Introduction

Internal carotid artery stenosis (ICAS) is a leading cause of ischemic stroke and a significant contributor to vascular cognitive impairment [1, 2]. While management strategies for symptomatic stenoses are well-established, the optimal approach for asymptomatic ICAS remains contentious [3, 4]. This uncertainty stems, in part, from an incomplete understanding of the underlying pathophysiological processes that lead to brain injury in the absence of overt clinical symptoms [5, 6]. Specifically, a key unresolved dilemma centers on the precise mechanisms linking macrovascular hemodynamic impairment to downstream diffuse white matter injury and cognitive decline.

The pathophysiology of cerebral injury in ICAS is multifactorial, involving a complex interplay between macro- and microvascular pathways [6, 7]. The stenotic lesion itself causes hemodynamic impairment, leading to chronic cerebral hypoperfusion and a deleterious metabolic state [8]. Concurrently, the systemic atherosclerotic process—driven by established vascular risk factors such as hypertension, diabetes, hyperlipidemia, and smoking—not only underlies the index macrovascular stenosis but also directly promotes cerebral small vessel disease (CSVD) [9,10,11,12]. These microvascular insults promote endothelial dysfunction, vessel wall stiffening, and impaired autoregulation, creating a diffuse substrate for parenchymal injury [9].

We propose that a key, yet insufficiently explored, consequence of this dual macro- and microvascular pathology may be a disruption of cerebral interstitial fluid (ISF) homeostasis, primarily mediated through impairment of the glymphatic system—a brain-wide perivascular network essential for ISF clearance and metabolic waste removal [13]. Glymphatic dysfunction has been increasingly implicated in the pathophysiology of various cerebrovascular diseases, including CSVD and Moyamoya disease [14,15,16]. Importantly, ISF dysregulation is not merely a passive marker of vascular injury; reflecting impaired glymphatic clearance, it may act as an active pathophysiological driver that exacerbates white matter vulnerability by facilitating neuroinflammation, compromising metabolic waste clearance, and disrupting fluid homeostasis, thereby promoting the initiation and progression of white matter hyperintensities (WMH) [14, 17, 18]. In the setting of carotid stenosis, diminished arterial pulsatility—a key driver of glymphatic flow—has been shown in animal models to lead to glymphatic dysfunction [19]. Thus, in ICAS, the convergence of macrovascular hemodynamic compromise and microvascular injury may synergistically impair glymphatic function, resulting in ISF accumulation.

Critically, advances in diffusion MRI now allow in vivo quantification of ISF accumulation via the free water in white matter (FW-WM) fraction. This imaging biomarker provides a unique, non-invasive window into glymphatic dysfunction and early tissue damage, potentially preceding overt structural changes [20,21,22].

Based on this rationale, we hypothesize that elevated FW-WM serves as a pivotal mechanistic link, connecting combined macro- and microvascular pathology to diffuse brain injury in asymptomatic extracranial ICAS. To test this hypothesis and elucidate the pathophysiological relationships, we employed a cross-sectional multimodal MRI study with the following objectives: (1) to characterize the dual pathology of focal hemodynamic compromise and diffuse microstructural injury; (2) to investigate FW-WM as a central biomarker connecting these pathological processes; and (3) to evaluate the association of these imaging alterations on cognitive function.

Materials and methods

Participants

This prospective, observational study received approval from the Ethics Committee of Jining NO.1 People’s Hospital Affiliated to Shandong First Medical University (Approval No: 2025-IIT-kuai019). All investigative procedures were conducted in strict adherence to the principles outlined in the Declaration of Helsinki and relevant national ethical guidelines. Written informed consent was obtained from every participant prior to their enrollment in the study.

