory environment 28 , 29 , which may further impede glymphatic flow. Such chronic inflammation may lead to reactive astrogliosis resulting in AQP4 depolarization and reduced AQP4 expression, consequently impeding CSF–ISF exchange 30 . Glymphatic Dysfunction in AD Progression: Novel Associations between Compartments and AD Stages Our compartment-based analysis revealed stage-dependent alterations, with pAs and WM changes emerging in early phases of AD, and pVs alterations becoming more pronounced in later stages. This pattern aligns with prior experimental and modeling studies in mice, which show that the pAs plays a disproportionately greater role in solute exchange between CSF and brain parenchyma compared with the pVs, owing to its greater abundance, larger caliber, and higher net CSF inflow 31 , 32 . In early stages of AD, reduced pAs-FWAD, potentially influenced by aging, poor sleep, and CMS-related factors 18 , 33 , likely reflects diminished arterial pulsatility and other driving forces that propel glymphatic inflow 34 . Such impairments would slow the clearance of soluble waste products, including Aβ and tau, leading to their gradual accumulation in the interstitial space 3 , 35 . In contrast, alterations in pVs functionality become more prominent in later stages of disease. These changes likely represent downstream consequences of chronic glymphatic failure, characterized by neuroinflammation and elevated flow resistance from Aβ and tau deposition and extracellular water retention 36 , 37 . Supporting this interpretation, previous studies have shown that cerebral amyloid angiopathy and Aβ accumulation in pAs can promote extracellular fluid buildup 38 , possibly through reactive astrogliosis 39 and chronic inflammation 2 . Astrocytic dysfunction, particularly of aquaporin-4 (AQP4) channels on astrocyte endfeet, may further disrupt water and ion homeostasis, exacerbating clearance deficits 39 , 40 . These stage-specific alterations provide a mechanistic explanation for the predictive patterns observed in our data. Because pAs dysfunction arises earlier, its metrics likely capture preclinical glymphatic deficits associated with initial amyloid accumulation, whereas pVs changes reflect later-stage pathology and vascular comorbidities. Integrating pAs, WM, and pVs features thus enables a comprehensive characterization of both early and late glymphatic failure. This multimodal approach achieved high accuracy in classifying individuals across the AD spectrum (CN, MCI and AD) and in predicting early Aβ accumulation—comparable to tau PET performance in non-demented individuals and superior to combined amyloid and tau PET features in differentiating MCI from CN. These results highlight the promise of DTI-based glymphatic markers as a cost-effective, noninvasive tool for staging and risk stratification in AD. Furthermore, our regional mapping successfully localized alterations of pAs and WM function in the AD-signature and precuneus regions, highlighting their association with early disease progression. This regional mapping approach may also help delineate disease-specific glymphatic dysfunction patterns in other neurovascular or neurodegenerative disorders, such as cerebral small vessel disease or α-synucleinopathies. For instance, pAs abnormalities may cluster around cerebral microbleeds in small vessel disease or localize to the substantia nigra and basal ganglia in Parkinson’s disease, where α-synuclein aggregation contributes to neurovascular and autonomic dysregulation. Vascular–Glymphatic–Amyloid–Cognition Pathway Our serial mediation analysis revealed a significant indirect pathway linking diastolic blood pressure to cognitive performance via pAs-FWAD and amyloid burden in the precuneus. This supports the concept that vascular risk factors—particularly elevated blood pressure—can impair periarterial function, which in turn accelerates amyloid deposition and its downstream neurotoxic effects. Notably, pAs FWAD reductions are consistent with early-stage glymphatic dysfunction, likely driven by diminished arterial pulsatility and other flow-driving forces, as described in our stage-dependent model. Although the total and direct effects of blood pressure on cognition were non-significant, the robust indirect pathway underscores the mechanistic relevance of vascular–glymphatic–amyloid interactions. These results align with prior animal findings indicating that vascular dysfunction, particularly hypertension, contributes to impaired glymphatic clearance 41 , 42 , promoting early amyloid accumulation in regions such as the precuneus—a hub of the default mode network particularly vulnerable in AD 43,44 . Moreover, the mediation pattern suggests that interventions aimed at preserving or restoring pAs function (e.g., optimizing blood pressure control, improving sleep quality, and enhancing vascular pulsatility) may delay the onset of amyloid-related neurotoxicity and cognitive decline. Such strategies could offer a mechanistically targeted approach to risk reduction, complementing amyloid- or tau-directed therapies and potentially serving as early preventive measures in at-risk populations. Limitations and Future Directions Diffusion MRI with high b -values reflects water mobility and free-water accumulation, providing an index of long-term glymphatic integrity rather than real-time CSF flow dynamics. The pAs diffusivity metrics were derived from semi-automated segmentation; although sensitivity analyses using the largest pAs volumes were performed, the influence of segmentation variability on measurement accuracy remains to be fully clarified. Potential confounders beyond age, sex, APOE genotype, and education—such as socioeconomic status, which is closely linked to sleep quality and cardiovascular health—were not included in this analysis. Moreover, the ADNI dataset lacks detailed sleep information, limiting the assessment of how sleep modulates glymphatic function in AD progression. The spatial resolution of diffusion MRI (1.5–2 mm) in the HCP and ADNI datasets may introduce partial volume effects, potentially affecting the precision of pAs free-water volume and diffusivity estimates, particularly when the pAs diameter 45 segmented on T1- or T2-weighted MRI falls below the diffusion voxel size. This limitation was mitigated by focusing on the four largest pAs in each ROI. We observed that the mean pAs diameters for all ROIs exceeded the resolution of DWI data. Furthermore, FWVF values were significantly higher in pAs than in WM, as expected ( Table S3 ). Finally, given the cross-sectional design, causal relationships cannot be inferred. Longitudinal studies incorporating advanced imaging (e.g., contrast-enhanced MRI) and mechanistic validation are warranted to delineate the temporal dynamics of glymphatic impairment. Future investigations should also test whether targeted interventions, such as sleep optimization or vascular health management, can restore glymphatic function and mitigate downstream neurodegeneration. Conclusion This study identifies stage-dependent glymphatic alterations across aging and Alzheimer’s disease (AD) progression, revealing distinct compartmental contributions to disease evolution. In cognitively normal individuals, periarterial dysfunction was associated with vascular risk and sleep disturbances, whereas pAs and WM impairments preceded amyloid accumulation. In contrast, perivenous deficits emerged later, accompanying protein aggregation and cognitive decline. Regional mapping localized early pAs and WM alterations to regions vulnerable to cardiometabolic and sleep-related stress, underscoring their potential as sensitive, spatially specific biomarkers of preclinical disease. Moreover, the vascular–glymphatic–amyloid–cognition pathway analysis demonstrated that vascular health influences cognition indirectly through its effects on glymphatic function and amyloid burden. Collectively, these findings position the pAs as a critical locus of early vulnerability and a promising target for intervention. Strategies that improve cardiovascular health, enhance sleep quality, and restore glymphatic clearance may hold potential to delay amyloid pathology and preserve cognitive function. Methods Study participants We analyzed two independent cohorts. The first cohort included 579 cognitively normal (CN) adults aged > 36 years from the HCP-Aging Lifespan Release 2.0 11,46 . Participants were recruited at Washington University in St. Louis, the University of Minnesota, Massachusetts General Hospital, and the University of California, Los Angeles. Exclusion criteria included the use of special educational services, MRI contraindications, history of major medical illness, head injury, endocrine or psychiatric disorders, and neurodevelopmental disorders. The second dataset used in this study was obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). ADNI was launched in 2003 as a public – private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI is to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessments can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD). For this study, ADNI participants were included only if they had multi-shell DWI, Fluid Attenuated Inversion Recovery (FLAIR) MRI, T1-weighted MRI, amyloid and tau PET, and covariate data (age, sex, education, and APOE ε4 genotype). Because multi-shell DWI is available exclusively in ADNI phases 3 and 4, the sample was restricted to these phases to leverage multi-shell diffusion data for compartment-specific modeling. AD dementia diagnosis followed NINCDS–ADRDA criteria 47 . MCI was diagnosed by a comprehensive evaluation of Mini-Mental State Examination (MMSE) scores (24–30), impaired delayed recall on the Wechsler Memory Scale Logical Memory II, Clinical Dementia Rating–Sum of Boxes (CDR-SB) ≥ 0.5, preserved daily functioning, and absence of dementia. Cognitively normal (CN) controls were defined as having MMSE ≥ 24 and CDR-SB = 0 or 0.5. The final ADNI sample comprised 10 AD dementia, 68 MCI, and 95 CN participants. Both the ADNI and HCP-Aging studies were approved by institutional review boards of all participating centers, and written informed consent was obtained from all participants. MRI and PET imaging HCP-A participants were scanned on a Siemens 3T Prisma system equipped with an 80 mT/m gradient coil and a Siemens 32-channel head coil. Diffusion-weighted images (DWI) were acquired using a single-shot spin-echo EPI sequence with the following parameters: TR = 4700 ms, TE = 64 ms, multiband factor = 2, GRAPPA R = 2, 6/8 partial Fourier, 6 b 0 volumes, 45 diffusion directions with b = 1500 s/mm², and 45 with b = 3000 s/mm². A total of 90 slices were acquired with 1.5 mm thickness and no gap. The in-plane resolution was 1.5 × 1.5 mm, zero-filled to 0.75 × 0.75 mm, with a field of view (FOV) of 220 mm. T1-weighted multi-echo MPRAGE scans were also acquired with TR = 2500 ms, inversion time (TI) = 1000 ms, TE = 1.8/3.6/5.4/7.2 ms, 4 echoes per line of k-space, voxel size = 0.8 mm isotropic, FOV = 256 × 240 × 166 mm, matrix = 320 × 300 × 208, and flip angle = 8°, following the Human Connectome Project–Aging protocol 46 , 48 . In the ADNI-3 and ADNI-4 dataset, DWI imaging was performed with an echo planar imaging sequence using: 14 b 0 volumes, six diffusion directions with b = 500 s/mm², 56 with b = 1000 s/mm², and 56 with b = 2000 s/mm² (TE = 61.8–83.1 ms, TR = 9500–14,200 ms, matrix = 256 × 256, FOV = 350 × 350 mm, and slice thickness = 2.0 mm), resulting in voxel size of 2 - mm isotropic. T1-weighted images were acquired using a spoiled gradient echo sequence with the following parameters: TR = 6.96–7.67 ms, TE = 2.83–3.17 ms, matrix = 256 × 256, FOV = 260 × 260 or 270 × 270 mm, and slice thickness = 1.2 mm. Fluid-attenuated inversion recovery (FLAIR) imaging was obtained with the following parameters: TE = 147.90–153.90 ms; TR = 11,000 ms; TI = 2250 ms; matrix = 256×256; FOV = 220 × 220 mm; and slice thickness = 5 mm. All the images were obtained from the ADNI database ( http://adni.loni.usc.edu/ ). The Aβ (florbetapir [ 18 F – AV45] or florbetaben [FBB]) and tau (flortaucipir [ 18 F – AV1451]) PET data were also available for the studied subjects. Aβ and tau SUVRs were calculated using the UC Berkeley amyloid and tau PET processing pipeline. Details of the processing protocol can be found in the open-source documentation ( https://adni.loni.usc.edu/updated-uc-berkeley-amyloid-pet-methods/ ) The overall pipeline for estimation of multimodal MRI features of glymphatic function is illustrated in Supplementary Fig. S5 . Image processing and pAs segmentation, and false positive exclusion Three-dimensional T1-weighted (T1w) volumetric images were preprocessed using the Montreal Neurological Institute (MNI) CIVET pipeline, which included brain extraction, bias-field correction, and tissue classification 49 . Binary brain tissue masks were generated by CIVET. After excluding dura mater, cerebral gray matter, cerebellum, and brainstem, cerebral WM was defined for pAs segmentation. To enhance visibility, histogram equalization was applied to the T1w images, followed by non-local means denoising to reduce noise, as described in a prior study 50 . pAs were then segmented using a pre-trained nnUNet model as previously described 51 . All segmentations were visually inspected for quality by neurologists (J.H., K.W.K., J.W.S., and T.W.Y.) and trained raters (A.S.K. and J.C.), with manual corrections applied when necessary. For the HCP-A dataset, combined T1w and T2w images were used to improve visual inspection where needed. To minimize contamination of pAs segmentation by WM hyperintensities (WMHs), falsely segmented WMHs and lacunes were excluded by evaluating morphology (linear vs. round), size, clustering, and anatomical location. In the ADNI dataset, available FLAIR images, which optimally delineate WMHs, were additionally used. WMHs were segmented from multimodal T1w and FLAIR images using the U-Net with highlighted foreground technique 52 . These FLAIR-derived WMH masks were then non-linearly registered to the corresponding T1w images, and WMH voxels were removed from the pAs annotations to ensure robust segmentation. For analyses of CMS effects on glymphatic function in cognitively normal participants, pAs volumes segmented from HCP-A were mapped onto nine functional regions of interest (ROIs). These ROIs were defined on the cortical surface using the Automated Anatomical Labeling (AAL) template, following prior work 53 , and included the frontoparietal network, dorsal attention, ventral attention/language, sensorimotor, default mode network, auditory, visual, salience, and limbic network. For analyses of AD progression, ROIs corresponding to the AD-signature, including bilateral hippocampi, temporal poles, and entorhinal, inferior temporal, superior frontal, inferior parietal, and anterior cingulate cortices, were also defined using the AAL template. The precuneus, implicated in early AD pathology 43 , 44 , was separately delineated for targeted analyses. ROIs were propagated into adjacent superficial white matter using k -nearest neighbors (k-NN) clustering, constrained to a depth of 5 mm from the gray matter–WM boundary using Laplacian distance mapping ( Supplementary Fig. S6 ). A pAs volume was assigned to an ROI if it overlapped with the ROI mask. Diffusion-weighted MRI-based glymphatic metrics: FWVF, FWAD, and ALPS index calculation Multi-shell diffusion-weighted imaging (DWI) data were corrected for head motion, eddy current distortions, and EPI-related susceptibility artifacts using FSL’s EDDY 54 . Unlike conventional free-water elimination DTI (FWE-DTI) 55 , which constrains the free-water compartment to be isotropic, the Tissue–Tensor Imaging (TTI) framework 16 accounts for directional diffusion within perivascular and extracellular fluid spaces. This enables estimation of fluid-compartment diffusivity measures (FWAD, FWRD, FWMD, FWFA) alongside tissue-compartment metrics (tAD, tRD, tMD, tFA), where fAD reflects the axial diffusivity of the fluid tensor and is interpreted as an indicator of directionality in perivascular fluid movement 56 . Detailed methods are described in the Supplementary Methods . Individual b 0 images were linearly and then nonlinearly registered to each participant’s T1-weighted (T1w) space using ANTs with a mutual information similarity metric 57 . The resulting DWI-derived maps were subsequently warped to each subject’s T1w space using the deformation fields from this registration, aligning DTI-based metrics to the anatomical space where pAs annotations and the functional network ROIs and AD-related ROIs were defined ( Supplementary Fig. S5 ). This framework leveraged the anatomical role of pAs as cerebrospinal fluid (CSF)–filled conduits. Within each segmented pAs, free-water axial diffusivity (FWAD) was quantified as an index of fluid mobility, whereas free-water volume fraction (FWVF) was calculated to capture local CSF accumulation that may hinder fluid transport. FWAD and FWVF values were averaged across voxels within each pAs mask in T1w space. For complementary validation, FWFA, FWMD, and FWRD were also computed within the same pAs regions. Validation results are presented in Supplementary Fig. S3 . To assess free-water burden in the WM parenchyma, all WMH and pAs voxels were excluded from the WM mask. The mean FWVF of the remaining WM voxels within each ROI was then calculated and defined as the ROI’s WM-FWVF. Finally, interstitial fluid efflux along pVs was evaluated using the DTI-ALPS, as described in Supplementary Methods . Because many pAs detected on high-resolution T1w MRI have diameters smaller than the effective DWI voxel size, diffusion measures within pAs masks may be affected by partial volume contamination. To mitigate this, we restricted analyses to the largest pAs within each ROI, determining the optimal number ( n ) that maximized measurement stability and sensitivity to clinical variables ( Supplementary Fig. S1 ). The summed absolute t-values from correlations of FWAD and FWVF with age, PSQI, and CMS factors (adjusted for age, sex, and education) were highest when four pAs per ROI were included, indicating an optimal trade-off between sensitivity and sample retention. At this threshold, subject inclusion remained high (95% at four pAs vs. 87% at six pAs, particularly within the auditory network). Therefore, the four largest pAs per ROI were used for all subsequent analyses. Cognitive assessments At the time of HCP-A recruitment, participants were in good health, without a diagnosed history of neurologic or major psychiatric disorder, symptomatic