Preoperative multimodal CT for selection of acute anterior circulation occlusion stroke patients for mechanical thrombectomy

Objective By developing and comparing various preoperative multimodal CT-based models to predict prognosis after successful recanalization in acute anterior circulation occlusive stroke, this study aimed to ascertain the prognostic assessment of patients following successful recanalization. Methods Patients with acute anterior circulation large vessel occlusion who underwent mechanical thrombectomy and achieved successful recanalization were consecutively enrolled from Sanming First Hospital Affiliated to Fujian Medical University between January 2022 and October 2024. Based on the 90-day modified Rankin Scale (mRS) scores, the patients were categorized into a favorable clinical outcome group (mRS 0–3) and an unfavorable clinical outcome group (mRS 4–6). Following the identification of statistically significant variables from the analyzed clinical and imaging data, optimal cut-off values were computed. Five multivariate logistic regression models were subsequently constructed: Clinical-Imaging (C-I), Clinical-non-Perfusion (C-NP), Clinical-non-Angiography (C-NC), Clinical-only (C), and Imaging-only (I). Models performance were compared using receiver operating characteristic (ROC) curve analysis, with comparisons based on the area under the curve (AUC), sensitivity, and specificity. Using the Delong test for model comparison and selection, a nomogram of the optimal model was developed. Model performance was subsequently assessed by means of a confusion matrix and 5-fold cross-validation. Results Of the 131 enrolled patients, 77 (58.78%) were classified into the favorable clinical outcome group and 54 (41.22%) into the unfavorable clinical outcome group. Statistically significant differences were identified in age (59.5 years), preoperative blood glucose (PBG) (7.21 mmol/L), National institutes of health stroke scale (NIHSS) (18.5), Alberta Stroke Program Early CT Score (ASPECTS) (7.5), collateral score (3.5), infarct core volume (19.25 mL), and hypoperfusion volume (180.65 mL). The AUCs for the five models were C-I Model (0.851, 95% CI: 0.785–0.917), C-NC Model (0.865, 95% CI: 0.804–0.926), C-NP Model (0.861, 95% CI: 0.798–0.923), C Model (0.713, 95% CI: 0.626–0.801) and I Model (0.772, 95% CI: 0.688–0.855). The accuracy of the optimal C-NP Model was 0.771, and the mean AUC from 5-fold cross-validation was 0.828. Conclusion The combined model revealed strong predictive power regarding outcomes following successful mechanical thrombectomy. Furthermore, the C-NP Model holds greater practical utility, as it enables clinicians to quickly assess the prognosis after a successful intervention.