Prediction model for vasospasm via snuffbox distal radial artery approach
Related Experiment Video
Updated: Aug 5, 2026
A Volumetric Method for Quantification of Cerebral Vasospasm in a Murine Model of Subarachnoid Hemorrhage
Published on: July 28, 2018
Lei Wang1, Dan Liu1, Hao Zhang1
1Department of Cardiology, Shihezi People's Hospital, Shihezi, China.
Objective:
To integrate demographic, preoperative ultrasonographic (anatomical and hemodynamic), and intraoperative variables for developing and validating a predictive model of vascular spasm associated with the distal transradial access (dTRA) via the anatomical snuffbox.
Methods:
A retrospective analysis was performed on 340 patients who underwent dTRA interventional procedures (January 2022-January 2026). Patients were randomly allocated to training (n = 238) and validation (n = 102) cohorts at a 7:3 ratio. In the training cohort, 58 (24.4%) developed vascular spasm versus 180 (75.6%) without. Univariate analysis and LASSO regression were applied for variable selection. Three machine learning models-multivariate logistic regression, random forest, and support vector machine-were subsequently constructed. Model performance was assessed using the area under the receiver operating characteristic curve calibration curves (AUC) and decision curve analysis. A flow-based nomogram was established, and SHapley Additive exPlanations (SHAP) analysis was performed for model interpretation.
Results:
Baseline characteristics were comparable between training and validation sets (all p > 0.05). Univariate analysis identified seven factors associated with dTRA‑related vasospasm (p < 0.05). LASSO and multivariate logistic regression confirmed that advanced age, male sex, and larger radial artery diameters were independent protective factors, while higher peak systolic velocity, more puncture attempts, and larger sheath size were independent risk factors (all p < 0.05). The random forest model performed best: training AUC = 0.839 (95% confidence interval (CI): 0.763-0.914), validation AUC = 0.773 (95% CI: 0.668-0.878), with good calibration and net clinical benefit. SHAP analysis showed that puncture attempts and age were the most influential variables, with feature contributions consistent with regression results.
Conclusion:
The multimodal random forest model, which integrates demographic, vascular ultrasonographic, and procedural variables, effectively identifies patients at high risk of dTRA-related vascular spasm, providing an objective quantitative tool for preoperative risk stratification and individualized surgical planning.