stroke or dementia. The self-reported Pittsburgh Sleep Quality Index (PSQI) questionnaire and Montreal Cognitive Assessment (MoCA) scores were included in this study. For participants from the ADNI cohort, global cognitive function was assessed using both Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) 58 . CMS profiles and AD biomarkers The HCP-A metadata provided comprehensive CMS profiles, including height, weight, blood pressure, fasting glucose, and lipid biomarkers. BMI was calculated as weight (kg) divided by height squared (m²), based on measurements obtained during participant interviews. Systolic and diastolic blood pressures were recorded with participants seated and mean arterial pressure (MAP) was derived as: MAP = (2 × diastolic blood pressure + systolic blood pressure) / 3. Metabolic biomarkers, including high-density lipoprotein (HDL), hemoglobin A1c (HbA1c), plasma glucose, triglycerides, and total cholesterol, were quantified from fasting blood samples. The ADNI dataset provided a small number of cardiometabolic measures, including height, weight, and systolic and diastolic blood pressures, allowing BMI and MAP to be calculated using the same formulas applied in HCP-A. In addition to these cardiometabolic variables, ADNI participants were also characterized by Aβ status (positive or negative) determined by amyloid PET 59 , enabling subgroup comparisons and evaluation of glymphatic alterations in relation to AD pathology. Statistical analysis All statistical analyses were conducted using Python (v3.9) and relevant packages including statsmodels and scipy. To examine associations between glymphatic fluid metrics and either age or CMS in the HCP cohort, linear regression models were fitted for each metric separately. Independent variables included age, BMI, systolic and diastolic blood pressure, mean arterial pressure (MAP), HDL, HbA1c, triglycerides, and glucose. Age and sex were adjusted as covariates (only sex was included for correlations with age). Regression coefficients (standardized β-values) and p-values were extracted and visualized in scatterplots. False discovery rate (FDR) correction using the Benjamini-Hochberg procedure was applied across multiple comparisons within each anatomical compartment to control for type I error. Significance was defined as FDR-corrected p < 0.05, and shaded areas in plots represent 95% confidence intervals. Variables with non-significant associations were labeled as “NS”. For the ADNI dataset, group-level differences in glymphatic dysfunction were examined using analysis of covariance (ANCOVA) across diagnostic groups (CN, MCI, and AD) and across amyloid status groups (A − and A+), adjusting for age, sex, APOE genotype (ε4 vs. non-carrier), and years of education. Post hoc pairwise comparisons were conducted using Tukey’s HSD test. Violin plots depict the distribution of each metric within groups, and statistically significant differences are marked with asterisks (*p < 0.05, **p < 0.01, ***p < 0.001). For features that did not conform to a normal distribution, nonparametric rank-ANCOVA and Wilcoxon rank-sum tests were applied for post hoc comparisons. To evaluate the discriminative ability of glymphatic MRI metrics for Aβ-PET positivity and clinical diagnosis, supervised learning models were trained using a set of linear and nonlinear classifiers, including support vector machines (SVMs; linear and radial basis function kernels), extreme gradient boosting (XGBoost), and generalized linear models (GLMs). Input features were defined from three glymphatic compartments: (1) pAs features, including FWVF and FWAD within the precuneus, AD-signature, and whole brain; (2) WM features, including FWVF within the precuneus, AD-signature, and whole brain; and (3) the pVs feature, represented by the DTI-ALPS index. For molecular imaging biomarkers, Aβ-PET features included SUVRs in the precuneus, AD-signature, and whole brain, and tau-PET features included SUVRs in the same regions. All models were trained and evaluated using stratified 10-fold cross-validation (number of splits = 10, random state = 42). Classification performance was quantified using the area under the receiver operating characteristic curve (AUC). Pairwise DeLong tests were conducted to assess statistical differences in AUC values between models, and corresponding AUCs are reported in Fig. 3 . Detailed performance results for all classifiers are provided in Table S4 − S6 . Mediation analyses were performed using a standard three-path model to test indirect effects (a × b) of age or CMS variables on outcome measures (MoCA, or MMSE) through glymphatic fluid metrics and Aβ-PET SUVR as mediators. Path coefficients (β), indirect effect estimates, and bias-corrected 95% confidence intervals were reported. Significance of indirect paths was determined via bootstrap resampling (n = 5,000). All variables were normalized to ensure comparability across predictors, and the significance of indirect effects was determined based on the bootstrapped confidence intervals that indirect effects are considered significant if the CI does not include zero. Declarations Acknowledgements Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. This work was supported by the Office of the Director, National Institutes of Health under Award Numbers U01AG024904 and S10OD032285, and the Foundation of the American Society of Neuroradiology Grant Program (1960682569). Additional support was provided by the University of Southern California Generative Artificial Intelligence Grant, and the University of Southern California Department of Biomedical Engineering Mann Pilot Grant (PR0002274). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Figure 1a was created using BioRender.com. Ethical statements This study used publicly available, de-identified data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Human Connectome Project in Aging (HCP-A). For ADNI, the study was approved by the Institutional Review Boards of all participating institutions, and all participants provided written informed consent. For HCP-A, the study was approved by the Institutional Review Boards of participating institutions, including Washington University in St. Louis, the University of Minnesota, Massachusetts General Hospital, and the University of California, Los Angeles, and all participants provided written informed consent. Because both datasets are multi-site studies, Institutional Review Board approval was obtained separately at each participating institution, and a single centralized approval number is not applicable. The present study involved only secondary analysis of de-identified data and did not require additional Institutional Review Board approval.. Author Contribution Y.C., H.Z., H.K., S.C., J.H., K.W.C., N.J., P.M.T., and D.J.J.W. contributed to the conception and design of the study; Y.C., H.Z., A.S.K., Q.S., S.C., J.H., K.W.C., H.C., G.P., J.C., J.W.S., and M.K. contributed to the analysis of data; Y.C., H.K., A.S.K., Q.S., S.C., H.C., and M.K. contributed to drafting the text or preparing the figures; E.Y.J., D.C.L., H.K., H.C.C., and A.W.T. contributed to revision of the manuscript; A.W.T. and H.K. contributed to funding acquisition. Data Availability The data that support the findings of this study are available within the article and its supplementary information. Imaging data were obtained from publicly available resources, including the Alzheimer’s Disease Neuroimaging Initiative (ADNI; http://adni.loni.usc.edu/ ) and the Human Connectome Project via the NIH NIMH Data Archive (https://nda.nih.gov/ccf ). Details of ADNI SUVR PET processing are available at https://adni.loni.usc.edu/updated-uc-berkeley-amyloid-pet-methods/ . Region-of-interest (ROI) templates are available at https://github.com/Yaqiongchai/ROI-template . Binary periarterial space segmentation masks are available from the corresponding author upon reasonable request. References Nedergaard, M. & Goldman, S. A. Glymphatic failure as a final common pathway to dementia. Science 370, 50–56, doi: 10.1126/science.abb8739 (2020). Iliff, J. J. et al. A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. Sci Transl Med 4, 147ra111, doi: 10.1126/scitranslmed.3003748 (2012). Harrison, I. F. et al. Impaired glymphatic function and clearance of tau in an Alzheimer's disease model. Brain 143, 2576–2593, doi: 10.1093/brain/awaa179 (2020). Taoka, T. et al. Evaluation of glymphatic system activity with the diffusion MR technique: diffusion tensor image analysis along the perivascular space (DTI-ALPS) in Alzheimer's disease cases. Jpn J Radiol 35, 172–178, doi: 10.1007/s11604-017-0617-z (2017). Gertje, E. C., van Westen, D., Panizo, C., Mattsson-Carlgren, N. & Hansson, O. Association of Enlarged Perivascular Spaces and Measures of Small Vessel and Alzheimer Disease. Neurology 96, e193-e202, doi: 10.1212/WNL.0000000000011046 (2021). Tatekawa, H., Matsushita, S. & Miki, Y. Reply to the letter to the editor: Improved reproducibility of diffusion tensor image analysis along the perivascular space (DTI-ALPS) index calculated by manual and automated methods. Jpn J Radiol 41, 1035–1036, doi: 10.1007/s11604-023-01455-6 (2023). Hayden, M. R. A Closer Look at the Perivascular Unit in the Development of Enlarged Perivascular Spaces in Obesity, Metabolic Syndrome, and Type 2 Diabetes Mellitus. Biomedicines 12, doi: 10.3390/biomedicines12010096 (2024). Acree, L. S., Montgomery, P. S. & Gardner, A. W. The influence of obesity on arterial compliance in adult men and women. Vasc Med 12, 183–188, doi: 10.1177/1358863X07079323 (2007). Shang, Y. et al. Diffusion Tensor Imaging Analysis Along the Perivascular Space (DTI-ALPS) Demonstrates That Sleep Disorders Exacerbate Glymphatic Circulatory Impairment and Cognitive Impairment in Patients with Alzheimer's Disease. Nat Sci Sleep 16, 2205–2215, doi: 10.2147/NSS.S496607 (2024). Dumont, M. et al. Free Water in White Matter Differentiates MCI and AD From Control Subjects. Front Aging Neurosci 11, 270, doi: 10.3389/fnagi.2019.00270 (2019). Bookheimer, S. Y. et al. The Lifespan Human Connectome Project in Aging: An overview. Neuroimage 185, 335–348, doi: 10.1016/j.neuroimage.2018.10.009 (2019). Iwatsubo, T. [Alzheimer's disease Neuroimaging Initiative (ADNI)]. Nihon Rinsho 69 Suppl 8, 570–574 (2011). Wardlaw, J. M. et al. Perivascular spaces in the brain: anatomy, physiology and pathology. Nat Rev Neurol 16, 137–153, doi: 10.1038/s41582-020-0312-z (2020). Lynch, K. M., Custer, R. M., Sepehrband, F., Toga, A. W. & Choupan, J. Perivascular and parenchymal fluid characteristics are related to age and cognitive performance across the lifespan. Imaging Neurosci (Camb) 3, doi: 10.1162/imag_a_00549 (2025). Pasternak, O., Sochen, N., Gur, Y., Intrator, N. & Assaf, Y. Free water elimination and mapping from diffusion MRI. Magn Reson Med 62, 717–730, doi: 10.1002/mrm.22055 (2009). Sepehrband, F. et al. Perivascular space fluid contributes to diffusion tensor imaging changes in white matter. Neuroimage 197, 243–254, doi: 10.1016/j.neuroimage.2019.04.070 (2019). Mestre, H. et al. Flow of cerebrospinal fluid is driven by arterial pulsations and is reduced in hypertension. Nature Communications 9, 4878, doi: 10.1038/s41467-018-07318-3 (2018). Mitchell, G. F. et al. Arterial stiffness, pressure and flow pulsatility and brain structure and function: the Age, Gene/Environment Susceptibility – Reykjavik Study. Brain 134, 3398–3407, doi: 10.1093/brain/awr253 (2011). Guo, F. et al. Assessing cerebral microvascular volumetric with high-resolution 4D cerebral blood volume MRI at 7 T. Nat Cardiovasc Res 4, 1424–1438, doi: 10.1038/s44161-025-00722-1 (2025). Macauley, S. L. et al. Hyperglycemia modulates extracellular amyloid-β concentrations and neuronal activity in vivo. J Clin Invest 125, 2463–2467, doi: 10.1172/jci79742 (2015). P, X. et al. The role of LRP1 in Aβ efflux transport across the blood-brain barrier and cognitive dysfunction in diabetes mellitus. Neurochem Int 160, 105417, doi: 10.1016/j.neuint.2022.105417 (2022). Jiang, Q. et al. Impairment of the glymphatic system after diabetes. J Cereb Blood Flow Metab 37, 1326–1337, doi: 10.1177/0271678x16654702 (2017). Yen, C. H. et al. Angiotensin Receptor Blockades Effect on Peripheral Muscular and Central Aortic Arterial Stiffness: A Meta-Analysis of Randomized Controlled Trials and Systematic Review. Acta Cardiol Sin 30, 98–107 (2014). Strandberg, T. E. & Tienari, P. J. Contributions of vascular and Alzheimer's disease pathology to dementia. Alzheimers Dement 15, 1004–1005, doi: 10.1016/j.jalz.2019.04.004 (2019). Xie, L. et al. Sleep drives metabolite clearance from the adult brain. Science 342, 373–377, doi: 10.1126/science.1241224 (2013). Ma, J. et al. Effects of sleep on the glymphatic functioning and multimodal human brain network affecting memory in older adults. Mol Psychiatry , doi: 10.1038/s41380-024-02778-0 (2024). Baril, A.-A. et al. Lighter sleep is associated with higher enlarged perivascular spaces burden in middle-aged and elderly individuals. Sleep Medicine 100, 558–564, doi: https://doi.org/10.1016/j.sleep.2022.10.006 (2022). Nadjar, A., Wigren, H. M. & Tremblay, M. E. Roles of Microglial Phagocytosis and Inflammatory Mediators in the Pathophysiology of Sleep Disorders. Front Cell Neurosci 11, 250, doi: 10.3389/fncel.2017.00250 (2017). Bickford, P. C., Flowers, A. & Grimmig, B. Aging leads to altered microglial function that reduces brain resiliency increasing vulnerability to neurodegenerative diseases. Exp Gerontol 94, 4–8, doi: 10.1016/j.exger.2017.01.027 (2017). Cai, Y. et al. The relationship between inflammation, impaired glymphatic system, and neurodegenerative disorders: A vicious cycle. Neurobiology of Disease 192, 106426, doi: https://doi.org/10.1016/j.nbd.2024.106426 (2024). Smets, N. G., van der Panne, S. A., Strijkers, G. J. & Bakker, E. Perivascular spaces around arteries exceed perivenous spaces in the mouse brain. Sci Rep 14, 17655, doi: 10.1038/s41598-024-67885-y (2024). Albargothy, N. J. et al. Convective influx/glymphatic system: tracers injected into the CSF enter and leave the brain along separate periarterial basement membrane pathways. Acta Neuropathol 136, 139–152, doi: 10.1007/s00401-018-1862-7 (2018). Kress, B. T. et al. Impairment of paravascular clearance pathways in the aging brain. Ann Neurol 76, 845–861, doi: 10.1002/ana.24271 (2014). Xie, L. et al. Higher intracranial arterial pulsatility is associated with presumed imaging markers of the glymphatic system: An explorative study. NeuroImage 288, 120524, doi: https://doi.org/10.1016/j.neuroimage.2024.120524 (2024). Xin, S. H., Tan, L., Cao, X., Yu, J. T. & Tan, L. Clearance of Amyloid Beta and Tau in Alzheimer's Disease: from Mechanisms to Therapy. Neurotox Res 34, 733–748, doi: 10.1007/s12640-018-9895-1 (2018). Huang, S. Y. et al. Glymphatic system dysfunction predicts amyloid deposition, neurodegeneration, and clinical progression in Alzheimer's disease. Alzheimers Dement 20, 3251–3269, doi: 10.1002/alz.13789 (2024). Kamagata, K. et al. Association of MRI Indices of Glymphatic System With Amyloid Deposition and Cognition in Mild Cognitive Impairment and Alzheimer Disease. Neurology 99, e2648-e2660, doi: 10.1212/wnl.0000000000201300 (2022). Beschorner, N. & Nedergaard, M. Glymphatic system dysfunction in neurodegenerative diseases. Curr Opin Neurol 37, 182–188, doi: 10.1097/wco.0000000000001252 (2024). Lee, H. G., Lee, J. H., Flausino, L. E. & Quintana, F. J. Neuroinflammation: An astrocyte perspective. Sci Transl Med 15, eadi7828, doi: 10.1126/scitranslmed.adi7828 (2023). Wang, S. et al. Ion Channel Dysfunction in Astrocytes in Neurodegenerative Diseases. Front Physiol 13, 814285, doi: 10.3389/fphys.2022.814285 (2022). Li, M. et al. Impaired Glymphatic Function and Pulsation Alterations in a Mouse Model of Vascular Cognitive Impairment. Front Aging Neurosci 13, 788519, doi: 10.3389/fnagi.2021.788519 (2021). Xia, Y. et al. The glymphatic system was impaired in spontaneously hypertensive rats. Scientific Reports 15, 18321, doi: 10.1038/s41598-025-02054-3 (2025). Villeneuve, S. et al. Existing Pittsburgh Compound-B positron emission tomography thresholds are too high: statistical and pathological evaluation. Brain 138, 2020–2033, doi: 10.1093/brain/awv112 (2015). Palmqvist, S. et al. Earliest accumulation of β-amyloid occurs within the default-mode network and concurrently affects brain connectivity. Nat Commun 8, 1214, doi: 10.1038/s41467-017-01150-x (2017). Wang, X. et al. Skeleton-based cerebrovascular quantitative analysis. BMC Med Imaging 16, 68, doi: 10.1186/s12880-016-0170-8 (2016). Glasser, M. F. et al. The minimal preprocessing pipelines for the Human Connectome Project. Neuroimage 80, 105–124, doi: 10.1016/j.neuroimage.2013.04.127 (2013). McKhann, G. et al. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 34, 939–944, doi: 10.1212/wnl.34.7.939 (1984). 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, doi: 10.1016/j.neuroimage.2018.09.060 (2018). Lepage, C. et al. CIVET-Macaque: An automated pipeline for MRI-based cortical surface generation and cortical thickness in macaques. Neuroimage 227, 117622, doi: 10.1016/j.neuroimage.2020.117622 (2021). Barisano, G. et al. Imaging perivascular space structure and function using brain MRI. Neuroimage 257, 119329, doi: 10.1016/j.neuroimage.2022.119329 (2022). Chai, Y., Hedong Zhang, Gilsoon Park, Erika Lopez, Cong Zang, Jongmok Ha, Omar Elhawary, and Hosung Kim. in In Medical Imaging with Deep Learning (Nashvile, TN, USA, 2023). Park, G., Hong, J., Duffy, B. A., Lee, J. M. & Kim, H. White matter hyperintensities segmentation using the ensemble U-Net with multi-scale highlighting foregrounds. Neuroimage 237, 118140, doi: 10.1016/j.neuroimage.2021.118140 (2021). Haddad, E. et al. P-Wave Duration Is Associated With Aging Patterns in Structural Brain Networks. J Am Heart Assoc 13, e035881, doi: 10.1161/jaha.124.035881 (2024). Zhang, Q., Coolen, B. F., Versluis, M. J., Strijkers, G. J. & Nederveen, A. J. Diffusion-prepared stimulated-echo turbo spin echo (DPsti-TSE): An eddy current-insensitive sequence for three-dimensional high-resolution and undistorted diffusion-weighted imaging. NMR Biomed 30, doi: 10.1002/nbm.3719 (2017). Hoy, A. R., Koay, C. G., Kecskemeti, S. R. & Alexander, A. L. Optimization of a free water elimination two-compartment model for diffusion tensor imaging. Neuroimage 103, 323–333, doi: 10.1016/j.neuroimage.2014.09.053 (2014). Cabeen, R. P., David H. Laidlaw, and Arthur W. Toga. in ISMRM-ESMRMB Abstracts (2018). Avants, B. B. et al. