The feasibility of dual-energy CT to predict the probability of symptomatic intracerebral haemorrhage after successful mechanical thrombectomy

Abstract

Aim: To study the ability of dual-energy computed tomography (DECT) after successful mechanical thrombectomy (MT) to predict symptomatic intracerebral haemorrhage (sICH) in anterior circulation acute ischaemic stroke (AIS).

Materials and methods: From June 2018 to February 2020, 102 AIS patients with DECT performed immediately after successful MT were enrolled prospectively. According to the presence of iodine contrast media extravasation (ICME) on DECT and subsequent sICH development, patients were classified into four groups. The neurological outcome was compared among groups. Imaging parameters, together with clinical factors, were investigated for sICH prediction based on a linear logistic regression model after class-imbalance resolved by Synthetic Minority Sampling Technique (SMOTE) method.

Results: Among 102 patients, patients (14.7%, 15/102) with the presence of sICH experienced worse outcomes than others without sICH (p<0.001). No case without ICME was observed with sICH development (0/102). The parameters derived from DECT have excellent performance for sICH prediction after successful MT, which is better than clinical predictive model boosted data (area under the curve [AUC]: DECT 0.87 versus clinical prediction 0.65), cross-validation results (AUC: DECT 0.87 versus clinical prediction 0.65), and original data (AUC: DECT 0.85 versus clinical prediction 0.68). By combining clinical and radiological parameters, the predictive performance for sICH could be further improved with an AUC of 0.90 (95% CI: 0.85-0.96).

Conclusions: Based on DECT parameters acquired immediately after successful MT, the present model was more efficient than the clinical model for accurate prediction of sICH. Rho and ICME volume appeared to be the best parameters for predicting sICH using DECT.