Acute stroke risk prediction model based on dual-energy CTA-derived carotid plaque, perivascular adipose tissue characteristics, and serum lipid parameters: a dual-center study.

Abstract

Background: Acute stroke is a major global cause of mortality and disability. Accurate prediction of stroke risk is crucial for effective clinical management. This study aimed to develop a multidimensional prediction model for acute stroke using carotid plaque characteristics, lumen parameters, perivascular adipose tissue (PVAT) quantitative metrics derived from dual-energy computed tomography angiography (DE-CTA), and serum lipid biomarkers.

Methods: This retrospective dual-center study enrolled 212 patients who underwent DE-CTA and MRI between January 2023 and October 2024, comprising a training cohort (137 patients) and an external validation cohort (75 patients). Quantitative parameters including carotid plaque features (composition and intraplaque parameters), lumen metrics, PVAT quantitative indices, and serum lipid levels were collected. Patients with ipsilateral acute anterior circulation infarcts identified on MRI were classified as symptomatic (STA), and those without infarcts as asymptomatic (ATA). Variables were selected via univariate analysis and LASSO regression to construct a multivariate logistic regression model. Model performance was evaluated by ROC analysis, confusion matrix, calibration curves, and clinical decision curves, followed by external validation.

Results: External validation of the final model showed an area under the ROC curve (AUC) of 0.810, with a sensitivity of 80.8% and specificity of 65.3%, indicating robust predictive performance and good clinical applicability.

Conclusions: The multidimensional predictive model integrating DE-CTA-derived carotid plaque features, PVAT metrics, and serum lipid parameters effectively predicts acute stroke risk, providing a reliable quantitative tool for early screening and clinical intervention.