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage 54, 2033–2044, doi: 10.1016/j.neuroimage.2010.09.025 (2011). Smith, T., Gildeh, N. & Holmes, C. The Montreal Cognitive Assessment: validity and utility in a memory clinic setting. The Canadian Journal of Psychiatry 52, 329–332 (2007). Josephs, K. A., Weigand, S. D. & Whitwell, J. L. Characterizing Amyloid-Positive Individuals With Normal Tau PET Levels After 5 Years: An ADNI Study. Neurology 98, e2282-e2292, doi: 10.1212/wnl.0000000000200287 (2022). Additional Declarations No competing interests reported. Supplementary Files Supplementarynpjsubmit.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 18 May, 2026 Reviewers agreed at journal 08 May, 2026 Reviewers invited by journal 23 Apr, 2026 Editor assigned by journal 21 Apr, 2026 Submission checks completed at journal 16 Apr, 2026 First submitted to journal 10 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9374985","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":633613193,"identity":"f5a068be-9f59-42f3-9c95-7f49d7a2d8ec","order_by":0,"name":"Yaqiong Chai","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Yaqiong","middleName":"","lastName":"Chai","suffix":""},{"id":633613194,"identity":"ad657d4b-8c58-4c73-8057-de1f9db3701d","order_by":1,"name":"Hedong Zhang","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Hedong","middleName":"","lastName":"Zhang","suffix":""},{"id":633613195,"identity":"cf89fdff-7eb2-49fd-a7a4-7ab0057e91dc","order_by":2,"name":"Andrew S. Kim","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"S.","lastName":"Kim","suffix":""},{"id":633613196,"identity":"02bfa825-b009-48a0-9d0f-3b2eb6bdd1a9","order_by":3,"name":"Quanhao Sun","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Quanhao","middleName":"","lastName":"Sun","suffix":""},{"id":633613197,"identity":"7da8a207-ffea-46ed-b628-915e582a8b8b","order_by":4,"name":"Shuting Chen","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Shuting","middleName":"","lastName":"Chen","suffix":""},{"id":633613198,"identity":"8ba79a76-f783-4288-b4b2-4d2f555726b2","order_by":5,"name":"Jongmok Ha","email":"","orcid":"","institution":"Samsung Medical Center, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jongmok","middleName":"","lastName":"Ha","suffix":""},{"id":633613199,"identity":"16834d03-b258-40b7-b633-1ce73d878824","order_by":6,"name":"Kyung Wook Kang","email":"","orcid":"","institution":"Chonnam National University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Kyung","middleName":"Wook","lastName":"Kang","suffix":""},{"id":633613200,"identity":"dace79ec-5010-4032-9f57-efb92ad45bdb","order_by":7,"name":"Harmony Cen","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Harmony","middleName":"","lastName":"Cen","suffix":""},{"id":633613201,"identity":"495ef503-52bd-4cc0-bb43-f08859291e16","order_by":8,"name":"Jung Won Shin","email":"","orcid":"","institution":"Samsung Medical Center, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jung","middleName":"Won","lastName":"Shin","suffix":""},{"id":633613202,"identity":"f7e5bca1-8255-4c8b-8b15-32a4a8f1486a","order_by":9,"name":"Gilsoon Park","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Gilsoon","middleName":"","lastName":"Park","suffix":""},{"id":633613203,"identity":"c9cdcaaf-4e1a-44a5-bc98-8e802a01a260","order_by":10,"name":"Jiyeah Choi","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Jiyeah","middleName":"","lastName":"Choi","suffix":""},{"id":633613204,"identity":"82fb0452-c523-4f72-b68e-6b9bdf6601bf","order_by":11,"name":"Minji Kim","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Minji","middleName":"","lastName":"Kim","suffix":""},{"id":633613205,"identity":"34997b5d-f87f-4a1c-b8d9-cac79c8dee85","order_by":12,"name":"Neda Jahanshad","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Neda","middleName":"","lastName":"Jahanshad","suffix":""},{"id":633613206,"identity":"01a895d6-5cd7-465e-83c9-0de80d52259b","order_by":13,"name":"Paul M. Thompson","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"M.","lastName":"Thompson","suffix":""},{"id":633613207,"identity":"40211e13-26bc-4ee9-86ac-113832c7e2fe","order_by":14,"name":"Helena C. Chui","email":"","orcid":"","institution":"University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Helena","middleName":"C.","lastName":"Chui","suffix":""},{"id":633613208,"identity":"3be5981e-a72f-476f-8839-67d7fe68efa4","order_by":15,"name":"Danny JJ Wang","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Danny","middleName":"JJ","lastName":"Wang","suffix":""},{"id":633613209,"identity":"151d40b8-c591-42d8-b0d9-296124f97dcd","order_by":16,"name":"Diane C. Lim","email":"","orcid":"","institution":"University of Miami","correspondingAuthor":false,"prefix":"","firstName":"Diane","middleName":"C.","lastName":"Lim","suffix":""},{"id":633613210,"identity":"ef1ec593-bf0c-46a8-bf0e-a61435c505a9","order_by":17,"name":"Eun Yeon Joo","email":"","orcid":"","institution":"Samsung Medical Center, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Eun","middleName":"Yeon","lastName":"Joo","suffix":""},{"id":633613211,"identity":"744ae2d8-e211-488c-b679-b22837da438c","order_by":18,"name":"Arthur W. Toga","email":"","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Arthur","middleName":"W.","lastName":"Toga","suffix":""},{"id":633613212,"identity":"cbbc38e5-13e8-43a1-88ff-8e7978a5385d","order_by":19,"name":"Hosung Kim","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAu0lEQVRIiWNgGAWjYBACAwiVwMMPYTCToEWygVQtDAYHiNVizt7+TLrgV5qM8Y30hx8YKqwTGwhpsew5YyY9sy+Hx+xGjrEEw5l0wloMbuSwSfP2VIC0sDEwth0mRkv6M7AW4xnpzxgY/xGlJcFMmudHDo+BRIIZA2MDMVrOnDG25m1I45E488ZYIuFYujFhLcfbH97m+ZNsz98ODLEPNdayBLWAAWMblJFAlHIw+EO80lEwCkbBKBiBAAAXiTuTcsiu/QAAAABJRU5ErkJggg==","orcid":"","institution":"Keck School of Medicine of University of Southern California","correspondingAuthor":true,"prefix":"","firstName":"Hosung","middleName":"","lastName":"Kim","suffix":""}],"badges":[],"createdAt":"2026-04-10 05:53:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9374985/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9374985/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108736283,"identity":"a0c74af9-f1be-4024-ac08-23c04a21e739","added_by":"auto","created_at":"2026-05-07 20:14:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3531926,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual framework and whole-brain associations of glymphatic MRI metrics. a. In healthy individuals, cerebrospinal fluid (CSF) flows along periarterial spaces (pAs), facilitating perivascular CSF influx into the white matter (WM) parenchyma. This enables efficient interstitial solute clearance through perivenous spaces (pVs) toward lymphatic vessels. Low free water volume fraction (FWVF) in both pAs and WM, along with preserved axial diffusivity (FWAD) and higher ALPS index, reflects intact glymphatic flow. b. In individuals with vascular or metabolic risk, arterial stiffening impairs periarterial CSF influx, reflected by reduced FWAD and increased FWVF in pAs. Elevated FWVF in WM impedes interstitial fluid (ISF) flow and CSF-ISF exchange. The diminished ALPS index indicates reduced ISF efflux in pVs. c. We analyzed whole brain fluid diffusion metrics for healthy subjects in HCP-Aging cohort. Whole brain pAs FWAD correlates with cardiometabolic syndrome (CMS) risk factors, and cognitive measure (MoCA). pAs- and WM-FWVF correlated with age and cardiometabolic syndrome (CMS) risk factors. ALPS correlates only with age. d. Regional correlation of pAs-FWAD and pAs- and WM-FWVF with age, CMS, and MoCA. All analyses were adjusted for age and sex, except the correlation with age (adjusted for sex only).\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-9374985/v1/01810dfd8285979b90dd47f7.png"},{"id":108807640,"identity":"1eb8afda-07b6-4eca-9381-fbe694110b75","added_by":"auto","created_at":"2026-05-08 15:31:00","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":316437,"visible":true,"origin":"","legend":"\u003cp\u003eGroup differences in glymphatic MRI measures across periarterial, white matter, and perivenous compartments. Violin plots show group-wise distributions of the measures: a. pAs-FWAD and FWVF within the precuneus and AD-signature regions; b. WM-FWVF within the precuneus and AD-signature regions; c. DTI-ALPS index in the perivenous space. Significant group differences were assessed using ANCOVA followed by Tukey’s HSD post hoc tests. Asterisks denote statistical significance (*p \u0026lt; 0.05, **p \u0026lt; 0.01, ***p \u0026lt; 0.001). All analyses were adjusted for adjusting for age, sex, education, and APOE carrier status.\u003c/p\u003e\n\u003cp\u003eAbbreviations: pAs: periarterial spaces; WM: white matter; FWVF = free water volume fraction; FWAD = free water axial diffusivity; DTI-ALPS = DTI along perivascular space.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9374985/v1/6fd09252a4eb2cb3a006f4f0.png"},{"id":108736285,"identity":"a955bbf0-2322-4882-8aa8-949d162e8ab1","added_by":"auto","created_at":"2026-05-07 20:14:54","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":231230,"visible":true,"origin":"","legend":"\u003cp\u003ePredictive performance of glymphatic MRI metrics for Aβ positivity and diagnostic classification. Ten-fold stratified cross-validation was used for all models. a. Aβ-PET positivity classification. Receiver operating characteristic (ROC) curves and area-under-the-curve (AUC) values show the performance of glymphatic metrics (pAs, pVs, WM, pAs+WM, and combined features) to predict Aβ-PET positivity across all participants (CN + MCI + AD) and within the non-demented subgroup. Integrated models achieved the highest performance (AUC = 0.817–0.823), comparable to tau PET prediction (AUC = 0.796–0.812). b. Diagnostic classification. ROC curves and AUCs demonstrating performance of glymphatic metrics for distinguishing CN from MCI+AD and CN from MCI, compared with Aβ, tau, and combined PET-based models. The integrated glymphatic model outperformed Aβ or tau PET individually (AUC = 0.812–0.821 vs. 0.688–0.745).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9374985/v1/1f2f079684c88ce5038607d0.png"},{"id":108736286,"identity":"93c07e51-4e0d-430d-ac14-f7946fb717f9","added_by":"auto","created_at":"2026-05-07 20:14:54","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":79033,"visible":true,"origin":"","legend":"\u003cp\u003eMediation model of the effect of diastolic blood pressure (BP) on cognition (MoCA score) via precuneus periarterial microstructure and amyloid β burden (Aβ SUVR). Standardized path coefficients (β) are shown next to each arrow; * indicates p \u0026lt; 0.05. Diastolic BP positively predicted free water axial diffusivity (FWAD) in the precuneus (a₁ = 0.00002, p \u0026lt; 0.05), which in turn was inversely associated with Aβ SUVR in the precuneus (a₂ = –64.01, p \u0026lt; 0.05). Higher Aβ SUVR was linked to poorer cognition (b = –2.733, p \u0026lt; 0.05). The total effect of BP on MoCA (c = –0.031, p = ns) and the direct effect controlling for both mediators (c′ = –0.039, p = ns) are shown. The indirect (mediated) effect, calculated as a₁ × a₂ × b, was significant (0.0032*, 95% CI [0.0002, 0.0080]).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9374985/v1/02dc05aaaca87d2a0c26c599.png"},{"id":108809988,"identity":"daea93eb-21a7-4d86-b4e9-1bcae6d148ee","added_by":"auto","created_at":"2026-05-08 15:56:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5010465,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9374985/v1/b342c0a0-3249-4858-8147-62f425b85f41.pdf"},{"id":108736287,"identity":"d683a70a-d9c6-4aea-abcb-7128093ea82d","added_by":"auto","created_at":"2026-05-07 20:14:54","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":3527968,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarynpjsubmit.docx","url":"https://assets-eu.researchsquare.com/files/rs-9374985/v1/09fb97b2ecbb3cb4044952e5.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multimodal Free-Water Imaging Links Cardiometabolic Risk to Periarterial Dysfunction and Amyloid Accumulation in Early Alzheimer’s","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAlzheimer\u0026rsquo;s disease (AD) is defined by progressive impairment of memory and cognitive function and by the pathological accumulation of amyloid-β (Aβ) plaques and tau neurofibrillary tangles, hallmarks central to disease pathogenesis. Emerging evidence indicates that impaired clearance of Aβ via the glymphatic system significantly contributes to AD onset and progression\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Insufficient removal of tau aggregates may further accelerate cognitive decline\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe glymphatic system supports brain waste clearance through three principal phases: cerebrospinal fluid (CSF) influx into periarterial spaces (pAs), bulk solute exchange with interstitial fluid (ISF) in brain parenchyma, and efflux via perivenous spaces (pVs) toward lymphatic drainage. Yet, the relative vulnerability and distinct contributions of these three components to protein clearance impairments in AD, especially in humans, remain poorly understood.\u003c/p\u003e \u003cp\u003eCurrent \u003cem\u003ein vivo\u003c/em\u003e human imaging methods of the glymphatic system predominantly rely on diffusion tensor imaging along the perivascular space (DTI-ALPS)\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, which quantifies water diffusivity within pVs near paraventricular white matter. While the ALPS index is significantly reduced in AD, its sensitivity to glymphatic dysfunction in early AD stages is limited\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Studies in individuals with mild cognitive impairment (MCI) or subjective cognitive decline have yielded inconsistent results, partly due to the ALPS index\u0026rsquo;s regional specificity. It primarily captures changes near the lateral ventricles and may miss early alterations in regions more vulnerable to initial amyloid pathology, such as the precuneus, or tau pathology, such as the entorhinal cortex\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAdditionally, emerging hypotheses propose distinct underlying mechanisms of glymphatic dysfunction across disease stages. For instance, early-stage impairment may stem from reduced periarterial CSF influx due to arterial stiffening and diminished pulsatility, resulting in pressure reversal, impaired solute clearance, and cerebral amyloid angiopathy. In contrast, disruptions detected by ALPS in the perivenous system may reflect downstream consequences of extensive parenchymal Aβ accumulation in later disease stages.\u003c/p\u003e \u003cp\u003eMoreover, the presence of other comorbidities may additively impair the glymphatic system (e.g., hypertensive arteriolosclerosis may decrease pulsatility in pAS, while obstructive sleep apnea or traumatic bleeding into the subarachnoid space may reduce CSF absorption in the pVs). For instance, pAs function, reliant on arterial compliance, may be more vulnerable to vascular risk factors such as hypertension, hyperlipidemia, diabetes and obesity\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Conversely, pVs dynamics are influenced by respiratory mechanics and may be adversely affected by sleep-disordered breathing, which commonly precedes cognitive impairment and may exacerbate glymphatic dysfunction\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWhile increased free water volume fraction (FWVF) in white matter (WM) has been proposed as a potential imaging correlate of glymphatic dysfunction\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e, the field still lacks a direct, validated in vivo imaging framework that can evaluate water diffusivity across all three compartments of the glymphatic pathway. This creates a critical knowledge gap regarding glymphatic dysfunction mechanisms across various AD stages, age groups, and comorbidities.\u003c/p\u003e \u003cp\u003eTo address this gap, we leveraged two complementary imaging cohorts: the Human Connectome Project in Aging (HCP-A) to examine glymphatic vulnerability to cardiometabolic risk factors and sleep in cognitively normal older adults\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, and the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI) to characterize glymphatic alterations associated with Aβ and tau pathology, clinical disease severity, and cognitive impairment\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. This dual-cohort approach allowed us to evaluate both early vulnerability and AD-specific dysfunction in a stage-sensitive manner.\u003c/p\u003e \u003cp\u003eIn this study, we set out to quantify glymphatic clearance across the three core compartments using multimodal MRI. Specifically, we aimed to: (1) identify glymphatic changes linked to aging, sleep, and cardiometabolic syndrome (CMS) in cognitively normal (CN) adults; (2) determine glymphatic components most strongly associated with Aβ and tau pathology; (3) characterize glymphatic disruption across CN, mild cognitive impairment (MCI), and AD cohorts; (4) evaluate the predictive value of glymphatic markers in classifying MCI or MCI/AD from CN individuals; and (5) map regional patterns of glymphatic dysfunction linked to aging and CMS, and AD progression.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy participants\u003c/h2\u003e \u003cp\u003eFor the analysis of free-water diffusivity using multi-shell DWI and bi-tensor modeling, we analyzed 579 cognitively normal subjects (age range: 36\u0026ndash;100 years; 58.6 \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\pm\\:\\)\u003c/span\u003e\u003c/span\u003e 14.1 years; 312 females) from the HCP-A Release 2.0 and 10 AD dementia patients, 68 MCI patients, and 95 CN controls from the ADNI-3 and ADNI-4 datasets (age range: 36\u0026ndash;100 years; 76.5 \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\pm\\:\\)\u003c/span\u003e\u003c/span\u003e 7.3 years; 78 females).\u003c/p\u003e \u003cp\u003eInitially, 687 subjects in HCP-A had MRI data available. However, 63 participants were excluded due to poor data quality following quality assurance protocols; 48 were excluded due to lack of the CMS and sleep data; 28 were excluded due to segmentation errors; and 2 were excluded due to anatomical abnormalities, resulting in 546 subjects from the HCP-A cohort in the present analysis. The demographic and clinical characteristics of the HCP-A and ADNI samples are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, respectively.