Between January 2025 and October 2025, we prospectively screened and recruited consecutive patients diagnosed with unilateral, asymptomatic, severe extracranial ICAS from the Departments of Neurosurgery and Neurology at our institution. The inclusion criteria were meticulously defined as follows: (1) radiographic confirmation of unilateral severe stenosis (≥ 70%) or complete occlusion at the origin of the internal carotid artery, as per the criteria established by the North American Symptomatic Carotid Endarterectomy Trial (NASCET); (2) an asymptomatic clinical status, defined by the absence of any history of cerebrovascular events (e.g., stroke) or transient ischemic attacks (TIAs) within the 6 months preceding enrollment, and no objective evidence of focal neurological deficits upon comprehensive clinical examination. It is noteworthy that the presence of non-specific symptoms such as dizziness or headache, or the incidental radiological identification of silent old infarct foci, were not grounds for exclusion; (3) completion of a comprehensive multimodal MRI protocol, as detailed below. Exclusion criteria comprised: (1) acute ischemic stroke and TIA presentation; (2) co-existent stenosis (> 30%) in the contralateral internal carotid artery or any other hemodynamically significant stenosis in major intracranial arteries; (3) prior neurosurgical interventions; (4) pre-existing major neurological or psychiatric comorbidities, such as neurodegenerative disorders (e.g., Alzheimer’s disease) and major depressive disorder that could independently affect cognitive function or brain structure; (5) diagnosis of non-atherosclerotic vasculopathies, such as vasculitis or moyamoya disease; (6) poor image quality rendering quantitative analysis unreliable; (7) MRI contraindications. Overall, 40 patients with unilateral severe asymptomatic ICAS were included in the study. And 40 HCs matched for age and gender were recruited according to the following criteria: (1) no history of neurological, psychiatric, or cognitive disorders; (2) no clinically significant carotid or intracranial artery stenosis; (3) no MRI contraindications.

Clinical and neuropsychological assessments

A systematic and standardized approach was employed to collect demographic and clinical data from all participants. This included: (1) age and gender; (2) detailed documentation of vascular risk factors, including hypertension, diabetes mellitus, coronary artery disease, plasma homocysteine levels, and history of smoking and alcohol consumption. Cognitive function was specifically evaluated in the ICAS patient cohort using the Montreal Cognitive Assessment (MoCA).

MRI protocol

All MRI data were acquired on a 3.0 T Philips Elition scanner with a 32-channel head coil. The imaging protocol included the following sequences:

3D-T1-Weighted Imaging (3D-T1WI): Sagittal acquisition; TR/TE = 6.6/3.0 ms; slice thickness = 1 mm; matrix = 240 × 240; FOV = 240 × 240 × 170 mm³; voxel size = 1.0 × 1.0 × 1.0 mm³; slices = 170; number of signal averages (NSA) = 1.

3D-T2-FLAIR: Sagittal acquisition; TR/TE = 4800/389 ms; slice thickness = 2 mm; matrix = 356 × 357; FOV = 250 × 250 × 150 mm³; voxel size = 0.7 × 0.7 × 2 mm³; slices = 150; NSA = 2.

Diffusion Tensor Imaging (DTI): Axial acquisition using spin-echo echo-planar imaging with 128 diffusion directions; b-values = 0, 1000 s/mm²; TR/TE = 2699/72 ms; slice thickness = 2.2 mm; matrix = 104 × 100; FOV = 224 × 224 × 145 mm³; voxel size = 2.2 mm³; slices = 66; NSA = 1.

Arterial Spin Labeling (ASL): 3D pseudo-continuous arterial spin labeling (pCASL) using Gradient and Spin echo (GRASE) readout and two PLDs (1500 ms and 2500 ms); TR/TE = 3642/12 ms (PLD = 1500) and 4642/12 ms (PLD = 2500); slice thickness = 6 mm; matrix = 64 × 62; FOV = 240 × 240 × 120 mm³; voxel size = 3.75 × 3.75 × 6 mm³; slices = 20; NSA = 1. The pCASL labeling duration was 1800 ms. In total, 8 dynamics were acquired. For the first dynamic, no labeling was applied, resulting in pure M0 images. The remaining 7 dynamics were used to acquire control-label pairs. A background suppression pulse set with 4 pulses (applied at 629, 1753, 2612, and 3099 ms for PLD = 1500 ms; and at 939, 2428, 3494, and 4073 ms for PLD = 2500 ms) was used to suppress the background tissue.

Image processing and analysis

WMH segmentation and calculation

Utilizing the uAI Intelligent Imaging Analysis System, a two-dimensional deep learning model based on the VB-Net algorithm was used for the segmentation and volumetric measurement of WMH on 3D-T2FLAIR images [23].

