4,068 posters, 219 videos, 101 topics, 69 sessions, 14,365 authors, 6,128 institutions
ePostersLive by SciGen Technologies S.A. All rights reserved.
Alzheimer's Association International Conference 2026
July 12-15, 2026 | London, United Kingdom, and Online
4740
Automated Quantification Of Enlarged Perivascular Spaces And Associations With Demographics And Cognition
Poster Presenter
Authors
Part of Topic
BACKGROUND
•Enlarged perivascular spaces (ePVS) are a hallmark of cerebral small vessel disease and CAA.
•Impaired glymphatic clearance associated with ePVS may contribute to neuroinflammation and neurodegeneration.
•ePVS assessment relies mostly on visual review, although quantitative MRI approaches using T1- and T2-weighted imaging are being developed.
•We aimed to evaluate the reproducibility of a T1-only quantitative method and to investigate associations between ePVS burden, clinical characteristics and cognitive performance across the AD continuum.
METHODS
Data
•1433 ADNI subjects were included: 839 cognitively normal (CN), 473 with MCI and 121 with AD.
•199 additional MRI scans from 46 ADNI-2 subjects were analyzed for test-retest reproducibility (accelerated vs. non-accelerated 3DT1 scans).
•Cognitive performance was assessed using MMSE, ADAS-Cog and CDR-SB.
Image analysis (Fig. 1)
•3DT1 data were segmented with a deep-learning method1 trained to detect ePVS.
•Midbrain ePVS were additionally assessed using Frangi filtering to improve sensitivity.
•ePVS count/volume were calculated in the basal ganglia, cerebral white-matter, midbrain and whole brain.
Statistical analysis
Test-retest
•Test-retest reproducibility was assessed using Pearson correlation (r) and intraclass correlation coefficient (ICC(2,1), two-way random, absolute agreement).
•Absolute agreement was further evaluated using Bland–Altman analysis, including mean bias and 95% limits of agreement.
•For midbrain counts, where values were low and discrete, reproducibility was additionally summarized using exact and ±1 agreement.
Associations with clinical characteristics
•Baseline associations between ePVS and demographic, genetic and clinical variables were assessed.
•ePVS counts and volumes were derived for the basal ganglia (BG), white matter (WM), midbrain (MB) and whole brain (Total).
•ePVS volumes and counts were log-transformed to account for skewness (except midbrain volume due to small magnitude).
•Multivariable linear regression models included age, sex, intracranial volume (ICV), APOE ε4 carrier status and cognitive measures as predictors.
•Model performance was compared across regions and metrics using R².
•Group differences (CN, MCI, AD) were assessed using ANOVA with post-hoc tests, and effect sizes were quantified using η².
•To assess independent effects of sex and head size, ICV was residualized with respect to sex.
•Standardized regression coefficients were computed to compare relative predictor importance.
RESULTS
Test-retest (Figs. 2 and 3)
•Reproducibility was excellent for basal ganglia, white matter and total measures, with high correlations (r = 0.92-0.98, p < 0.001) and ICC values (0.89-0.97).
•Volume measures demonstrated particularly high reliability across all regions (ICC ≥ 0.79), with near-perfect agreement in volume (ICC ≈ 0.96-0.97).
•Bland-Altman analysis showed small biases and no systematic differences.
•Midbrain measures showed lower reproducibility for counts (r = 0.49, ICC = 0.47), with 38% exact agreement and 80% agreement within ±1 count, reflecting low and discrete burden, while volume retained good reliability (r = 0.79, ICC = 0.79).
Associations with clinical characteristics
Regional comparison (Fig. 4 top left)
Model performance varied markedly by region.
•Basal ganglia ePVS volume showed the strongest associations (R² = 0.17).
•Midbrain showed moderate performance (R² ≈ 0.11)
•White matter and total measures showed weak associations (R² ≈ 0.02-0.04).
•Volume and count metrics performed similarly across regions.
Age effects (Fig. 4 top right)
Age was strongly associated with BG ePVS volume (R² = 0.13, p < 0.001), with a clear positive relationship.
Clinical group differences (Fig. 4 bottom left)
Group differences were statistically significant (ANOVA p < 0.001), but effect size was very small (η² = 0.006).
•MCI showed slightly higher ePVS than CN.
•No differences were observed for AD.
In multivariable models:
•MCI vs CN remained significant (β ≈ 0.09, p < 0.001)
•AD showed no association.
Determinants of ePVS (Fig. 4 bottom right)
In the multivariate analysis using standardized coefficients:
•Age was the dominant predictor (β = 0.33).
•ICV showed a moderate positive association (β = 0.18)
•Sex showed a small independent effect after accounting for head size (β = 0.08).
•APOE ε4 carrier status showed a small negative association (β = −0.08, p < 0.05).
•Cognitive associations were weak and inconsistent:
•CDR-SB showed a small positive association (β = 0.12, p < 0.001).
•MMSE and ADAS-Cog were not associated.
CONCLUSIONS
•Although 3D T1-weighted MRI data alone may not be optimal for ePVS assessment, a T1-only quantitative method showed good reproducibility and robust associations with biological factors.
•ePVS burden is primarily driven by aging, with secondary contributions from head size and sex, and only weak associations with APOE ε4 status and cognitive measures.
•Basal ganglia volume provides the most sensitive measure in this analysis.
•Future work will investigate potential longitudinal associations.
REFERENCES
1)Huang P. et al., Development and validation of a perivascular space segmentation method in multi-center datasets, Neuroimage. 2024.
Comments
Email is required but is not shown publicly.
Loading comments...