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics of HCP-Aging participants\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN\u0026thinsp;=\u0026thinsp;546\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e58.61 (14.08)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWomen, n(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e312 (57.14)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e27.10 (4.84)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSystolic (mmHg)\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e129.05 (16.42)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiastolic (mmHg)\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e79.29 (10.45)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean arterial pressure (mmHg)\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e95.00 (14.34)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHigh density lipoprotein (mg/dL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e60.05 (16.47)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHemoglobin A1c (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e5.36 (0.53)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal cholesterol (mg/dL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e198.69 (36.41)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTriglycerides (mg/dL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e108.41 (68.34)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFasting-glucose (mg/dL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e101.91 (17.94)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePSQI\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4.68 (2.69)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMoCA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e26.40 (2.55)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"2\"\u003eData are presented as mean (s.d.) or number (percentage). \u003csup\u003ea\u003c/sup\u003eBMI was missing for two subjects. \u003csup\u003eb\u003c/sup\u003eBlood pressure was missing for five subjects. \u003csup\u003ec\u003c/sup\u003ePSQI was missing for one subject.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics of ADNI participants\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAge, years\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAll (n\u0026thinsp;=\u0026thinsp;173)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCN (n\u0026thinsp;=\u0026thinsp;95)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMCI (n\u0026thinsp;=\u0026thinsp;68)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAD (n\u0026thinsp;=\u0026thinsp;10)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e76.49 (7.33)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e76.40 (7.48)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e76.44 (7.00)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e77.70 (8.60)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWomen (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e78 (45.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45 (47.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28 (41.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5 (50.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAPOE-ε4 carriers, n (%)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e54 (42.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24 (36.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e24 (43.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6 (85.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEducation, years\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16.73(2.31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.99 (2.33)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16.35 (2.65)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e16.90 (2.01)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMMSE\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27.90 (2.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e28.81 (1.32)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e27.26 (2.89)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e23.50 (4.88)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMoCA\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.45 (4.15)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26.06 (2.45)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22.75 (4.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e17.86 (3.13)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAβ-PET measures\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;158\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAmyloid positive (n, %)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.45 (0.50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.33 (0.47)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.58 (0.50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.00 (0.00)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAβ SUVR (whole brain)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.20 (0.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.12 (0.23)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.27 (0.28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.49 (0.34)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAβ SUVR (precuneus)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.29 (0.32)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.21 (0.28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.37 (0.32)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.59 (0.37)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAβ SUVR (AD-Signature)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.16 (0.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.08 (0.23)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.24 (0.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.45 (0.36)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTau-PET measures\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;160\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTau SUVR (whole brain)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.19 (0.23)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.13 (0.12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.23 (0.29)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.48 (0.38)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTau SUVR (precuneus)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.19 (0.23)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.13 (0.12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.23 (0.29)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.48 (0.38)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTau SUVR (AD-Signature)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.32 (0.32)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.21 (0.13)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.40 (0.37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.96 (0.49)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI\u003csup\u003ee\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28.23 (11.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30.02 (15.89)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e26.22 (3.88)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e26.69 (6.97)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSystolic Blood Pressure\u003csup\u003ef\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e136.91 (18.77)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e135.97 (17.42)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e138.03 (19.98)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e138.00 (23.67)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiastolic Blood Pressure\u003csup\u003ef\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73.53 (9.60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e73.65 (9.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e73.15 (10.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e75.00 (11.21)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eData are presented as mean (s.d.) or number (percentage). Aβ-PET positivity was determined as composite AV45 SUVR\u0026thinsp;\u0026gt;\u0026thinsp;0.1.11 or FBB SUVR\u0026thinsp;\u0026gt;\u0026thinsp;1.478. \u003csup\u003ea\u003c/sup\u003eAPOE-ε4 was missing for 30 subjects in CN, 13 in MCI, and 3 in AD. \u003csup\u003eb\u003c/sup\u003eEducation was missing for five subjects in CN and two in MCI. \u003csup\u003ec\u003c/sup\u003eMoCA was missing for two subjects in CN, seven subjects in MCI and three in AD. \u003csup\u003ed\u003c/sup\u003eMMSE was missing for two participants in MCI. \u003csup\u003ee\u003c/sup\u003eBMI was missing for two subjects in AD, 30 in CN and 17 in MCI. \u003csup\u003ef\u003c/sup\u003eBlood pressures were missing for two subjects in CN.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMultimodal MRI metrics to understand glymphatic function across the three compartments\u003c/h3\u003e\n\u003cp\u003eWe estimated the following complementary metrics: 1) pAs diffusivity, quantified using free-water axial diffusivity (FWAD) and additional diffusion parameters (radial diffusivity, mean diffusivity), along with pAs-FWVF; 2) WM-FWVF, reflecting parenchymal fluid accumulation and ISF\u0026ndash;CSF exchange efficiency; and 3) DTI-ALPS index, reflecting perivenous efflux integrity.\u003c/p\u003e \u003cp\u003eMotivated by prior evidence that MRI-visible perivascular spaces are predominantly periarteriolar in nature\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, we quantified pAs diffusivity within perivascular spaces segmented from T1-weighted MRI. Multi-shell DWI data were analyzed using tissue tensor imaging (TTI), an advanced bi-tensor approach that separates tissue and free-water components while allowing anisotropic diffusion in the free-water compartment\u003csup\u003e\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. In contrast to conventional free-water\u0026ndash;eliminated DTI, TTI enables directional characterization of fluid diffusivity along CSF-filled conduits such as the pAs (\u003cb\u003eSupplementary Methods\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e provides a conceptual illustration of glymphatic alterations in healthy versus pathological states based on our findings in the following sections (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e). In healthy brains, robust CSF mobility along the pAs (high FWAD), low FWVF in pAs and WM, and high pVs-ALPS indices signify effective glymphatic circulation. In contrast, pathological states exhibit diminished pAs FWAD (impaired influx), elevated FWVF (fluid retention and greater flow impedance), and reduced ALPS (impaired pVs-mediated efflux), consistent with progressive glymphatic failure (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003ea\u0026ndash;b).\u003c/p\u003e \u003cp\u003eTo minimize partial-volume contamination in pAs diffusivity measurements, it was essential that the selected pAs segments be sufficiently large relative to the spatial resolution of the DWI data. By selecting the four largest pAs per region of interest (ROI), we observed that the mean pAs diameters in both HCP-A and ADNI cohorts exceeded the corresponding DWI voxel sizes (\u003cb\u003eTables S1\u0026ndash;S2\u003c/b\u003e; HCP-A: pAs diameter\u0026thinsp;=\u0026thinsp;1.7\u0026ndash;2.2 mm vs. 1.5 mm isotropic DWI; ADNI: 2.7\u0026ndash;3.0 mm vs. 2.0 mm isotropic DWI). We further validated that this choice provided optimal stability and sensitivity of associations with clinical variables compared with other choices of \u003cem\u003en\u003c/em\u003e largest pAs (see \u003cb\u003eMethods\u003c/b\u003e and \u003cb\u003eSupplementary Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e), supporting the use of the four-largest-pAs threshold in subsequent analyses.\u003c/p\u003e\n\u003ch3\u003eAssociation of whole white matter glymphatic measures with aging, sleep, cognition, and CMS in HCP-A\u003c/h3\u003e\n\u003cp\u003eWhole WM glymphatic measures were examined in healthy aging cohorts after adjusting for effects of age and sex. In the periarterial compartment (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003ec; \u003cb\u003eSupplementary Fig. S2\u003c/b\u003e), FWAD exhibited significant negative correlations with multiple CMS indicators, including body mass index (BMI; β = \u0026minus;\u0026thinsp;0.36, p\u0026thinsp;=\u0026thinsp;2.8\u0026times;10⁻\u003csup\u003e17\u003c/sup\u003e), systolic blood pressure (β = \u0026minus;\u0026thinsp;0.18, p\u0026thinsp;=\u0026thinsp;1.0\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e), diastolic blood pressure (β = \u0026minus;\u0026thinsp;0.17, p\u0026thinsp;=\u0026thinsp;1.2\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e), mean arterial pressure (MAP; β = \u0026minus;\u0026thinsp;0.19, p\u0026thinsp;=\u0026thinsp;3.0\u0026times;10⁻⁵), hemoglobin A1c (HbA1c; β = \u0026minus;\u0026thinsp;0.16, p\u0026thinsp;=\u0026thinsp;5.1\u0026times;10\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;4\u003c/b\u003e\u003c/sup\u003e), and fasting glucose (β = \u0026minus;\u0026thinsp;0.15, p\u0026thinsp;=\u0026thinsp;0.001), while showing a positive correlation with high-density lipoprotein (HDL; β\u0026thinsp;=\u0026thinsp;0.17, p\u0026thinsp;=\u0026thinsp;5.1\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e). FWAD was not associated with age (p\u0026thinsp;=\u0026thinsp;0.10). The effects of fasting glucose and blood pressure on pAs-FWAD were significant (p\u0026thinsp;=\u0026thinsp;5.1\u0026times;10⁻\u003csup\u003e6\u003c/sup\u003e and 6.7\u0026times;10⁻\u003csup\u003e4\u003c/sup\u003e, respectively), but after adding antihypertensive and diabetic medication use as covariates, these correlations were no longer significant (p\u0026thinsp;=\u0026thinsp;0.236 and 0.968). On the other hand, FWVF increased with age (β\u0026thinsp;=\u0026thinsp;0.48, p\u0026thinsp;=\u0026thinsp;1.2\u0026times;10⁻\u003csup\u003e24\u003c/sup\u003e), and was positively associated with BMI (β\u0026thinsp;=\u0026thinsp;0.19, p\u0026thinsp;=\u0026thinsp;1.2\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e) and triglycerides (β\u0026thinsp;=\u0026thinsp;0.17, p\u0026thinsp;=\u0026thinsp;2.0\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e), but negatively correlated with HDL (β = \u0026minus;\u0026thinsp;0.15, p\u0026thinsp;=\u0026thinsp;1.1\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e). All p-values reported after false discovery rate correction.\u003c/p\u003e \u003cp\u003eAfter adjusting for age (but not sex), cognitive performance (MoCA) was positively correlated with FWAD (β\u0026thinsp;=\u0026thinsp;0.11, p\u0026thinsp;=\u0026thinsp;0.02). This relationship was observed in males (β\u0026thinsp;=\u0026thinsp;0.11, p\u0026thinsp;=\u0026thinsp;0.02) but not in females. No significant correlations between global pAs metrics and sleep quality (PSQI) were found.\u003c/p\u003e \u003cp\u003eComparisons with other diffusivity metrics, including free-water radial diffusivity (FWRD), mean diffusivity (FWMD), and fractional anisotropy (FWFA) revealed similar directional trends but weaker associations (\u003cb\u003eSupplementary Fig. S3\u003c/b\u003e). Fewer CMS factors were significantly correlated with these metrics, and their absolute β-values were consistently smaller than those for FWAD. Thus, FWAD was selected as the representative marker of glymphatic diffusivity for subsequent analyses.\u003c/p\u003e \u003cp\u003eIn WM parenchyma, FWVF positively correlated with age (β\u0026thinsp;=\u0026thinsp;0.55, p\u0026thinsp;=\u0026thinsp;4.9x10\u003csup\u003e-47\u003c/sup\u003e), BMI (β\u0026thinsp;=\u0026thinsp;0.17, p\u0026thinsp;=\u0026thinsp;8.7x10\u003csup\u003e-6\u003c/sup\u003e), and triglyceride levels (β\u0026thinsp;=\u0026thinsp;0.18, p\u0026thinsp;=\u0026thinsp;8.7x10\u003csup\u003e-6\u003c/sup\u003e), and negatively with HDL (β = \u0026minus;\u0026thinsp;0.11, p\u0026thinsp;=\u0026thinsp;0.01). No correlation with PSQI was found. This pattern was similar to that of FWVF in pAs.\u003c/p\u003e \u003cp\u003eFinally, the ALPS index, reflecting perivenous efflux, showed a robust negative association with age (β = \u0026minus;\u0026thinsp;0.37, p\u0026thinsp;=\u0026thinsp;3.3x10\u003csup\u003e-20\u003c/sup\u003e) but no correlation with PSQI, MoCA, or CMS factors.\u003c/p\u003e \u003cp\u003e \u003cb\u003eRegional association of glymphatic metrics with aging, sleep, cognition, and CMS in HCP-A.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAs we divided the entire WM into nine functional ROIs (see \u003cb\u003eMethods\u003c/b\u003e), region-specific analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003ed) revealed differential spatial patterns of glymphatic vulnerability in pAs and WM parenchyma to CMS factors. Note that no regional pattern is available for pVs ALPS as this index is a single global measure.\u003c/p\u003e \u003cp\u003eAcross nine ROIs, age showed widespread positive associations with FWVF in both pAs and WM parenchyma, spanning the frontoparietal (FPN), dorsal attention, ventral language, sensorimotor, auditory, visual, salience, default mode (DMN), and limbic networks (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). No regional FWAD associations with age were detected.\u003c/p\u003e \u003cp\u003eOf the CMS variables, higher BMI was associated with lower FWAD and higher FWVF in pAs across all ROIs, paralleling the WM-FWVF pattern, though effects were most pronounced in the DMN, salience, sensorimotor, dorsal attention, and visual networks. Higher diastolic blood pressure was linked to widespread lower FWAD across networks and to higher FWVF restricted within the dorsal attention network (pAs) and FPN (WM).\u003c/p\u003e \u003cp\u003eHigher triglyceride levels were associated with lower pAs-FWAD in the sensorimotor, FPN, visual, and DMN regions, and with higher FWVF across all WM networks except the limbic region. Furthermore, as in \u003cb\u003eSupplementary Fig. S4\u003c/b\u003e, lower HDL cholesterol correlated with lower FWAD and higher FWVF in both pAs and WM, particularly in frontal and temporal cortices. Elevated HbA1c and fasting glucose were also linked to widespread reductions in FWAD, primarily along periarterial pathways. Sleep quality (PSQI) showed a positive correlation with FWAD in the dorsal attention network and a negative correlation with periarterial FWVF in the visual network.\u003c/p\u003e \u003cp\u003eCognitive performance (MoCA) positively correlated with pAs-FWAD within the dorsal attention, salience, auditory, and default mode networks, whereas no significant relationships were observed for WM-FWVF or pVs-ALPS.