FW and FW-Corrected fractional anisotropy (fwcFA) mapping

The FW fraction was computed using well-established scripts from the Mark VCID project (https://markvcid.partners.org/markvcid2-protocols-resources). Methodologically, this approach quantifies free water content by fitting the bi-tensor free-water elimination model, which decomposes the diffusion MRI signal into two distinct compartments: an isotropic compartment representing fast-diffusing free water and an anisotropic compartment representing tissue diffusion. This process concurrently generates both an FW map and a free water-corrected Fractional Anisotropy (fwcFA) map [24]. Initially, standard FA maps were generated for each subject using FSL’s dtifit tool. These individual FA maps were then co-registered to the standard FSL FA template (FMRIB58_FA_1mm) in MNI space through a combination of linear and non-linear transformations. The resulting transformation parameters were subsequently applied to the native-space FW and fwcFA maps to bring them into the standardized MNI space. To restrict the analysis specifically to white matter (WM), a WM mask was created by thresholding the FSL FA template at a value of 0.3. This WM mask was then applied to each registered FW and fwcFA map, ensuring that subsequent calculations pertained exclusively to white matter voxels. Finally, the mean values of FW and fwcFA within this WM mask were extracted for group-level analysis (Fig. 1).

Fig. 1
Fig. 1
Full size image

The processing workflow of free water in white matter (FW-WM) and free water-corrected fractional anisotropy (fwcFA) maps

ASL Data processing and analysis (Fig. 2)

Fig. 2
Fig. 2
Full size image

ASL data processing workflow

Cerebral blood flow (CBF) quantification and subsequent analyses were performed using a combination of Advanced Normalization Tools (ANTs) [25] and SPM12. The processing pipeline encompassed image preprocessing, region of interest (ROI) definition, and hemodynamic parameter extraction, as outlined below.

Image preprocessing

For each participant, CBF maps were first calculated from the acquired ASL data at PLD 1500 ms and 2500 ms based on the general kinetic model [26]: The formula used was:

\(\>{\rm{CBF = }}{{6000 \cdot \lambda \cdot {\rm{(S}}{{\rm{I}}_{{\rm{control}}}}{\rm{ - S}}{{\rm{I}}_{{\rm{label}}}}{\rm{)}} \cdot {{\rm{e}}^{{{{\rm{PLD}}} \over {{{\rm{T}}_{{\rm{1,blood}}}}}}}}} \over {2 \cdot {\rm{\alpha}} \cdot {{\rm{T}}_{{\rm{1,blood}}}} \cdot {\rm{S}}{{\rm{I}}_{{\rm{PD}}}} \cdot {\rm{(1 - }}{{\rm{e}}^{{\tau \over {{{\rm{T}}_{{\rm{1,blood}}}}}}}}{\rm{)}}}}\left[ {{\rm{ml/100g/min}}} \right]\)  

Where: λ = brain-blood partition coefficient (0.9 mL/g); SIcontrol and SIlabel = signal intensities of control and label images; T1, blood = longitudinal relaxation time of blood (1350 ms); α = labeling efficiency (0.85); SIPD = proton density‑weighted signal; τ = labeling duration (1800 ms). PLD = post‑labeling delay (1500 or 2500 ms).

The preprocessing of derived CBF maps involved three sequential steps:

(1) Spatial Coregistration: Individual CBF maps were first coregistered to their corresponding T1-weighted structural images, thereby aligning them to a common anatomical space and accounting for any subject motion between the ASL and T1 acquisitions.

(2) Spatial Normalization: The symmetric normalization (SyN) algorithm implemented in ANTs was used to perform nonlinear registration of each subject’s T1-weighted image to the Montreal Neurological Institute (MNI152) standard T1-weighted template. The resulting transformation was subsequently applied to warp the CBF maps into MNI standard space.

