基于CT血管造影影像组学特征构建高血脂合并急性脑梗死患者临床预后早期预测模型OA

Objective:To investigate the construction of an early clinical prognosis model for patients with hyperlipidemia and acute cerebral infarction(ACI)on the basis of the characteristics of computed tomography angiography(CTA)radiomics.Methods:A retrospective cohort study was used to select clinical and CTA imaging data of 160 patients with hyperlipidemia complicated ACI at Hengshui Hospital of Traditional Chinese Medicine from February 2021 to February 2024.They were divided into train set(112 cases)and test set(48 cases)as the ratio of 7 to 3.The score of the modified Rankins Scale(mRS)at the 3rd month after surgery was used as standard of prognosis,and patients with mRS<3 were included in favorable prognosis group(67 cases),and patients with mRS≥3 were divided into poor prognosis group(93 cases).Two physicians of the department of imaging adopted blind method to outline infarct focus of blood supply area of responsible vessels as region of interest(ROI),and to use PyRadiomics platform to extract initial features of image omics.The Lasso regression operator analysis was used to screen key characteristics to construct lab of image omics.Three models of machine learning of image omics[support vector machine(SVM),random forest(RF)and K-proximity(KNN)]were constructed on the basis of training set.The predictive efficiency of three models were verified by the test set,and the area under curve(AUC)of receiver operating characteristics(ROC)curve,and Matthews correlation coefficient(MCC)were adopted to analyze the efficiency of predictive models.Results:Twelve key charactersitics were screened from 78 extracted CTA image characteristics.In predictive accuracy,AUC and MCC of test set,these indicators of SVM model were respectively 84.0%,0.932(95%CI:0.892～0.971)and 0.732,and these indicators of RF model were respectively 0.906,0.946(0.903～0.988)and 0.803.The predictive efficiencies both of SVM model and RF model were better than that of KNN model.Conclusions:The early predictive model based on characteristics of CTA radiomics for clinical prognosis of patients with hyperlipidemia and ACI can effectively predict the risk of clinically adverse prognosis of patients with hyperlipidemia and ACI,and the predictive efficiencies of RF and SVM models are better in three constructed models.