\u003c/p\u003e\n\u003ch3\u003eGlymphatic alterations associated with Alzheimer's pathology and clinical stages in ADNI\u003c/h3\u003e\n\u003cp\u003eWe examined group differences in glymphatic metrics across CN, MCI, and AD participants from the ADNI dataset (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e), with age, sex, years of education, and APOE genotype (allele ε4 vs. others) as covariates. pAs-FWAD was significantly lower in Aβ-positive compared to Aβ-negative individuals within both the precuneus (p\u0026thinsp;=\u0026thinsp;0.038) and AD-signature regions (p\u0026thinsp;=\u0026thinsp;0.003). pAs-FWVF in these same regions was significantly higher in the Aβ-positive group than the Aβ-negative (p\u0026thinsp;=\u0026thinsp;0.008 in precuneus and 0.014 in AD-signature regions). Similarly, WM-FWVF was significantly higher in Aβ-positive participants within the precuneus (p\u0026thinsp;=\u0026thinsp;0.041) and AD-signature regions (p\u0026thinsp;=\u0026thinsp;0.009). However, the global pAs-FWAD, pAs-FWVF, and WM-FWVF showed no significant differences between Aβ-positive and Aβ-negative groups. The ALPS index showed no significant difference between Aβ groups, though a non-significant trend toward higher values in Aβ-positive individuals was observed.\u003c/p\u003e \u003cp\u003eWhen stratified by diagnostic group, pAs and WM-FWVF in the whole brain, precuneus and AD-signature regions showed significant group differences between all pairs of CN, MCI and AD groups (whole brain pAs/FWVF: p\u0026thinsp;=\u0026thinsp;0.002/0.011; precuneus: p\u0026thinsp;=\u0026thinsp;0.002/0.011; AD-signature: p\u0026thinsp;=\u0026thinsp;0.003/0.009). pAs-FWAD in the whole brain, precuneus, and AD-signature regions exhibited significant differences observed between CN and MCI (whole brain: p\u0026thinsp;=\u0026thinsp;0.006; precuneus: p\u0026thinsp;=\u0026thinsp;0.002; AD-signature: p\u0026thinsp;=\u0026thinsp;0.003). We observed significant differences in ALPS between CN and AD (p\u0026thinsp;=\u0026thinsp;0.009) and between MCI and AD (p\u0026thinsp;=\u0026thinsp;0.014) but not between CN and MCI (p\u0026thinsp;=\u0026thinsp;0.738).\u003c/p\u003e\n\u003ch3\u003eClassification of Aβ positivity and clinical diagnosis\u003c/h3\u003e\n\u003cp\u003eWe next evaluated the ability of glymphatic metrics to classify Aβ-PET positivity using pairwise DeLong tests. The integrated glymphatic model combining all three components achieved the highest discriminative performance (AUC\u0026thinsp;=\u0026thinsp;0.817 when including subjects with all stages of CN\u0026thinsp;+\u0026thinsp;MCI\u0026thinsp;+\u0026thinsp;AD, AUC\u0026thinsp;=\u0026thinsp;0.823 for those with non-demented stages of CN\u0026thinsp;+\u0026thinsp;MCI only), significantly outperforming models using WM-only features (AUC\u0026thinsp;=\u0026thinsp;0.633 for all and 0.626 for non-demented, ΔAUC\u0026thinsp;=\u0026thinsp;0.184\u0026ndash;0.197; p\u0026thinsp;=\u0026thinsp;3.2\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e) and ALPS (AUC\u0026thinsp;=\u0026thinsp;0.606 for all and 0.612 for non-demented, ΔAUC\u0026thinsp;=\u0026thinsp;0.211; p\u0026thinsp;=\u0026thinsp;1.3\u0026times;10⁻⁵). A combined pAs\u0026thinsp;+\u0026thinsp;WM model (AUC\u0026thinsp;=\u0026thinsp;0.802 for all, and 0.803 for non-demented) also performed significantly better than either WM alone (ΔAUC\u0026thinsp;=\u0026thinsp;0.169\u0026ndash;0.177; p\u0026thinsp;=\u0026thinsp;1.6\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;2\u003c/sup\u003e) or ALPS (ΔAUC\u0026thinsp;=\u0026thinsp;0.191\u0026ndash;0.196; p\u0026thinsp;=\u0026thinsp;2.7\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e). Among single-compartment models, pAs (AUC\u0026thinsp;=\u0026thinsp;0.74 in all and 0.727 in non-demented) outperformed both WM (ΔAUC\u0026thinsp;=\u0026thinsp;0.101\u0026ndash;0.107; p\u0026thinsp;=\u0026thinsp;0.031) and ALPS (ΔAUC\u0026thinsp;=\u0026thinsp;0.115\u0026ndash;0.134; p\u0026thinsp;=\u0026thinsp;0.001). Although the integrated model achieved the numerically highest AUC, its improvement over pAs\u0026thinsp;+\u0026thinsp;WM (ΔAUC\u0026thinsp;=\u0026thinsp;0.015\u0026ndash;0.02; p\u0026thinsp;\u0026gt;\u0026thinsp;0.1) or pAs alone (ΔAUC\u0026thinsp;=\u0026thinsp;0.077\u0026ndash;0.096; p\u0026thinsp;=\u0026thinsp;0.06) did not reach statistical significance. These findings indicate that pAs features provide the most robust discriminative signal, while WM integration enhances stability, and full glymphatic integration yields only marginal gains.\u003c/p\u003e \u003cp\u003eFor comparison with a PET biomarker, three combined features of the precuneus, AD-signature and whole brain tau-PET SUVR yielded AUCs of 0.812 (all participants) and 0.796 (non-demented), displaying significant difference in the discriminative performance between integrated glymphatic measures and established molecular imaging biomarkers.\u003c/p\u003e \u003cp\u003eIn differentiating clinical diagnostic groups (CN vs. MCI\u0026thinsp;+\u0026thinsp;AD and CN vs. MCI; Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e), the integrated glymphatic features again achieved the highest overall accuracy (AUC\u0026thinsp;=\u0026thinsp;0.812\u0026ndash;0.821), significantly outperforming both pVs-only (ΔAUC\u0026thinsp;=\u0026thinsp;0.200-0.208; p\u0026thinsp;=\u0026thinsp;3.1\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e) and WM-only (ΔAUC\u0026thinsp;=\u0026thinsp;0.186\u0026ndash;0.188; p\u0026thinsp;=\u0026thinsp;8.0\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e) models. Among individual components, pAs features (AUC\u0026thinsp;=\u0026thinsp;0.795\u0026ndash;0.802) exhibited markedly higher discriminative ability than pVs (ΔAUC\u0026thinsp;=\u0026thinsp;0.181\u0026ndash;0.191; p\u0026thinsp;=\u0026thinsp;3.7\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e) and WM (ΔAUC\u0026thinsp;=\u0026thinsp;0.169; p\u0026thinsp;=\u0026thinsp;0.04). Adding WM to pAs modestly improved performance (AUC\u0026thinsp;=\u0026thinsp;0.787\u0026ndash;0.789) but achieved a marginal significance (ΔAUC\u0026thinsp;=\u0026thinsp;0.008\u0026ndash;0.013; p\u0026thinsp;=\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eFor comparison with AD pathological biomarkers, combined Aβ- and tau-PET SUVR measures achieved lower AUCs (CN vs. MCI\u0026thinsp;+\u0026thinsp;AD: Aβ\u0026thinsp;=\u0026thinsp;0.669, tau\u0026thinsp;=\u0026thinsp;0.743; CN vs. MCI: Aβ\u0026thinsp;=\u0026thinsp;0.648, tau\u0026thinsp;=\u0026thinsp;0.698) than the integrated glymphatic features (CN vs. MCI\u0026thinsp;+\u0026thinsp;AD: ΔAUC\u0026thinsp;=\u0026thinsp;0.076; p\u0026thinsp;=\u0026thinsp;0.031; CN vs. MCI: ΔAUC\u0026thinsp;=\u0026thinsp;0.124; p\u0026thinsp;=\u0026thinsp;0.009). Combining tau with Aβ did not substantially improve performance compared to tau only.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eMediation analysis: blood pressure, glymphatic function, amyloid burden, and cognition\u003c/h2\u003e \u003cp\u003eWe conducted a serial mediation analysis to examine whether the relationship between elevated blood pressure, a key factor that dysregulates arterial pulsatility, and cognitive decline (MoCA) was mediated by glymphatic dysfunction and amyloid burden (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Diastolic blood pressure (BP) was positively associated with periarterial FWAD in the precuneus (a₁ = 0.00002, p\u0026thinsp;=\u0026thinsp;0.03), which, in turn, was negatively associated with Aβ SUVR (a₂ = \u0026minus;\u0026thinsp;64.01, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Higher Aβ SUVR was associated with poorer cognitive performance (b = \u0026minus;\u0026thinsp;2.733, p\u0026thinsp;=\u0026thinsp;0.02). The total effect of diastolic BP on MoCA was not significant (c = \u0026minus;\u0026thinsp;0.031, p\u0026thinsp;=\u0026thinsp;0.39), and the direct effect after accounting for both mediators remained non-significant (c\u0026rsquo; = \u0026minus;\u0026thinsp;0.039, p\u0026thinsp;=\u0026thinsp;0.27). However, the indirect effect through both mediators was significant (0.0032, 95% CI, 0.0002, 0.008), indicating that elevated diastolic BP contributes to cognitive decline indirectly by reducing periarterial water mobility (lower FWAD) and subsequently promoting greater Aβ accumulation on PET.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we used multi-compartment glymphatic metrics derived from a bi-tensor model using multi-shell diffusion MRI to characterize glymphatic alterations across pAs, WM parenchyma, and pVs in two complementary cohorts. Our principal findings are fourfolds: first, measures reflecting increased pAs free-water volume and reduced pAs water mobility were significantly linked to aging, CMS factors, and cognition in healthy aging adults and were also observed in Aβ-positive and cognitively impaired participants, with early changes observed in the precuneus and AD-signature regions. Second, pVs diffusivity decline was associated with aging and was most marked at later clinical AD stages, suggesting a downstream parenchymal and efflux failure following early periarterial compromise. Third, integrated glymphatic metrics combining periarterial, WM, and perivenous glymphatic features classified Aβ-PET positivity, with performance comparable to that of tau PET and clinical diagnosis, and outperformed combined tau and amyloid PET measures. Finally, a serial mediation analysis implicated a directional association linking higher vascular risk factors, pAs glymphatic dysfunction, Aβ tissue accumulation, and cognitive decline, with a statistically significant indirect effect. Together, using multi-compartment diffusion MRI in two large cohorts, our results support a stage-dependent compartmental model of glymphatic dysfunction in human aging and AD.\u003c/p\u003e\n\u003ch3\u003eImpact of Aging and Cardiovascular Risk Factors on Glymphatic Function\u003c/h3\u003e\n\u003cp\u003eAging is a well-established contributor to glymphatic system decline, and our findings from pAs-FWVF, WM-FWVF, and pVs-ALPS measurements are consistent with this framework. These metrics revealed age-related deterioration in pAs, pVs, and WM microstructural integrity, supporting the notion that advancing age compromises all three compartments of glymphatic system.\u003c/p\u003e \u003cp\u003eBeyond aging, specific CMS risk factors demonstrated distinct associations with pAs functionality, a marker closely linked to arterial pulsatility\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. In particular, reduced pAs-FWAD was strongly associated with elevated diastolic blood pressure and fasting glucose levels, suggesting that vascular stiffening and metabolic dysregulation jointly compromise the pulsatile forces driving periarterial CSF transport. Hypertension alters the hemodynamic profile of arterial pulsation in ways detrimental to perivascular flow. In animal models, two-photon microscopy has shown that CSF transport along periarterial channels is driven by rhythmic arterial wall motion, which becomes markedly attenuated under hypertensive conditions due to reduced arterial compliance despite increased pulsation amplitude\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. A 7T MRI human study similarly revealed elevated microvascular volumetric pulsatility in deep WM among hypertensive and older adults\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, consistent with the increased pulsation amplitude observed in animal models. Collectively, these findings suggest that hypertension amplifies local vascular stress yet disrupts the coordinated wall motion required for efficient glymphatic transport.\u003c/p\u003e \u003cp\u003eChronic hyperglycemia, as seen in diabetes and poorly controlled glucose levels, may exacerbate glymphatic dysfunction through mechanisms such as microangiopathy, oxidative stress, and alterations in the perivascular basement membrane composition, all of which can impede interstitial solute clearance\u003csup\u003e\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. This interpretation aligns with animal studies demonstrating that experimental diabetes leads to perivascular AQP4 mislocalization and reduced glymphatic transport efficiency\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eNotably, the associations between fasting glucose, blood pressure, and pAs-FWAD were no longer significant after adjusting for antihypertensive and antidiabetic medication use. This attenuation implies that pharmacological control of vascular and metabolic risk factors may help preserve periarterial fluid mobility. Effective management of blood pressure and glucose has been shown to improve cerebrovascular compliance and reduce arterial stiffness\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e, which may in turn stabilize the mechanical driving forces underlying glymphatic transport.\u003c/p\u003e \u003cp\u003eOther CMS-related factors, including lower HDL cholesterol, elevated BMI, and higher triglyceride levels were associated with both lower pAs-FWAD and higher pAs- and WM-FWVF, but not with pVs-ALPS. These patterns suggest a combination of microvascular compromise and extracellular space expansion, possibly reflecting chronic low-grade inflammation and tissue remodeling. The convergence of these vascular\u0026ndash;metabolic insults may not only hinder clearance but also promote amyloid accumulation and white matter damage over time\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTaken together, our findings support the concept that both aging and CMS components exert additive and potentially synergistic effects on glymphatic function, with periarterial pathways appearing especially vulnerable to hemodynamic and metabolic stressors.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eRole of Sleep Disturbances in Glymphatic Dysfunction\u003c/h2\u003e \u003cp\u003eSleep disturbances critically affect glymphatic clearance and, consequently, AD-related neurodegeneration. In our study, poor sleep quality, indicated by high PSQI scores, was associated with pAs-FWAD in the frontoparietal network and FWVF in the visual network. Animal studies have robustly demonstrated that sleep deprivation impairs glymphatic clearance\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, though human evidence remains limited. Recent work by Ma et. al. indicates that sleep quality directly influences glymphatic function and brain networks critical for visual memory in older adults\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Likewise, lighter sleep in middle-age and elderly individuals has been linked to an increased burden of enlarged perivascular spaces\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Aging exacerbates the adverse effects of poor sleep by impairing microglial and macrophage function, fostering a pro-inflammatory environment\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e,\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e, which may further impede glymphatic flow. Such chronic inflammation may lead to reactive astrogliosis resulting in AQP4 depolarization and reduced AQP4 expression, consequently impeding CSF\u0026ndash;ISF exchange\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eGlymphatic Dysfunction in AD Progression: Novel Associations between Compartments and AD Stages\u003c/h2\u003e \u003cp\u003eOur compartment-based analysis revealed stage-dependent alterations, with pAs and WM changes emerging in early phases of AD, and pVs alterations becoming more pronounced in later stages. This pattern aligns with prior experimental and modeling studies in mice, which show that the pAs plays a disproportionately greater role in solute exchange between CSF and brain parenchyma compared with the pVs, owing to its greater abundance, larger caliber, and higher net CSF inflow\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn early stages of AD, reduced pAs-FWAD, potentially influenced by aging, poor sleep, and CMS-related factors\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e, likely reflects diminished arterial pulsatility and other driving forces that propel glymphatic inflow\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Such impairments would slow the clearance of soluble waste products, including Aβ and tau, leading to their gradual accumulation in the interstitial space\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. In contrast, alterations in pVs functionality become more prominent in later stages of disease. These changes likely represent downstream consequences of chronic glymphatic failure, characterized by neuroinflammation and elevated flow resistance from Aβ and tau deposition and extracellular water retention\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSupporting this interpretation, previous studies have shown that cerebral amyloid angiopathy and Aβ accumulation in pAs can promote extracellular fluid buildup\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, possibly through reactive astrogliosis\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e and chronic inflammation\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Astrocytic dysfunction, particularly of aquaporin-4 (AQP4) channels on astrocyte endfeet, may further disrupt water and ion homeostasis, exacerbating clearance deficits\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThese stage-specific alterations provide a mechanistic explanation for the predictive patterns observed in our data. Because pAs dysfunction arises earlier, its metrics likely capture preclinical glymphatic deficits associated with initial amyloid accumulation, whereas pVs changes reflect later-stage pathology and vascular comorbidities. Integrating pAs, WM, and pVs features thus enables a comprehensive characterization of both early and late glymphatic failure. This multimodal approach achieved high accuracy in classifying individuals across the AD spectrum (CN, MCI and AD) and in predicting early Aβ accumulation\u0026mdash;comparable to tau PET performance in non-demented individuals and superior to combined amyloid and tau PET features in differentiating MCI from CN. These results highlight the promise of DTI-based glymphatic markers as a cost-effective, noninvasive tool for staging and risk stratification in AD.