(3) Tissue-Specific ROI Masking: A standardized cerebrovascular territory atlas [27] in MNI space was resampled to the study template resolution. To enable separate analysis of cortical and subcortical perfusion, the atlas regions for the anterior (ACA), middle (MCA), and posterior (PCA) cerebral arteries were restricted to gray matter (GM) and white matter (WM) by intersection with binary tissue masks. These masks were generated by thresholding (> 0.3) the probabilistic tissue maps provided in the TPM folder of SPM12 [28]. Subsequent CBF analyses were performed as follows: for the 1500 ms PLD, analyses were conducted exclusively within the GM masks; for the 2500 ms PLD, analyses were carried out separately for the GM and WM compartments.

Hemodynamic parameter extraction

For each GM- and WM- restricted vascular territory ROI, two key parameters were calculated from the normalized CBF maps:

Mean CBF: The average of all voxel-wise CBF values within the ROI.

Spatial Coefficient of Variation (sCoV): To quantify intra-regional perfusion heterogeneity, sCoV was defined as the ratio of the standard deviation to the mean CBF within the ROI [29]:

$$\>{\rm{sCoV = }}{{{\rm{\sigma CBF}}} \over {{\rm{\mu CBF}}}}$$

where σ and μ represent the standard deviation and the mean CBF within the ROI, respectively.

Lateralization index calculation

To robustly quantify the asymmetry of cerebral perfusion resulting from the unilateral stenosis, a Lateralization Index for CBF (LI-CBF) was calculated for the ACA, MCA and PCA territories using the following formula[ 6]:

$${\rm{Lateralization - Index = }}{{{\rm{2*(right\ or\ ipsilateral - left\ or\ contralateral)}}} \over {{\rm{(right\ or\ ipsilateral + left\ or\ contralateral)}}}}$$

Statistical analysis

All analyses were performed using IBM SPSS Statistics (v25.0, Armonk, NY). Data are presented as mean ± standard deviation for normally distributed variables, median with interquartile range (Q1, Q3) for non-normally distributed variables, and as frequency (percentage) for categorical variables. The Shapiro–Wilk test was used to assess normality.

Group comparisons between patients and healthy controls were conducted using independent samples t‑tests or Mann–Whitney U tests for continuous variables, and the chi‑square test or Fisher’s exact test for categorical variables. Within‑patient hemispheric comparisons were performed with paired t‑tests or Wilcoxon signed‑rank tests.

To test the hypothesis that FW-WM integrates focal hemodynamic impairment with diffuse white matter injury, multivariable linear regression analyses were performed. In these models, FW-WM served as the dependent variable, with LI-CBF-ACA-GM (hemodynamic marker) and either WMH volume or fwcFA (white matter injury markers) as independent variables. All models were adjusted for the covariates age, sex, hypertension, diabetes, and plasma homocysteine level.

Partial correlation analyses, controlling for age, sex, hypertension, diabetes, and homocysteine (plus education for cognitive analyses), were used to examine associations among neuroimaging variables and with MoCA scores within the patient group.

A two‑tailed p‑value < 0.05 was considered statistically significant.

Results

Participant characteristics of the study cohorts

40 patients with asymptomatic unilateral severe ICAS and 40 age- and sex-matched HCs were included. As summarized in Table 1, the two groups were comparable in terms of age, sex, smoking, alcohol consumption, and the prevalence of diabetes and coronary disease (all p > 0.05). However, as anticipated in a population with significant atherosclerotic disease, the patient group had a significantly higher prevalence of hypertension (70.0% vs. 47.5%, p = 0.041) and a greater proportion of individuals with elevated homocysteine levels (50.0% vs. 27.5%, p = 0.039). The global cognitive performance of the patient cohort, as measured by the MoCA, showed notable impairment (mean score 18.85 ± 4.17), consistent with the spectrum of vascular cognitive impairment.

