基于增强CT影像组学与机器学习算法预测肿块型肝内胆管癌微血管侵犯OACSTPCD

Objective To study the radiomics and clinical characteristics of intratumoral microvascular invasion in mass-forming intrahepatic cholangiocarcinoma(ICC)and establish a predictive model with the best machine learning algorithm.Methods We retrospectively collected the data from 75 patients with ICC who were treated at the First Affiliated Hospital of Wenzhou Medical University and 29 patients with ICC who were treated at the Second Affiliated Hospital of Wenzhou Medical University between Jan.2015 and Feb.2023.Radiomic features were extracted from contrast-enhanced CT images,and various machine learning methods were used for analysis.The best radiomics machine learning method was combined with clinical data to establish a predictive model,which was then validated.Results Among various radiomic machine learning methods,naive Bayes classification based on portal venous phase imaging features performed relatively well,the area under the cure(AUC)was 0.818.By combining two selected clinical features(tumor arterial penetration,CEA＞5 ng/mL),a predictive model was established with the AUC of 0.883 in the training group and 0.891 in the testing group.In the training group,sensitivity was 0.978,while specificity was 0.656,and in the testing group,sensitivity was 0.909,while specificity was 0.700.Conclusion The model based on radiomic machine learning combined with clinical data from contrast-enhanced CT has good diagnostic value for predicting microvascular invasion status in ICC.