\u003c/p\u003e \u003cp\u003eFurthermore, our regional mapping successfully localized alterations of pAs and WM function in the AD-signature and precuneus regions, highlighting their association with early disease progression. This regional mapping approach may also help delineate disease-specific glymphatic dysfunction patterns in other neurovascular or neurodegenerative disorders, such as cerebral small vessel disease or α-synucleinopathies. For instance, pAs abnormalities may cluster around cerebral microbleeds in small vessel disease or localize to the substantia nigra and basal ganglia in Parkinson\u0026rsquo;s disease, where α-synuclein aggregation contributes to neurovascular and autonomic dysregulation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eVascular\u0026ndash;Glymphatic\u0026ndash;Amyloid\u0026ndash;Cognition Pathway\u003c/h2\u003e \u003cp\u003eOur serial mediation analysis revealed a significant indirect pathway linking diastolic blood pressure to cognitive performance via pAs-FWAD and amyloid burden in the precuneus. This supports the concept that vascular risk factors\u0026mdash;particularly elevated blood pressure\u0026mdash;can impair periarterial function, which in turn accelerates amyloid deposition and its downstream neurotoxic effects. Notably, pAs FWAD reductions are consistent with early-stage glymphatic dysfunction, likely driven by diminished arterial pulsatility and other flow-driving forces, as described in our stage-dependent model. Although the total and direct effects of blood pressure on cognition were non-significant, the robust indirect pathway underscores the mechanistic relevance of vascular\u0026ndash;glymphatic\u0026ndash;amyloid interactions.\u003c/p\u003e \u003cp\u003eThese results align with prior animal findings indicating that vascular dysfunction, particularly hypertension, contributes to impaired glymphatic clearance\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e, promoting early amyloid accumulation in regions such as the precuneus\u0026mdash;a hub of the default mode network particularly vulnerable in AD\u003csup\u003e43,44\u003c/sup\u003e. Moreover, the mediation pattern suggests that interventions aimed at preserving or restoring pAs function (e.g., optimizing blood pressure control, improving sleep quality, and enhancing vascular pulsatility) may delay the onset of amyloid-related neurotoxicity and cognitive decline. Such strategies could offer a mechanistically targeted approach to risk reduction, complementing amyloid- or tau-directed therapies and potentially serving as early preventive measures in at-risk populations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eLimitations and Future Directions\u003c/h2\u003e \u003cp\u003eDiffusion MRI with high \u003cem\u003eb\u003c/em\u003e-values reflects water mobility and free-water accumulation, providing an index of long-term glymphatic integrity rather than real-time CSF flow dynamics.\u003c/p\u003e \u003cp\u003eThe pAs diffusivity metrics were derived from semi-automated segmentation; although sensitivity analyses using the largest pAs volumes were performed, the influence of segmentation variability on measurement accuracy remains to be fully clarified.\u003c/p\u003e \u003cp\u003ePotential confounders beyond age, sex, APOE genotype, and education\u0026mdash;such as socioeconomic status, which is closely linked to sleep quality and cardiovascular health\u0026mdash;were not included in this analysis. Moreover, the ADNI dataset lacks detailed sleep information, limiting the assessment of how sleep modulates glymphatic function in AD progression.\u003c/p\u003e \u003cp\u003eThe spatial resolution of diffusion MRI (1.5\u0026ndash;2 mm) in the HCP and ADNI datasets may introduce partial volume effects, potentially affecting the precision of pAs free-water volume and diffusivity estimates, particularly when the pAs diameter\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e segmented on T1- or T2-weighted MRI falls below the diffusion voxel size. This limitation was mitigated by focusing on the four largest pAs in each ROI. We observed that the mean pAs diameters for all ROIs exceeded the resolution of DWI data. Furthermore, FWVF values were significantly higher in pAs than in WM, as expected (\u003cb\u003eTable S3\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eFinally, given the cross-sectional design, causal relationships cannot be inferred. Longitudinal studies incorporating advanced imaging (e.g., contrast-enhanced MRI) and mechanistic validation are warranted to delineate the temporal dynamics of glymphatic impairment. Future investigations should also test whether targeted interventions, such as sleep optimization or vascular health management, can restore glymphatic function and mitigate downstream neurodegeneration.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study identifies stage-dependent glymphatic alterations across aging and Alzheimer\u0026rsquo;s disease (AD) progression, revealing distinct compartmental contributions to disease evolution. In cognitively normal individuals, periarterial dysfunction was associated with vascular risk and sleep disturbances, whereas pAs and WM impairments preceded amyloid accumulation. In contrast, perivenous deficits emerged later, accompanying protein aggregation and cognitive decline. Regional mapping localized early pAs and WM alterations to regions vulnerable to cardiometabolic and sleep-related stress, underscoring their potential as sensitive, spatially specific biomarkers of preclinical disease. Moreover, the vascular\u0026ndash;glymphatic\u0026ndash;amyloid\u0026ndash;cognition pathway analysis demonstrated that vascular health influences cognition indirectly through its effects on glymphatic function and amyloid burden. Collectively, these findings position the pAs as a critical locus of early vulnerability and a promising target for intervention. Strategies that improve cardiovascular health, enhance sleep quality, and restore glymphatic clearance may hold potential to delay amyloid pathology and preserve cognitive function.\u003c/p\u003e "},{"header":"Methods","content":"\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003cdiv id=\"Sec17\" class=\"Section3\"\u003e \u003ch2\u003eStudy participants\u003c/h2\u003e \u003cp\u003eWe analyzed two independent cohorts. The first cohort included 579 cognitively normal (CN) adults aged\u0026thinsp;\u0026gt;\u0026thinsp;36 years from the HCP-Aging Lifespan Release 2.0\u003csup\u003e11,46\u003c/sup\u003e. Participants were recruited at Washington University in St. Louis, the University of Minnesota, Massachusetts General Hospital, and the University of California, Los Angeles. Exclusion criteria included the use of special educational services, MRI contraindications, history of major medical illness, head injury, endocrine or psychiatric disorders, and neurodevelopmental disorders.\u003c/p\u003e \u003cp\u003eThe second dataset used in this study was obtained from the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). ADNI was launched in 2003 as a public\u003cb\u003e\u0026ndash;\u003c/b\u003eprivate partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI is to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessments can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer\u0026rsquo;s disease (AD).\u003c/p\u003e \u003cp\u003eFor this study, ADNI participants were included only if they had multi-shell DWI, Fluid Attenuated Inversion Recovery (FLAIR) MRI, T1-weighted MRI, amyloid and tau PET, and covariate data (age, sex, education, and APOE ε4 genotype). Because multi-shell DWI is available exclusively in ADNI phases 3 and 4, the sample was restricted to these phases to leverage multi-shell diffusion data for compartment-specific modeling.\u003c/p\u003e \u003cp\u003eAD dementia diagnosis followed NINCDS\u0026ndash;ADRDA criteria\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. MCI was diagnosed by a comprehensive evaluation of Mini-Mental State Examination (MMSE) scores (24\u0026ndash;30), impaired delayed recall on the Wechsler Memory Scale Logical Memory II, Clinical Dementia Rating\u0026ndash;Sum of Boxes (CDR-SB)\u0026thinsp;\u0026ge;\u0026thinsp;0.5, preserved daily functioning, and absence of dementia. Cognitively normal (CN) controls were defined as having MMSE\u0026thinsp;\u0026ge;\u0026thinsp;24 and CDR-SB\u0026thinsp;=\u0026thinsp;0 or 0.5. The final ADNI sample comprised 10 AD dementia, 68 MCI, and 95 CN participants.\u003c/p\u003e \u003cp\u003e Both the ADNI and HCP-Aging studies were approved by institutional review boards of all participating centers, and written informed consent was obtained from all participants.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eMRI and PET imaging\u003c/h2\u003e \u003cp\u003eHCP-A participants were scanned on a Siemens 3T Prisma system equipped with an 80 mT/m gradient coil and a Siemens 32-channel head coil. Diffusion-weighted images (DWI) were acquired using a single-shot spin-echo EPI sequence with the following parameters: TR\u0026thinsp;=\u0026thinsp;4700 ms, TE\u0026thinsp;=\u0026thinsp;64 ms, multiband factor\u0026thinsp;=\u0026thinsp;2, GRAPPA R\u0026thinsp;=\u0026thinsp;2, 6/8 partial Fourier, 6 b\u003csub\u003e0\u003c/sub\u003e volumes, 45 diffusion directions with b\u0026thinsp;=\u0026thinsp;1500 s/mm\u0026sup2;, and 45 with b\u0026thinsp;=\u0026thinsp;3000 s/mm\u0026sup2;. A total of 90 slices were acquired with 1.5 mm thickness and no gap. The in-plane resolution was 1.5\u003cb\u003e\u0026times;\u003c/b\u003e1.5 mm, zero-filled to 0.75\u003cb\u003e\u0026times;\u003c/b\u003e0.75 mm, with a field of view (FOV) of 220 mm.\u003c/p\u003e \u003cp\u003eT1-weighted multi-echo MPRAGE scans were also acquired with TR\u0026thinsp;=\u0026thinsp;2500 ms, inversion time (TI)\u0026thinsp;=\u0026thinsp;1000 ms, TE\u0026thinsp;=\u0026thinsp;1.8/3.6/5.4/7.2 ms, 4 echoes per line of k-space, voxel size\u0026thinsp;=\u0026thinsp;0.8 mm isotropic, FOV\u0026thinsp;=\u0026thinsp;256\u003cb\u003e\u0026times;\u003c/b\u003e240\u003cb\u003e\u0026times;\u003c/b\u003e166 mm, matrix\u0026thinsp;=\u0026thinsp;320\u003cb\u003e\u0026times;\u003c/b\u003e300\u003cb\u003e\u0026times;\u003c/b\u003e208, and flip angle\u0026thinsp;=\u0026thinsp;8\u0026deg;, following the Human Connectome Project\u0026ndash;Aging protocol\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn the ADNI-3 and ADNI-4 dataset, DWI imaging was performed with an echo planar imaging sequence using: 14 b\u003csub\u003e0\u003c/sub\u003e volumes, six diffusion directions with b\u0026thinsp;=\u0026thinsp;500 s/mm\u0026sup2;, 56 with b\u0026thinsp;=\u0026thinsp;1000 s/mm\u0026sup2;, and 56 with b\u0026thinsp;=\u0026thinsp;2000 s/mm\u0026sup2; (TE\u0026thinsp;=\u0026thinsp;61.8\u0026ndash;83.1 ms, TR\u0026thinsp;=\u0026thinsp;9500\u0026ndash;14,200 ms, matrix\u0026thinsp;=\u0026thinsp;256\u003cb\u003e\u0026times;\u003c/b\u003e256, FOV\u0026thinsp;=\u0026thinsp;350\u003cb\u003e\u0026times;\u003c/b\u003e350 mm, and slice thickness\u0026thinsp;=\u0026thinsp;2.0 mm), resulting in voxel size of 2\u003cb\u003e-\u003c/b\u003emm isotropic. T1-weighted images were acquired using a spoiled gradient echo sequence with the following parameters: TR\u0026thinsp;=\u0026thinsp;6.96\u0026ndash;7.67 ms, TE\u0026thinsp;=\u0026thinsp;2.83\u0026ndash;3.17 ms, matrix\u0026thinsp;=\u0026thinsp;256\u003cb\u003e\u0026times;\u003c/b\u003e256, FOV\u0026thinsp;=\u0026thinsp;260\u003cb\u003e\u0026times;\u003c/b\u003e260 or 270\u003cb\u003e\u0026times;\u003c/b\u003e270 mm, and slice thickness\u0026thinsp;=\u0026thinsp;1.2 mm. Fluid-attenuated inversion recovery (FLAIR) imaging was obtained with the following parameters: TE\u0026thinsp;=\u0026thinsp;147.90\u0026ndash;153.90 ms; TR\u0026thinsp;=\u0026thinsp;11,000 ms; TI\u0026thinsp;=\u0026thinsp;2250 ms; matrix\u0026thinsp;=\u0026thinsp;256\u0026times;256; FOV\u0026thinsp;=\u0026thinsp;220\u003cb\u003e\u0026times;\u003c/b\u003e220 mm; and slice thickness\u0026thinsp;=\u0026thinsp;5 mm. All the images were obtained from the ADNI database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://adni.loni.usc.edu/\u003c/span\u003e\u003cspan address=\"http://adni.loni.usc.edu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe Aβ (florbetapir [\u003csup\u003e18\u003c/sup\u003eF\u003cb\u003e\u0026ndash;\u003c/b\u003eAV45] or florbetaben [FBB]) and tau (flortaucipir [\u003csup\u003e18\u003c/sup\u003eF\u003cb\u003e\u0026ndash;\u003c/b\u003eAV1451]) PET data were also available for the studied subjects. Aβ and tau SUVRs were calculated using the UC Berkeley amyloid and tau PET processing pipeline. Details of the processing protocol can be found in the open-source documentation (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://adni.loni.usc.edu/updated-uc-berkeley-amyloid-pet-methods/\u003c/span\u003e\u003cspan address=\"https://adni.loni.usc.edu/updated-uc-berkeley-amyloid-pet-methods/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003cp\u003eThe overall pipeline for estimation of multimodal MRI features of glymphatic function is illustrated in \u003cb\u003eSupplementary Fig. S5\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eImage processing and pAs segmentation, and false positive exclusion\u003c/h2\u003e \u003cp\u003eThree-dimensional T1-weighted (T1w) volumetric images were preprocessed using the Montreal Neurological Institute (MNI) CIVET pipeline, which included brain extraction, bias-field correction, and tissue classification\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. Binary brain tissue masks were generated by CIVET. After excluding dura mater, cerebral gray matter, cerebellum, and brainstem, cerebral WM was defined for pAs segmentation.\u003c/p\u003e \u003cp\u003eTo enhance visibility, histogram equalization was applied to the T1w images, followed by non-local means denoising to reduce noise, as described in a prior study\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. pAs were then segmented using a pre-trained nnUNet model as previously described\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e. All segmentations were visually inspected for quality by neurologists (J.H., K.W.K., J.W.S., and T.W.Y.) and trained raters (A.S.K. and J.C.), with manual corrections applied when necessary. For the HCP-A dataset, combined T1w and T2w images were used to improve visual inspection where needed.\u003c/p\u003e \u003cp\u003eTo minimize contamination of pAs segmentation by WM hyperintensities (WMHs), falsely segmented WMHs and lacunes were excluded by evaluating morphology (linear vs. round), size, clustering, and anatomical location. In the ADNI dataset, available FLAIR images, which optimally delineate WMHs, were additionally used. WMHs were segmented from multimodal T1w and FLAIR images using the U-Net with highlighted foreground technique\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. These FLAIR-derived WMH masks were then non-linearly registered to the corresponding T1w images, and WMH voxels were removed from the pAs annotations to ensure robust segmentation.\u003c/p\u003e \u003cp\u003eFor analyses of CMS effects on glymphatic function in cognitively normal participants, pAs volumes segmented from HCP-A were mapped onto nine functional regions of interest (ROIs). These ROIs were defined on the cortical surface using the Automated Anatomical Labeling (AAL) template, following prior work\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e, and included the frontoparietal network, dorsal attention, ventral attention/language, sensorimotor, default mode network, auditory, visual, salience, and limbic network.\u003c/p\u003e \u003cp\u003eFor analyses of AD progression, ROIs corresponding to the AD-signature, including bilateral hippocampi, temporal poles, and entorhinal, inferior temporal, superior frontal, inferior parietal, and anterior cingulate cortices, were also defined using the AAL template. The precuneus, implicated in early AD pathology\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e, was separately delineated for targeted analyses.\u003c/p\u003e \u003cp\u003eROIs were propagated into adjacent superficial white matter using \u003cem\u003ek\u003c/em\u003e-nearest neighbors (k-NN) clustering, constrained to a depth of 5 mm from the gray matter\u0026ndash;WM boundary using Laplacian distance mapping (\u003cb\u003eSupplementary Fig. S6\u003c/b\u003e). A pAs volume was assigned to an ROI if it overlapped with the ROI mask.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eDiffusion-weighted MRI-based glymphatic metrics: FWVF, FWAD, and ALPS index calculation\u003c/h2\u003e \u003cp\u003eMulti-shell diffusion-weighted imaging (DWI) data were corrected for head motion, eddy current distortions, and EPI-related susceptibility artifacts using FSL\u0026rsquo;s EDDY\u003csup\u003e\u003cb\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/b\u003e\u003c/sup\u003e. Unlike conventional free-water elimination DTI (FWE-DTI)\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e, which constrains the free-water compartment to be isotropic, the Tissue\u0026ndash;Tensor Imaging (TTI) framework\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e accounts for directional diffusion within perivascular and extracellular fluid spaces. This enables estimation of fluid-compartment diffusivity measures (FWAD, FWRD, FWMD, FWFA) alongside tissue-compartment metrics (tAD, tRD, tMD, tFA), where fAD reflects the axial diffusivity of the fluid tensor and is interpreted as an indicator of directionality in perivascular fluid movement\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. Detailed methods are described in the \u003cb\u003eSupplementary Methods\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eIndividual b\u003csub\u003e0\u003c/sub\u003e images were linearly and then nonlinearly registered to each participant\u0026rsquo;s T1-weighted (T1w) space using ANTs with a mutual information similarity metric\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e. The resulting DWI-derived maps were subsequently warped to each subject\u0026rsquo;s T1w space using the deformation fields from this registration, aligning DTI-based metrics to the anatomical space where pAs annotations and the functional network ROIs and AD-related ROIs were defined (\u003cb\u003eSupplementary Fig. S5\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eThis framework leveraged the anatomical role of pAs as cerebrospinal fluid (CSF)\u0026ndash;filled conduits. Within each segmented pAs, free-water axial diffusivity (FWAD) was quantified as an index of fluid mobility, whereas free-water volume fraction (FWVF) was calculated to capture local CSF accumulation that may hinder fluid transport. FWAD and FWVF values were averaged across voxels within each pAs mask in T1w space. For complementary validation, FWFA, FWMD, and FWRD were also computed within the same pAs regions. Validation results are presented in \u003cb\u003eSupplementary Fig. S3\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eTo assess free-water burden in the WM parenchyma, all WMH and pAs voxels were excluded from the WM mask. The mean FWVF of the remaining WM voxels within each ROI was then calculated and defined as the ROI\u0026rsquo;s WM-FWVF. Finally, interstitial fluid efflux along pVs was evaluated using the DTI-ALPS, as described in \u003cb\u003eSupplementary Methods\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eBecause many pAs detected on high-resolution T1w MRI have diameters smaller than the effective DWI voxel size, diffusion measures within pAs masks may be affected by partial volume contamination. To mitigate this, we restricted analyses to the largest pAs within each ROI, determining the optimal number (\u003cem\u003en\u003c/em\u003e) that maximized measurement stability and sensitivity to clinical variables (\u003cb\u003eSupplementary Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e). The summed absolute t-values from correlations of FWAD and FWVF with age, PSQI, and CMS factors (adjusted for age, sex, and education) were highest when four pAs per ROI were included, indicating an optimal trade-off between sensitivity and sample retention. At this threshold, subject inclusion remained high (95% at four pAs vs. 87% at six pAs, particularly within the auditory network). Therefore, the four largest pAs per ROI were used for all subsequent analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eCognitive assessments\u003c/h2\u003e \u003cp\u003eAt the time of HCP-A recruitment, participants were in good health, without a diagnosed history of neurologic or major psychiatric disorder, symptomatic stroke or dementia. The self-reported Pittsburgh Sleep Quality Index (PSQI) questionnaire and Montreal Cognitive Assessment (MoCA) scores were included in this study.