Table 1 Demographic and clinical features of patients group and HCs
Full size table

Imaging findings of the study cohorts

Compared to HCs, patients exhibited significant alterations across multiple neuroimaging metrics (Table 2). Patients exhibited significantly elevated FW-WM (0.20 ± 0.03 vs. 0.18 ± 0.02, p < 0.001) and reduced fwcFA (0.45 ± 0.014 vs. 0.46 ± 0.013, p = 0.017), alongside a greater burden of WMH (median volume: 4.41 cm³ vs. 2.43 cm³, p = 0.003). Perfusion asymmetry, as indicated by the LI-CBF, was more pronounced in patients. This was evidenced by significantly lower values in the ACA and MCA GM territories at a PLD of 1500 ms (p = 0.003 and p = 0.001, respectively), as well as in the MCA GM territory at a PLD of 2500 ms (p = 0.040). A significant reduction was also observed in LI-CBF-ACA-WM at the 2500 ms PLD (p = 0.005). In contrast, LI-CBF values in the PCA territories did not differ significantly between groups at either PLD (all p > 0.05). Representative examples of these CBF patterns are shown in Supplementary Figure S1.

Table 2 Neuroimaging characteristics of patients group and HCs
Full size table

Analysis of spatial coefficient of variation (sCoV) at PLD 1500 ms

To further assess the spatial heterogeneity of perfusion, potentially reflecting differences in arterial transit efficiency, we calculated the spatial coefficient of variation (sCoV) of CBF within the grey matter of each vascular territory for the PLD 1500 ms data. Within the patient group, sCoV was significantly higher in the ipsilateral hemisphere compared to the contralateral hemisphere in both the ACA-GM (median ipsilateral: 39.86% vs. contralateral: 37.07%, p < 0.001) and MCA-GM (median ipsilateral: 42.58% vs. contralateral: 33.61%, p < 0.001) territories. And no significant interhemispheric difference was observed in the PCA-GM territory (p = 0.872) (Table 3).

Table 3 The results for sCoV within patients group
Full size table

Within-group comparisons, correlations, and regression models

An analysis within the patient group revealed a distinct dual pattern of pathology between the ipsilateral and contralateral hemispheres (Table 4). Focal alterations were observed ipsilaterally, characterized by significantly higher FW-WM (p = 0.002) and lower CBF in the ACA and MCA GM at a PLD of 1500 ms (p = 0.001 and p < 0.001, respectively). These perfusion deficits persisted in the MCA territory at a PLD of 2500 ms, affecting both GM and WM (both p < 0.001). In contrast, neither the WMH volume nor the fwcFA showed significant interhemispheric differences (p = 0.253 and p = 0.249, respectively), suggesting these measures reflect a more diffuse pathological process.

Table 4 Neuroimaging characteristics within patients group
Full size table

Partial correlation analyses, controlling for age, sex, hypertension, diabetes, and homocysteine, revealed significant interrelationships among neuroimaging metrics. Within the ipsilateral hemisphere, FW-WM was positively correlated with WMH (r = 0.521, p = 0.001) but inversely correlated with LI-CBF-ACA-GM at PLD 1500ms and 2500ms (r = -0.464, p = 0.005 and r = -0.416, p = 0.013, respectively). At the global level, FW-WM correlated positively with total WMH volume (r = 0.344, p = 0.043) and showed a strong negative correlation with fwcFA (r = -0.599, p < 0.001).

To further investigate the independent contributions of hemodynamic impairment and white matter injury to FW-WM, we performed multivariable linear regression analyses at both the ipsilateral and global levels. Four models were constructed with FW-WM as the dependent variable, and LI-CBF-ACA-GM and white matter injury (either WMH or fwcFA) as independent variables, adjusted for age, sex, hypertension, diabetes, and homocysteine.

At the ipsilateral level, in the model using PLD 1500ms, ipsilateral FW-WM showed a significant positive association with WMH volume (β = 0.403, p = 0.012) and a borderline significant negative association with LI-CBF-ACA-GM (β = -0.319, p = 0.047). Similarly, in the model using PLD 2500ms, ipsilateral FW-WM was significantly positively associated with WMH volume (β = 0.426, p = 0.009) but not significantly associated with LI-CBF-ACA-GM (β = -0.277, p = 0.090).

At the global level, global FW-WM was not significantly associated with LI-CBF-ACA-GM in either the PLD 1500ms model (β = -0.158, p = 0.326) or the PLD 2500ms model (β = -0.113, p = 0.508). However, it demonstrated strong and significant negative associations with global fwcFA in both models (PLD 1500ms: β = -0.574, p = 0.002; PLD 2500ms: β = -0.590, p = 0.003).

