Prediction of hemorrhagic transformation via pre-treatment CT radiomics in acute ischemic stroke patients receiving endovascular therapy
Xuehua Wen, Yanan Xiao, Xingfei Hu, Junfa Chen, Fahuan Song
Abstract
Objective: This study aimed to extract radiomics features (RFs) from pre-treatment CT scans in patients with acute ischemic stroke (AIS), and to establish a radiomics model to predict hemorrhagic transformation (HT) after endovascular therapy (EVT). Methods: A total of 105 patients who were diagnosed with AIS [with occlusion of the M1 segment of the middle cerebral artery (MCA) and/or internal carotid artery] and received EVT were enrolled. They were randomly divided into the development cohort (n = 73) and the validation cohort (n = 32). The clinicoradiological data of all patients, including pre-treatment cranial CT without contrast enhancement, CT perfusion, and CT angiography, were obtained. The MCA territory on pre-treatment CT images was segmented to extract RFs associated with HT after EVT. Then, a CT radiomics model based on the selected RFs was constructed to predict HT after EVT. Results: The sensitivity, specificity, and area under the curve of the CT radiomics model for predicting HT after EVT based on pre-treatment CT RFs was 0.806, 0.649, and 0.781 (95% confidence interval (CI): 0.675–0.886), respectively, in the development cohort. The sensitivity, specificity, and area under the curve in the validation cohort was 0.625, 0.875, and 0.797 (95% CI: 0.642–0.951), respectively. Conclusion: CT radiomics analysis is a valuable tool for predicting HT in AIS patients receiving EVT. It may guide the selection of patients in practice and improve procedural safety and effectiveness. Advances in knowledge: Identification of the importance of pre-treatment CT radiomics in the prediction of HT in AIS patients after EVT.