\u003c/p\u003e \u003cp\u003eFor participants from the ADNI cohort, global cognitive function was assessed using both Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA)\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eCMS profiles and AD biomarkers\u003c/h2\u003e \u003cp\u003eThe HCP-A metadata provided comprehensive CMS profiles, including height, weight, blood pressure, fasting glucose, and lipid biomarkers. BMI was calculated as weight (kg) divided by height squared (m\u0026sup2;), based on measurements obtained during participant interviews. Systolic and diastolic blood pressures were recorded with participants seated and mean arterial pressure (MAP) was derived as: MAP = (2 \u0026times; diastolic blood pressure\u0026thinsp;+\u0026thinsp;systolic blood pressure) / 3. Metabolic biomarkers, including high-density lipoprotein (HDL), hemoglobin A1c (HbA1c), plasma glucose, triglycerides, and total cholesterol, were quantified from fasting blood samples.\u003c/p\u003e \u003cp\u003eThe ADNI dataset provided a small number of cardiometabolic measures, including height, weight, and systolic and diastolic blood pressures, allowing BMI and MAP to be calculated using the same formulas applied in HCP-A. In addition to these cardiometabolic variables, ADNI participants were also characterized by Aβ status (positive or negative) determined by amyloid PET\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e, enabling subgroup comparisons and evaluation of glymphatic alterations in relation to AD pathology.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were conducted using Python (v3.9) and relevant packages including statsmodels and scipy.\u003c/p\u003e \u003cp\u003eTo examine associations between glymphatic fluid metrics and either age or CMS in the HCP cohort, linear regression models were fitted for each metric separately. Independent variables included age, BMI, systolic and diastolic blood pressure, mean arterial pressure (MAP), HDL, HbA1c, triglycerides, and glucose. Age and sex were adjusted as covariates (only sex was included for correlations with age). Regression coefficients (standardized β-values) and p-values were extracted and visualized in scatterplots. False discovery rate (FDR) correction using the Benjamini-Hochberg procedure was applied across multiple comparisons within each anatomical compartment to control for type I error. Significance was defined as FDR-corrected p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, and shaded areas in plots represent 95% confidence intervals. Variables with non-significant associations were labeled as \u0026ldquo;NS\u0026rdquo;.\u003c/p\u003e \u003cp\u003eFor the ADNI dataset, group-level differences in glymphatic dysfunction were examined using analysis of covariance (ANCOVA) across diagnostic groups (CN, MCI, and AD) and across amyloid status groups (A\u0026thinsp;\u0026minus;\u0026thinsp;and A+), adjusting for age, sex, APOE genotype (ε4 vs. non-carrier), and years of education. Post hoc pairwise comparisons were conducted using Tukey\u0026rsquo;s HSD test. Violin plots depict the distribution of each metric within groups, and statistically significant differences are marked with asterisks (*p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, **p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, ***p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). For features that did not conform to a normal distribution, nonparametric rank-ANCOVA and Wilcoxon rank-sum tests were applied for post hoc comparisons.\u003c/p\u003e \u003cp\u003eTo evaluate the discriminative ability of glymphatic MRI metrics for Aβ-PET positivity and clinical diagnosis, supervised learning models were trained using a set of linear and nonlinear classifiers, including support vector machines (SVMs; linear and radial basis function kernels), extreme gradient boosting (XGBoost), and generalized linear models (GLMs). Input features were defined from three glymphatic compartments: (1) pAs features, including FWVF and FWAD within the precuneus, AD-signature, and whole brain; (2) WM features, including FWVF within the precuneus, AD-signature, and whole brain; and (3) the pVs feature, represented by the DTI-ALPS index. For molecular imaging biomarkers, Aβ-PET features included SUVRs in the precuneus, AD-signature, and whole brain, and tau-PET features included SUVRs in the same regions.\u003c/p\u003e \u003cp\u003eAll models were trained and evaluated using stratified 10-fold cross-validation (number of splits\u0026thinsp;=\u0026thinsp;10, random state\u0026thinsp;=\u0026thinsp;42). Classification performance was quantified using the area under the receiver operating characteristic curve (AUC). Pairwise DeLong tests were conducted to assess statistical differences in AUC values between models, and corresponding AUCs are reported in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Detailed performance results for all classifiers are provided in \u003cb\u003eTable S4\u003c/b\u003e\u0026thinsp;\u0026minus;\u0026thinsp;\u003cb\u003eS6\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eMediation analyses were performed using a standard three-path model to test indirect effects (a \u0026times; b) of age or CMS variables on outcome measures (MoCA, or MMSE) through glymphatic fluid metrics and Aβ-PET SUVR as mediators. Path coefficients (β), indirect effect estimates, and bias-corrected 95% confidence intervals were reported. Significance of indirect paths was determined via bootstrap resampling (n\u0026thinsp;=\u0026thinsp;5,000). All variables were normalized to ensure comparability across predictors, and the significance of indirect effects was determined based on the bootstrapped confidence intervals that indirect effects are considered significant if the CI does not include zero.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData collection and sharing for this project was funded by the Alzheimer\u0026apos;s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer\u0026rsquo;s Association; Alzheimer\u0026rsquo;s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research \u0026amp; Development, LLC.; Johnson \u0026amp; Johnson Pharmaceutical Research \u0026amp; Development LLC.; Lumosity; Lundbeck; Merck \u0026amp; Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer\u0026rsquo;s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Office of the Director, National Institutes of Health under Award Numbers U01AG024904 and S10OD032285, and the Foundation of the American Society of Neuroradiology Grant Program (1960682569). Additional support was provided by the University of Southern California Generative Artificial Intelligence Grant, and the University of Southern California Department of Biomedical Engineering Mann Pilot Grant (PR0002274). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.\u003c/p\u003e\n\u003cp\u003eFigure 1a was created using BioRender.com.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical statements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study used publicly available, de-identified data from the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI) and the Human Connectome Project in Aging (HCP-A). For ADNI, the study was approved by the Institutional Review Boards of all participating institutions, and all participants provided written informed consent. For HCP-A, the study was approved by the Institutional Review Boards of participating institutions, including Washington University in St. Louis, the University of Minnesota, Massachusetts General Hospital, and the University of California, Los Angeles, and all participants provided written informed consent.\u003c/p\u003e\n\u003cp\u003eBecause both datasets are multi-site studies, Institutional Review Board approval was obtained separately at each participating institution, and a single centralized approval number is not applicable. The present study involved only secondary analysis of de-identified data and did not require additional Institutional Review Board approval..\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eY.C., H.Z., H.K., S.C., J.H., K.W.C., N.J., P.M.T., and D.J.J.W. contributed to the conception and design of the study; Y.C., H.Z., A.S.K., Q.S., S.C., J.H., K.W.C., H.C., G.P., J.C., J.W.S., and M.K. contributed to the analysis of data; Y.C., H.K., A.S.K., Q.S., S.C., H.C., and M.K. contributed to drafting the text or preparing the figures; E.Y.J., D.C.L., H.K., H.C.C., and A.W.T. contributed to revision of the manuscript; A.W.T. and H.K. contributed to funding acquisition.\u003c/p\u003e\n\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eThe data that support the findings of this study are available within the article and its supplementary information. Imaging data were obtained from publicly available resources, including the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI; http://adni.loni.usc.edu/ ) and the Human Connectome Project via the NIH NIMH Data Archive (https://nda.nih.gov/ccf ). Details of ADNI SUVR PET processing are available at https://adni.loni.usc.edu/updated-uc-berkeley-amyloid-pet-methods/ . Region-of-interest (ROI) templates are available at https://github.com/Yaqiongchai/ROI-template . Binary periarterial space segmentation masks are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eNedergaard, M. \u0026amp; Goldman, S. A. Glymphatic failure as a final common pathway to dementia. \u003cem\u003eScience\u003c/em\u003e 370, 50\u0026ndash;56, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/science.abb8739\u003c/span\u003e\u003cspan address=\"10.1126/science.abb8739\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIliff, J. J. \u003cem\u003eet al.\u003c/em\u003e A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. \u003cem\u003eSci Transl Med\u003c/em\u003e 4, 147ra111, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/scitranslmed.3003748\u003c/span\u003e\u003cspan address=\"10.1126/scitranslmed.3003748\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarrison, I. F. \u003cem\u003eet al.\u003c/em\u003e Impaired glymphatic function and clearance of tau in an Alzheimer's disease model. \u003cem\u003eBrain\u003c/em\u003e 143, 2576\u0026ndash;2593, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/brain/awaa179\u003c/span\u003e\u003cspan address=\"10.1093/brain/awaa179\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaoka, T. \u003cem\u003eet al.\u003c/em\u003e Evaluation of glymphatic system activity with the diffusion MR technique: diffusion tensor image analysis along the perivascular space (DTI-ALPS) in Alzheimer's disease cases. \u003cem\u003eJpn J Radiol\u003c/em\u003e 35, 172\u0026ndash;178, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11604-017-0617-z\u003c/span\u003e\u003cspan address=\"10.1007/s11604-017-0617-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGertje, E. C., van Westen, D., Panizo, C., Mattsson-Carlgren, N. \u0026amp; Hansson, O. Association of Enlarged Perivascular Spaces and Measures of Small Vessel and Alzheimer Disease. \u003cem\u003eNeurology\u003c/em\u003e 96, e193-e202, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1212/WNL.0000000000011046\u003c/span\u003e\u003cspan address=\"10.1212/WNL.0000000000011046\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTatekawa, H., Matsushita, S. \u0026amp; Miki, Y. Reply to the letter to the editor: Improved reproducibility of diffusion tensor image analysis along the perivascular space (DTI-ALPS) index calculated by manual and automated methods. \u003cem\u003eJpn J Radiol\u003c/em\u003e 41, 1035\u0026ndash;1036, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11604-023-01455-6\u003c/span\u003e\u003cspan address=\"10.1007/s11604-023-01455-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHayden, M. R. A Closer Look at the Perivascular Unit in the Development of Enlarged Perivascular Spaces in Obesity, Metabolic Syndrome, and Type 2 Diabetes Mellitus. \u003cem\u003eBiomedicines\u003c/em\u003e 12, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/biomedicines12010096\u003c/span\u003e\u003cspan address=\"10.3390/biomedicines12010096\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAcree, L. S., Montgomery, P. S. \u0026amp; Gardner, A. W. The influence of obesity on arterial compliance in adult men and women. \u003cem\u003eVasc Med\u003c/em\u003e 12, 183\u0026ndash;188, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/1358863X07079323\u003c/span\u003e\u003cspan address=\"10.1177/1358863X07079323\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShang, Y. \u003cem\u003eet al.\u003c/em\u003e Diffusion Tensor Imaging Analysis Along the Perivascular Space (DTI-ALPS) Demonstrates That Sleep Disorders Exacerbate Glymphatic Circulatory Impairment and Cognitive Impairment in Patients with Alzheimer's Disease. \u003cem\u003eNat Sci Sleep\u003c/em\u003e 16, 2205\u0026ndash;2215, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/NSS.S496607\u003c/span\u003e\u003cspan address=\"10.2147/NSS.S496607\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumont, M. \u003cem\u003eet al.\u003c/em\u003e Free Water in White Matter Differentiates MCI and AD From Control Subjects. \u003cem\u003eFront Aging Neurosci\u003c/em\u003e 11, 270, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnagi.2019.00270\u003c/span\u003e\u003cspan address=\"10.3389/fnagi.2019.00270\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBookheimer, S. Y. \u003cem\u003eet al.\u003c/em\u003e The Lifespan Human Connectome Project in Aging: An overview. \u003cem\u003eNeuroimage\u003c/em\u003e 185, 335\u0026ndash;348, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2018.10.009\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2018.10.009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIwatsubo, T. [Alzheimer's disease Neuroimaging Initiative (ADNI)]. \u003cem\u003eNihon Rinsho\u003c/em\u003e 69 Suppl 8, 570\u0026ndash;574 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWardlaw, J. M. \u003cem\u003eet al.\u003c/em\u003e Perivascular spaces in the brain: anatomy, physiology and pathology. \u003cem\u003eNat Rev Neurol\u003c/em\u003e 16, 137\u0026ndash;153, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41582-020-0312-z\u003c/span\u003e\u003cspan address=\"10.1038/s41582-020-0312-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLynch, K. M., Custer, R. M., Sepehrband, F., Toga, A. W. \u0026amp; Choupan, J. Perivascular and parenchymal fluid characteristics are related to age and cognitive performance across the lifespan. \u003cem\u003eImaging Neurosci (Camb)\u003c/em\u003e 3, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1162/imag_a_00549\u003c/span\u003e\u003cspan address=\"10.1162/imag_a_00549\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasternak, O., Sochen, N., Gur, Y., Intrator, N. \u0026amp; Assaf, Y. Free water elimination and mapping from diffusion MRI. \u003cem\u003eMagn Reson Med\u003c/em\u003e 62, 717\u0026ndash;730, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/mrm.22055\u003c/span\u003e\u003cspan address=\"10.1002/mrm.22055\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSepehrband, F. \u003cem\u003eet al.\u003c/em\u003e Perivascular space fluid contributes to diffusion tensor imaging changes in white matter. \u003cem\u003eNeuroimage\u003c/em\u003e 197, 243\u0026ndash;254, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2019.04.070\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2019.04.070\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMestre, H. \u003cem\u003eet al.\u003c/em\u003e Flow of cerebrospinal fluid is driven by arterial pulsations and is reduced in hypertension. \u003cem\u003eNature Communications\u003c/em\u003e 9, 4878, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41467-018-07318-3\u003c/span\u003e\u003cspan address=\"10.1038/s41467-018-07318-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMitchell, G. F. \u003cem\u003eet al.\u003c/em\u003e Arterial stiffness, pressure and flow pulsatility and brain structure and function: the Age, Gene/Environment Susceptibility \u0026ndash; Reykjavik Study. \u003cem\u003eBrain\u003c/em\u003e 134, 3398\u0026ndash;3407, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/brain/awr253\u003c/span\u003e\u003cspan address=\"10.1093/brain/awr253\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo, F. \u003cem\u003eet al.\u003c/em\u003e Assessing cerebral microvascular volumetric with high-resolution 4D cerebral blood volume MRI at 7 T. \u003cem\u003eNat Cardiovasc Res\u003c/em\u003e 4, 1424\u0026ndash;1438, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s44161-025-00722-1\u003c/span\u003e\u003cspan address=\"10.1038/s44161-025-00722-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacauley, S. L. \u003cem\u003eet al.