Vertebrobasilar artery geometry and quantitative high-resolution vessel wall MRI for stratifying vulnerable plaques in young and middle-aged adults with stroke risk factors

Objective To investigate cross-sectional associations of vertebrobasilar arterial geometry and quantitative high-resolution vessel-wall imaging (HR-VWI) features with an HR-VWI-defined vulnerable plaque phenotype in individuals with stroke risk factors, and to evaluate exploratory imaging models for plaque stratification. Methods This single-center cross-sectional study included 174 patients aged 18–59 years with stroke risk factors and measurable vertebrobasilar atherosclerotic plaques on HR-VWI. Patients were classified as having vulnerable or non-vulnerable plaques according to HR-VWI features: marked plaque enhancement, intraplaque hemorrhage, irregular plaque surface, or marked signal heterogeneity. Vertebrobasilar geometric parameters and quantitative plaque measurements were assessed. Age and body mass index (BMI) were included as covariates. Logistic regression and receiver operating characteristic analyses identified imaging factors associated with vulnerable plaque phenotype and compared age-and-BMI, age-and-BMI plus geometry, age-and-BMI plus quantitative HR-VWI, and combined imaging models. Results Sixty-nine patients had vulnerable plaques and 105 had non-vulnerable plaques. Vulnerable plaques were associated with greater basilar artery curvature index, larger vertebrobasilar junction angle, greater bilateral vertebral artery diameter difference, greater maximum basilar artery deviation, higher plaque burden, larger wall area, greater percentage area stenosis, and higher remodeling index. In the exploratory multivariable model, plaque burden, percentage area stenosis, remodeling index, and maximum basilar artery deviation remained associated with vulnerable plaque phenotype after adjustment for age and BMI. Development-cohort AUCs were 0.637 for age-and-BMI, 0.774 for age-and-BMI plus geometry, 0.883 for age-and-BMI plus quantitative HR-VWI, and 0.903 for combined imaging. The difference between combined imaging and quantitative HR-VWI was not significant (ΔAUC = 0.020; P = 0.162). At optimal thresholds, the combined model had higher specificity (89.52% vs. 76.19%) but lower sensitivity (78.26% vs. 86.96%). Conclusion Plaque burden, percentage area stenosis, remodeling index, and maximum basilar artery deviation were associated with the HR-VWI-defined vulnerable plaque phenotype after adjustment for age and BMI. Although combined imaging had the highest numerical AUC, adding vertebrobasilar geometry did not significantly improve discrimination beyond quantitative HR-VWI. Because predictors and phenotype were assessed during the same examination, these AUCs reflect within-examination classification rather than prediction of an independent biological or clinical endpoint. These exploratory findings should not be interpreted as predicting future posterior-circulation ischemic events.