Association with cognitive function

After controlling for age, sex, and education, worse cognitive performance on MoCA was significantly associated with higher global FW-WM (r = -0.435, p = 0.007). Better MoCA scores showed a trend toward correlation with higher LI-CBF-ACA-GM at PLD 1500ms (r = 0.287, p = 0.085). Better MoCA scores showed a trend toward correlation with lower global WMH volume (r = -0.296, p = 0.075). No significant correlation was found between fwcFA and MoCA scores (r = 0.202, p = 0.231). Critically, the significant association between elevated global FW-WM and worse MoCA performance persisted after adjusting for LI-CBF-ACA-GM (partial r = -0.363, p = 0.029). The significant association between elevated global FW-WM and worse MoCA performance persisted after adjusting for global WMH volume (partial r = -0.361, p = 0.030).

Discussion

This multimodal MRI study delineates a dual-pathology model in asymptomatic extracranial ICAS, characterized by focal hemodynamic impairment and diffuse white matter injury. Within this framework, FW-WM emerges as an integrative imaging biomarker, concurrently associated with macrovascular hypoperfusion, microstructural white matter damage, and cognitive decline. Its independent relationship with cognitive performance—after adjusting for established markers such as WMH volume or perfusion asymmetry—suggests that FW-WM may reflect a distinct and potentially dynamic pathological process underlying early brain injury in ICAS.

Our findings of lateralized hemodynamic impairment align with prior reports in high-grade carotid stenosis [30]. Notably, after adjusting for covariates, ipsilateral FW-WM was specifically associated with grey matter LI-CBF, but not with white matter LI-CBF, a discrepancy likely attributable to the low signal-to-noise ratio of ASL for quantifying white matter at conventional PLDs [31]. Multivariable regression confirmed the independent association of ipsilateral FW-WM with perfusion asymmetry at the shorter PLD of 1500 ms. This perfusion heterogeneity at the shorter PLD was further supported by our analysis of the sCoV at PLD 1500 ms, which revealed significantly higher spatial heterogeneity of perfusion within the ipsilateral ACA and MCA grey matter territories compared to the contralateral side, suggesting impaired arterial transit efficiency. The differences in LI-CBF effect sizes between PLD 1500 ms and PLD 2500 ms may reflect not only genuine perfusion deficits but also methodological sensitivity to arterial transit time (ATT) delays, which are common in severe stenosis [32, 33]. This hemodynamic inefficiency may signal early microvascular dysfunction, including impaired arteriolar compliance, diminished vascular reactivity, and altered pulsatility—factors that could predispose to ISF dysregulation by compromising blood-brain barrier integrity and glymphatic clearance [34,35,36]. Although our study cannot establish causality, the association between perfusion asymmetry (at short PLD) and elevated FW-WM supports the hypothesis that inefficient macro- to microvascular flow transition may be an early contributor to ISF accumulation. We found no significant association between anterior circulation CBF lateralization and global cognitive function (MoCA), highlighting the need for future studies to employ domain-specific neuropsychological assessments. Ultimately, longitudinal studies combining multi-PLD ASL for direct ATT quantification with free-water imaging are needed to establish the temporal and mechanistic links between vascular delivery efficiency, ISF dysregulation, and cognitive decline in ICAS.

Notably, the spatial distribution of these pathological markers reveals a sophisticated pattern of cerebral injury in extracranial ICAS: FW-WM showed a lateralized increase in the hemisphere ipsilateral to the stenosis, whereas WMH volume and fwcFA displayed bilateral symmetry. This dissociation may be understood by considering distinct yet overlapping pathogenic pathways [6, 7]. Ipsilateral FW-WM likely reflects acute, perfusion-driven ISF accumulation resulting directly from hemodynamic compromise downstream of the stenosis. In contrast, the bilateral nature of WMH and reduced fwcFA points toward a chronic, systemic process consistent with diffuse small vessel disease, characterized by arteriolosclerosis, endothelial dysfunction, and persistent blood-brain barrier leakage—processes often amplified by age and vascular risk factors [10,11,12,37]. Thus, although both processes originate from underlying atherosclerotic pathology, they differ in their primary mechanisms and spatiotemporal evolution: FW-WM appears more sensitive to acute, hemodynamically mediated fluid dysregulation, whereas WMH volume predominantly signifies chronic, cumulative parenchymal injury. This mechanistic distinction clarifies why FW-WM exhibits a lateralized pattern closely linked to the stenosis, while WMH manifests as a bilateral burden typical of diffuse microangiopathy.