\u003c/em\u003e Hyperglycemia modulates extracellular amyloid-β concentrations and neuronal activity in vivo. \u003cem\u003eJ Clin Invest\u003c/em\u003e 125, 2463\u0026ndash;2467, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1172/jci79742\u003c/span\u003e\u003cspan address=\"10.1172/jci79742\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eP, X. \u003cem\u003eet al.\u003c/em\u003e The role of LRP1 in Aβ efflux transport across the blood-brain barrier and cognitive dysfunction in diabetes mellitus. \u003cem\u003eNeurochem Int\u003c/em\u003e 160, 105417, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuint.2022.105417\u003c/span\u003e\u003cspan address=\"10.1016/j.neuint.2022.105417\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang, Q. \u003cem\u003eet al.\u003c/em\u003e Impairment of the glymphatic system after diabetes. \u003cem\u003eJ Cereb Blood Flow Metab\u003c/em\u003e 37, 1326\u0026ndash;1337, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/0271678x16654702\u003c/span\u003e\u003cspan address=\"10.1177/0271678x16654702\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYen, C. H. \u003cem\u003eet al.\u003c/em\u003e Angiotensin Receptor Blockades Effect on Peripheral Muscular and Central Aortic Arterial Stiffness: A Meta-Analysis of Randomized Controlled Trials and Systematic Review. \u003cem\u003eActa Cardiol Sin\u003c/em\u003e 30, 98\u0026ndash;107 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStrandberg, T. E. \u0026amp; Tienari, P. J. Contributions of vascular and Alzheimer's disease pathology to dementia. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 15, 1004\u0026ndash;1005, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jalz.2019.04.004\u003c/span\u003e\u003cspan address=\"10.1016/j.jalz.2019.04.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXie, L. \u003cem\u003eet al.\u003c/em\u003e Sleep drives metabolite clearance from the adult brain. \u003cem\u003eScience\u003c/em\u003e 342, 373\u0026ndash;377, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/science.1241224\u003c/span\u003e\u003cspan address=\"10.1126/science.1241224\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa, J. \u003cem\u003eet al.\u003c/em\u003e Effects of sleep on the glymphatic functioning and multimodal human brain network affecting memory in older adults. \u003cem\u003eMol Psychiatry\u003c/em\u003e, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41380-024-02778-0\u003c/span\u003e\u003cspan address=\"10.1038/s41380-024-02778-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaril, A.-A. \u003cem\u003eet al.\u003c/em\u003e Lighter sleep is associated with higher enlarged perivascular spaces burden in middle-aged and elderly individuals. \u003cem\u003eSleep Medicine\u003c/em\u003e 100, 558\u0026ndash;564, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.sleep.2022.10.006\u003c/span\u003e\u003cspan address=\"10.1016/j.sleep.2022.10.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNadjar, A., Wigren, H. M. \u0026amp; Tremblay, M. E. Roles of Microglial Phagocytosis and Inflammatory Mediators in the Pathophysiology of Sleep Disorders. \u003cem\u003eFront Cell Neurosci\u003c/em\u003e 11, 250, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fncel.2017.00250\u003c/span\u003e\u003cspan address=\"10.3389/fncel.2017.00250\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBickford, P. C., Flowers, A. \u0026amp; Grimmig, B. Aging leads to altered microglial function that reduces brain resiliency increasing vulnerability to neurodegenerative diseases. \u003cem\u003eExp Gerontol\u003c/em\u003e 94, 4\u0026ndash;8, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.exger.2017.01.027\u003c/span\u003e\u003cspan address=\"10.1016/j.exger.2017.01.027\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai, Y. \u003cem\u003eet al.\u003c/em\u003e The relationship between inflammation, impaired glymphatic system, and neurodegenerative disorders: A vicious cycle. \u003cem\u003eNeurobiology of Disease\u003c/em\u003e 192, 106426, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.nbd.2024.106426\u003c/span\u003e\u003cspan address=\"10.1016/j.nbd.2024.106426\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmets, N. G., van der Panne, S. A., Strijkers, G. J. \u0026amp; Bakker, E. Perivascular spaces around arteries exceed perivenous spaces in the mouse brain. \u003cem\u003eSci Rep\u003c/em\u003e 14, 17655, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-67885-y\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-67885-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlbargothy, N. J. \u003cem\u003eet al.\u003c/em\u003e Convective influx/glymphatic system: tracers injected into the CSF enter and leave the brain along separate periarterial basement membrane pathways. \u003cem\u003eActa Neuropathol\u003c/em\u003e 136, 139\u0026ndash;152, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00401-018-1862-7\u003c/span\u003e\u003cspan address=\"10.1007/s00401-018-1862-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKress, B. T. \u003cem\u003eet al.\u003c/em\u003e Impairment of paravascular clearance pathways in the aging brain. \u003cem\u003eAnn Neurol\u003c/em\u003e 76, 845\u0026ndash;861, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/ana.24271\u003c/span\u003e\u003cspan address=\"10.1002/ana.24271\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXie, L. \u003cem\u003eet al.\u003c/em\u003e Higher intracranial arterial pulsatility is associated with presumed imaging markers of the glymphatic system: An explorative study. \u003cem\u003eNeuroImage\u003c/em\u003e 288, 120524, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.neuroimage.2024.120524\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2024.120524\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXin, S. H., Tan, L., Cao, X., Yu, J. T. \u0026amp; Tan, L. Clearance of Amyloid Beta and Tau in Alzheimer's Disease: from Mechanisms to Therapy. \u003cem\u003eNeurotox Res\u003c/em\u003e 34, 733\u0026ndash;748, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12640-018-9895-1\u003c/span\u003e\u003cspan address=\"10.1007/s12640-018-9895-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang, S. Y. \u003cem\u003eet al.\u003c/em\u003e Glymphatic system dysfunction predicts amyloid deposition, neurodegeneration, and clinical progression in Alzheimer's disease. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 20, 3251\u0026ndash;3269, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/alz.13789\u003c/span\u003e\u003cspan address=\"10.1002/alz.13789\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKamagata, K. \u003cem\u003eet al.\u003c/em\u003e Association of MRI Indices of Glymphatic System With Amyloid Deposition and Cognition in Mild Cognitive Impairment and Alzheimer Disease. \u003cem\u003eNeurology\u003c/em\u003e 99, e2648-e2660, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1212/wnl.0000000000201300\u003c/span\u003e\u003cspan address=\"10.1212/wnl.0000000000201300\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeschorner, N. \u0026amp; Nedergaard, M. Glymphatic system dysfunction in neurodegenerative diseases. \u003cem\u003eCurr Opin Neurol\u003c/em\u003e 37, 182\u0026ndash;188, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/wco.0000000000001252\u003c/span\u003e\u003cspan address=\"10.1097/wco.0000000000001252\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, H. G., Lee, J. H., Flausino, L. E. \u0026amp; Quintana, F. J. Neuroinflammation: An astrocyte perspective. \u003cem\u003eSci Transl Med\u003c/em\u003e 15, eadi7828, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/scitranslmed.adi7828\u003c/span\u003e\u003cspan address=\"10.1126/scitranslmed.adi7828\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, S. \u003cem\u003eet al.\u003c/em\u003e Ion Channel Dysfunction in Astrocytes in Neurodegenerative Diseases. \u003cem\u003eFront Physiol\u003c/em\u003e 13, 814285, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fphys.2022.814285\u003c/span\u003e\u003cspan address=\"10.3389/fphys.2022.814285\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, M. \u003cem\u003eet al.\u003c/em\u003e Impaired Glymphatic Function and Pulsation Alterations in a Mouse Model of Vascular Cognitive Impairment. \u003cem\u003eFront Aging Neurosci\u003c/em\u003e 13, 788519, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnagi.2021.788519\u003c/span\u003e\u003cspan address=\"10.3389/fnagi.2021.788519\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia, Y. \u003cem\u003eet al.\u003c/em\u003e The glymphatic system was impaired in spontaneously hypertensive rats. \u003cem\u003eScientific Reports\u003c/em\u003e 15, 18321, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-025-02054-3\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-02054-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVilleneuve, S. \u003cem\u003eet al.\u003c/em\u003e Existing Pittsburgh Compound-B positron emission tomography thresholds are too high: statistical and pathological evaluation. \u003cem\u003eBrain\u003c/em\u003e 138, 2020\u0026ndash;2033, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/brain/awv112\u003c/span\u003e\u003cspan address=\"10.1093/brain/awv112\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalmqvist, S. \u003cem\u003eet al.\u003c/em\u003e Earliest accumulation of β-amyloid occurs within the default-mode network and concurrently affects brain connectivity. \u003cem\u003eNat Commun\u003c/em\u003e 8, 1214, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41467-017-01150-x\u003c/span\u003e\u003cspan address=\"10.1038/s41467-017-01150-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, X. \u003cem\u003eet al.\u003c/em\u003e Skeleton-based cerebrovascular quantitative analysis. \u003cem\u003eBMC Med Imaging\u003c/em\u003e 16, 68, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12880-016-0170-8\u003c/span\u003e\u003cspan address=\"10.1186/s12880-016-0170-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGlasser, M. F. \u003cem\u003eet al.\u003c/em\u003e The minimal preprocessing pipelines for the Human Connectome Project. \u003cem\u003eNeuroimage\u003c/em\u003e 80, 105\u0026ndash;124, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2013.04.127\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2013.04.127\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcKhann, G. \u003cem\u003eet al.\u003c/em\u003e Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. \u003cem\u003eNeurology\u003c/em\u003e 34, 939\u0026ndash;944, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1212/wnl.34.7.939\u003c/span\u003e\u003cspan address=\"10.1212/wnl.34.7.939\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1984).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarms, M. P. \u003cem\u003eet al.\u003c/em\u003e Extending the Human Connectome Project across ages: Imaging protocols for the Lifespan Development and Aging projects. \u003cem\u003eNeuroimage\u003c/em\u003e 183, 972\u0026ndash;984, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2018.09.060\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2018.09.060\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLepage, C. \u003cem\u003eet al.\u003c/em\u003e CIVET-Macaque: An automated pipeline for MRI-based cortical surface generation and cortical thickness in macaques. \u003cem\u003eNeuroimage\u003c/em\u003e 227, 117622, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2020.117622\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2020.117622\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarisano, G. \u003cem\u003eet al.\u003c/em\u003e Imaging perivascular space structure and function using brain MRI. \u003cem\u003eNeuroimage\u003c/em\u003e 257, 119329, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2022.119329\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2022.119329\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChai, Y., Hedong Zhang, Gilsoon Park, Erika Lopez, Cong Zang, Jongmok Ha, Omar Elhawary, and Hosung Kim. in \u003cem\u003eIn Medical Imaging with Deep Learning\u003c/em\u003e (Nashvile, TN, USA, 2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark, G., Hong, J., Duffy, B. A., Lee, J. M. \u0026amp; Kim, H. White matter hyperintensities segmentation using the ensemble U-Net with multi-scale highlighting foregrounds. \u003cem\u003eNeuroimage\u003c/em\u003e 237, 118140, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2021.118140\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2021.118140\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaddad, E. \u003cem\u003eet al.\u003c/em\u003e P-Wave Duration Is Associated With Aging Patterns in Structural Brain Networks. \u003cem\u003eJ Am Heart Assoc\u003c/em\u003e 13, e035881, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/jaha.124.035881\u003c/span\u003e\u003cspan address=\"10.1161/jaha.124.035881\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, Q., Coolen, B. F., Versluis, M. J., Strijkers, G. J. \u0026amp; Nederveen, A. J. Diffusion-prepared stimulated-echo turbo spin echo (DPsti-TSE): An eddy current-insensitive sequence for three-dimensional high-resolution and undistorted diffusion-weighted imaging. \u003cem\u003eNMR Biomed\u003c/em\u003e 30, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/nbm.3719\u003c/span\u003e\u003cspan address=\"10.1002/nbm.3719\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoy, A. R., Koay, C. G., Kecskemeti, S. R. \u0026amp; Alexander, A. L. Optimization of a free water elimination two-compartment model for diffusion tensor imaging. \u003cem\u003eNeuroimage\u003c/em\u003e 103, 323\u0026ndash;333, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2014.09.053\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2014.09.053\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCabeen, R. P., David H. Laidlaw, and Arthur W. Toga. in \u003cem\u003eISMRM-ESMRMB Abstracts\u003c/em\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAvants, B. B. \u003cem\u003eet al.\u003c/em\u003e A reproducible evaluation of ANTs similarity metric performance in brain image registration. \u003cem\u003eNeuroimage\u003c/em\u003e 54, 2033\u0026ndash;2044, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.neuroimage.2010.09.025\u003c/span\u003e\u003cspan address=\"10.1016/j.neuroimage.2010.09.025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith, T., Gildeh, N. \u0026amp; Holmes, C. The Montreal Cognitive Assessment: validity and utility in a memory clinic setting. \u003cem\u003eThe Canadian Journal of Psychiatry\u003c/em\u003e 52, 329\u0026ndash;332 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJosephs, K. A., Weigand, S. D. \u0026amp; Whitwell, J. L. Characterizing Amyloid-Positive Individuals With Normal Tau PET Levels After 5 Years: An ADNI Study. \u003cem\u003eNeurology\u003c/em\u003e 98, e2282-e2292, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1212/wnl.0000000000200287\u003c/span\u003e\u003cspan address=\"10.1212/wnl.0000000000200287\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-dementia","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Dementia](https://www.nature.com/npjdementia/)","snPcode":"44400","submissionUrl":"https://submission.springernature.com/new-submission/44400/3","title":"npj Dementia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9374985/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9374985/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe brain\u0026rsquo;s waste-clearance (glymphatic) system removes metabolic byproducts via periarterial influx, interstitial exchange, and perivenous efflux. Although dysfunction is implicated in Alzheimer\u0026rsquo;s disease (AD), current imaging markers emphasize perivenous changes and may overlook earlier periarterial impairment. We developed a diffusion MRI framework to quantify periarterial fluid mobility, white matter free water, and perivenous integrity, and applied it to 546 cognitively normal adults (HCP-Aging) and 173 participants across the AD spectrum (ADNI). Periarterial mobility was reduced with higher cardiometabolic risk and amyloid positivity, particularly in AD-vulnerable regions. Free water increased with aging and metabolic burden, whereas perivenous dysfunction was most pronounced in AD. Combined measures predicted amyloid positivity and cognitive impairment (AUC\u0026thinsp;=\u0026thinsp;0.82). Mediation analyses showed that blood pressure influenced cognition through periarterial dysfunction and amyloid burden. These findings support a staged, compartment-specific trajectory of glymphatic dysfunction, with early periarterial impairment representing a potential biomarker and therapeutic target.\u003c/p\u003e","manuscriptTitle":"Multimodal Free-Water Imaging Links Cardiometabolic Risk to Periarterial Dysfunction and Amyloid Accumulation in Early Alzheimer’s","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-07 20:14:50","doi":"10.21203/rs.3.rs-9374985/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"109526560184165207868513892477728851016","date":"2026-05-18T23:22:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"58590773946702051887766090321434442249","date":"2026-05-08T16:02:53+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-23T12:43:05+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-22T03:26:59+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-16T04:51:02+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Dementia","date":"2026-04-10T05:37:48+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-dementia","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Dementia](https://www.nature.com/npjdementia/)","snPcode":"44400","submissionUrl":"https://submission.springernature.com/new-submission/44400/3","title":"npj Dementia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"87651038-31e1-4b3d-9fd6-ba67179a0387","owner":[],"postedDate":"May 7th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"109526560184165207868513892477728851016","date":"2026-05-18T23:22:15+00:00","index":33,"fulltext":""},{"type":"reviewerAgreed","content":"58590773946702051887766090321434442249","date":"2026-05-08T16:02:53+00:00","index":23,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":67437090,"name":"Health sciences/Biomarkers"},{"id":67437091,"name":"Health sciences/Diseases"},{"id":67437092,"name":"Health sciences/Neurology"},{"id":67437093,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2026-05-07T20:14:50+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-07 20:14:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9374985","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9374985","identity":"rs-9374985","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}