Identification and scoring of factors for ischemic stroke related to cervical arterial dissection: a high-resolution magnetic resonance imaging study.

Background: Cervical arterial dissection (CAD) is the primary cause of ischemic stroke among young and middle-aged individuals. This study aimed to identify the high-resolution magnetic resonance imaging (HR-MRI) features associated with ischemic stroke in patients with CAD, thereby providing a foundation for future multi-center validation studies. Methods: This study retrospectively enrolled patients diagnosed with CAD who underwent HR-MRI. CAD was classified into two types: type I lesions, which exhibit only intramural hematoma, and type II lesions, characterized by the presence of an intimal flap and/or double lumen, which may be accompanied by other dissection imaging manifestations. To account for within-patient correlation from multiple lesions, generalized estimating equations (GEE) with an exchangeable working correlation structure were used to compare clinical and imaging characteristics between stroke and non-stroke groups, and between type I and type II lesions. Univariable and multivariable GEE analyses identified factors independently associated with stroke. Using these identified factors, we developed a straightforward score. The discriminative performance of the total score was evaluated through receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). The incremental value of multi-index assessment compared to a single index was calculated by using net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Internal validation was performed with patient-level bootstrap resampling (B=1,000) to obtain optimism-corrected estimates. Finally, stroke incidence was compared across score groups with Bonferroni correction. Results: This analysis included a total of 108 patients, with 125 affected vessels. Type I was more frequently associated with moderate to severe stenosis, and the incidence of intraluminal thrombus and stroke was higher compared to Type II. Multivariate analysis indicated that type I [odds ratio (OR) =3.261, P=0.033], moderate to severe stenosis (OR =3.404, P=0.006), and intraluminal thrombus (OR =3.413, P=0.036) were independent factors for stroke in patients with CAD. The total score integrating these three features showed good discriminative performance for ischemic stroke, with an optimism-corrected area under the curve (AUC) of 0.747 [95% confidence interval (CI): 0.678-0.810], accuracy =0.711, and sensitivity =0.781. Compared with type I alone, combining type I with intraluminal thrombus and moderate-to-severe stenosis significantly improved the assessment of CAD-related stroke [NRI (optimism-corrected) =0.776, 95% CI: 0.533-1.026, P<0.001, IDI (optimism-corrected) =0.107, 95% CI: 0.071-0.142, P<0.001]. Furthermore, as the total score increased, the incidence rate of stroke demonstrated a significant upward trend. Conclusions: The presence of type I, moderate to severe stenosis, and intraluminal thrombus were closely associated with the occurrence of CAD-related stroke. The exploratory score constructed based on these characteristics still requires further external validation before being used for assessing CAD-related stroke.