Plaque-RADS and Neovascularization for Risk Stratification of Carotid Stenosis: A Bimodal Imaging Study

BACKGROUND AND PURPOSE: The Carotid Plaque-RADS (Reporting and Data System) score is a comprehensive imaging-based framework for assessing carotid plaque vulnerability. Intraplaque neovascularization (IPN) is closely linked to atherosclerotic plaque progression. However, the interrelationship between Plaque-RADS and IPN, and their respective roles in patients with symptomatic carotid stenosis, remain incompletely understood. MATERIALS AND METHODS: We retrospectively enrolled 128 patients with unilateral carotid artery stenosis who underwent carotid magnetic resonance vessel wall imaging and contrast-enhanced ultrasound for integrated plaque characterization. Logistic regression was used to construct predictive models. Model performance was evaluated via the area under the receiver operating characteristic curve (AUC) with DeLong test for comparisons; net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were calculated to assess incremental predictive value. RESULTS: Patients with symptomatic stenosis had significantly higher Plaque-RADS scores (P = 0.001), IPN grades (P = 0.002), and plaque volumes (P = 0.011) than those with remote symptomatic stenosis or asymptomatic stenosis. Both Plaque-RADS score and IPN grade showed significant inverse correlations with duration since symptom onset (Plaque-RADS: ρ = –0.296, P = 0.003; IPN: ρ = –0.282, P = 0.001). Five predictive models were constructed: three single-indicator models (Plaque-RADS only, IPN only, and plaque volume only), a combined model (Plaque-RADS and IPN), and a comprehensive model incorporating Plaque-RADS, IPN, and plaque volume. The combined model achieved an AUC of 0.751 (95% CI: 0.661–0.841), significantly outperforming single-indicator models (plaque volume: 0.688, P=0.038; IPN only: 0.684, P=0.042; Plaque-RADS only: 0.644, P=0.021) but comparable to the comprehensive model (0.770; P=0.247). The NRI and IDI of the combined model showed significant incremental value over single-indicator models (all P