Building on evidence linking free water accumulation to perivascular space dilation and WMH progression in CSVD [14, 38], this study aimed to identify the determinants of FW-WM in asymptomatic extracranial ICAS. Multivariable linear regression revealed that white matter injury—whether measured as WMH volume or reduced fwcFA—constituted the strongest independent correlate of FW-WM. Although univariate analysis indicated a correlation between lateralized CBF reduction and FW-WM, this association was attenuated after adjusting for WMH burden (β = -0.319, p = 0.047). Similarly, at the global level, perfusion asymmetry showed negligible independent contribution to FW-WM when fwcFA was included in the model. These results suggest that hemodynamic impairment influences FW-WM largely through its effect on white matter integrity. Thus, FW-WM may serve as an imaging phenotype that more directly reflects parenchymal compromise, linking hemodynamic stress to subsequent structural damage.

A key finding of this study is that elevated global FW-WM was associated with poorer cognitive performance on the MoCA, and this association remained significant after separately adjusting for either WMH volume or perfusion asymmetry. This suggests that FW-WM captures a pathological process distinct from these conventional markers, potentially explaining the documented discrepancy between traditional imaging findings and cognitive status in asymptomatic ICAS [39, 40]. Furthermore, although WMH burden is strongly age-dependent and encompasses mixed vascular and non-vascular etiologies, the independent link between FW-WM and cognition—even after controlling for WMH—implies that FW-WM reflects pathology beyond age-related white matter changes. Specifically, while WMH largely represents chronic, irreversible tissue injury characteristic [41, 42], FW-WM may reflect dynamic fluid dysregulation and glymphatic impairment [14, 43, 44], which could more directly influence neural function and cognitive networks. Consequently, the measurement of FW-WM holds promise for the identification of so-called “asymptomatic” patients who exhibit significant cerebral microenvironment disruption, indicating a potentially higher risk of progressive cognitive decline.

This study has several limitations. First, its cross-sectional design precludes causal interpretation of the observed associations. Longitudinal studies are needed to determine whether elevated FW-WM precedes white matter injury and cognitive decline, and to formally test its potential mediating role. Second, although the sample size was adequate for primary analyses, it limited subgroup explorations based on stenosis severity or occlusion status, and future studies with larger and more balanced samples are warranted to explore pathophysiological differences across the spectrum of carotid disease severity. Third, The generation of gray and white matter masks through thresholding of probabilistic tissue maps does not completely account for partial volume effects. Residual partial volume contamination, particularly at the cortical boundaries and in periventricular regions, may influence the absolute quantification of white matter CBF and the precise interpretation of gray matter-white matter perfusion contrast. Nevertheless, since these effects are likely to be symmetrical across hemispheres in our cohort, they are not expected to substantially bias the primary study outcomes concerning the lateralization of perfusion. Fourth, we did not systematically evaluate Circle of Willis variants, which are crucial for understanding individual posterior circulation vulnerability and reserve. Finally, we did not perform sub-territorial analyses to correlate local perfusion asymmetry with local WMH burden within specific vascular territories. Future studies employing spatially refined segmentation—separating FW within WMH from normal-appearing white matter and mapping lesions to individual vascular sub-territories—would help to disentangle these contributions and elucidate more precise structure-function relationships in carotid stenosis.

Conclusion

In summary, this study identifies FW-WM as an integrative imaging biomarker that captures both focal hemodynamic impairment and diffuse microstructural injury in asymptomatic extracranial ICAS. Its independent association with cognitive decline positions FW-WM as a key pathophysiological link between carotid stenosis and early clinical dysfunction. Prospective multimodal studies are warranted to validate its prognostic utility in stratifying the risk of progressive white matter injury and cognitive decline in this population.