Predicting periprocedural complications risk in intracranial angioplasty and stenting from integrated high-resolution vessel wall imaging radiomics and clinical characteristics.

Purpose: To develop and validate an integrated model based on MR high-resolution vessel wall imaging (HR-VWI) radiomics and clinical features to preoperatively assess periprocedural complications (PC) risk in patients with intracranial atherosclerotic disease (ICAD) undergoing percutaneous transluminal angioplasty and stenting (PTAS).

Methods: This multicenter retrospective study enrolled 601 PTAS patients (PC+, n = 84; PC -, n = 517) from three centers. Patients were divided into training (n = 336), validation (n = 144), and test (n = 121) cohorts. All patients underwent preoperative HR-VWI (precontrast T1-weighted [T1] and postcontrast T1-weighted [T1CE] sequences). We extracted 2,396 radiomic features and selected clinical variables via multivariate logistic regression. Radiomics, clinical and integrated model were developed. Model performance was evaluated using areas under the curve (AUC) and DeLong test. Decision Curve Analysis (DCA) was used to evaluate the net benefit of each model.

Results: Age was the sole independent clinical predictor (OR = 1.06, p = 0.001). The integrated model demonstrated favorable predictive performance in the training cohort (AUC: 0.93, 95% CI [0.88, 0.96]), validation cohort (AUC: 0.87, 95% CI [0.74, 0.99]), and test cohort (AUC: 0.87, 95% CI [0.78, 0.95]). It significantly outperformed all clinical models (AUC range: 0.59-0.73; all p < 0.05) and showed performance comparable to the optimal radiomics model (T1-T1CE model; AUC range: 0.80-0.91; all p > 0.05).Notably, the DCA curve indicated that the integrated model achieved the optimal clinical net benefit across the 0-90% threshold range in the test cohort.

Conclusion: The integrated model demonstrates clinical utility for preoperative PC risk stratification in PTAS patients.