Radiomics-based machine learning model for predicting clinically ineffective reperfusion in acute ischaemic stroke patients after endovascular treatment.

Background: Patients with acute ischaemic stroke (AIS) undergoing endovascular treatment may have a poor prognosis, even with successful recanalization. This study aims to evaluate a machine learning model based on CT-thrombosis radiomics to assess clinically ineffective reperfusion (CIR) after endovascular treatment (EVT) in patients with AIS. Methods: A total of 144 patients from two centres were included in this study, spanning from December 2021 to October 2024. The participants were randomly divided into a training set (70%) and a test set (30%). Patient outcomes were defined as clinically ineffective reperfusion (thrombolysis in cerebral infarction, TICI ≥2b, three-month post-surgery modified Rankin Scale, mRS ≥3) and effective reperfusion (TICI ≥2b, three-month post-surgery mRS <3). A total of 1,702 features were extracted from the intrathrombus and perithrombus regions. The minimum redundancy maximum relevance (mRMR) and least absolute shrinkage and selection operator (LASSO) algorithm were used for feature selection to construct the machine learning model, with the AUC of the receiver operating characteristic (ROC) curve used for model evaluation. Results: In the test set, the random forest (RF) model demonstrated the highest diagnostic performance among all the models (RF_INTRA AUC = 0.78, RF_PERI AUC = 0.76, RF_F AUC = 0.83). Conclusion: The machine learning model based on intrathrombus and perithrombus radiomics features can accurately predict clinically ineffective reperfusion in patients after EVT. However, further study is needed to validate these findings in larger, independent cohorts and explore the broader clinical applicability of the model.